diff --git "a/trainer_state.json" "b/trainer_state.json" deleted file mode 100644--- "a/trainer_state.json" +++ /dev/null @@ -1,198346 +0,0 @@ -{ - "best_metric": null, - "best_model_checkpoint": null, - "epoch": 0.11161450379742982, - "global_step": 33000, - "is_hyper_param_search": false, - "is_local_process_zero": true, - "is_world_process_zero": true, - "log_history": [ - { - "epoch": 0.0, - "learning_rate": 0.00019999966177365895, - "loss": 5.9733, - "step": 1 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019999932354731786, - "loss": 4.1251, - "step": 2 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999989853209768, - "loss": 3.4323, - "step": 3 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019999864709463575, - "loss": 2.977, - "step": 4 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019999830886829466, - "loss": 2.8536, - "step": 5 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999979706419536, - "loss": 2.6541, - "step": 6 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019999763241561254, - "loss": 2.528, - "step": 7 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019999729418927146, - "loss": 2.561, - "step": 8 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999969559629304, - "loss": 2.5316, - "step": 9 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019999661773658934, - "loss": 2.4726, - "step": 10 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019999627951024825, - "loss": 2.4404, - "step": 11 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999959412839072, - "loss": 2.5051, - "step": 12 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019999560305756614, - "loss": 2.3783, - "step": 13 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019999526483122505, - "loss": 2.4299, - "step": 14 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199994926604884, - "loss": 2.359, - "step": 15 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019999458837854293, - "loss": 2.3862, - "step": 16 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019999425015220185, - "loss": 2.401, - "step": 17 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999939119258608, - "loss": 2.4827, - "step": 18 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019999357369951973, - "loss": 2.3976, - "step": 19 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019999323547317864, - "loss": 2.3235, - "step": 20 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019999289724683758, - "loss": 2.4198, - "step": 21 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019999255902049652, - "loss": 2.312, - "step": 22 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019999222079415544, - "loss": 2.3964, - "step": 23 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019999188256781438, - "loss": 2.3543, - "step": 24 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019999154434147332, - "loss": 2.2414, - "step": 25 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019999120611513224, - "loss": 2.2467, - "step": 26 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019999086788879118, - "loss": 2.3079, - "step": 27 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019999052966245012, - "loss": 2.3848, - "step": 28 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019999019143610906, - "loss": 2.327, - "step": 29 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199989853209768, - "loss": 2.4258, - "step": 30 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019998951498342694, - "loss": 2.4113, - "step": 31 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019998917675708585, - "loss": 2.246, - "step": 32 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999888385307448, - "loss": 2.3391, - "step": 33 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019998850030440374, - "loss": 2.3297, - "step": 34 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019998816207806265, - "loss": 2.4049, - "step": 35 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999878238517216, - "loss": 2.3637, - "step": 36 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019998748562538053, - "loss": 2.2526, - "step": 37 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019998714739903945, - "loss": 2.3123, - "step": 38 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999868091726984, - "loss": 2.2254, - "step": 39 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019998647094635733, - "loss": 2.3191, - "step": 40 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019998613272001624, - "loss": 2.3166, - "step": 41 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019998579449367519, - "loss": 2.2293, - "step": 42 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019998545626733413, - "loss": 2.323, - "step": 43 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019998511804099304, - "loss": 2.3741, - "step": 44 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019998477981465198, - "loss": 2.3037, - "step": 45 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019998444158831092, - "loss": 2.2681, - "step": 46 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019998410336196984, - "loss": 2.2601, - "step": 47 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019998376513562878, - "loss": 2.184, - "step": 48 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019998342690928772, - "loss": 2.2409, - "step": 49 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019998308868294663, - "loss": 2.2313, - "step": 50 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019998275045660557, - "loss": 2.2668, - "step": 51 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019998241223026452, - "loss": 2.2698, - "step": 52 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019998207400392343, - "loss": 2.3098, - "step": 53 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019998173577758237, - "loss": 2.2727, - "step": 54 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999813975512413, - "loss": 2.2584, - "step": 55 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019998105932490023, - "loss": 2.2613, - "step": 56 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019998072109855917, - "loss": 2.1284, - "step": 57 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999803828722181, - "loss": 2.1589, - "step": 58 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019998004464587702, - "loss": 2.216, - "step": 59 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019997970641953596, - "loss": 2.2743, - "step": 60 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999793681931949, - "loss": 2.3149, - "step": 61 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019997902996685382, - "loss": 2.2433, - "step": 62 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019997869174051276, - "loss": 2.3639, - "step": 63 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999783535141717, - "loss": 2.2522, - "step": 64 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019997801528783061, - "loss": 2.2565, - "step": 65 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019997767706148956, - "loss": 2.1465, - "step": 66 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999773388351485, - "loss": 2.2467, - "step": 67 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999770006088074, - "loss": 2.0428, - "step": 68 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019997666238246635, - "loss": 2.2198, - "step": 69 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999763241561253, - "loss": 2.1548, - "step": 70 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999759859297842, - "loss": 2.1721, - "step": 71 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019997564770344315, - "loss": 2.1973, - "step": 72 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999753094771021, - "loss": 2.0651, - "step": 73 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199974971250761, - "loss": 2.0281, - "step": 74 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019997463302441994, - "loss": 2.1349, - "step": 75 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019997429479807889, - "loss": 2.1041, - "step": 76 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999739565717378, - "loss": 2.0383, - "step": 77 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019997361834539674, - "loss": 2.1886, - "step": 78 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019997328011905568, - "loss": 2.3179, - "step": 79 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999729418927146, - "loss": 2.1407, - "step": 80 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019997260366637354, - "loss": 2.2062, - "step": 81 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019997226544003248, - "loss": 2.275, - "step": 82 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999719272136914, - "loss": 2.161, - "step": 83 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019997158898735033, - "loss": 2.2849, - "step": 84 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019997125076100927, - "loss": 2.2406, - "step": 85 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019997091253466822, - "loss": 2.2017, - "step": 86 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019997057430832716, - "loss": 2.2042, - "step": 87 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019997023608198607, - "loss": 2.2018, - "step": 88 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199969897855645, - "loss": 2.3154, - "step": 89 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019996955962930395, - "loss": 2.2339, - "step": 90 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019996922140296287, - "loss": 2.1915, - "step": 91 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999688831766218, - "loss": 2.1834, - "step": 92 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019996854495028075, - "loss": 2.0172, - "step": 93 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999682067239397, - "loss": 2.1447, - "step": 94 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999678684975986, - "loss": 2.1743, - "step": 95 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019996753027125755, - "loss": 2.2437, - "step": 96 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999671920449165, - "loss": 2.1837, - "step": 97 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999668538185754, - "loss": 2.275, - "step": 98 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019996651559223434, - "loss": 2.284, - "step": 99 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019996617736589328, - "loss": 2.0941, - "step": 100 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999658391395522, - "loss": 2.204, - "step": 101 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019996550091321114, - "loss": 2.1735, - "step": 102 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019996516268687008, - "loss": 2.2171, - "step": 103 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199964824460529, - "loss": 2.0399, - "step": 104 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019996448623418793, - "loss": 2.2344, - "step": 105 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019996414800784688, - "loss": 2.1115, - "step": 106 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999638097815058, - "loss": 2.1768, - "step": 107 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019996347155516473, - "loss": 2.249, - "step": 108 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019996313332882367, - "loss": 2.1946, - "step": 109 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019996279510248259, - "loss": 2.1924, - "step": 110 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019996245687614153, - "loss": 2.1125, - "step": 111 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019996211864980047, - "loss": 2.2375, - "step": 112 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019996178042345938, - "loss": 2.1352, - "step": 113 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019996144219711832, - "loss": 2.1934, - "step": 114 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019996110397077726, - "loss": 2.1254, - "step": 115 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019996076574443618, - "loss": 2.1428, - "step": 116 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019996042751809512, - "loss": 2.154, - "step": 117 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019996008929175406, - "loss": 2.1618, - "step": 118 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019995975106541298, - "loss": 2.1912, - "step": 119 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019995941283907192, - "loss": 2.2085, - "step": 120 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019995907461273086, - "loss": 2.24, - "step": 121 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019995873638638977, - "loss": 2.0309, - "step": 122 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999583981600487, - "loss": 2.1547, - "step": 123 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019995805993370765, - "loss": 2.1181, - "step": 124 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019995772170736657, - "loss": 2.2295, - "step": 125 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999573834810255, - "loss": 2.0635, - "step": 126 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019995704525468445, - "loss": 2.1258, - "step": 127 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019995670702834336, - "loss": 2.1078, - "step": 128 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999563688020023, - "loss": 2.2913, - "step": 129 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019995603057566125, - "loss": 2.0787, - "step": 130 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019995569234932016, - "loss": 2.1033, - "step": 131 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999553541229791, - "loss": 2.0832, - "step": 132 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019995501589663804, - "loss": 2.1737, - "step": 133 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019995467767029696, - "loss": 2.2305, - "step": 134 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999543394439559, - "loss": 2.2178, - "step": 135 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019995400121761484, - "loss": 2.0741, - "step": 136 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019995366299127375, - "loss": 2.1699, - "step": 137 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999533247649327, - "loss": 2.1868, - "step": 138 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019995298653859164, - "loss": 2.1613, - "step": 139 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019995264831225055, - "loss": 2.122, - "step": 140 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019995231008590952, - "loss": 2.2076, - "step": 141 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019995197185956843, - "loss": 2.169, - "step": 142 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019995163363322737, - "loss": 2.1509, - "step": 143 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019995129540688631, - "loss": 2.1904, - "step": 144 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019995095718054523, - "loss": 2.1456, - "step": 145 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019995061895420417, - "loss": 2.1309, - "step": 146 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999502807278631, - "loss": 2.1714, - "step": 147 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019994994250152202, - "loss": 2.1179, - "step": 148 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019994960427518097, - "loss": 2.056, - "step": 149 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999492660488399, - "loss": 2.072, - "step": 150 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019994892782249882, - "loss": 2.2005, - "step": 151 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019994858959615776, - "loss": 1.9705, - "step": 152 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999482513698167, - "loss": 2.0816, - "step": 153 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019994791314347562, - "loss": 2.16, - "step": 154 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019994757491713456, - "loss": 2.0555, - "step": 155 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999472366907935, - "loss": 2.1196, - "step": 156 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999468984644524, - "loss": 2.092, - "step": 157 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019994656023811135, - "loss": 1.9955, - "step": 158 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999462220117703, - "loss": 2.2194, - "step": 159 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999458837854292, - "loss": 2.2023, - "step": 160 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019994554555908815, - "loss": 2.0786, - "step": 161 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999452073327471, - "loss": 2.0986, - "step": 162 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199944869106406, - "loss": 2.0123, - "step": 163 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019994453088006495, - "loss": 2.1087, - "step": 164 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999441926537239, - "loss": 2.0548, - "step": 165 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999438544273828, - "loss": 2.0114, - "step": 166 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019994351620104174, - "loss": 2.0024, - "step": 167 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019994317797470068, - "loss": 2.1048, - "step": 168 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999428397483596, - "loss": 2.108, - "step": 169 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019994250152201854, - "loss": 2.1524, - "step": 170 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019994216329567748, - "loss": 2.1122, - "step": 171 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999418250693364, - "loss": 2.0919, - "step": 172 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019994148684299534, - "loss": 2.1548, - "step": 173 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019994114861665428, - "loss": 2.0768, - "step": 174 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999408103903132, - "loss": 1.963, - "step": 175 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019994047216397213, - "loss": 2.0291, - "step": 176 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019994013393763107, - "loss": 2.224, - "step": 177 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019993979571129001, - "loss": 2.0497, - "step": 178 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019993945748494893, - "loss": 2.0526, - "step": 179 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019993911925860787, - "loss": 2.0637, - "step": 180 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999387810322668, - "loss": 2.0681, - "step": 181 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019993844280592572, - "loss": 1.9559, - "step": 182 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019993810457958467, - "loss": 1.9792, - "step": 183 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999377663532436, - "loss": 2.1107, - "step": 184 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019993742812690252, - "loss": 2.1538, - "step": 185 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019993708990056146, - "loss": 2.1603, - "step": 186 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999367516742204, - "loss": 2.0376, - "step": 187 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019993641344787932, - "loss": 2.116, - "step": 188 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019993607522153826, - "loss": 2.1186, - "step": 189 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999357369951972, - "loss": 2.03, - "step": 190 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019993539876885611, - "loss": 2.0829, - "step": 191 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019993506054251505, - "loss": 2.0339, - "step": 192 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199934722316174, - "loss": 2.1648, - "step": 193 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999343840898329, - "loss": 2.1118, - "step": 194 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019993404586349185, - "loss": 2.1775, - "step": 195 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999337076371508, - "loss": 2.094, - "step": 196 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019993336941080973, - "loss": 2.1395, - "step": 197 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019993303118446867, - "loss": 2.0895, - "step": 198 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999326929581276, - "loss": 2.0729, - "step": 199 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019993235473178653, - "loss": 1.9618, - "step": 200 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019993201650544547, - "loss": 2.0848, - "step": 201 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019993167827910439, - "loss": 1.9137, - "step": 202 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019993134005276333, - "loss": 2.0896, - "step": 203 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019993100182642227, - "loss": 2.0557, - "step": 204 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019993066360008118, - "loss": 2.0246, - "step": 205 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019993032537374012, - "loss": 1.9769, - "step": 206 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019992998714739906, - "loss": 2.1241, - "step": 207 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019992964892105798, - "loss": 1.9877, - "step": 208 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019992931069471692, - "loss": 2.1518, - "step": 209 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019992897246837586, - "loss": 2.0611, - "step": 210 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019992863424203477, - "loss": 2.1196, - "step": 211 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019992829601569372, - "loss": 2.1946, - "step": 212 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019992795778935266, - "loss": 2.131, - "step": 213 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019992761956301157, - "loss": 2.0337, - "step": 214 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999272813366705, - "loss": 1.9468, - "step": 215 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019992694311032945, - "loss": 2.1084, - "step": 216 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019992660488398837, - "loss": 2.0624, - "step": 217 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999262666576473, - "loss": 2.0985, - "step": 218 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019992592843130625, - "loss": 2.0201, - "step": 219 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019992559020496516, - "loss": 2.0481, - "step": 220 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999252519786241, - "loss": 2.0226, - "step": 221 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019992491375228305, - "loss": 2.0437, - "step": 222 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019992457552594196, - "loss": 2.1284, - "step": 223 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999242372996009, - "loss": 2.0974, - "step": 224 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019992389907325984, - "loss": 2.1002, - "step": 225 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019992356084691876, - "loss": 2.0161, - "step": 226 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999232226205777, - "loss": 2.0709, - "step": 227 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019992288439423664, - "loss": 2.0921, - "step": 228 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019992254616789555, - "loss": 2.0577, - "step": 229 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999222079415545, - "loss": 2.2253, - "step": 230 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019992186971521343, - "loss": 2.0385, - "step": 231 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019992153148887235, - "loss": 2.0107, - "step": 232 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999211932625313, - "loss": 2.1987, - "step": 233 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019992085503619023, - "loss": 2.1294, - "step": 234 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019992051680984914, - "loss": 2.0173, - "step": 235 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019992017858350809, - "loss": 2.0658, - "step": 236 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019991984035716703, - "loss": 2.1124, - "step": 237 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019991950213082594, - "loss": 1.995, - "step": 238 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019991916390448488, - "loss": 2.04, - "step": 239 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019991882567814382, - "loss": 2.1323, - "step": 240 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019991848745180274, - "loss": 2.0986, - "step": 241 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019991814922546168, - "loss": 2.0648, - "step": 242 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019991781099912062, - "loss": 2.131, - "step": 243 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019991747277277953, - "loss": 2.1524, - "step": 244 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019991713454643847, - "loss": 2.0726, - "step": 245 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019991679632009742, - "loss": 2.0277, - "step": 246 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019991645809375633, - "loss": 2.0652, - "step": 247 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019991611986741527, - "loss": 2.0408, - "step": 248 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999157816410742, - "loss": 2.1124, - "step": 249 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019991544341473313, - "loss": 2.0864, - "step": 250 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019991510518839207, - "loss": 1.8862, - "step": 251 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199914766962051, - "loss": 2.1129, - "step": 252 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019991442873570995, - "loss": 2.1113, - "step": 253 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999140905093689, - "loss": 2.1647, - "step": 254 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019991375228302783, - "loss": 2.2165, - "step": 255 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019991341405668675, - "loss": 2.1683, - "step": 256 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999130758303457, - "loss": 2.0884, - "step": 257 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019991273760400463, - "loss": 2.0471, - "step": 258 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019991239937766354, - "loss": 1.9271, - "step": 259 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019991206115132248, - "loss": 2.0365, - "step": 260 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019991172292498142, - "loss": 2.0779, - "step": 261 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019991138469864034, - "loss": 2.0399, - "step": 262 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019991104647229928, - "loss": 2.1314, - "step": 263 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019991070824595822, - "loss": 1.9989, - "step": 264 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019991037001961713, - "loss": 2.0614, - "step": 265 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019991003179327608, - "loss": 2.1055, - "step": 266 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019990969356693502, - "loss": 2.0883, - "step": 267 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019990935534059393, - "loss": 2.0638, - "step": 268 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019990901711425287, - "loss": 2.0537, - "step": 269 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999086788879118, - "loss": 2.0363, - "step": 270 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019990834066157073, - "loss": 2.0682, - "step": 271 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019990800243522967, - "loss": 2.1532, - "step": 272 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999076642088886, - "loss": 1.9857, - "step": 273 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019990732598254752, - "loss": 2.0577, - "step": 274 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019990698775620646, - "loss": 2.0986, - "step": 275 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999066495298654, - "loss": 2.1018, - "step": 276 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019990631130352432, - "loss": 2.1615, - "step": 277 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019990597307718326, - "loss": 2.1373, - "step": 278 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999056348508422, - "loss": 2.0259, - "step": 279 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019990529662450112, - "loss": 2.0535, - "step": 280 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019990495839816006, - "loss": 1.8688, - "step": 281 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199904620171819, - "loss": 2.1781, - "step": 282 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999042819454779, - "loss": 2.1036, - "step": 283 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019990394371913685, - "loss": 1.9399, - "step": 284 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999036054927958, - "loss": 1.8992, - "step": 285 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999032672664547, - "loss": 2.1933, - "step": 286 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019990292904011365, - "loss": 2.0939, - "step": 287 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999025908137726, - "loss": 2.1048, - "step": 288 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999022525874315, - "loss": 2.0274, - "step": 289 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019990191436109045, - "loss": 2.2158, - "step": 290 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999015761347494, - "loss": 2.1181, - "step": 291 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999012379084083, - "loss": 2.0921, - "step": 292 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019990089968206724, - "loss": 2.0459, - "step": 293 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019990056145572618, - "loss": 2.0066, - "step": 294 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001999002232293851, - "loss": 2.1238, - "step": 295 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019989988500304404, - "loss": 2.1294, - "step": 296 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019989954677670298, - "loss": 2.0224, - "step": 297 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998992085503619, - "loss": 2.0424, - "step": 298 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019989887032402084, - "loss": 2.0409, - "step": 299 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019989853209767978, - "loss": 2.0398, - "step": 300 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998981938713387, - "loss": 2.1096, - "step": 301 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019989785564499763, - "loss": 2.0602, - "step": 302 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019989751741865657, - "loss": 2.0056, - "step": 303 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998971791923155, - "loss": 2.1174, - "step": 304 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019989684096597443, - "loss": 2.0823, - "step": 305 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019989650273963337, - "loss": 1.9642, - "step": 306 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019989616451329228, - "loss": 2.1224, - "step": 307 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019989582628695122, - "loss": 2.1064, - "step": 308 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998954880606102, - "loss": 2.0862, - "step": 309 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998951498342691, - "loss": 1.9699, - "step": 310 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019989481160792805, - "loss": 2.0964, - "step": 311 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199894473381587, - "loss": 1.9842, - "step": 312 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998941351552459, - "loss": 2.0563, - "step": 313 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019989379692890484, - "loss": 2.1615, - "step": 314 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019989345870256379, - "loss": 1.9875, - "step": 315 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998931204762227, - "loss": 1.9565, - "step": 316 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019989278224988164, - "loss": 2.1689, - "step": 317 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019989244402354058, - "loss": 2.0431, - "step": 318 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998921057971995, - "loss": 2.0465, - "step": 319 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019989176757085844, - "loss": 2.0677, - "step": 320 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019989142934451738, - "loss": 2.0345, - "step": 321 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998910911181763, - "loss": 2.0229, - "step": 322 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019989075289183523, - "loss": 2.0487, - "step": 323 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019989041466549417, - "loss": 2.1542, - "step": 324 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998900764391531, - "loss": 2.0493, - "step": 325 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019988973821281203, - "loss": 2.1935, - "step": 326 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019988939998647097, - "loss": 2.0688, - "step": 327 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019988906176012988, - "loss": 1.9943, - "step": 328 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019988872353378883, - "loss": 2.0764, - "step": 329 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019988838530744777, - "loss": 2.1879, - "step": 330 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019988804708110668, - "loss": 2.1964, - "step": 331 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019988770885476562, - "loss": 2.0272, - "step": 332 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019988737062842456, - "loss": 1.9451, - "step": 333 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019988703240208348, - "loss": 2.1256, - "step": 334 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019988669417574242, - "loss": 1.9027, - "step": 335 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019988635594940136, - "loss": 2.0499, - "step": 336 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019988601772306027, - "loss": 2.0664, - "step": 337 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019988567949671921, - "loss": 2.0153, - "step": 338 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019988534127037816, - "loss": 2.0179, - "step": 339 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019988500304403707, - "loss": 2.0034, - "step": 340 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199884664817696, - "loss": 2.0309, - "step": 341 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019988432659135495, - "loss": 2.0709, - "step": 342 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019988398836501387, - "loss": 2.0321, - "step": 343 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998836501386728, - "loss": 1.999, - "step": 344 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019988331191233175, - "loss": 2.054, - "step": 345 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019988297368599066, - "loss": 2.0167, - "step": 346 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998826354596496, - "loss": 2.1531, - "step": 347 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019988229723330854, - "loss": 1.9503, - "step": 348 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019988195900696746, - "loss": 1.89, - "step": 349 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998816207806264, - "loss": 2.0224, - "step": 350 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019988128255428534, - "loss": 1.9361, - "step": 351 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019988094432794425, - "loss": 2.0178, - "step": 352 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998806061016032, - "loss": 1.9231, - "step": 353 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019988026787526214, - "loss": 1.9749, - "step": 354 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019987992964892105, - "loss": 1.8635, - "step": 355 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019987959142258, - "loss": 2.1706, - "step": 356 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019987925319623893, - "loss": 1.9783, - "step": 357 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019987891496989785, - "loss": 2.0293, - "step": 358 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998785767435568, - "loss": 1.882, - "step": 359 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019987823851721573, - "loss": 2.0575, - "step": 360 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019987790029087464, - "loss": 2.1755, - "step": 361 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019987756206453359, - "loss": 2.1811, - "step": 362 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019987722383819253, - "loss": 1.947, - "step": 363 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019987688561185144, - "loss": 2.0453, - "step": 364 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019987654738551038, - "loss": 2.1032, - "step": 365 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019987620915916935, - "loss": 1.9674, - "step": 366 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019987587093282826, - "loss": 2.1371, - "step": 367 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998755327064872, - "loss": 2.0831, - "step": 368 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019987519448014615, - "loss": 2.0925, - "step": 369 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019987485625380506, - "loss": 1.9922, - "step": 370 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199874518027464, - "loss": 2.1789, - "step": 371 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019987417980112294, - "loss": 2.1864, - "step": 372 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019987384157478186, - "loss": 2.0456, - "step": 373 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998735033484408, - "loss": 1.8753, - "step": 374 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019987316512209974, - "loss": 1.9939, - "step": 375 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019987282689575865, - "loss": 1.9528, - "step": 376 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998724886694176, - "loss": 1.9484, - "step": 377 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019987215044307653, - "loss": 1.9537, - "step": 378 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019987181221673545, - "loss": 2.1228, - "step": 379 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998714739903944, - "loss": 1.972, - "step": 380 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019987113576405333, - "loss": 2.049, - "step": 381 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019987079753771225, - "loss": 2.0395, - "step": 382 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019987045931137119, - "loss": 2.1139, - "step": 383 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019987012108503013, - "loss": 2.0621, - "step": 384 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019986978285868904, - "loss": 2.0738, - "step": 385 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019986944463234798, - "loss": 2.0719, - "step": 386 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019986910640600692, - "loss": 2.0019, - "step": 387 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019986876817966584, - "loss": 2.1366, - "step": 388 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019986842995332478, - "loss": 1.9908, - "step": 389 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019986809172698372, - "loss": 1.9814, - "step": 390 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019986775350064263, - "loss": 1.9815, - "step": 391 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019986741527430158, - "loss": 2.0751, - "step": 392 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019986707704796052, - "loss": 2.0132, - "step": 393 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019986673882161943, - "loss": 1.8854, - "step": 394 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019986640059527837, - "loss": 1.9881, - "step": 395 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998660623689373, - "loss": 1.9644, - "step": 396 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019986572414259623, - "loss": 1.9256, - "step": 397 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019986538591625517, - "loss": 2.178, - "step": 398 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998650476899141, - "loss": 2.0578, - "step": 399 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019986470946357302, - "loss": 2.0917, - "step": 400 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019986437123723196, - "loss": 2.0974, - "step": 401 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998640330108909, - "loss": 1.9364, - "step": 402 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019986369478454982, - "loss": 1.9394, - "step": 403 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019986335655820876, - "loss": 2.1514, - "step": 404 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998630183318677, - "loss": 1.9198, - "step": 405 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019986268010552662, - "loss": 2.0411, - "step": 406 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019986234187918556, - "loss": 1.9441, - "step": 407 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998620036528445, - "loss": 1.8918, - "step": 408 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998616654265034, - "loss": 2.0205, - "step": 409 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019986132720016235, - "loss": 1.9205, - "step": 410 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998609889738213, - "loss": 2.0386, - "step": 411 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998606507474802, - "loss": 1.8677, - "step": 412 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019986031252113915, - "loss": 1.9898, - "step": 413 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998599742947981, - "loss": 1.9326, - "step": 414 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199859636068457, - "loss": 1.9612, - "step": 415 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019985929784211595, - "loss": 1.9536, - "step": 416 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998589596157749, - "loss": 2.0676, - "step": 417 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998586213894338, - "loss": 1.9085, - "step": 418 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019985828316309274, - "loss": 2.0512, - "step": 419 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019985794493675168, - "loss": 2.0224, - "step": 420 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998576067104106, - "loss": 1.8575, - "step": 421 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019985726848406957, - "loss": 2.1921, - "step": 422 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019985693025772848, - "loss": 1.9409, - "step": 423 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019985659203138742, - "loss": 1.9096, - "step": 424 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019985625380504636, - "loss": 2.0409, - "step": 425 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019985591557870528, - "loss": 1.9798, - "step": 426 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019985557735236422, - "loss": 2.1268, - "step": 427 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019985523912602316, - "loss": 2.0168, - "step": 428 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019985490089968207, - "loss": 1.9651, - "step": 429 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199854562673341, - "loss": 1.9599, - "step": 430 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019985422444699995, - "loss": 2.06, - "step": 431 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019985388622065887, - "loss": 1.9492, - "step": 432 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998535479943178, - "loss": 2.0082, - "step": 433 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019985320976797675, - "loss": 1.9849, - "step": 434 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019985287154163566, - "loss": 1.9224, - "step": 435 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998525333152946, - "loss": 2.0695, - "step": 436 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019985219508895355, - "loss": 1.9767, - "step": 437 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019985185686261246, - "loss": 1.9545, - "step": 438 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998515186362714, - "loss": 2.0331, - "step": 439 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019985118040993034, - "loss": 1.9905, - "step": 440 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019985084218358926, - "loss": 1.9374, - "step": 441 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998505039572482, - "loss": 2.0785, - "step": 442 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019985016573090714, - "loss": 1.9805, - "step": 443 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019984982750456608, - "loss": 1.9762, - "step": 444 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199849489278225, - "loss": 2.0102, - "step": 445 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019984915105188394, - "loss": 1.973, - "step": 446 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019984881282554288, - "loss": 2.034, - "step": 447 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998484745992018, - "loss": 2.0176, - "step": 448 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019984813637286073, - "loss": 2.0537, - "step": 449 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019984779814651967, - "loss": 1.9435, - "step": 450 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998474599201786, - "loss": 1.8241, - "step": 451 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019984712169383753, - "loss": 1.9874, - "step": 452 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019984678346749647, - "loss": 2.0618, - "step": 453 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019984644524115538, - "loss": 2.1344, - "step": 454 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019984610701481432, - "loss": 2.1084, - "step": 455 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019984576878847327, - "loss": 1.9881, - "step": 456 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019984543056213218, - "loss": 2.0749, - "step": 457 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019984509233579112, - "loss": 2.1655, - "step": 458 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019984475410945006, - "loss": 1.7967, - "step": 459 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019984441588310898, - "loss": 2.1419, - "step": 460 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019984407765676792, - "loss": 1.9385, - "step": 461 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019984373943042686, - "loss": 1.9544, - "step": 462 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019984340120408577, - "loss": 2.1265, - "step": 463 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019984306297774471, - "loss": 2.1143, - "step": 464 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019984272475140366, - "loss": 2.0412, - "step": 465 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019984238652506257, - "loss": 2.0602, - "step": 466 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998420482987215, - "loss": 2.0837, - "step": 467 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019984171007238045, - "loss": 2.1164, - "step": 468 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019984137184603937, - "loss": 2.0105, - "step": 469 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998410336196983, - "loss": 1.9976, - "step": 470 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019984069539335725, - "loss": 1.8243, - "step": 471 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019984035716701616, - "loss": 2.0637, - "step": 472 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998400189406751, - "loss": 1.7974, - "step": 473 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019983968071433404, - "loss": 2.0007, - "step": 474 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019983934248799296, - "loss": 1.9775, - "step": 475 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998390042616519, - "loss": 2.0848, - "step": 476 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019983866603531084, - "loss": 1.9423, - "step": 477 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019983832780896978, - "loss": 1.8775, - "step": 478 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019983798958262872, - "loss": 1.9308, - "step": 479 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019983765135628764, - "loss": 2.0348, - "step": 480 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019983731312994658, - "loss": 2.0771, - "step": 481 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019983697490360552, - "loss": 2.0322, - "step": 482 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019983663667726443, - "loss": 1.9868, - "step": 483 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019983629845092337, - "loss": 1.9959, - "step": 484 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019983596022458232, - "loss": 1.9597, - "step": 485 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019983562199824123, - "loss": 1.9148, - "step": 486 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019983528377190017, - "loss": 2.0017, - "step": 487 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998349455455591, - "loss": 2.0513, - "step": 488 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019983460731921803, - "loss": 1.9508, - "step": 489 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019983426909287697, - "loss": 2.0352, - "step": 490 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998339308665359, - "loss": 2.0491, - "step": 491 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019983359264019482, - "loss": 1.8922, - "step": 492 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019983325441385376, - "loss": 2.0568, - "step": 493 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998329161875127, - "loss": 1.8767, - "step": 494 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019983257796117162, - "loss": 2.096, - "step": 495 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019983223973483056, - "loss": 1.8955, - "step": 496 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998319015084895, - "loss": 2.0011, - "step": 497 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019983156328214841, - "loss": 1.8872, - "step": 498 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019983122505580736, - "loss": 1.8866, - "step": 499 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998308868294663, - "loss": 1.9423, - "step": 500 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998305486031252, - "loss": 2.0268, - "step": 501 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019983021037678415, - "loss": 1.9603, - "step": 502 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998298721504431, - "loss": 2.0531, - "step": 503 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199829533924102, - "loss": 1.878, - "step": 504 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019982919569776095, - "loss": 1.9032, - "step": 505 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998288574714199, - "loss": 1.9436, - "step": 506 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998285192450788, - "loss": 1.9853, - "step": 507 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019982818101873774, - "loss": 1.9443, - "step": 508 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019982784279239669, - "loss": 2.0179, - "step": 509 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998275045660556, - "loss": 2.1365, - "step": 510 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019982716633971454, - "loss": 2.0552, - "step": 511 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019982682811337348, - "loss": 1.9423, - "step": 512 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998264898870324, - "loss": 2.0864, - "step": 513 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019982615166069134, - "loss": 2.0085, - "step": 514 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019982581343435028, - "loss": 1.9237, - "step": 515 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998254752080092, - "loss": 1.9586, - "step": 516 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019982513698166813, - "loss": 1.9516, - "step": 517 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019982479875532707, - "loss": 1.8511, - "step": 518 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199824460528986, - "loss": 1.9659, - "step": 519 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019982412230264493, - "loss": 1.9647, - "step": 520 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019982378407630387, - "loss": 1.8034, - "step": 521 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019982344584996278, - "loss": 1.8911, - "step": 522 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019982310762362173, - "loss": 2.0519, - "step": 523 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019982276939728067, - "loss": 2.1095, - "step": 524 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019982243117093958, - "loss": 1.9367, - "step": 525 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019982209294459852, - "loss": 2.0866, - "step": 526 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019982175471825746, - "loss": 1.9373, - "step": 527 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019982141649191638, - "loss": 2.0943, - "step": 528 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019982107826557532, - "loss": 2.1287, - "step": 529 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019982074003923426, - "loss": 1.9691, - "step": 530 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998204018128932, - "loss": 1.9046, - "step": 531 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019982006358655212, - "loss": 1.9817, - "step": 532 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019981972536021106, - "loss": 2.0341, - "step": 533 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019981938713387, - "loss": 1.9676, - "step": 534 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019981904890752894, - "loss": 2.0387, - "step": 535 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019981871068118788, - "loss": 2.0826, - "step": 536 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998183724548468, - "loss": 1.9484, - "step": 537 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019981803422850573, - "loss": 2.0582, - "step": 538 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019981769600216468, - "loss": 2.0279, - "step": 539 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998173577758236, - "loss": 1.951, - "step": 540 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019981701954948253, - "loss": 1.8365, - "step": 541 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019981668132314147, - "loss": 2.0301, - "step": 542 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019981634309680039, - "loss": 1.9125, - "step": 543 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019981600487045933, - "loss": 1.9657, - "step": 544 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019981566664411827, - "loss": 2.0104, - "step": 545 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019981532841777718, - "loss": 2.0056, - "step": 546 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019981499019143612, - "loss": 2.0802, - "step": 547 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019981465196509506, - "loss": 1.9531, - "step": 548 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019981431373875398, - "loss": 1.984, - "step": 549 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019981397551241292, - "loss": 2.0562, - "step": 550 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019981363728607186, - "loss": 1.9849, - "step": 551 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019981329905973078, - "loss": 1.9206, - "step": 552 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019981296083338972, - "loss": 1.9433, - "step": 553 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019981262260704866, - "loss": 1.9292, - "step": 554 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019981228438070757, - "loss": 2.1156, - "step": 555 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998119461543665, - "loss": 2.0783, - "step": 556 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019981160792802545, - "loss": 1.8687, - "step": 557 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019981126970168437, - "loss": 2.0066, - "step": 558 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998109314753433, - "loss": 1.9583, - "step": 559 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019981059324900225, - "loss": 2.0632, - "step": 560 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019981025502266116, - "loss": 1.9925, - "step": 561 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998099167963201, - "loss": 2.0857, - "step": 562 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019980957856997905, - "loss": 1.8284, - "step": 563 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019980924034363796, - "loss": 1.822, - "step": 564 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998089021172969, - "loss": 2.0047, - "step": 565 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019980856389095584, - "loss": 1.8899, - "step": 566 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019980822566461476, - "loss": 1.9768, - "step": 567 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998078874382737, - "loss": 1.9918, - "step": 568 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019980754921193264, - "loss": 2.0272, - "step": 569 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019980721098559155, - "loss": 1.9476, - "step": 570 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998068727592505, - "loss": 2.0123, - "step": 571 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019980653453290944, - "loss": 1.9434, - "step": 572 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019980619630656835, - "loss": 1.9412, - "step": 573 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998058580802273, - "loss": 1.9162, - "step": 574 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019980551985388623, - "loss": 1.9446, - "step": 575 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019980518162754515, - "loss": 1.9809, - "step": 576 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001998048434012041, - "loss": 2.0824, - "step": 577 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019980450517486303, - "loss": 2.0234, - "step": 578 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019980416694852194, - "loss": 2.0293, - "step": 579 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019980382872218088, - "loss": 1.9739, - "step": 580 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019980349049583982, - "loss": 1.837, - "step": 581 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019980315226949874, - "loss": 2.0688, - "step": 582 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019980281404315768, - "loss": 1.8954, - "step": 583 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019980247581681662, - "loss": 2.0246, - "step": 584 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019980213759047553, - "loss": 2.0843, - "step": 585 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019980179936413448, - "loss": 1.8904, - "step": 586 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019980146113779342, - "loss": 2.0551, - "step": 587 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019980112291145233, - "loss": 1.9521, - "step": 588 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019980078468511127, - "loss": 2.0411, - "step": 589 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019980044645877024, - "loss": 1.847, - "step": 590 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019980010823242915, - "loss": 1.9453, - "step": 591 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997997700060881, - "loss": 1.8346, - "step": 592 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019979943177974704, - "loss": 2.0273, - "step": 593 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019979909355340595, - "loss": 1.9145, - "step": 594 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997987553270649, - "loss": 1.9497, - "step": 595 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019979841710072383, - "loss": 2.0276, - "step": 596 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019979807887438275, - "loss": 2.0711, - "step": 597 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997977406480417, - "loss": 2.0179, - "step": 598 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019979740242170063, - "loss": 2.0021, - "step": 599 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019979706419535954, - "loss": 2.0822, - "step": 600 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019979672596901848, - "loss": 2.0356, - "step": 601 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019979638774267743, - "loss": 2.0962, - "step": 602 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019979604951633634, - "loss": 2.0301, - "step": 603 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019979571128999528, - "loss": 1.8209, - "step": 604 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019979537306365422, - "loss": 1.8448, - "step": 605 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019979503483731314, - "loss": 1.9946, - "step": 606 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019979469661097208, - "loss": 1.9962, - "step": 607 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019979435838463102, - "loss": 1.9927, - "step": 608 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019979402015828993, - "loss": 1.9894, - "step": 609 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019979368193194887, - "loss": 2.0343, - "step": 610 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019979334370560781, - "loss": 2.0315, - "step": 611 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019979300547926673, - "loss": 2.0699, - "step": 612 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019979266725292567, - "loss": 1.8152, - "step": 613 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997923290265846, - "loss": 1.8728, - "step": 614 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019979199080024352, - "loss": 1.9474, - "step": 615 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019979165257390247, - "loss": 1.9079, - "step": 616 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997913143475614, - "loss": 1.8826, - "step": 617 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019979097612122032, - "loss": 1.9618, - "step": 618 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019979063789487926, - "loss": 1.8733, - "step": 619 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997902996685382, - "loss": 1.9392, - "step": 620 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019978996144219712, - "loss": 2.0088, - "step": 621 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019978962321585606, - "loss": 1.9767, - "step": 622 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199789284989515, - "loss": 1.9203, - "step": 623 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019978894676317391, - "loss": 1.9456, - "step": 624 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019978860853683285, - "loss": 2.0347, - "step": 625 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997882703104918, - "loss": 1.9138, - "step": 626 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997879320841507, - "loss": 2.0262, - "step": 627 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019978759385780965, - "loss": 1.9096, - "step": 628 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997872556314686, - "loss": 1.9331, - "step": 629 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997869174051275, - "loss": 2.0177, - "step": 630 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019978657917878645, - "loss": 1.9256, - "step": 631 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997862409524454, - "loss": 1.9873, - "step": 632 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997859027261043, - "loss": 1.9684, - "step": 633 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019978556449976324, - "loss": 1.953, - "step": 634 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019978522627342219, - "loss": 1.9434, - "step": 635 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997848880470811, - "loss": 1.9601, - "step": 636 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019978454982074004, - "loss": 2.047, - "step": 637 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019978421159439898, - "loss": 1.9972, - "step": 638 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997838733680579, - "loss": 1.9724, - "step": 639 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019978353514171684, - "loss": 2.0615, - "step": 640 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019978319691537578, - "loss": 1.9975, - "step": 641 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997828586890347, - "loss": 2.0452, - "step": 642 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019978252046269363, - "loss": 1.855, - "step": 643 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019978218223635257, - "loss": 1.9491, - "step": 644 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997818440100115, - "loss": 2.1169, - "step": 645 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019978150578367046, - "loss": 2.0199, - "step": 646 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997811675573294, - "loss": 2.0716, - "step": 647 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997808293309883, - "loss": 1.942, - "step": 648 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019978049110464725, - "loss": 1.9655, - "step": 649 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997801528783062, - "loss": 1.7844, - "step": 650 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997798146519651, - "loss": 1.9326, - "step": 651 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019977947642562405, - "loss": 2.0477, - "step": 652 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199779138199283, - "loss": 1.9569, - "step": 653 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997787999729419, - "loss": 1.9607, - "step": 654 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019977846174660085, - "loss": 2.1283, - "step": 655 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019977812352025979, - "loss": 1.883, - "step": 656 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997777852939187, - "loss": 2.0697, - "step": 657 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019977744706757764, - "loss": 2.0584, - "step": 658 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019977710884123658, - "loss": 2.0025, - "step": 659 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997767706148955, - "loss": 2.0662, - "step": 660 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019977643238855444, - "loss": 1.9412, - "step": 661 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019977609416221338, - "loss": 1.855, - "step": 662 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997757559358723, - "loss": 1.957, - "step": 663 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019977541770953123, - "loss": 1.8974, - "step": 664 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019977507948319018, - "loss": 1.861, - "step": 665 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997747412568491, - "loss": 2.079, - "step": 666 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019977440303050803, - "loss": 1.9628, - "step": 667 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019977406480416697, - "loss": 1.989, - "step": 668 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019977372657782589, - "loss": 1.8578, - "step": 669 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019977338835148483, - "loss": 1.9314, - "step": 670 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019977305012514377, - "loss": 1.9861, - "step": 671 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019977271189880268, - "loss": 1.9258, - "step": 672 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019977237367246162, - "loss": 1.9842, - "step": 673 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019977203544612056, - "loss": 1.974, - "step": 674 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019977169721977948, - "loss": 1.8281, - "step": 675 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019977135899343842, - "loss": 1.8448, - "step": 676 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019977102076709736, - "loss": 2.0186, - "step": 677 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019977068254075627, - "loss": 2.0447, - "step": 678 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019977034431441522, - "loss": 1.9465, - "step": 679 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019977000608807416, - "loss": 1.8989, - "step": 680 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019976966786173307, - "loss": 1.9997, - "step": 681 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199769329635392, - "loss": 1.8363, - "step": 682 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019976899140905095, - "loss": 1.9591, - "step": 683 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019976865318270987, - "loss": 2.1292, - "step": 684 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997683149563688, - "loss": 1.9706, - "step": 685 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019976797673002775, - "loss": 2.1602, - "step": 686 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019976763850368666, - "loss": 2.0206, - "step": 687 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997673002773456, - "loss": 1.947, - "step": 688 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019976696205100455, - "loss": 1.8413, - "step": 689 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019976662382466346, - "loss": 1.9852, - "step": 690 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997662855983224, - "loss": 1.8583, - "step": 691 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019976594737198134, - "loss": 2.0499, - "step": 692 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019976560914564026, - "loss": 1.9588, - "step": 693 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997652709192992, - "loss": 1.8789, - "step": 694 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019976493269295814, - "loss": 1.9939, - "step": 695 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019976459446661705, - "loss": 1.8911, - "step": 696 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199764256240276, - "loss": 1.8251, - "step": 697 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019976391801393493, - "loss": 1.9954, - "step": 698 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019976357978759385, - "loss": 1.8902, - "step": 699 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997632415612528, - "loss": 1.8233, - "step": 700 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019976290333491173, - "loss": 2.0035, - "step": 701 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019976256510857065, - "loss": 1.9489, - "step": 702 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997622268822296, - "loss": 2.0103, - "step": 703 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019976188865588853, - "loss": 1.8976, - "step": 704 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019976155042954747, - "loss": 1.9271, - "step": 705 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997612122032064, - "loss": 2.0012, - "step": 706 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019976087397686532, - "loss": 2.0562, - "step": 707 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019976053575052426, - "loss": 2.0173, - "step": 708 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997601975241832, - "loss": 1.9777, - "step": 709 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019975985929784212, - "loss": 1.858, - "step": 710 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019975952107150106, - "loss": 1.9191, - "step": 711 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019975918284516, - "loss": 1.9458, - "step": 712 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019975884461881894, - "loss": 1.9032, - "step": 713 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019975850639247786, - "loss": 1.9952, - "step": 714 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997581681661368, - "loss": 1.8827, - "step": 715 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019975782993979574, - "loss": 2.0235, - "step": 716 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019975749171345465, - "loss": 2.0221, - "step": 717 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997571534871136, - "loss": 1.9551, - "step": 718 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019975681526077254, - "loss": 1.812, - "step": 719 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019975647703443145, - "loss": 2.1017, - "step": 720 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997561388080904, - "loss": 1.9824, - "step": 721 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019975580058174933, - "loss": 1.9314, - "step": 722 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019975546235540825, - "loss": 2.0244, - "step": 723 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997551241290672, - "loss": 1.9493, - "step": 724 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019975478590272613, - "loss": 1.966, - "step": 725 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019975444767638504, - "loss": 1.942, - "step": 726 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019975410945004398, - "loss": 2.0014, - "step": 727 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019975377122370292, - "loss": 1.8799, - "step": 728 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019975343299736184, - "loss": 1.9176, - "step": 729 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019975309477102078, - "loss": 1.9475, - "step": 730 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019975275654467972, - "loss": 2.075, - "step": 731 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019975241831833864, - "loss": 1.8512, - "step": 732 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019975208009199758, - "loss": 2.0743, - "step": 733 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019975174186565652, - "loss": 1.9461, - "step": 734 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019975140363931543, - "loss": 1.9555, - "step": 735 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019975106541297437, - "loss": 1.902, - "step": 736 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019975072718663331, - "loss": 1.8957, - "step": 737 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019975038896029223, - "loss": 1.9968, - "step": 738 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019975005073395117, - "loss": 1.8345, - "step": 739 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997497125076101, - "loss": 1.9926, - "step": 740 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019974937428126902, - "loss": 1.9395, - "step": 741 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019974903605492797, - "loss": 1.9603, - "step": 742 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997486978285869, - "loss": 2.0278, - "step": 743 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019974835960224582, - "loss": 1.8057, - "step": 744 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019974802137590476, - "loss": 1.9504, - "step": 745 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997476831495637, - "loss": 2.0321, - "step": 746 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019974734492322262, - "loss": 1.9538, - "step": 747 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019974700669688156, - "loss": 1.996, - "step": 748 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997466684705405, - "loss": 1.9069, - "step": 749 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997463302441994, - "loss": 1.9314, - "step": 750 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019974599201785835, - "loss": 1.9786, - "step": 751 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997456537915173, - "loss": 2.0727, - "step": 752 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997453155651762, - "loss": 2.0587, - "step": 753 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019974497733883515, - "loss": 2.0802, - "step": 754 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997446391124941, - "loss": 1.836, - "step": 755 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199744300886153, - "loss": 1.9151, - "step": 756 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019974396265981195, - "loss": 1.8716, - "step": 757 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997436244334709, - "loss": 1.9399, - "step": 758 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019974328620712983, - "loss": 2.0286, - "step": 759 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019974294798078877, - "loss": 2.0953, - "step": 760 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019974260975444768, - "loss": 1.8654, - "step": 761 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019974227152810663, - "loss": 2.0397, - "step": 762 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019974193330176557, - "loss": 1.9867, - "step": 763 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019974159507542448, - "loss": 1.9975, - "step": 764 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019974125684908342, - "loss": 2.0528, - "step": 765 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019974091862274236, - "loss": 1.9572, - "step": 766 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019974058039640128, - "loss": 1.9628, - "step": 767 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019974024217006022, - "loss": 1.94, - "step": 768 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019973990394371916, - "loss": 1.819, - "step": 769 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019973956571737807, - "loss": 1.9103, - "step": 770 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019973922749103701, - "loss": 1.87, - "step": 771 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019973888926469596, - "loss": 1.9177, - "step": 772 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019973855103835487, - "loss": 1.9769, - "step": 773 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997382128120138, - "loss": 1.9236, - "step": 774 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019973787458567275, - "loss": 1.9161, - "step": 775 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019973753635933167, - "loss": 1.9713, - "step": 776 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997371981329906, - "loss": 1.9482, - "step": 777 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019973685990664955, - "loss": 1.9351, - "step": 778 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019973652168030846, - "loss": 1.9701, - "step": 779 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997361834539674, - "loss": 1.9343, - "step": 780 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019973584522762634, - "loss": 1.9036, - "step": 781 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019973550700128526, - "loss": 2.0407, - "step": 782 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997351687749442, - "loss": 2.0747, - "step": 783 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019973483054860314, - "loss": 1.9933, - "step": 784 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019973449232226205, - "loss": 1.9452, - "step": 785 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199734154095921, - "loss": 1.854, - "step": 786 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019973381586957994, - "loss": 1.8806, - "step": 787 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019973347764323885, - "loss": 2.0617, - "step": 788 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997331394168978, - "loss": 1.9255, - "step": 789 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019973280119055673, - "loss": 1.861, - "step": 790 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019973246296421565, - "loss": 1.9719, - "step": 791 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997321247378746, - "loss": 1.962, - "step": 792 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019973178651153353, - "loss": 2.0679, - "step": 793 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019973144828519244, - "loss": 1.9911, - "step": 794 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019973111005885139, - "loss": 2.0097, - "step": 795 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019973077183251033, - "loss": 2.0489, - "step": 796 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019973043360616927, - "loss": 1.9711, - "step": 797 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019973009537982818, - "loss": 2.0008, - "step": 798 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019972975715348712, - "loss": 2.0507, - "step": 799 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019972941892714606, - "loss": 1.8638, - "step": 800 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019972908070080498, - "loss": 1.9982, - "step": 801 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019972874247446392, - "loss": 1.9376, - "step": 802 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019972840424812286, - "loss": 1.9215, - "step": 803 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019972806602178177, - "loss": 1.9712, - "step": 804 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019972772779544072, - "loss": 1.8764, - "step": 805 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019972738956909966, - "loss": 1.9915, - "step": 806 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019972705134275857, - "loss": 1.9553, - "step": 807 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997267131164175, - "loss": 2.1104, - "step": 808 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019972637489007645, - "loss": 1.9608, - "step": 809 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019972603666373537, - "loss": 1.9804, - "step": 810 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997256984373943, - "loss": 2.0171, - "step": 811 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019972536021105325, - "loss": 2.0319, - "step": 812 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019972502198471216, - "loss": 2.0937, - "step": 813 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997246837583711, - "loss": 2.008, - "step": 814 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019972434553203005, - "loss": 1.8835, - "step": 815 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019972400730568899, - "loss": 1.9227, - "step": 816 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019972366907934793, - "loss": 1.8879, - "step": 817 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019972333085300684, - "loss": 1.9002, - "step": 818 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019972299262666578, - "loss": 2.0218, - "step": 819 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019972265440032472, - "loss": 1.9708, - "step": 820 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019972231617398364, - "loss": 1.9583, - "step": 821 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019972197794764258, - "loss": 1.9097, - "step": 822 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019972163972130152, - "loss": 1.954, - "step": 823 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019972130149496043, - "loss": 1.9199, - "step": 824 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019972096326861938, - "loss": 1.9503, - "step": 825 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019972062504227832, - "loss": 2.0563, - "step": 826 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019972028681593723, - "loss": 1.8328, - "step": 827 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019971994858959617, - "loss": 1.952, - "step": 828 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997196103632551, - "loss": 1.9054, - "step": 829 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019971927213691403, - "loss": 1.9603, - "step": 830 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019971893391057297, - "loss": 2.0201, - "step": 831 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997185956842319, - "loss": 1.8683, - "step": 832 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019971825745789082, - "loss": 1.9168, - "step": 833 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019971791923154976, - "loss": 1.8975, - "step": 834 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997175810052087, - "loss": 1.9709, - "step": 835 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019971724277886762, - "loss": 1.9421, - "step": 836 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019971690455252656, - "loss": 1.9253, - "step": 837 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997165663261855, - "loss": 1.9705, - "step": 838 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019971622809984442, - "loss": 1.8106, - "step": 839 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019971588987350336, - "loss": 1.897, - "step": 840 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997155516471623, - "loss": 1.9749, - "step": 841 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997152134208212, - "loss": 2.0231, - "step": 842 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019971487519448015, - "loss": 1.8466, - "step": 843 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997145369681391, - "loss": 1.9443, - "step": 844 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199714198741798, - "loss": 1.986, - "step": 845 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019971386051545695, - "loss": 2.0066, - "step": 846 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997135222891159, - "loss": 2.0044, - "step": 847 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997131840627748, - "loss": 1.9773, - "step": 848 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019971284583643375, - "loss": 1.9334, - "step": 849 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997125076100927, - "loss": 1.8838, - "step": 850 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997121693837516, - "loss": 1.9391, - "step": 851 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019971183115741054, - "loss": 1.8895, - "step": 852 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019971149293106948, - "loss": 1.8756, - "step": 853 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997111547047284, - "loss": 1.9889, - "step": 854 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019971081647838734, - "loss": 1.8506, - "step": 855 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019971047825204628, - "loss": 1.8947, - "step": 856 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997101400257052, - "loss": 1.946, - "step": 857 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019970980179936413, - "loss": 1.8528, - "step": 858 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019970946357302308, - "loss": 2.0483, - "step": 859 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199709125346682, - "loss": 1.963, - "step": 860 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019970878712034093, - "loss": 1.8695, - "step": 861 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019970844889399987, - "loss": 1.8457, - "step": 862 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019970811066765879, - "loss": 1.9546, - "step": 863 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019970777244131773, - "loss": 1.8137, - "step": 864 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019970743421497667, - "loss": 1.8856, - "step": 865 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019970709598863558, - "loss": 1.9838, - "step": 866 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019970675776229452, - "loss": 2.1443, - "step": 867 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019970641953595346, - "loss": 1.917, - "step": 868 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019970608130961238, - "loss": 1.8867, - "step": 869 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019970574308327132, - "loss": 1.9102, - "step": 870 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997054048569303, - "loss": 2.0038, - "step": 871 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997050666305892, - "loss": 1.7972, - "step": 872 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019970472840424814, - "loss": 1.8827, - "step": 873 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019970439017790708, - "loss": 1.9976, - "step": 874 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199704051951566, - "loss": 1.9009, - "step": 875 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019970371372522494, - "loss": 1.9376, - "step": 876 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019970337549888388, - "loss": 2.0911, - "step": 877 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997030372725428, - "loss": 1.9425, - "step": 878 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019970269904620174, - "loss": 1.8644, - "step": 879 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019970236081986068, - "loss": 1.9744, - "step": 880 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997020225935196, - "loss": 1.9709, - "step": 881 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019970168436717853, - "loss": 1.9523, - "step": 882 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019970134614083747, - "loss": 1.9323, - "step": 883 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001997010079144964, - "loss": 1.8935, - "step": 884 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019970066968815533, - "loss": 1.9332, - "step": 885 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019970033146181427, - "loss": 1.7972, - "step": 886 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019969999323547318, - "loss": 1.9354, - "step": 887 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019969965500913212, - "loss": 2.1459, - "step": 888 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019969931678279107, - "loss": 1.9157, - "step": 889 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019969897855644998, - "loss": 1.8559, - "step": 890 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019969864033010892, - "loss": 1.8796, - "step": 891 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019969830210376786, - "loss": 1.9261, - "step": 892 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019969796387742678, - "loss": 1.8823, - "step": 893 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019969762565108572, - "loss": 1.8885, - "step": 894 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019969728742474466, - "loss": 1.9611, - "step": 895 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019969694919840357, - "loss": 1.9225, - "step": 896 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019969661097206251, - "loss": 1.783, - "step": 897 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019969627274572146, - "loss": 2.0008, - "step": 898 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019969593451938037, - "loss": 1.8814, - "step": 899 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996955962930393, - "loss": 1.9831, - "step": 900 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019969525806669825, - "loss": 1.8488, - "step": 901 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019969491984035717, - "loss": 1.8735, - "step": 902 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996945816140161, - "loss": 1.8916, - "step": 903 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019969424338767505, - "loss": 2.0051, - "step": 904 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019969390516133396, - "loss": 2.0235, - "step": 905 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996935669349929, - "loss": 1.8442, - "step": 906 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019969322870865184, - "loss": 1.8985, - "step": 907 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019969289048231076, - "loss": 1.8577, - "step": 908 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996925522559697, - "loss": 1.8767, - "step": 909 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019969221402962864, - "loss": 1.9647, - "step": 910 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019969187580328755, - "loss": 1.9257, - "step": 911 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996915375769465, - "loss": 1.9687, - "step": 912 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019969119935060544, - "loss": 1.8171, - "step": 913 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019969086112426435, - "loss": 1.9108, - "step": 914 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996905228979233, - "loss": 1.788, - "step": 915 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019969018467158223, - "loss": 1.9889, - "step": 916 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019968984644524115, - "loss": 1.8891, - "step": 917 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996895082189001, - "loss": 1.9842, - "step": 918 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019968916999255903, - "loss": 1.9754, - "step": 919 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019968883176621794, - "loss": 1.9432, - "step": 920 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019968849353987688, - "loss": 1.8644, - "step": 921 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019968815531353583, - "loss": 1.8897, - "step": 922 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019968781708719474, - "loss": 1.8629, - "step": 923 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019968747886085368, - "loss": 1.9639, - "step": 924 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019968714063451262, - "loss": 2.0377, - "step": 925 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019968680240817154, - "loss": 1.9411, - "step": 926 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996864641818305, - "loss": 2.0232, - "step": 927 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019968612595548945, - "loss": 1.9583, - "step": 928 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019968578772914836, - "loss": 1.9616, - "step": 929 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996854495028073, - "loss": 1.8171, - "step": 930 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019968511127646624, - "loss": 1.8551, - "step": 931 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019968477305012516, - "loss": 2.0569, - "step": 932 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996844348237841, - "loss": 1.915, - "step": 933 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019968409659744304, - "loss": 1.9401, - "step": 934 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019968375837110195, - "loss": 1.8389, - "step": 935 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996834201447609, - "loss": 2.0751, - "step": 936 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019968308191841983, - "loss": 1.8912, - "step": 937 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019968274369207875, - "loss": 1.9667, - "step": 938 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996824054657377, - "loss": 1.9724, - "step": 939 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019968206723939663, - "loss": 1.7139, - "step": 940 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019968172901305554, - "loss": 2.0036, - "step": 941 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019968139078671449, - "loss": 1.8578, - "step": 942 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019968105256037343, - "loss": 1.7973, - "step": 943 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019968071433403234, - "loss": 1.9233, - "step": 944 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019968037610769128, - "loss": 1.9025, - "step": 945 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019968003788135022, - "loss": 2.0039, - "step": 946 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019967969965500914, - "loss": 1.9692, - "step": 947 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019967936142866808, - "loss": 1.9187, - "step": 948 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019967902320232702, - "loss": 1.928, - "step": 949 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019967868497598593, - "loss": 1.9921, - "step": 950 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019967834674964487, - "loss": 1.8889, - "step": 951 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019967800852330382, - "loss": 1.9128, - "step": 952 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019967767029696273, - "loss": 1.9811, - "step": 953 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019967733207062167, - "loss": 1.9225, - "step": 954 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996769938442806, - "loss": 1.8595, - "step": 955 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019967665561793953, - "loss": 2.0474, - "step": 956 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019967631739159847, - "loss": 1.9859, - "step": 957 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996759791652574, - "loss": 1.9067, - "step": 958 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019967564093891632, - "loss": 1.9479, - "step": 959 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019967530271257526, - "loss": 1.7623, - "step": 960 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996749644862342, - "loss": 1.8939, - "step": 961 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019967462625989312, - "loss": 1.9114, - "step": 962 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019967428803355206, - "loss": 1.8759, - "step": 963 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199673949807211, - "loss": 1.9715, - "step": 964 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019967361158086992, - "loss": 1.924, - "step": 965 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019967327335452886, - "loss": 2.0324, - "step": 966 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996729351281878, - "loss": 1.9937, - "step": 967 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996725969018467, - "loss": 1.8336, - "step": 968 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019967225867550565, - "loss": 1.7873, - "step": 969 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996719204491646, - "loss": 1.9434, - "step": 970 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996715822228235, - "loss": 1.8568, - "step": 971 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019967124399648245, - "loss": 2.0011, - "step": 972 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996709057701414, - "loss": 1.9617, - "step": 973 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996705675438003, - "loss": 2.05, - "step": 974 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019967022931745925, - "loss": 2.0453, - "step": 975 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019966989109111819, - "loss": 1.8656, - "step": 976 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996695528647771, - "loss": 1.8141, - "step": 977 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019966921463843604, - "loss": 1.8974, - "step": 978 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019966887641209498, - "loss": 1.8117, - "step": 979 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996685381857539, - "loss": 1.9643, - "step": 980 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019966819995941284, - "loss": 1.8884, - "step": 981 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019966786173307178, - "loss": 1.9644, - "step": 982 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019966752350673072, - "loss": 1.9644, - "step": 983 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019966718528038966, - "loss": 2.0188, - "step": 984 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996668470540486, - "loss": 2.1158, - "step": 985 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019966650882770752, - "loss": 1.874, - "step": 986 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019966617060136646, - "loss": 1.8918, - "step": 987 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996658323750254, - "loss": 1.8496, - "step": 988 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996654941486843, - "loss": 2.1146, - "step": 989 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019966515592234325, - "loss": 1.8907, - "step": 990 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996648176960022, - "loss": 1.8877, - "step": 991 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996644794696611, - "loss": 1.8196, - "step": 992 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019966414124332005, - "loss": 1.9164, - "step": 993 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199663803016979, - "loss": 1.9489, - "step": 994 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996634647906379, - "loss": 1.9292, - "step": 995 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019966312656429685, - "loss": 1.9888, - "step": 996 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996627883379558, - "loss": 1.8694, - "step": 997 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996624501116147, - "loss": 1.8366, - "step": 998 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019966211188527364, - "loss": 1.8291, - "step": 999 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019966177365893258, - "loss": 1.9383, - "step": 1000 - }, - { - "epoch": 0.0, - "eval_loss": 1.3967848557692308, - "mrc_em": 6.0, - "mrc_f1": 11.03030303030303, - "nli_acc": 0.35, - "step": 1000, - "sts_f1": 0.4615384615384615, - "ynat_f1": 0.26114203257060403 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996614354325915, - "loss": 1.9921, - "step": 1001 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019966109720625044, - "loss": 1.9286, - "step": 1002 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019966075897990938, - "loss": 2.0872, - "step": 1003 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996604207535683, - "loss": 1.7144, - "step": 1004 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019966008252722724, - "loss": 1.9327, - "step": 1005 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019965974430088618, - "loss": 1.954, - "step": 1006 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996594060745451, - "loss": 1.9155, - "step": 1007 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019965906784820403, - "loss": 2.0901, - "step": 1008 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019965872962186297, - "loss": 1.9421, - "step": 1009 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996583913955219, - "loss": 1.8816, - "step": 1010 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019965805316918083, - "loss": 1.9014, - "step": 1011 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019965771494283977, - "loss": 1.9886, - "step": 1012 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019965737671649868, - "loss": 1.8961, - "step": 1013 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019965703849015762, - "loss": 1.8609, - "step": 1014 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019965670026381657, - "loss": 1.9128, - "step": 1015 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019965636203747548, - "loss": 1.8792, - "step": 1016 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019965602381113442, - "loss": 1.8966, - "step": 1017 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019965568558479336, - "loss": 1.9893, - "step": 1018 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019965534735845228, - "loss": 1.9097, - "step": 1019 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019965500913211122, - "loss": 1.8599, - "step": 1020 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019965467090577016, - "loss": 1.8929, - "step": 1021 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019965433267942907, - "loss": 1.9045, - "step": 1022 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199653994453088, - "loss": 1.9593, - "step": 1023 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019965365622674695, - "loss": 1.7983, - "step": 1024 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019965331800040587, - "loss": 1.9367, - "step": 1025 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996529797740648, - "loss": 2.0655, - "step": 1026 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019965264154772375, - "loss": 1.8532, - "step": 1027 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019965230332138266, - "loss": 1.9282, - "step": 1028 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996519650950416, - "loss": 1.8485, - "step": 1029 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019965162686870055, - "loss": 2.0562, - "step": 1030 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019965128864235946, - "loss": 1.969, - "step": 1031 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996509504160184, - "loss": 1.9293, - "step": 1032 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019965061218967734, - "loss": 2.007, - "step": 1033 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019965027396333626, - "loss": 1.7491, - "step": 1034 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996499357369952, - "loss": 1.98, - "step": 1035 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019964959751065414, - "loss": 1.7682, - "step": 1036 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019964925928431305, - "loss": 1.8933, - "step": 1037 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199648921057972, - "loss": 1.9841, - "step": 1038 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019964858283163094, - "loss": 1.8081, - "step": 1039 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019964824460528988, - "loss": 1.979, - "step": 1040 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019964790637894882, - "loss": 1.9011, - "step": 1041 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019964756815260773, - "loss": 1.9171, - "step": 1042 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019964722992626667, - "loss": 1.8862, - "step": 1043 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019964689169992561, - "loss": 1.8669, - "step": 1044 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019964655347358453, - "loss": 1.8324, - "step": 1045 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019964621524724347, - "loss": 1.8772, - "step": 1046 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996458770209024, - "loss": 1.9306, - "step": 1047 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019964553879456132, - "loss": 1.7231, - "step": 1048 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019964520056822027, - "loss": 1.9699, - "step": 1049 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996448623418792, - "loss": 1.8803, - "step": 1050 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019964452411553812, - "loss": 1.9907, - "step": 1051 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019964418588919706, - "loss": 1.9269, - "step": 1052 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199643847662856, - "loss": 1.9991, - "step": 1053 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019964350943651492, - "loss": 1.9628, - "step": 1054 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019964317121017386, - "loss": 2.0494, - "step": 1055 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996428329838328, - "loss": 1.8065, - "step": 1056 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019964249475749171, - "loss": 1.7516, - "step": 1057 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019964215653115066, - "loss": 1.9196, - "step": 1058 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996418183048096, - "loss": 1.8115, - "step": 1059 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996414800784685, - "loss": 1.959, - "step": 1060 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019964114185212745, - "loss": 1.7998, - "step": 1061 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996408036257864, - "loss": 1.9082, - "step": 1062 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996404653994453, - "loss": 1.7272, - "step": 1063 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019964012717310425, - "loss": 1.9251, - "step": 1064 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996397889467632, - "loss": 1.933, - "step": 1065 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019963945072042213, - "loss": 1.9945, - "step": 1066 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019963911249408104, - "loss": 2.0841, - "step": 1067 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019963877426773999, - "loss": 1.766, - "step": 1068 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019963843604139893, - "loss": 1.8715, - "step": 1069 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019963809781505784, - "loss": 2.0222, - "step": 1070 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019963775958871678, - "loss": 1.7361, - "step": 1071 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019963742136237572, - "loss": 1.8406, - "step": 1072 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019963708313603464, - "loss": 1.7735, - "step": 1073 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019963674490969358, - "loss": 1.9677, - "step": 1074 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019963640668335252, - "loss": 1.8519, - "step": 1075 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019963606845701143, - "loss": 2.0177, - "step": 1076 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019963573023067037, - "loss": 1.7989, - "step": 1077 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019963539200432932, - "loss": 2.0598, - "step": 1078 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019963505377798823, - "loss": 1.9022, - "step": 1079 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019963471555164717, - "loss": 1.8368, - "step": 1080 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996343773253061, - "loss": 1.9999, - "step": 1081 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019963403909896503, - "loss": 1.7713, - "step": 1082 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019963370087262397, - "loss": 1.9506, - "step": 1083 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996333626462829, - "loss": 1.9603, - "step": 1084 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019963302441994182, - "loss": 1.7847, - "step": 1085 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019963268619360076, - "loss": 1.8909, - "step": 1086 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996323479672597, - "loss": 1.9976, - "step": 1087 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019963200974091862, - "loss": 1.9781, - "step": 1088 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019963167151457756, - "loss": 1.9926, - "step": 1089 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996313332882365, - "loss": 1.8879, - "step": 1090 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019963099506189541, - "loss": 1.8478, - "step": 1091 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019963065683555436, - "loss": 1.8786, - "step": 1092 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996303186092133, - "loss": 1.898, - "step": 1093 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996299803828722, - "loss": 1.9755, - "step": 1094 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019962964215653115, - "loss": 2.0055, - "step": 1095 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996293039301901, - "loss": 1.8487, - "step": 1096 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019962896570384903, - "loss": 1.9463, - "step": 1097 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019962862747750798, - "loss": 1.9404, - "step": 1098 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996282892511669, - "loss": 1.7563, - "step": 1099 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019962795102482583, - "loss": 1.8961, - "step": 1100 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019962761279848477, - "loss": 1.7717, - "step": 1101 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019962727457214369, - "loss": 1.6769, - "step": 1102 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019962693634580263, - "loss": 1.9833, - "step": 1103 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019962659811946157, - "loss": 1.8615, - "step": 1104 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019962625989312048, - "loss": 1.7735, - "step": 1105 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019962592166677942, - "loss": 1.8323, - "step": 1106 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019962558344043836, - "loss": 1.7895, - "step": 1107 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019962524521409728, - "loss": 1.8405, - "step": 1108 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019962490698775622, - "loss": 1.9516, - "step": 1109 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019962456876141516, - "loss": 1.8927, - "step": 1110 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019962423053507407, - "loss": 1.8666, - "step": 1111 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019962389230873302, - "loss": 1.9853, - "step": 1112 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019962355408239196, - "loss": 1.9138, - "step": 1113 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019962321585605087, - "loss": 2.0184, - "step": 1114 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996228776297098, - "loss": 2.0006, - "step": 1115 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019962253940336875, - "loss": 1.9545, - "step": 1116 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019962220117702767, - "loss": 1.9769, - "step": 1117 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996218629506866, - "loss": 1.928, - "step": 1118 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019962152472434555, - "loss": 1.9999, - "step": 1119 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019962118649800446, - "loss": 1.9129, - "step": 1120 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996208482716634, - "loss": 2.048, - "step": 1121 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019962051004532235, - "loss": 1.8335, - "step": 1122 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019962017181898126, - "loss": 1.9505, - "step": 1123 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996198335926402, - "loss": 1.7848, - "step": 1124 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019961949536629914, - "loss": 1.9381, - "step": 1125 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019961915713995806, - "loss": 1.8338, - "step": 1126 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199618818913617, - "loss": 1.8811, - "step": 1127 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019961848068727594, - "loss": 1.9297, - "step": 1128 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019961814246093485, - "loss": 1.9322, - "step": 1129 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996178042345938, - "loss": 1.7595, - "step": 1130 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019961746600825273, - "loss": 1.8354, - "step": 1131 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019961712778191165, - "loss": 1.9103, - "step": 1132 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996167895555706, - "loss": 1.7702, - "step": 1133 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019961645132922953, - "loss": 2.0576, - "step": 1134 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019961611310288845, - "loss": 1.7796, - "step": 1135 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019961577487654739, - "loss": 1.8048, - "step": 1136 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019961543665020633, - "loss": 1.8631, - "step": 1137 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019961509842386524, - "loss": 1.958, - "step": 1138 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019961476019752418, - "loss": 1.9243, - "step": 1139 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019961442197118312, - "loss": 1.9523, - "step": 1140 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019961408374484204, - "loss": 1.9104, - "step": 1141 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019961374551850098, - "loss": 1.9834, - "step": 1142 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019961340729215992, - "loss": 1.8655, - "step": 1143 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019961306906581883, - "loss": 1.99, - "step": 1144 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019961273083947778, - "loss": 1.925, - "step": 1145 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019961239261313672, - "loss": 1.8752, - "step": 1146 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019961205438679563, - "loss": 1.8175, - "step": 1147 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019961171616045457, - "loss": 1.7881, - "step": 1148 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996113779341135, - "loss": 1.8631, - "step": 1149 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019961103970777245, - "loss": 1.8801, - "step": 1150 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019961070148143137, - "loss": 2.0025, - "step": 1151 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019961036325509034, - "loss": 1.93, - "step": 1152 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019961002502874925, - "loss": 1.8753, - "step": 1153 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996096868024082, - "loss": 1.8244, - "step": 1154 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019960934857606713, - "loss": 1.822, - "step": 1155 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019960901034972605, - "loss": 1.8942, - "step": 1156 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199608672123385, - "loss": 1.8642, - "step": 1157 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019960833389704393, - "loss": 1.954, - "step": 1158 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019960799567070284, - "loss": 1.895, - "step": 1159 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019960765744436178, - "loss": 1.9002, - "step": 1160 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019960731921802073, - "loss": 2.0335, - "step": 1161 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019960698099167964, - "loss": 1.8314, - "step": 1162 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019960664276533858, - "loss": 1.8792, - "step": 1163 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019960630453899752, - "loss": 1.8658, - "step": 1164 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019960596631265644, - "loss": 1.9683, - "step": 1165 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019960562808631538, - "loss": 1.9863, - "step": 1166 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019960528985997432, - "loss": 1.9556, - "step": 1167 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019960495163363323, - "loss": 2.0062, - "step": 1168 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019960461340729217, - "loss": 1.9963, - "step": 1169 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019960427518095111, - "loss": 2.0351, - "step": 1170 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019960393695461003, - "loss": 1.8918, - "step": 1171 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019960359872826897, - "loss": 1.9254, - "step": 1172 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996032605019279, - "loss": 1.887, - "step": 1173 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019960292227558682, - "loss": 1.7929, - "step": 1174 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019960258404924577, - "loss": 1.8612, - "step": 1175 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996022458229047, - "loss": 1.8591, - "step": 1176 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019960190759656362, - "loss": 1.8882, - "step": 1177 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019960156937022256, - "loss": 1.8665, - "step": 1178 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996012311438815, - "loss": 1.8981, - "step": 1179 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019960089291754042, - "loss": 1.8195, - "step": 1180 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019960055469119936, - "loss": 1.9167, - "step": 1181 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001996002164648583, - "loss": 2.0566, - "step": 1182 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995998782385172, - "loss": 1.9817, - "step": 1183 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019959954001217615, - "loss": 1.7609, - "step": 1184 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995992017858351, - "loss": 1.8776, - "step": 1185 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199598863559494, - "loss": 1.9174, - "step": 1186 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019959852533315295, - "loss": 1.8204, - "step": 1187 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995981871068119, - "loss": 1.8465, - "step": 1188 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995978488804708, - "loss": 1.8643, - "step": 1189 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019959751065412975, - "loss": 1.8964, - "step": 1190 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995971724277887, - "loss": 1.9615, - "step": 1191 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995968342014476, - "loss": 1.9958, - "step": 1192 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019959649597510654, - "loss": 1.97, - "step": 1193 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019959615774876548, - "loss": 1.8722, - "step": 1194 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995958195224244, - "loss": 1.8625, - "step": 1195 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019959548129608334, - "loss": 1.8921, - "step": 1196 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019959514306974228, - "loss": 1.8127, - "step": 1197 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995948048434012, - "loss": 1.8633, - "step": 1198 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019959446661706014, - "loss": 1.8481, - "step": 1199 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019959412839071908, - "loss": 1.6695, - "step": 1200 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199593790164378, - "loss": 1.9438, - "step": 1201 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019959345193803693, - "loss": 1.9112, - "step": 1202 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019959311371169587, - "loss": 1.895, - "step": 1203 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995927754853548, - "loss": 1.9713, - "step": 1204 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019959243725901373, - "loss": 1.8101, - "step": 1205 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019959209903267267, - "loss": 1.9409, - "step": 1206 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019959176080633158, - "loss": 1.9303, - "step": 1207 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019959142257999055, - "loss": 1.8362, - "step": 1208 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995910843536495, - "loss": 1.8739, - "step": 1209 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995907461273084, - "loss": 1.864, - "step": 1210 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019959040790096735, - "loss": 1.8808, - "step": 1211 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995900696746263, - "loss": 1.9444, - "step": 1212 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995897314482852, - "loss": 2.0089, - "step": 1213 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019958939322194414, - "loss": 1.7225, - "step": 1214 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019958905499560309, - "loss": 1.8974, - "step": 1215 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199588716769262, - "loss": 2.1063, - "step": 1216 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019958837854292094, - "loss": 1.8643, - "step": 1217 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019958804031657988, - "loss": 1.9271, - "step": 1218 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995877020902388, - "loss": 1.7252, - "step": 1219 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019958736386389774, - "loss": 1.9061, - "step": 1220 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019958702563755668, - "loss": 1.8397, - "step": 1221 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995866874112156, - "loss": 1.953, - "step": 1222 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019958634918487453, - "loss": 1.9083, - "step": 1223 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019958601095853347, - "loss": 2.0271, - "step": 1224 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995856727321924, - "loss": 1.9448, - "step": 1225 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019958533450585133, - "loss": 1.9186, - "step": 1226 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019958499627951027, - "loss": 1.9714, - "step": 1227 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019958465805316919, - "loss": 1.9668, - "step": 1228 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019958431982682813, - "loss": 1.8426, - "step": 1229 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019958398160048707, - "loss": 1.7283, - "step": 1230 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019958364337414598, - "loss": 1.9829, - "step": 1231 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019958330514780492, - "loss": 1.6623, - "step": 1232 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019958296692146386, - "loss": 1.836, - "step": 1233 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019958262869512278, - "loss": 1.9433, - "step": 1234 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019958229046878172, - "loss": 1.914, - "step": 1235 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019958195224244066, - "loss": 1.9171, - "step": 1236 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019958161401609957, - "loss": 1.9897, - "step": 1237 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019958127578975852, - "loss": 1.9906, - "step": 1238 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019958093756341746, - "loss": 1.843, - "step": 1239 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019958059933707637, - "loss": 1.8331, - "step": 1240 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995802611107353, - "loss": 1.8003, - "step": 1241 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019957992288439425, - "loss": 1.9715, - "step": 1242 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019957958465805317, - "loss": 1.8638, - "step": 1243 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995792464317121, - "loss": 1.9551, - "step": 1244 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019957890820537105, - "loss": 1.8778, - "step": 1245 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019957856997902996, - "loss": 1.8675, - "step": 1246 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995782317526889, - "loss": 1.8975, - "step": 1247 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019957789352634785, - "loss": 1.8596, - "step": 1248 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019957755530000676, - "loss": 1.8279, - "step": 1249 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995772170736657, - "loss": 1.8995, - "step": 1250 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019957687884732464, - "loss": 1.8555, - "step": 1251 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019957654062098356, - "loss": 1.7856, - "step": 1252 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995762023946425, - "loss": 1.8785, - "step": 1253 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019957586416830144, - "loss": 1.6919, - "step": 1254 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019957552594196035, - "loss": 1.9398, - "step": 1255 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995751877156193, - "loss": 1.8996, - "step": 1256 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019957484948927823, - "loss": 1.8848, - "step": 1257 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019957451126293715, - "loss": 1.8745, - "step": 1258 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995741730365961, - "loss": 1.8556, - "step": 1259 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019957383481025503, - "loss": 1.9224, - "step": 1260 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019957349658391394, - "loss": 2.0046, - "step": 1261 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019957315835757289, - "loss": 1.8366, - "step": 1262 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019957282013123183, - "loss": 1.9718, - "step": 1263 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019957248190489077, - "loss": 1.8507, - "step": 1264 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995721436785497, - "loss": 1.8942, - "step": 1265 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019957180545220865, - "loss": 1.7538, - "step": 1266 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019957146722586756, - "loss": 1.9892, - "step": 1267 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995711289995265, - "loss": 1.9963, - "step": 1268 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019957079077318545, - "loss": 1.8967, - "step": 1269 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019957045254684436, - "loss": 1.8922, - "step": 1270 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995701143205033, - "loss": 1.9845, - "step": 1271 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019956977609416224, - "loss": 1.821, - "step": 1272 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019956943786782116, - "loss": 1.8206, - "step": 1273 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995690996414801, - "loss": 1.8719, - "step": 1274 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019956876141513904, - "loss": 1.9254, - "step": 1275 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019956842318879795, - "loss": 1.9672, - "step": 1276 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995680849624569, - "loss": 1.8667, - "step": 1277 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019956774673611584, - "loss": 1.9204, - "step": 1278 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019956740850977475, - "loss": 1.8451, - "step": 1279 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995670702834337, - "loss": 1.8955, - "step": 1280 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019956673205709263, - "loss": 1.9896, - "step": 1281 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019956639383075155, - "loss": 1.7097, - "step": 1282 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995660556044105, - "loss": 1.8914, - "step": 1283 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019956571737806943, - "loss": 1.9358, - "step": 1284 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019956537915172834, - "loss": 1.9182, - "step": 1285 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019956504092538728, - "loss": 1.8407, - "step": 1286 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019956470269904622, - "loss": 1.9355, - "step": 1287 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019956436447270514, - "loss": 1.8537, - "step": 1288 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019956402624636408, - "loss": 1.7731, - "step": 1289 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019956368802002302, - "loss": 1.9486, - "step": 1290 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019956334979368193, - "loss": 1.8041, - "step": 1291 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019956301156734088, - "loss": 1.8068, - "step": 1292 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019956267334099982, - "loss": 1.8926, - "step": 1293 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019956233511465873, - "loss": 1.8554, - "step": 1294 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019956199688831767, - "loss": 1.8164, - "step": 1295 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995616586619766, - "loss": 1.9021, - "step": 1296 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019956132043563553, - "loss": 2.0477, - "step": 1297 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019956098220929447, - "loss": 1.9131, - "step": 1298 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995606439829534, - "loss": 1.8452, - "step": 1299 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019956030575661232, - "loss": 1.8748, - "step": 1300 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019955996753027126, - "loss": 2.0397, - "step": 1301 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995596293039302, - "loss": 1.8069, - "step": 1302 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019955929107758912, - "loss": 1.9138, - "step": 1303 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019955895285124806, - "loss": 1.8968, - "step": 1304 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199558614624907, - "loss": 1.7032, - "step": 1305 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019955827639856592, - "loss": 1.8711, - "step": 1306 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019955793817222486, - "loss": 1.7769, - "step": 1307 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995575999458838, - "loss": 1.9944, - "step": 1308 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995572617195427, - "loss": 1.968, - "step": 1309 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019955692349320165, - "loss": 1.7909, - "step": 1310 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995565852668606, - "loss": 1.824, - "step": 1311 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995562470405195, - "loss": 1.7957, - "step": 1312 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019955590881417845, - "loss": 1.9305, - "step": 1313 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995555705878374, - "loss": 1.8114, - "step": 1314 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995552323614963, - "loss": 1.8846, - "step": 1315 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019955489413515525, - "loss": 1.7874, - "step": 1316 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995545559088142, - "loss": 1.8876, - "step": 1317 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995542176824731, - "loss": 1.7833, - "step": 1318 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019955387945613204, - "loss": 1.8548, - "step": 1319 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019955354122979098, - "loss": 1.9003, - "step": 1320 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019955320300344992, - "loss": 1.9727, - "step": 1321 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019955286477710887, - "loss": 1.8935, - "step": 1322 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019955252655076778, - "loss": 1.9351, - "step": 1323 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019955218832442672, - "loss": 1.8372, - "step": 1324 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019955185009808566, - "loss": 1.8434, - "step": 1325 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019955151187174458, - "loss": 2.0064, - "step": 1326 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019955117364540352, - "loss": 1.776, - "step": 1327 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019955083541906246, - "loss": 1.7911, - "step": 1328 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019955049719272137, - "loss": 1.9975, - "step": 1329 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019955015896638031, - "loss": 1.8652, - "step": 1330 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019954982074003926, - "loss": 1.9539, - "step": 1331 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995494825136982, - "loss": 1.8738, - "step": 1332 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995491442873571, - "loss": 1.8175, - "step": 1333 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019954880606101605, - "loss": 1.915, - "step": 1334 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199548467834675, - "loss": 1.937, - "step": 1335 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995481296083339, - "loss": 1.7865, - "step": 1336 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019954779138199285, - "loss": 1.9663, - "step": 1337 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995474531556518, - "loss": 2.0452, - "step": 1338 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995471149293107, - "loss": 1.8407, - "step": 1339 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019954677670296964, - "loss": 1.7547, - "step": 1340 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019954643847662859, - "loss": 1.8531, - "step": 1341 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995461002502875, - "loss": 1.9296, - "step": 1342 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019954576202394644, - "loss": 1.8565, - "step": 1343 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019954542379760538, - "loss": 1.8036, - "step": 1344 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995450855712643, - "loss": 1.9054, - "step": 1345 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019954474734492324, - "loss": 1.8537, - "step": 1346 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019954440911858218, - "loss": 2.0505, - "step": 1347 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995440708922411, - "loss": 1.8371, - "step": 1348 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019954373266590003, - "loss": 2.0135, - "step": 1349 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019954339443955897, - "loss": 1.8168, - "step": 1350 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995430562132179, - "loss": 1.9224, - "step": 1351 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019954271798687683, - "loss": 1.8745, - "step": 1352 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019954237976053577, - "loss": 1.9233, - "step": 1353 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019954204153419468, - "loss": 1.8653, - "step": 1354 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019954170330785363, - "loss": 2.0045, - "step": 1355 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019954136508151257, - "loss": 1.9046, - "step": 1356 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019954102685517148, - "loss": 1.8163, - "step": 1357 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019954068862883042, - "loss": 1.891, - "step": 1358 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019954035040248936, - "loss": 1.7759, - "step": 1359 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019954001217614828, - "loss": 1.9296, - "step": 1360 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019953967394980722, - "loss": 1.8918, - "step": 1361 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019953933572346616, - "loss": 1.9324, - "step": 1362 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019953899749712507, - "loss": 1.8073, - "step": 1363 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019953865927078401, - "loss": 1.8355, - "step": 1364 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019953832104444296, - "loss": 1.9706, - "step": 1365 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019953798281810187, - "loss": 1.864, - "step": 1366 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995376445917608, - "loss": 1.9463, - "step": 1367 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019953730636541975, - "loss": 1.8556, - "step": 1368 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019953696813907867, - "loss": 1.9945, - "step": 1369 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995366299127376, - "loss": 1.8543, - "step": 1370 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019953629168639655, - "loss": 1.8816, - "step": 1371 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019953595346005546, - "loss": 2.0537, - "step": 1372 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995356152337144, - "loss": 2.0039, - "step": 1373 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019953527700737334, - "loss": 1.9207, - "step": 1374 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019953493878103226, - "loss": 1.7665, - "step": 1375 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995346005546912, - "loss": 1.8898, - "step": 1376 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019953426232835014, - "loss": 1.8838, - "step": 1377 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019953392410200908, - "loss": 1.836, - "step": 1378 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019953358587566802, - "loss": 1.8521, - "step": 1379 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019953324764932694, - "loss": 1.9037, - "step": 1380 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019953290942298588, - "loss": 1.8133, - "step": 1381 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019953257119664482, - "loss": 1.8853, - "step": 1382 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019953223297030373, - "loss": 1.8996, - "step": 1383 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019953189474396267, - "loss": 1.9392, - "step": 1384 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019953155651762162, - "loss": 1.8595, - "step": 1385 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019953121829128053, - "loss": 1.8727, - "step": 1386 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019953088006493947, - "loss": 1.8358, - "step": 1387 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995305418385984, - "loss": 1.8832, - "step": 1388 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019953020361225733, - "loss": 1.7473, - "step": 1389 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019952986538591627, - "loss": 1.8286, - "step": 1390 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995295271595752, - "loss": 1.9458, - "step": 1391 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019952918893323412, - "loss": 1.8538, - "step": 1392 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019952885070689306, - "loss": 1.861, - "step": 1393 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199528512480552, - "loss": 1.9091, - "step": 1394 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019952817425421092, - "loss": 1.7837, - "step": 1395 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019952783602786986, - "loss": 1.7871, - "step": 1396 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995274978015288, - "loss": 1.8995, - "step": 1397 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019952715957518772, - "loss": 1.894, - "step": 1398 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019952682134884666, - "loss": 1.8899, - "step": 1399 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995264831225056, - "loss": 1.7574, - "step": 1400 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995261448961645, - "loss": 1.9948, - "step": 1401 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019952580666982345, - "loss": 1.8764, - "step": 1402 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995254684434824, - "loss": 1.8052, - "step": 1403 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995251302171413, - "loss": 1.893, - "step": 1404 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019952479199080025, - "loss": 1.7149, - "step": 1405 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995244537644592, - "loss": 1.846, - "step": 1406 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995241155381181, - "loss": 2.0208, - "step": 1407 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019952377731177705, - "loss": 1.8804, - "step": 1408 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019952343908543599, - "loss": 1.8718, - "step": 1409 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995231008590949, - "loss": 1.7898, - "step": 1410 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019952276263275384, - "loss": 1.9508, - "step": 1411 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019952242440641278, - "loss": 1.9443, - "step": 1412 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995220861800717, - "loss": 1.9188, - "step": 1413 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019952174795373064, - "loss": 1.8222, - "step": 1414 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019952140972738958, - "loss": 1.9052, - "step": 1415 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995210715010485, - "loss": 1.9561, - "step": 1416 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019952073327470743, - "loss": 1.8924, - "step": 1417 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019952039504836638, - "loss": 1.9475, - "step": 1418 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019952005682202532, - "loss": 2.0012, - "step": 1419 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019951971859568423, - "loss": 1.9382, - "step": 1420 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019951938036934317, - "loss": 1.8406, - "step": 1421 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995190421430021, - "loss": 1.8011, - "step": 1422 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019951870391666103, - "loss": 1.7764, - "step": 1423 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019951836569031997, - "loss": 1.9127, - "step": 1424 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995180274639789, - "loss": 1.9321, - "step": 1425 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019951768923763782, - "loss": 1.8929, - "step": 1426 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019951735101129676, - "loss": 1.842, - "step": 1427 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995170127849557, - "loss": 1.8398, - "step": 1428 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019951667455861462, - "loss": 1.8911, - "step": 1429 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019951633633227356, - "loss": 2.0175, - "step": 1430 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995159981059325, - "loss": 1.796, - "step": 1431 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019951565987959142, - "loss": 1.781, - "step": 1432 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019951532165325038, - "loss": 1.9367, - "step": 1433 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995149834269093, - "loss": 1.8485, - "step": 1434 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019951464520056824, - "loss": 1.8459, - "step": 1435 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019951430697422718, - "loss": 1.9, - "step": 1436 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995139687478861, - "loss": 1.7367, - "step": 1437 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019951363052154504, - "loss": 1.822, - "step": 1438 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019951329229520398, - "loss": 1.8054, - "step": 1439 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995129540688629, - "loss": 1.7672, - "step": 1440 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019951261584252183, - "loss": 1.7976, - "step": 1441 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019951227761618077, - "loss": 1.9582, - "step": 1442 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995119393898397, - "loss": 1.8741, - "step": 1443 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019951160116349863, - "loss": 1.9291, - "step": 1444 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019951126293715757, - "loss": 1.8814, - "step": 1445 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019951092471081648, - "loss": 1.8493, - "step": 1446 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019951058648447542, - "loss": 1.9328, - "step": 1447 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019951024825813437, - "loss": 1.7882, - "step": 1448 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019950991003179328, - "loss": 1.7946, - "step": 1449 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019950957180545222, - "loss": 1.8159, - "step": 1450 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019950923357911116, - "loss": 1.7148, - "step": 1451 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019950889535277008, - "loss": 1.8748, - "step": 1452 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019950855712642902, - "loss": 1.7323, - "step": 1453 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019950821890008796, - "loss": 1.7762, - "step": 1454 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019950788067374687, - "loss": 1.8678, - "step": 1455 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995075424474058, - "loss": 1.7746, - "step": 1456 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019950720422106475, - "loss": 1.8391, - "step": 1457 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019950686599472367, - "loss": 1.891, - "step": 1458 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995065277683826, - "loss": 1.9739, - "step": 1459 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019950618954204155, - "loss": 1.9782, - "step": 1460 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019950585131570046, - "loss": 1.8611, - "step": 1461 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995055130893594, - "loss": 1.8454, - "step": 1462 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019950517486301835, - "loss": 1.9725, - "step": 1463 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019950483663667726, - "loss": 1.8951, - "step": 1464 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995044984103362, - "loss": 1.9838, - "step": 1465 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019950416018399514, - "loss": 1.8764, - "step": 1466 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019950382195765406, - "loss": 1.865, - "step": 1467 - }, - { - "epoch": 0.0, - "learning_rate": 0.000199503483731313, - "loss": 1.8893, - "step": 1468 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019950314550497194, - "loss": 1.9259, - "step": 1469 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019950280727863085, - "loss": 1.9197, - "step": 1470 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995024690522898, - "loss": 1.8822, - "step": 1471 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019950213082594874, - "loss": 2.0199, - "step": 1472 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019950179259960765, - "loss": 1.8202, - "step": 1473 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995014543732666, - "loss": 1.848, - "step": 1474 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019950111614692553, - "loss": 1.7342, - "step": 1475 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019950077792058445, - "loss": 1.9218, - "step": 1476 - }, - { - "epoch": 0.0, - "learning_rate": 0.0001995004396942434, - "loss": 1.7787, - "step": 1477 - }, - { - "epoch": 0.0, - "learning_rate": 0.00019950010146790233, - "loss": 1.8392, - "step": 1478 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019949976324156124, - "loss": 1.9108, - "step": 1479 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019949942501522018, - "loss": 1.8879, - "step": 1480 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019949908678887912, - "loss": 1.9064, - "step": 1481 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019949874856253804, - "loss": 1.9958, - "step": 1482 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019949841033619698, - "loss": 1.8441, - "step": 1483 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019949807210985592, - "loss": 1.9056, - "step": 1484 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019949773388351484, - "loss": 1.7298, - "step": 1485 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019949739565717378, - "loss": 1.8564, - "step": 1486 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019949705743083272, - "loss": 1.8349, - "step": 1487 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019949671920449163, - "loss": 1.7678, - "step": 1488 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994963809781506, - "loss": 1.7357, - "step": 1489 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019949604275180954, - "loss": 1.886, - "step": 1490 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019949570452546846, - "loss": 2.0175, - "step": 1491 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994953662991274, - "loss": 1.897, - "step": 1492 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019949502807278634, - "loss": 1.8954, - "step": 1493 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019949468984644525, - "loss": 1.9095, - "step": 1494 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994943516201042, - "loss": 1.8761, - "step": 1495 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019949401339376313, - "loss": 1.823, - "step": 1496 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019949367516742205, - "loss": 1.8956, - "step": 1497 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199493336941081, - "loss": 1.7625, - "step": 1498 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019949299871473993, - "loss": 1.9774, - "step": 1499 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019949266048839884, - "loss": 1.8647, - "step": 1500 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019949232226205779, - "loss": 1.8094, - "step": 1501 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019949198403571673, - "loss": 1.7794, - "step": 1502 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019949164580937564, - "loss": 1.875, - "step": 1503 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019949130758303458, - "loss": 1.6871, - "step": 1504 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019949096935669352, - "loss": 1.8877, - "step": 1505 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019949063113035244, - "loss": 1.9389, - "step": 1506 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019949029290401138, - "loss": 1.8539, - "step": 1507 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019948995467767032, - "loss": 1.9144, - "step": 1508 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019948961645132923, - "loss": 1.8838, - "step": 1509 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019948927822498817, - "loss": 1.9858, - "step": 1510 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019948893999864712, - "loss": 1.8796, - "step": 1511 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019948860177230603, - "loss": 1.8004, - "step": 1512 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019948826354596497, - "loss": 1.8938, - "step": 1513 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994879253196239, - "loss": 1.7736, - "step": 1514 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019948758709328283, - "loss": 1.9286, - "step": 1515 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019948724886694177, - "loss": 1.8208, - "step": 1516 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994869106406007, - "loss": 1.8787, - "step": 1517 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019948657241425962, - "loss": 1.9609, - "step": 1518 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019948623418791856, - "loss": 1.8266, - "step": 1519 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994858959615775, - "loss": 1.9125, - "step": 1520 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019948555773523642, - "loss": 1.9485, - "step": 1521 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019948521950889536, - "loss": 1.9031, - "step": 1522 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994848812825543, - "loss": 1.9137, - "step": 1523 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019948454305621321, - "loss": 1.7483, - "step": 1524 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019948420482987216, - "loss": 1.884, - "step": 1525 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994838666035311, - "loss": 1.8349, - "step": 1526 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019948352837719, - "loss": 1.8344, - "step": 1527 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019948319015084895, - "loss": 1.8423, - "step": 1528 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994828519245079, - "loss": 1.961, - "step": 1529 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994825136981668, - "loss": 1.7416, - "step": 1530 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019948217547182575, - "loss": 1.9619, - "step": 1531 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994818372454847, - "loss": 1.7832, - "step": 1532 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994814990191436, - "loss": 1.8764, - "step": 1533 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019948116079280254, - "loss": 2.0549, - "step": 1534 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019948082256646149, - "loss": 1.8407, - "step": 1535 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994804843401204, - "loss": 1.8681, - "step": 1536 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019948014611377934, - "loss": 1.8282, - "step": 1537 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019947980788743828, - "loss": 1.8112, - "step": 1538 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994794696610972, - "loss": 1.9159, - "step": 1539 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019947913143475614, - "loss": 1.925, - "step": 1540 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019947879320841508, - "loss": 1.896, - "step": 1541 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199478454982074, - "loss": 1.9694, - "step": 1542 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019947811675573293, - "loss": 1.8311, - "step": 1543 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019947777852939187, - "loss": 1.8514, - "step": 1544 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019947744030305082, - "loss": 1.8631, - "step": 1545 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019947710207670976, - "loss": 2.0501, - "step": 1546 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994767638503687, - "loss": 1.9055, - "step": 1547 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994764256240276, - "loss": 1.8165, - "step": 1548 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019947608739768655, - "loss": 1.8067, - "step": 1549 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994757491713455, - "loss": 1.7996, - "step": 1550 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994754109450044, - "loss": 1.8239, - "step": 1551 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019947507271866335, - "loss": 1.8411, - "step": 1552 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994747344923223, - "loss": 1.9147, - "step": 1553 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994743962659812, - "loss": 1.7427, - "step": 1554 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019947405803964015, - "loss": 1.9948, - "step": 1555 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994737198132991, - "loss": 1.869, - "step": 1556 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199473381586958, - "loss": 1.8652, - "step": 1557 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019947304336061694, - "loss": 1.7643, - "step": 1558 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019947270513427588, - "loss": 1.7406, - "step": 1559 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994723669079348, - "loss": 1.9571, - "step": 1560 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019947202868159374, - "loss": 1.8456, - "step": 1561 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019947169045525268, - "loss": 1.8573, - "step": 1562 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994713522289116, - "loss": 1.7982, - "step": 1563 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019947101400257053, - "loss": 1.9119, - "step": 1564 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019947067577622948, - "loss": 1.8072, - "step": 1565 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994703375498884, - "loss": 1.7621, - "step": 1566 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019946999932354733, - "loss": 1.9575, - "step": 1567 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019946966109720627, - "loss": 1.7792, - "step": 1568 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019946932287086519, - "loss": 1.774, - "step": 1569 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019946898464452413, - "loss": 1.8808, - "step": 1570 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019946864641818307, - "loss": 1.9068, - "step": 1571 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019946830819184198, - "loss": 1.8717, - "step": 1572 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019946796996550092, - "loss": 1.8882, - "step": 1573 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019946763173915986, - "loss": 1.8951, - "step": 1574 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019946729351281878, - "loss": 1.7323, - "step": 1575 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019946695528647772, - "loss": 1.9054, - "step": 1576 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019946661706013666, - "loss": 1.7965, - "step": 1577 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019946627883379558, - "loss": 1.7813, - "step": 1578 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019946594060745452, - "loss": 2.0227, - "step": 1579 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019946560238111346, - "loss": 1.8276, - "step": 1580 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019946526415477237, - "loss": 1.9047, - "step": 1581 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994649259284313, - "loss": 1.7771, - "step": 1582 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019946458770209025, - "loss": 1.933, - "step": 1583 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019946424947574917, - "loss": 1.886, - "step": 1584 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994639112494081, - "loss": 1.8768, - "step": 1585 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019946357302306705, - "loss": 1.8213, - "step": 1586 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019946323479672596, - "loss": 1.8658, - "step": 1587 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994628965703849, - "loss": 2.0283, - "step": 1588 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019946255834404385, - "loss": 1.8465, - "step": 1589 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019946222011770276, - "loss": 1.9562, - "step": 1590 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994618818913617, - "loss": 1.9186, - "step": 1591 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019946154366502064, - "loss": 1.82, - "step": 1592 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019946120543867956, - "loss": 1.881, - "step": 1593 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994608672123385, - "loss": 1.8559, - "step": 1594 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019946052898599744, - "loss": 1.891, - "step": 1595 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019946019075965635, - "loss": 1.9574, - "step": 1596 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994598525333153, - "loss": 1.8666, - "step": 1597 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019945951430697424, - "loss": 2.0037, - "step": 1598 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019945917608063315, - "loss": 1.9873, - "step": 1599 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994588378542921, - "loss": 1.7198, - "step": 1600 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019945849962795106, - "loss": 1.7982, - "step": 1601 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019945816140160997, - "loss": 1.7188, - "step": 1602 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019945782317526891, - "loss": 1.8532, - "step": 1603 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019945748494892786, - "loss": 1.9109, - "step": 1604 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019945714672258677, - "loss": 1.8561, - "step": 1605 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994568084962457, - "loss": 1.8464, - "step": 1606 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019945647026990465, - "loss": 1.9152, - "step": 1607 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019945613204356357, - "loss": 1.8087, - "step": 1608 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994557938172225, - "loss": 1.7949, - "step": 1609 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019945545559088145, - "loss": 1.8028, - "step": 1610 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019945511736454036, - "loss": 1.8267, - "step": 1611 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994547791381993, - "loss": 1.7683, - "step": 1612 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019945444091185824, - "loss": 1.8278, - "step": 1613 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019945410268551716, - "loss": 1.7242, - "step": 1614 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994537644591761, - "loss": 1.9983, - "step": 1615 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019945342623283504, - "loss": 1.9402, - "step": 1616 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019945308800649395, - "loss": 1.7418, - "step": 1617 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994527497801529, - "loss": 1.8811, - "step": 1618 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019945241155381184, - "loss": 1.8105, - "step": 1619 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019945207332747075, - "loss": 2.0333, - "step": 1620 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994517351011297, - "loss": 2.0502, - "step": 1621 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019945139687478863, - "loss": 1.8603, - "step": 1622 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019945105864844755, - "loss": 1.9698, - "step": 1623 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994507204221065, - "loss": 1.763, - "step": 1624 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019945038219576543, - "loss": 1.844, - "step": 1625 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019945004396942434, - "loss": 1.8767, - "step": 1626 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019944970574308328, - "loss": 1.9864, - "step": 1627 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019944936751674223, - "loss": 1.8595, - "step": 1628 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019944902929040114, - "loss": 1.8575, - "step": 1629 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019944869106406008, - "loss": 1.8451, - "step": 1630 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019944835283771902, - "loss": 1.8555, - "step": 1631 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019944801461137794, - "loss": 1.9568, - "step": 1632 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019944767638503688, - "loss": 1.9689, - "step": 1633 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019944733815869582, - "loss": 1.9532, - "step": 1634 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019944699993235473, - "loss": 1.7458, - "step": 1635 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019944666170601367, - "loss": 1.7541, - "step": 1636 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019944632347967261, - "loss": 1.8212, - "step": 1637 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019944598525333153, - "loss": 1.9328, - "step": 1638 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019944564702699047, - "loss": 1.7832, - "step": 1639 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994453088006494, - "loss": 1.8547, - "step": 1640 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019944497057430832, - "loss": 1.9043, - "step": 1641 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019944463234796727, - "loss": 1.8925, - "step": 1642 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994442941216262, - "loss": 1.8786, - "step": 1643 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019944395589528512, - "loss": 1.9301, - "step": 1644 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019944361766894406, - "loss": 1.7019, - "step": 1645 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199443279442603, - "loss": 1.8772, - "step": 1646 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019944294121626192, - "loss": 1.7796, - "step": 1647 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019944260298992086, - "loss": 1.777, - "step": 1648 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994422647635798, - "loss": 1.8031, - "step": 1649 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019944192653723871, - "loss": 1.695, - "step": 1650 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019944158831089766, - "loss": 1.8826, - "step": 1651 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994412500845566, - "loss": 1.796, - "step": 1652 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994409118582155, - "loss": 1.7995, - "step": 1653 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019944057363187445, - "loss": 1.8269, - "step": 1654 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994402354055334, - "loss": 1.8535, - "step": 1655 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994398971791923, - "loss": 1.9288, - "step": 1656 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019943955895285127, - "loss": 1.9406, - "step": 1657 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994392207265102, - "loss": 1.9033, - "step": 1658 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019943888250016913, - "loss": 1.8964, - "step": 1659 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019943854427382807, - "loss": 1.9057, - "step": 1660 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019943820604748699, - "loss": 1.9562, - "step": 1661 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019943786782114593, - "loss": 1.7726, - "step": 1662 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019943752959480487, - "loss": 1.8357, - "step": 1663 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019943719136846378, - "loss": 1.8279, - "step": 1664 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019943685314212272, - "loss": 1.8362, - "step": 1665 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019943651491578166, - "loss": 2.0656, - "step": 1666 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019943617668944058, - "loss": 1.9201, - "step": 1667 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019943583846309952, - "loss": 1.8445, - "step": 1668 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019943550023675846, - "loss": 1.9179, - "step": 1669 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019943516201041737, - "loss": 1.8928, - "step": 1670 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019943482378407632, - "loss": 1.7252, - "step": 1671 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019943448555773526, - "loss": 1.8646, - "step": 1672 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019943414733139417, - "loss": 1.832, - "step": 1673 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994338091050531, - "loss": 1.9539, - "step": 1674 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019943347087871205, - "loss": 1.9176, - "step": 1675 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019943313265237097, - "loss": 1.7569, - "step": 1676 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994327944260299, - "loss": 1.7935, - "step": 1677 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019943245619968885, - "loss": 1.8096, - "step": 1678 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019943211797334776, - "loss": 1.7881, - "step": 1679 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994317797470067, - "loss": 1.8804, - "step": 1680 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019943144152066565, - "loss": 1.7554, - "step": 1681 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019943110329432456, - "loss": 1.8889, - "step": 1682 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994307650679835, - "loss": 1.8215, - "step": 1683 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019943042684164244, - "loss": 1.8098, - "step": 1684 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019943008861530138, - "loss": 1.9936, - "step": 1685 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994297503889603, - "loss": 1.8097, - "step": 1686 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019942941216261924, - "loss": 1.7269, - "step": 1687 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019942907393627818, - "loss": 1.98, - "step": 1688 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994287357099371, - "loss": 1.9464, - "step": 1689 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019942839748359603, - "loss": 2.029, - "step": 1690 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019942805925725498, - "loss": 1.9617, - "step": 1691 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994277210309139, - "loss": 1.6267, - "step": 1692 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019942738280457283, - "loss": 1.8661, - "step": 1693 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019942704457823177, - "loss": 1.9497, - "step": 1694 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019942670635189069, - "loss": 1.7926, - "step": 1695 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019942636812554963, - "loss": 1.9456, - "step": 1696 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019942602989920857, - "loss": 1.8189, - "step": 1697 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019942569167286748, - "loss": 1.6931, - "step": 1698 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019942535344652642, - "loss": 1.9279, - "step": 1699 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019942501522018536, - "loss": 1.793, - "step": 1700 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019942467699384428, - "loss": 1.9155, - "step": 1701 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019942433876750322, - "loss": 1.9946, - "step": 1702 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019942400054116216, - "loss": 1.7868, - "step": 1703 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019942366231482107, - "loss": 1.8781, - "step": 1704 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019942332408848002, - "loss": 1.7606, - "step": 1705 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019942298586213896, - "loss": 1.8568, - "step": 1706 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019942264763579787, - "loss": 1.9374, - "step": 1707 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994223094094568, - "loss": 1.8422, - "step": 1708 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019942197118311575, - "loss": 1.9321, - "step": 1709 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019942163295677467, - "loss": 1.8215, - "step": 1710 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994212947304336, - "loss": 1.8235, - "step": 1711 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019942095650409255, - "loss": 1.813, - "step": 1712 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019942061827775146, - "loss": 1.7612, - "step": 1713 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019942028005141043, - "loss": 1.7606, - "step": 1714 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019941994182506935, - "loss": 1.6454, - "step": 1715 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994196035987283, - "loss": 1.8045, - "step": 1716 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019941926537238723, - "loss": 1.7704, - "step": 1717 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019941892714604614, - "loss": 1.9219, - "step": 1718 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019941858891970508, - "loss": 1.8912, - "step": 1719 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019941825069336402, - "loss": 1.9892, - "step": 1720 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019941791246702294, - "loss": 1.9207, - "step": 1721 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019941757424068188, - "loss": 1.8327, - "step": 1722 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019941723601434082, - "loss": 1.8961, - "step": 1723 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019941689778799973, - "loss": 1.8998, - "step": 1724 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019941655956165868, - "loss": 1.7239, - "step": 1725 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019941622133531762, - "loss": 1.8178, - "step": 1726 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019941588310897653, - "loss": 1.8139, - "step": 1727 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019941554488263547, - "loss": 1.9952, - "step": 1728 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994152066562944, - "loss": 1.8214, - "step": 1729 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019941486842995333, - "loss": 1.8802, - "step": 1730 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019941453020361227, - "loss": 1.9637, - "step": 1731 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994141919772712, - "loss": 1.8512, - "step": 1732 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019941385375093012, - "loss": 1.88, - "step": 1733 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019941351552458906, - "loss": 1.8416, - "step": 1734 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199413177298248, - "loss": 1.7734, - "step": 1735 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019941283907190692, - "loss": 1.8393, - "step": 1736 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019941250084556586, - "loss": 1.741, - "step": 1737 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994121626192248, - "loss": 1.8987, - "step": 1738 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019941182439288372, - "loss": 1.9043, - "step": 1739 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019941148616654266, - "loss": 1.936, - "step": 1740 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994111479402016, - "loss": 1.906, - "step": 1741 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994108097138605, - "loss": 1.8786, - "step": 1742 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019941047148751945, - "loss": 1.7951, - "step": 1743 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994101332611784, - "loss": 1.8268, - "step": 1744 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994097950348373, - "loss": 2.0345, - "step": 1745 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019940945680849625, - "loss": 1.9378, - "step": 1746 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994091185821552, - "loss": 1.6588, - "step": 1747 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994087803558141, - "loss": 1.8876, - "step": 1748 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019940844212947305, - "loss": 1.7273, - "step": 1749 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199408103903132, - "loss": 1.8716, - "step": 1750 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994077656767909, - "loss": 1.7438, - "step": 1751 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019940742745044984, - "loss": 1.7389, - "step": 1752 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019940708922410878, - "loss": 1.699, - "step": 1753 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994067509977677, - "loss": 1.833, - "step": 1754 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019940641277142664, - "loss": 1.8185, - "step": 1755 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019940607454508558, - "loss": 2.0406, - "step": 1756 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994057363187445, - "loss": 1.8316, - "step": 1757 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019940539809240344, - "loss": 1.8711, - "step": 1758 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019940505986606238, - "loss": 1.8134, - "step": 1759 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994047216397213, - "loss": 1.8689, - "step": 1760 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019940438341338023, - "loss": 1.9067, - "step": 1761 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019940404518703917, - "loss": 1.8146, - "step": 1762 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994037069606981, - "loss": 1.8164, - "step": 1763 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019940336873435703, - "loss": 1.7447, - "step": 1764 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019940303050801597, - "loss": 1.8343, - "step": 1765 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019940269228167488, - "loss": 1.8219, - "step": 1766 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019940235405533382, - "loss": 1.8982, - "step": 1767 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019940201582899277, - "loss": 1.6188, - "step": 1768 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994016776026517, - "loss": 1.7354, - "step": 1769 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019940133937631065, - "loss": 1.935, - "step": 1770 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994010011499696, - "loss": 1.7433, - "step": 1771 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001994006629236285, - "loss": 1.8546, - "step": 1772 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019940032469728744, - "loss": 1.8137, - "step": 1773 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019939998647094639, - "loss": 1.9303, - "step": 1774 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993996482446053, - "loss": 1.8847, - "step": 1775 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019939931001826424, - "loss": 1.7863, - "step": 1776 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019939897179192318, - "loss": 1.8506, - "step": 1777 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993986335655821, - "loss": 1.8437, - "step": 1778 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019939829533924104, - "loss": 1.9733, - "step": 1779 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019939795711289998, - "loss": 1.7875, - "step": 1780 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993976188865589, - "loss": 1.7944, - "step": 1781 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019939728066021783, - "loss": 2.0036, - "step": 1782 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019939694243387677, - "loss": 1.8845, - "step": 1783 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993966042075357, - "loss": 1.9257, - "step": 1784 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019939626598119463, - "loss": 2.0166, - "step": 1785 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019939592775485357, - "loss": 1.7989, - "step": 1786 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019939558952851248, - "loss": 1.9164, - "step": 1787 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019939525130217143, - "loss": 1.8561, - "step": 1788 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019939491307583037, - "loss": 1.8148, - "step": 1789 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019939457484948928, - "loss": 1.9047, - "step": 1790 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019939423662314822, - "loss": 1.9254, - "step": 1791 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019939389839680716, - "loss": 1.9061, - "step": 1792 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019939356017046608, - "loss": 1.827, - "step": 1793 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019939322194412502, - "loss": 1.75, - "step": 1794 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019939288371778396, - "loss": 1.817, - "step": 1795 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019939254549144287, - "loss": 1.7326, - "step": 1796 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019939220726510181, - "loss": 1.9381, - "step": 1797 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019939186903876076, - "loss": 1.9337, - "step": 1798 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019939153081241967, - "loss": 1.8366, - "step": 1799 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993911925860786, - "loss": 1.7377, - "step": 1800 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019939085435973755, - "loss": 1.9473, - "step": 1801 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019939051613339647, - "loss": 1.9701, - "step": 1802 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993901779070554, - "loss": 1.9682, - "step": 1803 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019938983968071435, - "loss": 1.7879, - "step": 1804 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019938950145437326, - "loss": 1.8519, - "step": 1805 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993891632280322, - "loss": 1.9074, - "step": 1806 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019938882500169114, - "loss": 1.8986, - "step": 1807 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019938848677535006, - "loss": 1.8993, - "step": 1808 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199388148549009, - "loss": 1.7601, - "step": 1809 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019938781032266794, - "loss": 1.8469, - "step": 1810 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019938747209632685, - "loss": 1.7413, - "step": 1811 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993871338699858, - "loss": 1.848, - "step": 1812 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019938679564364474, - "loss": 1.7714, - "step": 1813 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019938645741730365, - "loss": 1.783, - "step": 1814 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993861191909626, - "loss": 1.9162, - "step": 1815 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019938578096462153, - "loss": 1.8673, - "step": 1816 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019938544273828045, - "loss": 1.9262, - "step": 1817 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993851045119394, - "loss": 1.9706, - "step": 1818 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019938476628559833, - "loss": 1.7608, - "step": 1819 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019938442805925724, - "loss": 1.8998, - "step": 1820 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019938408983291619, - "loss": 1.8073, - "step": 1821 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019938375160657513, - "loss": 1.9073, - "step": 1822 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019938341338023404, - "loss": 1.7754, - "step": 1823 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019938307515389298, - "loss": 1.9185, - "step": 1824 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019938273692755192, - "loss": 1.9135, - "step": 1825 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019938239870121086, - "loss": 1.7988, - "step": 1826 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993820604748698, - "loss": 1.8674, - "step": 1827 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019938172224852875, - "loss": 1.9293, - "step": 1828 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019938138402218766, - "loss": 1.8211, - "step": 1829 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993810457958466, - "loss": 1.7844, - "step": 1830 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019938070756950554, - "loss": 1.7842, - "step": 1831 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019938036934316446, - "loss": 1.9117, - "step": 1832 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993800311168234, - "loss": 1.7131, - "step": 1833 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019937969289048234, - "loss": 1.8645, - "step": 1834 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019937935466414125, - "loss": 1.9347, - "step": 1835 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993790164378002, - "loss": 1.8397, - "step": 1836 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019937867821145913, - "loss": 1.7218, - "step": 1837 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019937833998511805, - "loss": 1.9146, - "step": 1838 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199378001758777, - "loss": 1.8708, - "step": 1839 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019937766353243593, - "loss": 1.9049, - "step": 1840 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019937732530609485, - "loss": 1.8359, - "step": 1841 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019937698707975379, - "loss": 1.9401, - "step": 1842 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019937664885341273, - "loss": 1.8512, - "step": 1843 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019937631062707164, - "loss": 1.6085, - "step": 1844 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019937597240073058, - "loss": 1.8671, - "step": 1845 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019937563417438952, - "loss": 1.8838, - "step": 1846 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019937529594804844, - "loss": 1.873, - "step": 1847 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019937495772170738, - "loss": 2.0318, - "step": 1848 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019937461949536632, - "loss": 1.9601, - "step": 1849 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019937428126902523, - "loss": 1.9242, - "step": 1850 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019937394304268418, - "loss": 1.7387, - "step": 1851 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019937360481634312, - "loss": 1.911, - "step": 1852 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019937326659000203, - "loss": 1.8368, - "step": 1853 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019937292836366097, - "loss": 1.7509, - "step": 1854 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993725901373199, - "loss": 1.9419, - "step": 1855 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019937225191097883, - "loss": 1.8011, - "step": 1856 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019937191368463777, - "loss": 1.8517, - "step": 1857 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993715754582967, - "loss": 1.8313, - "step": 1858 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019937123723195562, - "loss": 1.7397, - "step": 1859 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019937089900561456, - "loss": 1.8399, - "step": 1860 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993705607792735, - "loss": 1.9579, - "step": 1861 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019937022255293242, - "loss": 1.8697, - "step": 1862 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019936988432659136, - "loss": 1.8475, - "step": 1863 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993695461002503, - "loss": 1.756, - "step": 1864 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019936920787390922, - "loss": 1.8965, - "step": 1865 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019936886964756816, - "loss": 1.8576, - "step": 1866 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993685314212271, - "loss": 1.9051, - "step": 1867 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199368193194886, - "loss": 1.8938, - "step": 1868 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019936785496854495, - "loss": 1.8301, - "step": 1869 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993675167422039, - "loss": 1.8929, - "step": 1870 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993671785158628, - "loss": 1.9947, - "step": 1871 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019936684028952175, - "loss": 1.7735, - "step": 1872 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993665020631807, - "loss": 1.6857, - "step": 1873 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993661638368396, - "loss": 1.7633, - "step": 1874 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019936582561049855, - "loss": 1.8781, - "step": 1875 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993654873841575, - "loss": 1.8398, - "step": 1876 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993651491578164, - "loss": 1.9035, - "step": 1877 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019936481093147534, - "loss": 1.7455, - "step": 1878 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019936447270513428, - "loss": 1.7548, - "step": 1879 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993641344787932, - "loss": 2.0543, - "step": 1880 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019936379625245214, - "loss": 1.795, - "step": 1881 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993634580261111, - "loss": 1.8485, - "step": 1882 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019936311979977002, - "loss": 1.9616, - "step": 1883 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019936278157342896, - "loss": 1.8312, - "step": 1884 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993624433470879, - "loss": 1.924, - "step": 1885 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019936210512074682, - "loss": 1.8517, - "step": 1886 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019936176689440576, - "loss": 1.7755, - "step": 1887 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993614286680647, - "loss": 1.7478, - "step": 1888 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993610904417236, - "loss": 1.873, - "step": 1889 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019936075221538255, - "loss": 1.8581, - "step": 1890 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993604139890415, - "loss": 1.8985, - "step": 1891 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993600757627004, - "loss": 1.8979, - "step": 1892 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019935973753635935, - "loss": 1.84, - "step": 1893 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993593993100183, - "loss": 1.8513, - "step": 1894 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993590610836772, - "loss": 2.0324, - "step": 1895 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019935872285733615, - "loss": 1.8805, - "step": 1896 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993583846309951, - "loss": 1.9493, - "step": 1897 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199358046404654, - "loss": 1.753, - "step": 1898 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019935770817831294, - "loss": 1.8079, - "step": 1899 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019935736995197188, - "loss": 1.8253, - "step": 1900 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993570317256308, - "loss": 1.9642, - "step": 1901 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019935669349928974, - "loss": 1.7827, - "step": 1902 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019935635527294868, - "loss": 1.8492, - "step": 1903 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993560170466076, - "loss": 1.8559, - "step": 1904 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019935567882026654, - "loss": 1.9028, - "step": 1905 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019935534059392548, - "loss": 1.7881, - "step": 1906 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993550023675844, - "loss": 1.8746, - "step": 1907 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019935466414124333, - "loss": 1.7154, - "step": 1908 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019935432591490227, - "loss": 1.9828, - "step": 1909 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993539876885612, - "loss": 1.7195, - "step": 1910 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019935364946222013, - "loss": 1.9149, - "step": 1911 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019935331123587907, - "loss": 1.6488, - "step": 1912 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019935297300953798, - "loss": 1.8448, - "step": 1913 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019935263478319692, - "loss": 1.7274, - "step": 1914 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019935229655685587, - "loss": 1.7636, - "step": 1915 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019935195833051478, - "loss": 1.8117, - "step": 1916 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019935162010417372, - "loss": 1.7385, - "step": 1917 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019935128187783266, - "loss": 1.9489, - "step": 1918 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019935094365149158, - "loss": 1.8217, - "step": 1919 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019935060542515052, - "loss": 1.953, - "step": 1920 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019935026719880946, - "loss": 1.8128, - "step": 1921 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019934992897246837, - "loss": 1.7606, - "step": 1922 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019934959074612731, - "loss": 1.6699, - "step": 1923 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019934925251978626, - "loss": 1.741, - "step": 1924 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019934891429344517, - "loss": 1.7387, - "step": 1925 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993485760671041, - "loss": 1.9973, - "step": 1926 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019934823784076305, - "loss": 1.7484, - "step": 1927 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019934789961442197, - "loss": 1.6749, - "step": 1928 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993475613880809, - "loss": 1.7288, - "step": 1929 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019934722316173985, - "loss": 1.679, - "step": 1930 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019934688493539876, - "loss": 1.8897, - "step": 1931 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993465467090577, - "loss": 1.6933, - "step": 1932 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019934620848271664, - "loss": 1.8756, - "step": 1933 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019934587025637556, - "loss": 1.8351, - "step": 1934 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993455320300345, - "loss": 1.8625, - "step": 1935 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019934519380369344, - "loss": 1.8462, - "step": 1936 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019934485557735235, - "loss": 1.9025, - "step": 1937 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019934451735101132, - "loss": 1.6752, - "step": 1938 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019934417912467024, - "loss": 1.9886, - "step": 1939 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019934384089832918, - "loss": 1.794, - "step": 1940 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019934350267198812, - "loss": 1.8379, - "step": 1941 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019934316444564703, - "loss": 1.753, - "step": 1942 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019934282621930597, - "loss": 1.7736, - "step": 1943 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019934248799296492, - "loss": 1.8738, - "step": 1944 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019934214976662383, - "loss": 1.8348, - "step": 1945 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019934181154028277, - "loss": 1.8812, - "step": 1946 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993414733139417, - "loss": 1.9729, - "step": 1947 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019934113508760063, - "loss": 1.8054, - "step": 1948 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019934079686125957, - "loss": 1.9708, - "step": 1949 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993404586349185, - "loss": 1.8545, - "step": 1950 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019934012040857742, - "loss": 1.8994, - "step": 1951 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019933978218223636, - "loss": 1.7363, - "step": 1952 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993394439558953, - "loss": 1.7427, - "step": 1953 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019933910572955425, - "loss": 1.7238, - "step": 1954 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019933876750321316, - "loss": 1.7947, - "step": 1955 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993384292768721, - "loss": 1.8126, - "step": 1956 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019933809105053104, - "loss": 1.7627, - "step": 1957 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019933775282418996, - "loss": 1.8518, - "step": 1958 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993374145978489, - "loss": 1.8815, - "step": 1959 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019933707637150784, - "loss": 1.8572, - "step": 1960 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019933673814516675, - "loss": 1.7899, - "step": 1961 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993363999188257, - "loss": 1.8981, - "step": 1962 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019933606169248463, - "loss": 1.9631, - "step": 1963 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019933572346614355, - "loss": 1.7697, - "step": 1964 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993353852398025, - "loss": 1.861, - "step": 1965 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019933504701346143, - "loss": 1.8595, - "step": 1966 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019933470878712034, - "loss": 1.9555, - "step": 1967 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019933437056077929, - "loss": 1.8614, - "step": 1968 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019933403233443823, - "loss": 1.8713, - "step": 1969 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019933369410809714, - "loss": 1.8629, - "step": 1970 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019933335588175608, - "loss": 1.9378, - "step": 1971 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019933301765541502, - "loss": 1.9018, - "step": 1972 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019933267942907394, - "loss": 1.7432, - "step": 1973 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019933234120273288, - "loss": 1.8512, - "step": 1974 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019933200297639182, - "loss": 1.8134, - "step": 1975 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019933166475005073, - "loss": 1.9675, - "step": 1976 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019933132652370967, - "loss": 1.877, - "step": 1977 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019933098829736862, - "loss": 1.8596, - "step": 1978 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019933065007102753, - "loss": 1.9954, - "step": 1979 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019933031184468647, - "loss": 1.8788, - "step": 1980 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993299736183454, - "loss": 1.8731, - "step": 1981 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019932963539200433, - "loss": 1.7582, - "step": 1982 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019932929716566327, - "loss": 1.8995, - "step": 1983 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993289589393222, - "loss": 1.9592, - "step": 1984 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019932862071298112, - "loss": 1.8813, - "step": 1985 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019932828248664006, - "loss": 1.8384, - "step": 1986 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199327944260299, - "loss": 1.8452, - "step": 1987 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019932760603395792, - "loss": 1.7148, - "step": 1988 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019932726780761686, - "loss": 1.9386, - "step": 1989 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993269295812758, - "loss": 1.7841, - "step": 1990 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019932659135493472, - "loss": 1.881, - "step": 1991 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019932625312859366, - "loss": 1.8171, - "step": 1992 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993259149022526, - "loss": 1.8157, - "step": 1993 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019932557667591154, - "loss": 1.9222, - "step": 1994 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019932523844957048, - "loss": 1.9306, - "step": 1995 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993249002232294, - "loss": 1.8846, - "step": 1996 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019932456199688833, - "loss": 1.8454, - "step": 1997 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019932422377054728, - "loss": 1.8027, - "step": 1998 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993238855442062, - "loss": 1.8348, - "step": 1999 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019932354731786513, - "loss": 1.8109, - "step": 2000 - }, - { - "epoch": 0.01, - "eval_loss": 1.3814978966346154, - "mrc_em": 4.0, - "mrc_f1": 12.666666666666664, - "nli_acc": 0.3, - "step": 2000, - "sts_f1": 0.2972972972972973, - "ynat_f1": 0.4472669941830114 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019932320909152407, - "loss": 1.9224, - "step": 2001 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019932287086518299, - "loss": 1.9194, - "step": 2002 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019932253263884193, - "loss": 1.8514, - "step": 2003 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019932219441250087, - "loss": 1.8805, - "step": 2004 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019932185618615978, - "loss": 1.8346, - "step": 2005 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019932151795981872, - "loss": 1.8285, - "step": 2006 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019932117973347766, - "loss": 1.8621, - "step": 2007 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019932084150713658, - "loss": 1.8088, - "step": 2008 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019932050328079552, - "loss": 1.753, - "step": 2009 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019932016505445446, - "loss": 1.7909, - "step": 2010 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019931982682811338, - "loss": 1.7614, - "step": 2011 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019931948860177232, - "loss": 1.9025, - "step": 2012 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019931915037543126, - "loss": 1.8299, - "step": 2013 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019931881214909017, - "loss": 1.8852, - "step": 2014 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993184739227491, - "loss": 1.9167, - "step": 2015 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019931813569640805, - "loss": 1.8131, - "step": 2016 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019931779747006697, - "loss": 1.8818, - "step": 2017 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993174592437259, - "loss": 1.8513, - "step": 2018 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019931712101738485, - "loss": 1.7211, - "step": 2019 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019931678279104376, - "loss": 1.7397, - "step": 2020 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993164445647027, - "loss": 1.8779, - "step": 2021 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019931610633836165, - "loss": 1.814, - "step": 2022 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019931576811202056, - "loss": 1.8604, - "step": 2023 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993154298856795, - "loss": 1.7838, - "step": 2024 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019931509165933844, - "loss": 1.8968, - "step": 2025 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019931475343299736, - "loss": 1.8579, - "step": 2026 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993144152066563, - "loss": 1.8539, - "step": 2027 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019931407698031524, - "loss": 1.8763, - "step": 2028 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019931373875397415, - "loss": 1.9849, - "step": 2029 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993134005276331, - "loss": 1.7084, - "step": 2030 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019931306230129204, - "loss": 1.9631, - "step": 2031 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019931272407495095, - "loss": 1.7905, - "step": 2032 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993123858486099, - "loss": 1.8818, - "step": 2033 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019931204762226883, - "loss": 1.8084, - "step": 2034 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019931170939592775, - "loss": 1.773, - "step": 2035 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993113711695867, - "loss": 1.8416, - "step": 2036 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019931103294324563, - "loss": 1.8178, - "step": 2037 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019931069471690457, - "loss": 1.884, - "step": 2038 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019931035649056348, - "loss": 1.8586, - "step": 2039 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019931001826422242, - "loss": 1.8955, - "step": 2040 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019930968003788137, - "loss": 1.8862, - "step": 2041 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019930934181154028, - "loss": 1.8075, - "step": 2042 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019930900358519922, - "loss": 1.8135, - "step": 2043 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019930866535885816, - "loss": 1.9413, - "step": 2044 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019930832713251708, - "loss": 1.7145, - "step": 2045 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019930798890617602, - "loss": 1.8873, - "step": 2046 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019930765067983496, - "loss": 1.8346, - "step": 2047 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019930731245349387, - "loss": 1.6772, - "step": 2048 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993069742271528, - "loss": 1.8903, - "step": 2049 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019930663600081175, - "loss": 1.9086, - "step": 2050 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993062977744707, - "loss": 1.807, - "step": 2051 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019930595954812964, - "loss": 2.0732, - "step": 2052 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019930562132178855, - "loss": 1.8023, - "step": 2053 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993052830954475, - "loss": 1.9509, - "step": 2054 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019930494486910643, - "loss": 1.9329, - "step": 2055 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019930460664276535, - "loss": 1.8153, - "step": 2056 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001993042684164243, - "loss": 1.776, - "step": 2057 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019930393019008323, - "loss": 1.6949, - "step": 2058 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019930359196374214, - "loss": 1.9837, - "step": 2059 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019930325373740108, - "loss": 1.9341, - "step": 2060 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019930291551106003, - "loss": 1.8208, - "step": 2061 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019930257728471894, - "loss": 1.8871, - "step": 2062 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019930223905837788, - "loss": 1.7482, - "step": 2063 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019930190083203682, - "loss": 1.7233, - "step": 2064 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019930156260569574, - "loss": 1.8782, - "step": 2065 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019930122437935468, - "loss": 1.7675, - "step": 2066 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019930088615301362, - "loss": 1.948, - "step": 2067 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019930054792667253, - "loss": 1.8876, - "step": 2068 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019930020970033147, - "loss": 1.7912, - "step": 2069 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019929987147399041, - "loss": 1.8908, - "step": 2070 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019929953324764933, - "loss": 1.855, - "step": 2071 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019929919502130827, - "loss": 1.8574, - "step": 2072 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992988567949672, - "loss": 1.6777, - "step": 2073 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019929851856862612, - "loss": 1.9627, - "step": 2074 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019929818034228507, - "loss": 1.7886, - "step": 2075 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199297842115944, - "loss": 1.7285, - "step": 2076 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019929750388960292, - "loss": 1.852, - "step": 2077 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019929716566326186, - "loss": 1.9807, - "step": 2078 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992968274369208, - "loss": 1.8548, - "step": 2079 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019929648921057972, - "loss": 1.7352, - "step": 2080 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019929615098423866, - "loss": 1.846, - "step": 2081 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992958127578976, - "loss": 1.8048, - "step": 2082 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019929547453155651, - "loss": 1.8337, - "step": 2083 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019929513630521546, - "loss": 1.9201, - "step": 2084 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992947980788744, - "loss": 1.7017, - "step": 2085 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992944598525333, - "loss": 1.88, - "step": 2086 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019929412162619225, - "loss": 1.7499, - "step": 2087 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992937833998512, - "loss": 1.8364, - "step": 2088 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992934451735101, - "loss": 1.873, - "step": 2089 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019929310694716905, - "loss": 1.8274, - "step": 2090 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199292768720828, - "loss": 1.8318, - "step": 2091 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992924304944869, - "loss": 1.8934, - "step": 2092 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019929209226814584, - "loss": 1.9331, - "step": 2093 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019929175404180479, - "loss": 1.7264, - "step": 2094 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992914158154637, - "loss": 1.8597, - "step": 2095 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019929107758912264, - "loss": 1.7861, - "step": 2096 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019929073936278158, - "loss": 1.6312, - "step": 2097 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992904011364405, - "loss": 1.7417, - "step": 2098 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019929006291009944, - "loss": 1.9139, - "step": 2099 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019928972468375838, - "loss": 1.9292, - "step": 2100 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992893864574173, - "loss": 1.8627, - "step": 2101 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019928904823107623, - "loss": 1.7543, - "step": 2102 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019928871000473517, - "loss": 1.8516, - "step": 2103 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992883717783941, - "loss": 1.8916, - "step": 2104 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019928803355205303, - "loss": 1.8381, - "step": 2105 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019928769532571197, - "loss": 1.8686, - "step": 2106 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992873570993709, - "loss": 1.8007, - "step": 2107 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019928701887302985, - "loss": 1.7084, - "step": 2108 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992866806466888, - "loss": 1.7934, - "step": 2109 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992863424203477, - "loss": 1.7421, - "step": 2110 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019928600419400665, - "loss": 1.8578, - "step": 2111 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992856659676656, - "loss": 1.7481, - "step": 2112 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992853277413245, - "loss": 1.8549, - "step": 2113 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019928498951498345, - "loss": 1.7799, - "step": 2114 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019928465128864239, - "loss": 1.7396, - "step": 2115 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992843130623013, - "loss": 1.8411, - "step": 2116 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019928397483596024, - "loss": 1.8919, - "step": 2117 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019928363660961918, - "loss": 1.8747, - "step": 2118 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992832983832781, - "loss": 1.8131, - "step": 2119 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019928296015693704, - "loss": 1.7932, - "step": 2120 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019928262193059598, - "loss": 1.8145, - "step": 2121 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992822837042549, - "loss": 1.8195, - "step": 2122 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019928194547791383, - "loss": 1.9303, - "step": 2123 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019928160725157278, - "loss": 1.8686, - "step": 2124 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992812690252317, - "loss": 1.8013, - "step": 2125 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019928093079889063, - "loss": 1.6694, - "step": 2126 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019928059257254957, - "loss": 1.8693, - "step": 2127 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019928025434620849, - "loss": 1.6496, - "step": 2128 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019927991611986743, - "loss": 1.8612, - "step": 2129 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019927957789352637, - "loss": 1.8631, - "step": 2130 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019927923966718528, - "loss": 1.8511, - "step": 2131 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019927890144084422, - "loss": 1.7593, - "step": 2132 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019927856321450316, - "loss": 1.9313, - "step": 2133 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019927822498816208, - "loss": 1.975, - "step": 2134 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019927788676182102, - "loss": 1.9225, - "step": 2135 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019927754853547996, - "loss": 1.7429, - "step": 2136 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019927721030913887, - "loss": 1.7685, - "step": 2137 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019927687208279782, - "loss": 1.882, - "step": 2138 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019927653385645676, - "loss": 1.8135, - "step": 2139 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019927619563011567, - "loss": 1.8468, - "step": 2140 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992758574037746, - "loss": 1.7801, - "step": 2141 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019927551917743355, - "loss": 1.8367, - "step": 2142 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019927518095109247, - "loss": 1.8887, - "step": 2143 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992748427247514, - "loss": 1.8863, - "step": 2144 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019927450449841035, - "loss": 1.9803, - "step": 2145 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019927416627206926, - "loss": 1.7434, - "step": 2146 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992738280457282, - "loss": 2.0598, - "step": 2147 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019927348981938715, - "loss": 1.8001, - "step": 2148 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019927315159304606, - "loss": 1.9878, - "step": 2149 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199272813366705, - "loss": 1.7466, - "step": 2150 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019927247514036394, - "loss": 1.8267, - "step": 2151 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019927213691402286, - "loss": 1.8899, - "step": 2152 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992717986876818, - "loss": 1.8044, - "step": 2153 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019927146046134074, - "loss": 1.8495, - "step": 2154 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019927112223499965, - "loss": 1.834, - "step": 2155 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992707840086586, - "loss": 1.7281, - "step": 2156 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019927044578231753, - "loss": 1.8424, - "step": 2157 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019927010755597645, - "loss": 1.756, - "step": 2158 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992697693296354, - "loss": 1.9947, - "step": 2159 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019926943110329433, - "loss": 1.7978, - "step": 2160 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019926909287695325, - "loss": 1.8456, - "step": 2161 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019926875465061219, - "loss": 1.8696, - "step": 2162 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019926841642427115, - "loss": 1.7874, - "step": 2163 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019926807819793007, - "loss": 1.793, - "step": 2164 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199267739971589, - "loss": 1.7748, - "step": 2165 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019926740174524795, - "loss": 1.7708, - "step": 2166 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019926706351890686, - "loss": 1.8612, - "step": 2167 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992667252925658, - "loss": 1.9282, - "step": 2168 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019926638706622475, - "loss": 1.7452, - "step": 2169 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019926604883988366, - "loss": 1.7778, - "step": 2170 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992657106135426, - "loss": 1.9224, - "step": 2171 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019926537238720154, - "loss": 1.9528, - "step": 2172 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019926503416086046, - "loss": 1.725, - "step": 2173 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992646959345194, - "loss": 1.9696, - "step": 2174 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019926435770817834, - "loss": 1.8185, - "step": 2175 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019926401948183725, - "loss": 1.922, - "step": 2176 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992636812554962, - "loss": 1.7274, - "step": 2177 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019926334302915514, - "loss": 1.7184, - "step": 2178 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019926300480281405, - "loss": 1.8818, - "step": 2179 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199262666576473, - "loss": 1.7866, - "step": 2180 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019926232835013193, - "loss": 1.8911, - "step": 2181 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019926199012379085, - "loss": 1.8267, - "step": 2182 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992616518974498, - "loss": 1.8338, - "step": 2183 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019926131367110873, - "loss": 1.7809, - "step": 2184 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019926097544476764, - "loss": 1.7399, - "step": 2185 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019926063721842658, - "loss": 1.7084, - "step": 2186 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019926029899208553, - "loss": 1.8656, - "step": 2187 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019925996076574444, - "loss": 1.8425, - "step": 2188 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019925962253940338, - "loss": 1.7888, - "step": 2189 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019925928431306232, - "loss": 1.8668, - "step": 2190 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019925894608672124, - "loss": 1.7504, - "step": 2191 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019925860786038018, - "loss": 1.784, - "step": 2192 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019925826963403912, - "loss": 1.8121, - "step": 2193 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019925793140769803, - "loss": 1.663, - "step": 2194 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019925759318135697, - "loss": 1.8207, - "step": 2195 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019925725495501591, - "loss": 1.8077, - "step": 2196 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019925691672867483, - "loss": 1.5875, - "step": 2197 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019925657850233377, - "loss": 1.8766, - "step": 2198 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992562402759927, - "loss": 1.7768, - "step": 2199 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019925590204965162, - "loss": 1.8127, - "step": 2200 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019925556382331057, - "loss": 1.838, - "step": 2201 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992552255969695, - "loss": 1.9813, - "step": 2202 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019925488737062842, - "loss": 1.7492, - "step": 2203 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019925454914428736, - "loss": 1.8119, - "step": 2204 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992542109179463, - "loss": 1.9113, - "step": 2205 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019925387269160522, - "loss": 1.7008, - "step": 2206 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019925353446526416, - "loss": 1.7633, - "step": 2207 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992531962389231, - "loss": 1.832, - "step": 2208 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199252858012582, - "loss": 1.8491, - "step": 2209 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019925251978624095, - "loss": 1.8888, - "step": 2210 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992521815598999, - "loss": 1.7595, - "step": 2211 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992518433335588, - "loss": 1.7066, - "step": 2212 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019925150510721775, - "loss": 1.8528, - "step": 2213 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992511668808767, - "loss": 1.9157, - "step": 2214 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992508286545356, - "loss": 1.7932, - "step": 2215 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019925049042819455, - "loss": 1.6103, - "step": 2216 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992501522018535, - "loss": 1.8348, - "step": 2217 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992498139755124, - "loss": 1.8701, - "step": 2218 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019924947574917137, - "loss": 1.953, - "step": 2219 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992491375228303, - "loss": 1.8296, - "step": 2220 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019924879929648923, - "loss": 1.7922, - "step": 2221 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019924846107014817, - "loss": 1.8407, - "step": 2222 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992481228438071, - "loss": 1.8097, - "step": 2223 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019924778461746602, - "loss": 1.8172, - "step": 2224 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019924744639112496, - "loss": 1.8616, - "step": 2225 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992471081647839, - "loss": 1.9376, - "step": 2226 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019924676993844282, - "loss": 1.778, - "step": 2227 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019924643171210176, - "loss": 1.9684, - "step": 2228 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992460934857607, - "loss": 1.9426, - "step": 2229 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019924575525941961, - "loss": 1.8315, - "step": 2230 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019924541703307856, - "loss": 1.8141, - "step": 2231 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992450788067375, - "loss": 1.8262, - "step": 2232 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992447405803964, - "loss": 1.7481, - "step": 2233 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019924440235405535, - "loss": 1.7615, - "step": 2234 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992440641277143, - "loss": 1.9069, - "step": 2235 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992437259013732, - "loss": 1.9162, - "step": 2236 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019924338767503215, - "loss": 1.7695, - "step": 2237 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992430494486911, - "loss": 1.8976, - "step": 2238 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019924271122235, - "loss": 1.7906, - "step": 2239 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019924237299600894, - "loss": 1.7941, - "step": 2240 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019924203476966789, - "loss": 1.8227, - "step": 2241 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992416965433268, - "loss": 1.7666, - "step": 2242 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019924135831698574, - "loss": 1.7172, - "step": 2243 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019924102009064468, - "loss": 1.845, - "step": 2244 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992406818643036, - "loss": 1.892, - "step": 2245 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019924034363796254, - "loss": 1.7168, - "step": 2246 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019924000541162148, - "loss": 1.7484, - "step": 2247 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992396671852804, - "loss": 1.8677, - "step": 2248 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019923932895893933, - "loss": 1.7555, - "step": 2249 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019923899073259827, - "loss": 1.8267, - "step": 2250 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992386525062572, - "loss": 1.7351, - "step": 2251 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019923831427991613, - "loss": 1.802, - "step": 2252 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019923797605357507, - "loss": 1.8371, - "step": 2253 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019923763782723399, - "loss": 1.8654, - "step": 2254 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019923729960089293, - "loss": 1.8191, - "step": 2255 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019923696137455187, - "loss": 1.7852, - "step": 2256 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019923662314821078, - "loss": 1.8382, - "step": 2257 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019923628492186972, - "loss": 1.8761, - "step": 2258 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019923594669552866, - "loss": 1.9292, - "step": 2259 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019923560846918758, - "loss": 1.7099, - "step": 2260 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019923527024284652, - "loss": 1.7846, - "step": 2261 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019923493201650546, - "loss": 1.8499, - "step": 2262 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019923459379016437, - "loss": 1.8027, - "step": 2263 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019923425556382332, - "loss": 1.6427, - "step": 2264 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019923391733748226, - "loss": 1.7882, - "step": 2265 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019923357911114117, - "loss": 1.9209, - "step": 2266 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992332408848001, - "loss": 1.8723, - "step": 2267 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019923290265845905, - "loss": 1.7168, - "step": 2268 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019923256443211797, - "loss": 1.9376, - "step": 2269 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992322262057769, - "loss": 1.7545, - "step": 2270 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019923188797943585, - "loss": 1.9117, - "step": 2271 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019923154975309476, - "loss": 1.9385, - "step": 2272 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992312115267537, - "loss": 1.8655, - "step": 2273 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019923087330041265, - "loss": 1.7537, - "step": 2274 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019923053507407159, - "loss": 1.783, - "step": 2275 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019923019684773053, - "loss": 1.8289, - "step": 2276 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019922985862138944, - "loss": 1.7872, - "step": 2277 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019922952039504838, - "loss": 1.8226, - "step": 2278 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019922918216870732, - "loss": 1.8198, - "step": 2279 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019922884394236624, - "loss": 1.8848, - "step": 2280 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019922850571602518, - "loss": 1.722, - "step": 2281 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019922816748968412, - "loss": 1.7198, - "step": 2282 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019922782926334303, - "loss": 1.848, - "step": 2283 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019922749103700198, - "loss": 1.8314, - "step": 2284 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019922715281066092, - "loss": 1.7452, - "step": 2285 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019922681458431983, - "loss": 1.8375, - "step": 2286 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019922647635797877, - "loss": 1.8203, - "step": 2287 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992261381316377, - "loss": 1.9267, - "step": 2288 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019922579990529663, - "loss": 1.786, - "step": 2289 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019922546167895557, - "loss": 1.7863, - "step": 2290 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992251234526145, - "loss": 1.8761, - "step": 2291 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019922478522627342, - "loss": 1.8212, - "step": 2292 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019922444699993236, - "loss": 1.7708, - "step": 2293 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992241087735913, - "loss": 1.841, - "step": 2294 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019922377054725022, - "loss": 1.8972, - "step": 2295 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019922343232090916, - "loss": 1.8833, - "step": 2296 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992230940945681, - "loss": 1.8839, - "step": 2297 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019922275586822702, - "loss": 1.8747, - "step": 2298 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019922241764188596, - "loss": 1.911, - "step": 2299 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992220794155449, - "loss": 1.8243, - "step": 2300 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992217411892038, - "loss": 1.7102, - "step": 2301 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019922140296286275, - "loss": 1.8898, - "step": 2302 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992210647365217, - "loss": 1.7973, - "step": 2303 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019922072651018064, - "loss": 2.0127, - "step": 2304 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019922038828383955, - "loss": 1.8946, - "step": 2305 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992200500574985, - "loss": 1.8534, - "step": 2306 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019921971183115743, - "loss": 1.7724, - "step": 2307 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019921937360481635, - "loss": 1.8515, - "step": 2308 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992190353784753, - "loss": 1.8438, - "step": 2309 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019921869715213423, - "loss": 1.9751, - "step": 2310 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019921835892579314, - "loss": 1.9215, - "step": 2311 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019921802069945208, - "loss": 1.7864, - "step": 2312 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019921768247311102, - "loss": 1.7674, - "step": 2313 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019921734424676994, - "loss": 1.9098, - "step": 2314 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019921700602042888, - "loss": 1.9054, - "step": 2315 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019921666779408782, - "loss": 1.814, - "step": 2316 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019921632956774673, - "loss": 1.7469, - "step": 2317 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019921599134140568, - "loss": 1.8424, - "step": 2318 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019921565311506462, - "loss": 1.87, - "step": 2319 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019921531488872353, - "loss": 1.7675, - "step": 2320 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019921497666238247, - "loss": 1.7387, - "step": 2321 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992146384360414, - "loss": 1.8226, - "step": 2322 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019921430020970033, - "loss": 1.7394, - "step": 2323 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019921396198335927, - "loss": 1.8855, - "step": 2324 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992136237570182, - "loss": 1.7591, - "step": 2325 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019921328553067712, - "loss": 1.8477, - "step": 2326 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019921294730433606, - "loss": 1.6951, - "step": 2327 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199212609077995, - "loss": 1.875, - "step": 2328 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019921227085165392, - "loss": 1.7907, - "step": 2329 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019921193262531286, - "loss": 1.8264, - "step": 2330 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992115943989718, - "loss": 1.7495, - "step": 2331 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019921125617263074, - "loss": 1.744, - "step": 2332 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019921091794628968, - "loss": 1.7162, - "step": 2333 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992105797199486, - "loss": 1.8548, - "step": 2334 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019921024149360754, - "loss": 1.8077, - "step": 2335 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019920990326726648, - "loss": 1.8283, - "step": 2336 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992095650409254, - "loss": 1.9182, - "step": 2337 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019920922681458434, - "loss": 1.7547, - "step": 2338 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019920888858824328, - "loss": 1.6971, - "step": 2339 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992085503619022, - "loss": 1.826, - "step": 2340 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019920821213556113, - "loss": 1.7535, - "step": 2341 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019920787390922007, - "loss": 1.7201, - "step": 2342 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199207535682879, - "loss": 1.8524, - "step": 2343 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019920719745653793, - "loss": 1.8598, - "step": 2344 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019920685923019687, - "loss": 1.8553, - "step": 2345 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019920652100385578, - "loss": 1.8864, - "step": 2346 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019920618277751473, - "loss": 1.9265, - "step": 2347 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019920584455117367, - "loss": 1.9369, - "step": 2348 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019920550632483258, - "loss": 1.86, - "step": 2349 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019920516809849152, - "loss": 1.8294, - "step": 2350 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019920482987215046, - "loss": 1.9066, - "step": 2351 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019920449164580938, - "loss": 1.7646, - "step": 2352 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019920415341946832, - "loss": 1.7246, - "step": 2353 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019920381519312726, - "loss": 1.8788, - "step": 2354 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019920347696678617, - "loss": 1.8343, - "step": 2355 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019920313874044511, - "loss": 1.7001, - "step": 2356 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019920280051410406, - "loss": 1.7602, - "step": 2357 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019920246228776297, - "loss": 1.9037, - "step": 2358 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992021240614219, - "loss": 1.7741, - "step": 2359 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019920178583508085, - "loss": 1.6692, - "step": 2360 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019920144760873977, - "loss": 1.8398, - "step": 2361 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992011093823987, - "loss": 1.7349, - "step": 2362 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019920077115605765, - "loss": 1.7351, - "step": 2363 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019920043292971656, - "loss": 1.808, - "step": 2364 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001992000947033755, - "loss": 1.734, - "step": 2365 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019919975647703444, - "loss": 1.8734, - "step": 2366 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019919941825069336, - "loss": 1.8899, - "step": 2367 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991990800243523, - "loss": 1.7619, - "step": 2368 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019919874179801124, - "loss": 1.6797, - "step": 2369 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019919840357167015, - "loss": 1.8227, - "step": 2370 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991980653453291, - "loss": 1.7534, - "step": 2371 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019919772711898804, - "loss": 1.8433, - "step": 2372 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019919738889264695, - "loss": 1.7871, - "step": 2373 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991970506663059, - "loss": 1.8181, - "step": 2374 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019919671243996483, - "loss": 1.7321, - "step": 2375 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019919637421362375, - "loss": 1.8004, - "step": 2376 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991960359872827, - "loss": 1.6523, - "step": 2377 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019919569776094163, - "loss": 1.8512, - "step": 2378 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019919535953460054, - "loss": 1.7942, - "step": 2379 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019919502130825948, - "loss": 1.9196, - "step": 2380 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019919468308191843, - "loss": 1.6121, - "step": 2381 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019919434485557734, - "loss": 1.666, - "step": 2382 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019919400662923628, - "loss": 1.8207, - "step": 2383 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019919366840289522, - "loss": 1.8614, - "step": 2384 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019919333017655414, - "loss": 1.6297, - "step": 2385 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019919299195021308, - "loss": 1.781, - "step": 2386 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019919265372387202, - "loss": 1.7677, - "step": 2387 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019919231549753096, - "loss": 1.8751, - "step": 2388 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991919772711899, - "loss": 1.8082, - "step": 2389 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019919163904484884, - "loss": 1.7999, - "step": 2390 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019919130081850776, - "loss": 1.5327, - "step": 2391 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991909625921667, - "loss": 1.7186, - "step": 2392 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019919062436582564, - "loss": 1.8171, - "step": 2393 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019919028613948455, - "loss": 1.7881, - "step": 2394 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991899479131435, - "loss": 1.8236, - "step": 2395 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019918960968680243, - "loss": 1.9493, - "step": 2396 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019918927146046135, - "loss": 1.8172, - "step": 2397 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991889332341203, - "loss": 1.7603, - "step": 2398 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019918859500777923, - "loss": 1.8738, - "step": 2399 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019918825678143814, - "loss": 1.9371, - "step": 2400 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019918791855509709, - "loss": 1.7838, - "step": 2401 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019918758032875603, - "loss": 1.7924, - "step": 2402 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019918724210241494, - "loss": 1.7649, - "step": 2403 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019918690387607388, - "loss": 1.6807, - "step": 2404 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019918656564973282, - "loss": 1.7095, - "step": 2405 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019918622742339174, - "loss": 1.7335, - "step": 2406 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019918588919705068, - "loss": 1.6338, - "step": 2407 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019918555097070962, - "loss": 1.9291, - "step": 2408 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019918521274436853, - "loss": 1.7431, - "step": 2409 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019918487451802747, - "loss": 1.7394, - "step": 2410 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019918453629168642, - "loss": 1.8537, - "step": 2411 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019918419806534533, - "loss": 1.7884, - "step": 2412 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019918385983900427, - "loss": 1.8719, - "step": 2413 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991835216126632, - "loss": 1.6861, - "step": 2414 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019918318338632213, - "loss": 1.7384, - "step": 2415 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019918284515998107, - "loss": 1.8245, - "step": 2416 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019918250693364, - "loss": 1.8678, - "step": 2417 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019918216870729892, - "loss": 1.7039, - "step": 2418 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019918183048095786, - "loss": 1.8655, - "step": 2419 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991814922546168, - "loss": 1.9464, - "step": 2420 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019918115402827572, - "loss": 1.8136, - "step": 2421 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019918081580193466, - "loss": 1.821, - "step": 2422 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991804775755936, - "loss": 1.9518, - "step": 2423 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019918013934925252, - "loss": 1.7593, - "step": 2424 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019917980112291146, - "loss": 1.8053, - "step": 2425 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991794628965704, - "loss": 1.7143, - "step": 2426 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991791246702293, - "loss": 1.8586, - "step": 2427 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019917878644388825, - "loss": 1.8767, - "step": 2428 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991784482175472, - "loss": 1.7007, - "step": 2429 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991781099912061, - "loss": 1.6865, - "step": 2430 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019917777176486505, - "loss": 1.9973, - "step": 2431 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199177433538524, - "loss": 1.6528, - "step": 2432 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991770953121829, - "loss": 1.782, - "step": 2433 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019917675708584185, - "loss": 1.8865, - "step": 2434 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019917641885950079, - "loss": 1.7701, - "step": 2435 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991760806331597, - "loss": 1.9849, - "step": 2436 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019917574240681864, - "loss": 1.733, - "step": 2437 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019917540418047758, - "loss": 1.8209, - "step": 2438 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991750659541365, - "loss": 1.9452, - "step": 2439 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019917472772779544, - "loss": 1.7728, - "step": 2440 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019917438950145438, - "loss": 1.7529, - "step": 2441 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991740512751133, - "loss": 1.9145, - "step": 2442 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019917371304877223, - "loss": 1.7369, - "step": 2443 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991733748224312, - "loss": 1.7592, - "step": 2444 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019917303659609012, - "loss": 1.5347, - "step": 2445 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019917269836974906, - "loss": 1.8057, - "step": 2446 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199172360143408, - "loss": 1.7241, - "step": 2447 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991720219170669, - "loss": 1.9396, - "step": 2448 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019917168369072585, - "loss": 1.7365, - "step": 2449 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991713454643848, - "loss": 1.648, - "step": 2450 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991710072380437, - "loss": 1.8567, - "step": 2451 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019917066901170265, - "loss": 1.9886, - "step": 2452 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991703307853616, - "loss": 1.8925, - "step": 2453 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991699925590205, - "loss": 1.9021, - "step": 2454 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019916965433267945, - "loss": 1.7847, - "step": 2455 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991693161063384, - "loss": 1.8337, - "step": 2456 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991689778799973, - "loss": 1.8381, - "step": 2457 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019916863965365624, - "loss": 1.9277, - "step": 2458 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019916830142731518, - "loss": 1.8203, - "step": 2459 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991679632009741, - "loss": 1.6963, - "step": 2460 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019916762497463304, - "loss": 1.9249, - "step": 2461 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019916728674829198, - "loss": 1.7712, - "step": 2462 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991669485219509, - "loss": 1.7937, - "step": 2463 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019916661029560984, - "loss": 1.6577, - "step": 2464 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019916627206926878, - "loss": 1.7811, - "step": 2465 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991659338429277, - "loss": 1.8263, - "step": 2466 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019916559561658663, - "loss": 1.8033, - "step": 2467 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019916525739024557, - "loss": 1.9079, - "step": 2468 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991649191639045, - "loss": 1.7578, - "step": 2469 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019916458093756343, - "loss": 1.8918, - "step": 2470 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019916424271122237, - "loss": 1.8312, - "step": 2471 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019916390448488128, - "loss": 1.9887, - "step": 2472 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019916356625854022, - "loss": 1.8114, - "step": 2473 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019916322803219917, - "loss": 1.8121, - "step": 2474 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019916288980585808, - "loss": 1.7895, - "step": 2475 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019916255157951702, - "loss": 1.7515, - "step": 2476 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019916221335317596, - "loss": 1.6878, - "step": 2477 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019916187512683488, - "loss": 1.8224, - "step": 2478 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019916153690049382, - "loss": 1.7072, - "step": 2479 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019916119867415276, - "loss": 1.6616, - "step": 2480 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019916086044781167, - "loss": 1.8486, - "step": 2481 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991605222214706, - "loss": 1.7951, - "step": 2482 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019916018399512955, - "loss": 1.8992, - "step": 2483 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019915984576878847, - "loss": 1.8877, - "step": 2484 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991595075424474, - "loss": 1.9004, - "step": 2485 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019915916931610635, - "loss": 1.8562, - "step": 2486 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019915883108976526, - "loss": 1.7653, - "step": 2487 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991584928634242, - "loss": 1.7939, - "step": 2488 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019915815463708315, - "loss": 1.8762, - "step": 2489 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019915781641074206, - "loss": 1.9964, - "step": 2490 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199157478184401, - "loss": 1.7942, - "step": 2491 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019915713995805994, - "loss": 1.8773, - "step": 2492 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019915680173171886, - "loss": 1.7826, - "step": 2493 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991564635053778, - "loss": 1.7601, - "step": 2494 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019915612527903674, - "loss": 1.6449, - "step": 2495 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019915578705269565, - "loss": 1.9613, - "step": 2496 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991554488263546, - "loss": 1.778, - "step": 2497 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019915511060001354, - "loss": 1.8219, - "step": 2498 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019915477237367245, - "loss": 2.003, - "step": 2499 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019915443414733142, - "loss": 1.8002, - "step": 2500 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019915409592099036, - "loss": 1.8054, - "step": 2501 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019915375769464927, - "loss": 1.7352, - "step": 2502 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019915341946830821, - "loss": 1.9289, - "step": 2503 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019915308124196716, - "loss": 2.0194, - "step": 2504 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019915274301562607, - "loss": 1.7491, - "step": 2505 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199152404789285, - "loss": 1.7513, - "step": 2506 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019915206656294395, - "loss": 1.7894, - "step": 2507 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019915172833660287, - "loss": 1.6802, - "step": 2508 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991513901102618, - "loss": 1.679, - "step": 2509 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019915105188392075, - "loss": 1.8721, - "step": 2510 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019915071365757966, - "loss": 1.7729, - "step": 2511 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991503754312386, - "loss": 1.7264, - "step": 2512 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019915003720489754, - "loss": 1.7926, - "step": 2513 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019914969897855646, - "loss": 1.791, - "step": 2514 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991493607522154, - "loss": 1.8707, - "step": 2515 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019914902252587434, - "loss": 1.6888, - "step": 2516 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019914868429953326, - "loss": 1.7146, - "step": 2517 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991483460731922, - "loss": 1.7728, - "step": 2518 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019914800784685114, - "loss": 1.8552, - "step": 2519 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019914766962051005, - "loss": 1.8443, - "step": 2520 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199147331394169, - "loss": 1.7565, - "step": 2521 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019914699316782793, - "loss": 1.8193, - "step": 2522 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019914665494148685, - "loss": 1.8573, - "step": 2523 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991463167151458, - "loss": 1.9244, - "step": 2524 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019914597848880473, - "loss": 1.7782, - "step": 2525 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019914564026246364, - "loss": 1.7903, - "step": 2526 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019914530203612259, - "loss": 1.8256, - "step": 2527 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019914496380978153, - "loss": 1.8507, - "step": 2528 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019914462558344044, - "loss": 1.8492, - "step": 2529 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019914428735709938, - "loss": 1.8875, - "step": 2530 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019914394913075832, - "loss": 1.6974, - "step": 2531 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019914361090441724, - "loss": 1.8029, - "step": 2532 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019914327267807618, - "loss": 1.7113, - "step": 2533 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019914293445173512, - "loss": 1.8799, - "step": 2534 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019914259622539403, - "loss": 1.7718, - "step": 2535 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019914225799905297, - "loss": 1.8952, - "step": 2536 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019914191977271192, - "loss": 1.8187, - "step": 2537 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019914158154637083, - "loss": 1.9508, - "step": 2538 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019914124332002977, - "loss": 1.7193, - "step": 2539 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991409050936887, - "loss": 1.8555, - "step": 2540 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019914056686734763, - "loss": 1.7707, - "step": 2541 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019914022864100657, - "loss": 1.6859, - "step": 2542 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991398904146655, - "loss": 1.7435, - "step": 2543 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019913955218832442, - "loss": 1.9126, - "step": 2544 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019913921396198336, - "loss": 1.6001, - "step": 2545 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991388757356423, - "loss": 1.6759, - "step": 2546 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019913853750930122, - "loss": 1.8133, - "step": 2547 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019913819928296016, - "loss": 1.8712, - "step": 2548 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991378610566191, - "loss": 1.7666, - "step": 2549 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019913752283027801, - "loss": 1.8157, - "step": 2550 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019913718460393696, - "loss": 1.9898, - "step": 2551 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991368463775959, - "loss": 1.9623, - "step": 2552 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991365081512548, - "loss": 1.7804, - "step": 2553 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019913616992491375, - "loss": 2.058, - "step": 2554 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991358316985727, - "loss": 1.9208, - "step": 2555 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019913549347223163, - "loss": 1.818, - "step": 2556 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019913515524589058, - "loss": 1.7641, - "step": 2557 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991348170195495, - "loss": 1.9106, - "step": 2558 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019913447879320843, - "loss": 1.7792, - "step": 2559 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019913414056686737, - "loss": 1.7793, - "step": 2560 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019913380234052629, - "loss": 1.8498, - "step": 2561 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019913346411418523, - "loss": 1.9092, - "step": 2562 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019913312588784417, - "loss": 1.922, - "step": 2563 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019913278766150308, - "loss": 1.7578, - "step": 2564 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019913244943516202, - "loss": 1.7319, - "step": 2565 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019913211120882096, - "loss": 1.8454, - "step": 2566 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019913177298247988, - "loss": 1.7606, - "step": 2567 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019913143475613882, - "loss": 1.8242, - "step": 2568 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019913109652979776, - "loss": 1.84, - "step": 2569 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019913075830345667, - "loss": 1.829, - "step": 2570 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019913042007711562, - "loss": 1.78, - "step": 2571 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019913008185077456, - "loss": 1.7838, - "step": 2572 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991297436244335, - "loss": 1.8606, - "step": 2573 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991294053980924, - "loss": 1.84, - "step": 2574 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019912906717175135, - "loss": 1.7886, - "step": 2575 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991287289454103, - "loss": 1.7968, - "step": 2576 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991283907190692, - "loss": 1.815, - "step": 2577 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019912805249272815, - "loss": 1.8212, - "step": 2578 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991277142663871, - "loss": 1.7321, - "step": 2579 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199127376040046, - "loss": 1.7335, - "step": 2580 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019912703781370495, - "loss": 1.8326, - "step": 2581 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991266995873639, - "loss": 1.7734, - "step": 2582 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991263613610228, - "loss": 1.9855, - "step": 2583 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019912602313468174, - "loss": 1.9088, - "step": 2584 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019912568490834068, - "loss": 1.7747, - "step": 2585 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991253466819996, - "loss": 1.779, - "step": 2586 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019912500845565854, - "loss": 1.8468, - "step": 2587 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019912467022931748, - "loss": 1.8256, - "step": 2588 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991243320029764, - "loss": 1.7402, - "step": 2589 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019912399377663533, - "loss": 1.8689, - "step": 2590 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019912365555029428, - "loss": 2.0249, - "step": 2591 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991233173239532, - "loss": 1.653, - "step": 2592 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019912297909761213, - "loss": 1.7494, - "step": 2593 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019912264087127107, - "loss": 1.7193, - "step": 2594 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019912230264492999, - "loss": 1.8969, - "step": 2595 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019912196441858893, - "loss": 1.9405, - "step": 2596 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019912162619224787, - "loss": 1.7906, - "step": 2597 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019912128796590678, - "loss": 1.7616, - "step": 2598 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019912094973956572, - "loss": 1.7732, - "step": 2599 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019912061151322466, - "loss": 1.6344, - "step": 2600 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019912027328688358, - "loss": 1.7268, - "step": 2601 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019911993506054252, - "loss": 1.8551, - "step": 2602 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019911959683420146, - "loss": 1.6073, - "step": 2603 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019911925860786038, - "loss": 1.7865, - "step": 2604 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019911892038151932, - "loss": 1.9605, - "step": 2605 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019911858215517826, - "loss": 1.6188, - "step": 2606 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019911824392883717, - "loss": 1.7746, - "step": 2607 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991179057024961, - "loss": 1.8393, - "step": 2608 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019911756747615505, - "loss": 1.8128, - "step": 2609 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019911722924981397, - "loss": 1.8353, - "step": 2610 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991168910234729, - "loss": 1.7738, - "step": 2611 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019911655279713185, - "loss": 1.6498, - "step": 2612 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991162145707908, - "loss": 1.7994, - "step": 2613 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019911587634444973, - "loss": 1.7855, - "step": 2614 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019911553811810865, - "loss": 1.6977, - "step": 2615 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991151998917676, - "loss": 1.9448, - "step": 2616 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019911486166542653, - "loss": 1.8064, - "step": 2617 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019911452343908544, - "loss": 1.7832, - "step": 2618 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019911418521274438, - "loss": 1.756, - "step": 2619 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019911384698640333, - "loss": 1.7459, - "step": 2620 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019911350876006224, - "loss": 1.8621, - "step": 2621 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019911317053372118, - "loss": 1.7572, - "step": 2622 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019911283230738012, - "loss": 2.0143, - "step": 2623 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019911249408103904, - "loss": 1.7301, - "step": 2624 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019911215585469798, - "loss": 1.8383, - "step": 2625 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019911181762835692, - "loss": 1.7954, - "step": 2626 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019911147940201583, - "loss": 1.8819, - "step": 2627 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019911114117567477, - "loss": 1.7739, - "step": 2628 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019911080294933371, - "loss": 1.7725, - "step": 2629 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019911046472299263, - "loss": 1.7884, - "step": 2630 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019911012649665157, - "loss": 1.8219, - "step": 2631 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991097882703105, - "loss": 1.7691, - "step": 2632 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019910945004396942, - "loss": 1.8542, - "step": 2633 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019910911181762837, - "loss": 1.7262, - "step": 2634 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991087735912873, - "loss": 1.8078, - "step": 2635 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019910843536494622, - "loss": 1.8113, - "step": 2636 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019910809713860516, - "loss": 1.8426, - "step": 2637 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991077589122641, - "loss": 1.8921, - "step": 2638 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019910742068592302, - "loss": 1.826, - "step": 2639 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019910708245958196, - "loss": 1.6707, - "step": 2640 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991067442332409, - "loss": 1.8581, - "step": 2641 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991064060068998, - "loss": 1.856, - "step": 2642 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019910606778055875, - "loss": 1.9574, - "step": 2643 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991057295542177, - "loss": 1.6792, - "step": 2644 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991053913278766, - "loss": 1.8576, - "step": 2645 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019910505310153555, - "loss": 1.7399, - "step": 2646 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991047148751945, - "loss": 1.7788, - "step": 2647 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991043766488534, - "loss": 1.9, - "step": 2648 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019910403842251235, - "loss": 1.9078, - "step": 2649 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991037001961713, - "loss": 1.6666, - "step": 2650 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001991033619698302, - "loss": 1.6324, - "step": 2651 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019910302374348914, - "loss": 1.8071, - "step": 2652 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019910268551714808, - "loss": 1.8298, - "step": 2653 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199102347290807, - "loss": 1.8485, - "step": 2654 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019910200906446594, - "loss": 1.9073, - "step": 2655 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019910167083812488, - "loss": 1.8425, - "step": 2656 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019910133261178382, - "loss": 1.7584, - "step": 2657 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019910099438544274, - "loss": 1.9709, - "step": 2658 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019910065615910168, - "loss": 1.7288, - "step": 2659 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019910031793276062, - "loss": 1.8088, - "step": 2660 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019909997970641953, - "loss": 1.7451, - "step": 2661 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019909964148007847, - "loss": 1.8802, - "step": 2662 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019909930325373741, - "loss": 1.826, - "step": 2663 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019909896502739633, - "loss": 1.9044, - "step": 2664 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019909862680105527, - "loss": 1.8697, - "step": 2665 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990982885747142, - "loss": 1.7785, - "step": 2666 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019909795034837312, - "loss": 1.8169, - "step": 2667 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990976121220321, - "loss": 1.7585, - "step": 2668 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199097273895691, - "loss": 1.9593, - "step": 2669 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019909693566934995, - "loss": 1.7647, - "step": 2670 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990965974430089, - "loss": 1.7921, - "step": 2671 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990962592166678, - "loss": 1.7625, - "step": 2672 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019909592099032674, - "loss": 1.7824, - "step": 2673 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019909558276398569, - "loss": 1.7728, - "step": 2674 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990952445376446, - "loss": 1.737, - "step": 2675 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019909490631130354, - "loss": 1.6992, - "step": 2676 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019909456808496248, - "loss": 1.7439, - "step": 2677 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990942298586214, - "loss": 1.7473, - "step": 2678 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019909389163228034, - "loss": 1.8674, - "step": 2679 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019909355340593928, - "loss": 1.9974, - "step": 2680 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990932151795982, - "loss": 1.7787, - "step": 2681 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019909287695325713, - "loss": 1.758, - "step": 2682 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019909253872691607, - "loss": 1.7415, - "step": 2683 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199092200500575, - "loss": 1.8306, - "step": 2684 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019909186227423393, - "loss": 1.6422, - "step": 2685 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019909152404789287, - "loss": 1.8383, - "step": 2686 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019909118582155179, - "loss": 1.6889, - "step": 2687 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019909084759521073, - "loss": 1.7945, - "step": 2688 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019909050936886967, - "loss": 1.8024, - "step": 2689 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019909017114252858, - "loss": 1.7533, - "step": 2690 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019908983291618752, - "loss": 1.715, - "step": 2691 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019908949468984646, - "loss": 1.6761, - "step": 2692 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019908915646350538, - "loss": 1.8948, - "step": 2693 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019908881823716432, - "loss": 1.8755, - "step": 2694 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019908848001082326, - "loss": 1.7637, - "step": 2695 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019908814178448217, - "loss": 1.8504, - "step": 2696 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019908780355814112, - "loss": 1.9535, - "step": 2697 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019908746533180006, - "loss": 1.8236, - "step": 2698 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019908712710545897, - "loss": 1.7099, - "step": 2699 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990867888791179, - "loss": 1.8465, - "step": 2700 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019908645065277685, - "loss": 1.8217, - "step": 2701 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019908611242643577, - "loss": 1.8662, - "step": 2702 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990857742000947, - "loss": 1.7694, - "step": 2703 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019908543597375365, - "loss": 1.7895, - "step": 2704 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019908509774741256, - "loss": 1.7858, - "step": 2705 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990847595210715, - "loss": 1.7969, - "step": 2706 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019908442129473045, - "loss": 1.6673, - "step": 2707 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019908408306838936, - "loss": 1.7143, - "step": 2708 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990837448420483, - "loss": 1.8513, - "step": 2709 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019908340661570724, - "loss": 1.7909, - "step": 2710 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019908306838936616, - "loss": 1.7765, - "step": 2711 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990827301630251, - "loss": 1.6041, - "step": 2712 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019908239193668404, - "loss": 1.8541, - "step": 2713 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019908205371034295, - "loss": 1.8925, - "step": 2714 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990817154840019, - "loss": 1.8056, - "step": 2715 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019908137725766083, - "loss": 1.7126, - "step": 2716 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019908103903131975, - "loss": 1.7953, - "step": 2717 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990807008049787, - "loss": 1.8573, - "step": 2718 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019908036257863763, - "loss": 1.8896, - "step": 2719 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019908002435229654, - "loss": 1.8383, - "step": 2720 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019907968612595549, - "loss": 1.8524, - "step": 2721 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019907934789961443, - "loss": 1.7099, - "step": 2722 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019907900967327334, - "loss": 1.8521, - "step": 2723 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019907867144693228, - "loss": 1.756, - "step": 2724 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019907833322059125, - "loss": 1.7715, - "step": 2725 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019907799499425016, - "loss": 1.8106, - "step": 2726 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990776567679091, - "loss": 1.7831, - "step": 2727 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019907731854156805, - "loss": 1.635, - "step": 2728 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019907698031522696, - "loss": 2.0198, - "step": 2729 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990766420888859, - "loss": 1.8055, - "step": 2730 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019907630386254484, - "loss": 1.7598, - "step": 2731 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019907596563620376, - "loss": 1.6691, - "step": 2732 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990756274098627, - "loss": 1.8122, - "step": 2733 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019907528918352164, - "loss": 1.9708, - "step": 2734 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019907495095718055, - "loss": 1.7965, - "step": 2735 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990746127308395, - "loss": 1.8882, - "step": 2736 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019907427450449844, - "loss": 1.7947, - "step": 2737 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019907393627815735, - "loss": 1.9819, - "step": 2738 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990735980518163, - "loss": 1.7844, - "step": 2739 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019907325982547523, - "loss": 1.7704, - "step": 2740 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019907292159913415, - "loss": 1.8514, - "step": 2741 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990725833727931, - "loss": 1.8803, - "step": 2742 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019907224514645203, - "loss": 1.7945, - "step": 2743 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019907190692011094, - "loss": 2.0678, - "step": 2744 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019907156869376988, - "loss": 1.9912, - "step": 2745 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019907123046742882, - "loss": 1.793, - "step": 2746 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019907089224108774, - "loss": 1.7297, - "step": 2747 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019907055401474668, - "loss": 1.7962, - "step": 2748 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019907021578840562, - "loss": 1.7246, - "step": 2749 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019906987756206453, - "loss": 1.7796, - "step": 2750 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019906953933572348, - "loss": 1.7453, - "step": 2751 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019906920110938242, - "loss": 1.7575, - "step": 2752 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019906886288304133, - "loss": 1.7386, - "step": 2753 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019906852465670027, - "loss": 1.8163, - "step": 2754 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990681864303592, - "loss": 1.7049, - "step": 2755 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019906784820401813, - "loss": 1.882, - "step": 2756 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019906750997767707, - "loss": 1.649, - "step": 2757 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199067171751336, - "loss": 1.8491, - "step": 2758 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019906683352499492, - "loss": 1.7355, - "step": 2759 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019906649529865386, - "loss": 2.0065, - "step": 2760 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990661570723128, - "loss": 1.6733, - "step": 2761 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019906581884597172, - "loss": 1.7079, - "step": 2762 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019906548061963066, - "loss": 1.8061, - "step": 2763 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990651423932896, - "loss": 1.5844, - "step": 2764 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019906480416694852, - "loss": 1.9366, - "step": 2765 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019906446594060746, - "loss": 1.859, - "step": 2766 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990641277142664, - "loss": 1.6802, - "step": 2767 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990637894879253, - "loss": 1.7466, - "step": 2768 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019906345126158425, - "loss": 1.804, - "step": 2769 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990631130352432, - "loss": 1.7347, - "step": 2770 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990627748089021, - "loss": 1.7871, - "step": 2771 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019906243658256105, - "loss": 1.822, - "step": 2772 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019906209835622, - "loss": 1.7021, - "step": 2773 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990617601298789, - "loss": 1.7298, - "step": 2774 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019906142190353785, - "loss": 1.8486, - "step": 2775 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990610836771968, - "loss": 1.8107, - "step": 2776 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990607454508557, - "loss": 1.6327, - "step": 2777 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019906040722451464, - "loss": 1.8477, - "step": 2778 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019906006899817358, - "loss": 1.8828, - "step": 2779 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990597307718325, - "loss": 1.7874, - "step": 2780 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019905939254549147, - "loss": 1.7208, - "step": 2781 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990590543191504, - "loss": 1.8461, - "step": 2782 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019905871609280932, - "loss": 1.8407, - "step": 2783 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019905837786646826, - "loss": 1.691, - "step": 2784 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990580396401272, - "loss": 1.812, - "step": 2785 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019905770141378612, - "loss": 1.8188, - "step": 2786 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019905736318744506, - "loss": 1.7659, - "step": 2787 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199057024961104, - "loss": 1.7821, - "step": 2788 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019905668673476291, - "loss": 1.7697, - "step": 2789 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019905634850842186, - "loss": 1.7644, - "step": 2790 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990560102820808, - "loss": 1.7537, - "step": 2791 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990556720557397, - "loss": 1.8368, - "step": 2792 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019905533382939865, - "loss": 1.821, - "step": 2793 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990549956030576, - "loss": 1.6861, - "step": 2794 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990546573767165, - "loss": 1.9497, - "step": 2795 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019905431915037545, - "loss": 1.7688, - "step": 2796 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990539809240344, - "loss": 1.8191, - "step": 2797 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990536426976933, - "loss": 1.7885, - "step": 2798 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019905330447135224, - "loss": 1.6725, - "step": 2799 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019905296624501119, - "loss": 1.791, - "step": 2800 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990526280186701, - "loss": 1.8715, - "step": 2801 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019905228979232904, - "loss": 1.7006, - "step": 2802 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019905195156598798, - "loss": 1.8466, - "step": 2803 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990516133396469, - "loss": 1.7963, - "step": 2804 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019905127511330584, - "loss": 1.8809, - "step": 2805 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019905093688696478, - "loss": 1.6566, - "step": 2806 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990505986606237, - "loss": 1.8285, - "step": 2807 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019905026043428263, - "loss": 1.8442, - "step": 2808 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019904992220794157, - "loss": 1.756, - "step": 2809 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990495839816005, - "loss": 1.9193, - "step": 2810 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019904924575525943, - "loss": 1.857, - "step": 2811 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019904890752891837, - "loss": 1.7682, - "step": 2812 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019904856930257728, - "loss": 1.9024, - "step": 2813 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019904823107623623, - "loss": 1.7759, - "step": 2814 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019904789284989517, - "loss": 1.9382, - "step": 2815 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019904755462355408, - "loss": 1.8589, - "step": 2816 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019904721639721302, - "loss": 1.6971, - "step": 2817 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019904687817087196, - "loss": 1.7103, - "step": 2818 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019904653994453088, - "loss": 1.7981, - "step": 2819 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019904620171818982, - "loss": 1.8565, - "step": 2820 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019904586349184876, - "loss": 1.628, - "step": 2821 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019904552526550767, - "loss": 1.7673, - "step": 2822 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019904518703916661, - "loss": 1.8308, - "step": 2823 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019904484881282556, - "loss": 1.7249, - "step": 2824 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019904451058648447, - "loss": 1.7535, - "step": 2825 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990441723601434, - "loss": 1.8898, - "step": 2826 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019904383413380235, - "loss": 1.7703, - "step": 2827 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019904349590746127, - "loss": 2.0104, - "step": 2828 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990431576811202, - "loss": 1.5682, - "step": 2829 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019904281945477915, - "loss": 1.7458, - "step": 2830 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019904248122843806, - "loss": 1.7544, - "step": 2831 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199042143002097, - "loss": 1.8229, - "step": 2832 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019904180477575594, - "loss": 1.8874, - "step": 2833 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019904146654941486, - "loss": 1.7595, - "step": 2834 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990411283230738, - "loss": 1.7963, - "step": 2835 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019904079009673274, - "loss": 1.8034, - "step": 2836 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019904045187039168, - "loss": 1.7567, - "step": 2837 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019904011364405062, - "loss": 1.9826, - "step": 2838 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019903977541770956, - "loss": 1.8744, - "step": 2839 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019903943719136848, - "loss": 1.6726, - "step": 2840 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019903909896502742, - "loss": 1.7224, - "step": 2841 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019903876073868636, - "loss": 1.789, - "step": 2842 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019903842251234527, - "loss": 1.7656, - "step": 2843 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019903808428600422, - "loss": 1.7955, - "step": 2844 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019903774605966316, - "loss": 1.9108, - "step": 2845 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019903740783332207, - "loss": 1.7394, - "step": 2846 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199037069606981, - "loss": 1.7283, - "step": 2847 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019903673138063995, - "loss": 1.9539, - "step": 2848 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019903639315429887, - "loss": 1.8005, - "step": 2849 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990360549279578, - "loss": 1.7038, - "step": 2850 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019903571670161675, - "loss": 1.766, - "step": 2851 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019903537847527566, - "loss": 1.8273, - "step": 2852 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990350402489346, - "loss": 1.8886, - "step": 2853 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019903470202259355, - "loss": 1.796, - "step": 2854 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019903436379625246, - "loss": 1.7344, - "step": 2855 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990340255699114, - "loss": 1.8507, - "step": 2856 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019903368734357034, - "loss": 1.8625, - "step": 2857 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019903334911722926, - "loss": 1.9305, - "step": 2858 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990330108908882, - "loss": 1.9134, - "step": 2859 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019903267266454714, - "loss": 1.7439, - "step": 2860 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019903233443820605, - "loss": 1.6169, - "step": 2861 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199031996211865, - "loss": 1.6785, - "step": 2862 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019903165798552393, - "loss": 1.7845, - "step": 2863 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019903131975918285, - "loss": 1.7555, - "step": 2864 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990309815328418, - "loss": 1.7192, - "step": 2865 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019903064330650073, - "loss": 1.8419, - "step": 2866 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019903030508015965, - "loss": 1.8855, - "step": 2867 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019902996685381859, - "loss": 1.8622, - "step": 2868 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019902962862747753, - "loss": 1.8509, - "step": 2869 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019902929040113644, - "loss": 1.8163, - "step": 2870 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019902895217479538, - "loss": 1.781, - "step": 2871 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019902861394845432, - "loss": 1.7865, - "step": 2872 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019902827572211324, - "loss": 1.7505, - "step": 2873 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019902793749577218, - "loss": 1.7267, - "step": 2874 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019902759926943112, - "loss": 1.7304, - "step": 2875 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019902726104309003, - "loss": 1.8032, - "step": 2876 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019902692281674898, - "loss": 1.7884, - "step": 2877 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019902658459040792, - "loss": 1.8151, - "step": 2878 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019902624636406683, - "loss": 1.8069, - "step": 2879 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019902590813772577, - "loss": 1.8285, - "step": 2880 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990255699113847, - "loss": 1.7546, - "step": 2881 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019902523168504363, - "loss": 1.7651, - "step": 2882 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019902489345870257, - "loss": 1.7432, - "step": 2883 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990245552323615, - "loss": 1.7723, - "step": 2884 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019902421700602042, - "loss": 1.8518, - "step": 2885 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019902387877967936, - "loss": 1.9542, - "step": 2886 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990235405533383, - "loss": 1.7942, - "step": 2887 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019902320232699722, - "loss": 1.851, - "step": 2888 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019902286410065616, - "loss": 1.7478, - "step": 2889 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990225258743151, - "loss": 1.8817, - "step": 2890 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019902218764797402, - "loss": 1.7081, - "step": 2891 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019902184942163296, - "loss": 1.7621, - "step": 2892 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990215111952919, - "loss": 1.6992, - "step": 2893 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019902117296895084, - "loss": 1.8448, - "step": 2894 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019902083474260978, - "loss": 1.7244, - "step": 2895 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990204965162687, - "loss": 1.8504, - "step": 2896 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019902015828992764, - "loss": 1.7208, - "step": 2897 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019901982006358658, - "loss": 1.7392, - "step": 2898 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990194818372455, - "loss": 1.8608, - "step": 2899 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019901914361090443, - "loss": 1.8188, - "step": 2900 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019901880538456337, - "loss": 1.7599, - "step": 2901 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990184671582223, - "loss": 1.7905, - "step": 2902 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019901812893188123, - "loss": 1.6561, - "step": 2903 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019901779070554017, - "loss": 1.6827, - "step": 2904 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019901745247919908, - "loss": 1.965, - "step": 2905 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019901711425285802, - "loss": 1.8191, - "step": 2906 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019901677602651697, - "loss": 1.9241, - "step": 2907 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019901643780017588, - "loss": 1.6481, - "step": 2908 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019901609957383482, - "loss": 1.6883, - "step": 2909 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019901576134749376, - "loss": 1.7425, - "step": 2910 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019901542312115268, - "loss": 2.0074, - "step": 2911 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019901508489481162, - "loss": 1.7809, - "step": 2912 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019901474666847056, - "loss": 1.7819, - "step": 2913 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019901440844212947, - "loss": 1.8901, - "step": 2914 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990140702157884, - "loss": 1.8225, - "step": 2915 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019901373198944735, - "loss": 1.9567, - "step": 2916 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019901339376310627, - "loss": 1.7789, - "step": 2917 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990130555367652, - "loss": 1.8251, - "step": 2918 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019901271731042415, - "loss": 1.8298, - "step": 2919 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019901237908408306, - "loss": 1.7179, - "step": 2920 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199012040857742, - "loss": 1.8863, - "step": 2921 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019901170263140095, - "loss": 1.8126, - "step": 2922 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019901136440505986, - "loss": 1.9028, - "step": 2923 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990110261787188, - "loss": 1.9045, - "step": 2924 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019901068795237774, - "loss": 1.7259, - "step": 2925 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019901034972603668, - "loss": 1.7922, - "step": 2926 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990100114996956, - "loss": 1.8525, - "step": 2927 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019900967327335454, - "loss": 1.8576, - "step": 2928 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019900933504701348, - "loss": 1.7075, - "step": 2929 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990089968206724, - "loss": 1.7724, - "step": 2930 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019900865859433134, - "loss": 1.9664, - "step": 2931 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019900832036799028, - "loss": 1.8178, - "step": 2932 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990079821416492, - "loss": 1.8781, - "step": 2933 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019900764391530813, - "loss": 1.8124, - "step": 2934 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019900730568896707, - "loss": 1.7931, - "step": 2935 - }, - { - "epoch": 0.01, - "learning_rate": 0.000199006967462626, - "loss": 1.8728, - "step": 2936 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019900662923628493, - "loss": 1.7486, - "step": 2937 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019900629100994387, - "loss": 1.8123, - "step": 2938 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019900595278360278, - "loss": 1.7549, - "step": 2939 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019900561455726172, - "loss": 1.9156, - "step": 2940 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019900527633092067, - "loss": 1.9017, - "step": 2941 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019900493810457958, - "loss": 1.7882, - "step": 2942 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019900459987823852, - "loss": 1.8115, - "step": 2943 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019900426165189746, - "loss": 1.8227, - "step": 2944 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019900392342555638, - "loss": 1.7876, - "step": 2945 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019900358519921532, - "loss": 1.8687, - "step": 2946 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019900324697287426, - "loss": 1.8441, - "step": 2947 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019900290874653317, - "loss": 1.6205, - "step": 2948 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019900257052019214, - "loss": 1.7429, - "step": 2949 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019900223229385106, - "loss": 1.8162, - "step": 2950 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019900189406751, - "loss": 1.746, - "step": 2951 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019900155584116894, - "loss": 1.7792, - "step": 2952 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019900121761482785, - "loss": 1.7817, - "step": 2953 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001990008793884868, - "loss": 1.8536, - "step": 2954 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019900054116214573, - "loss": 1.7949, - "step": 2955 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019900020293580465, - "loss": 1.7829, - "step": 2956 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989998647094636, - "loss": 1.7976, - "step": 2957 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019899952648312253, - "loss": 1.6495, - "step": 2958 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019899918825678144, - "loss": 1.8074, - "step": 2959 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019899885003044039, - "loss": 1.9297, - "step": 2960 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019899851180409933, - "loss": 1.8515, - "step": 2961 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019899817357775824, - "loss": 1.8813, - "step": 2962 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019899783535141718, - "loss": 1.7, - "step": 2963 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019899749712507612, - "loss": 1.8228, - "step": 2964 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019899715889873504, - "loss": 1.7116, - "step": 2965 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019899682067239398, - "loss": 1.7508, - "step": 2966 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019899648244605292, - "loss": 1.6719, - "step": 2967 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019899614421971183, - "loss": 1.9206, - "step": 2968 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019899580599337077, - "loss": 1.7304, - "step": 2969 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019899546776702972, - "loss": 1.5928, - "step": 2970 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019899512954068863, - "loss": 1.8997, - "step": 2971 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019899479131434757, - "loss": 1.8052, - "step": 2972 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989944530880065, - "loss": 1.7758, - "step": 2973 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019899411486166543, - "loss": 1.7079, - "step": 2974 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019899377663532437, - "loss": 1.6995, - "step": 2975 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989934384089833, - "loss": 1.7301, - "step": 2976 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019899310018264222, - "loss": 1.9935, - "step": 2977 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019899276195630116, - "loss": 1.8525, - "step": 2978 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989924237299601, - "loss": 1.8317, - "step": 2979 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019899208550361902, - "loss": 1.8252, - "step": 2980 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019899174727727796, - "loss": 1.7876, - "step": 2981 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989914090509369, - "loss": 1.8645, - "step": 2982 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019899107082459581, - "loss": 1.8816, - "step": 2983 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019899073259825476, - "loss": 1.8091, - "step": 2984 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989903943719137, - "loss": 1.8324, - "step": 2985 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989900561455726, - "loss": 1.8075, - "step": 2986 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019898971791923155, - "loss": 1.7888, - "step": 2987 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989893796928905, - "loss": 1.7502, - "step": 2988 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989890414665494, - "loss": 1.8152, - "step": 2989 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019898870324020835, - "loss": 1.5845, - "step": 2990 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989883650138673, - "loss": 1.778, - "step": 2991 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989880267875262, - "loss": 1.7697, - "step": 2992 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019898768856118514, - "loss": 1.7651, - "step": 2993 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019898735033484409, - "loss": 1.8051, - "step": 2994 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198987012108503, - "loss": 1.7463, - "step": 2995 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019898667388216194, - "loss": 1.9106, - "step": 2996 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019898633565582088, - "loss": 1.8707, - "step": 2997 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989859974294798, - "loss": 1.7605, - "step": 2998 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019898565920313874, - "loss": 1.7671, - "step": 2999 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019898532097679768, - "loss": 1.7474, - "step": 3000 - }, - { - "epoch": 0.01, - "eval_loss": 1.3555438701923077, - "mrc_em": 10.0, - "mrc_f1": 13.293650793650793, - "nli_acc": 0.27, - "step": 3000, - "sts_f1": 0.2972972972972973, - "ynat_f1": 0.39019346308915437 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989849827504566, - "loss": 1.7143, - "step": 3001 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019898464452411553, - "loss": 1.8734, - "step": 3002 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019898430629777447, - "loss": 1.8555, - "step": 3003 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989839680714334, - "loss": 1.7221, - "step": 3004 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019898362984509236, - "loss": 1.8316, - "step": 3005 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989832916187513, - "loss": 1.8046, - "step": 3006 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989829533924102, - "loss": 1.8253, - "step": 3007 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019898261516606915, - "loss": 1.6894, - "step": 3008 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989822769397281, - "loss": 1.7816, - "step": 3009 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198981938713387, - "loss": 1.8493, - "step": 3010 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019898160048704595, - "loss": 1.7391, - "step": 3011 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989812622607049, - "loss": 1.8276, - "step": 3012 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989809240343638, - "loss": 1.6874, - "step": 3013 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019898058580802275, - "loss": 1.706, - "step": 3014 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989802475816817, - "loss": 1.7256, - "step": 3015 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989799093553406, - "loss": 1.6843, - "step": 3016 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019897957112899954, - "loss": 1.7468, - "step": 3017 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019897923290265848, - "loss": 1.7173, - "step": 3018 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989788946763174, - "loss": 1.7273, - "step": 3019 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019897855644997634, - "loss": 1.8688, - "step": 3020 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019897821822363528, - "loss": 1.7665, - "step": 3021 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989778799972942, - "loss": 1.6928, - "step": 3022 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019897754177095313, - "loss": 1.791, - "step": 3023 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019897720354461208, - "loss": 1.8845, - "step": 3024 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198976865318271, - "loss": 1.8518, - "step": 3025 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019897652709192993, - "loss": 1.759, - "step": 3026 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019897618886558887, - "loss": 1.6929, - "step": 3027 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019897585063924779, - "loss": 1.61, - "step": 3028 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019897551241290673, - "loss": 1.7329, - "step": 3029 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019897517418656567, - "loss": 1.7024, - "step": 3030 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019897483596022458, - "loss": 1.6207, - "step": 3031 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019897449773388352, - "loss": 1.6174, - "step": 3032 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019897415950754246, - "loss": 1.9456, - "step": 3033 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019897382128120138, - "loss": 1.8135, - "step": 3034 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019897348305486032, - "loss": 1.8997, - "step": 3035 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019897314482851926, - "loss": 1.7676, - "step": 3036 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019897280660217818, - "loss": 1.7293, - "step": 3037 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019897246837583712, - "loss": 1.8687, - "step": 3038 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019897213014949606, - "loss": 1.6265, - "step": 3039 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019897179192315497, - "loss": 1.9576, - "step": 3040 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989714536968139, - "loss": 1.6673, - "step": 3041 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019897111547047285, - "loss": 1.85, - "step": 3042 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019897077724413177, - "loss": 1.8372, - "step": 3043 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989704390177907, - "loss": 1.8993, - "step": 3044 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019897010079144965, - "loss": 1.812, - "step": 3045 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019896976256510856, - "loss": 1.7809, - "step": 3046 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989694243387675, - "loss": 1.7527, - "step": 3047 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019896908611242645, - "loss": 1.767, - "step": 3048 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019896874788608536, - "loss": 1.8442, - "step": 3049 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989684096597443, - "loss": 1.6887, - "step": 3050 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019896807143340324, - "loss": 1.7467, - "step": 3051 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019896773320706216, - "loss": 1.8586, - "step": 3052 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989673949807211, - "loss": 1.6691, - "step": 3053 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019896705675438004, - "loss": 1.8012, - "step": 3054 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019896671852803895, - "loss": 1.8272, - "step": 3055 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989663803016979, - "loss": 1.8721, - "step": 3056 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019896604207535684, - "loss": 1.8306, - "step": 3057 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019896570384901575, - "loss": 1.6994, - "step": 3058 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989653656226747, - "loss": 1.7534, - "step": 3059 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019896502739633363, - "loss": 1.9476, - "step": 3060 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019896468916999255, - "loss": 1.5985, - "step": 3061 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019896435094365151, - "loss": 1.7692, - "step": 3062 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019896401271731046, - "loss": 1.8906, - "step": 3063 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019896367449096937, - "loss": 1.8487, - "step": 3064 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989633362646283, - "loss": 1.7684, - "step": 3065 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019896299803828725, - "loss": 1.8753, - "step": 3066 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019896265981194617, - "loss": 1.8469, - "step": 3067 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989623215856051, - "loss": 1.7845, - "step": 3068 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019896198335926405, - "loss": 1.6947, - "step": 3069 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019896164513292296, - "loss": 1.7879, - "step": 3070 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989613069065819, - "loss": 1.6341, - "step": 3071 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019896096868024084, - "loss": 1.8191, - "step": 3072 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019896063045389976, - "loss": 1.7955, - "step": 3073 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989602922275587, - "loss": 1.6803, - "step": 3074 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019895995400121764, - "loss": 1.8144, - "step": 3075 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019895961577487655, - "loss": 1.7351, - "step": 3076 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989592775485355, - "loss": 1.7866, - "step": 3077 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019895893932219444, - "loss": 1.848, - "step": 3078 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019895860109585335, - "loss": 1.825, - "step": 3079 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989582628695123, - "loss": 1.7611, - "step": 3080 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019895792464317123, - "loss": 1.7075, - "step": 3081 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019895758641683015, - "loss": 1.7158, - "step": 3082 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989572481904891, - "loss": 1.6942, - "step": 3083 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019895690996414803, - "loss": 1.8771, - "step": 3084 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019895657173780694, - "loss": 1.8064, - "step": 3085 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019895623351146588, - "loss": 1.7429, - "step": 3086 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019895589528512483, - "loss": 1.7558, - "step": 3087 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019895555705878374, - "loss": 1.9005, - "step": 3088 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019895521883244268, - "loss": 1.6814, - "step": 3089 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019895488060610162, - "loss": 1.7303, - "step": 3090 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019895454237976054, - "loss": 1.88, - "step": 3091 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019895420415341948, - "loss": 1.8905, - "step": 3092 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019895386592707842, - "loss": 1.7833, - "step": 3093 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019895352770073733, - "loss": 1.7318, - "step": 3094 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019895318947439627, - "loss": 1.7439, - "step": 3095 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019895285124805521, - "loss": 1.7922, - "step": 3096 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019895251302171413, - "loss": 2.0058, - "step": 3097 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019895217479537307, - "loss": 1.7742, - "step": 3098 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198951836569032, - "loss": 1.7562, - "step": 3099 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019895149834269092, - "loss": 1.7539, - "step": 3100 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019895116011634987, - "loss": 1.7612, - "step": 3101 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989508218900088, - "loss": 1.8159, - "step": 3102 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019895048366366772, - "loss": 1.8829, - "step": 3103 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019895014543732666, - "loss": 1.8677, - "step": 3104 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989498072109856, - "loss": 1.7451, - "step": 3105 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019894946898464452, - "loss": 1.86, - "step": 3106 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019894913075830346, - "loss": 1.7912, - "step": 3107 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989487925319624, - "loss": 1.6944, - "step": 3108 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019894845430562131, - "loss": 1.7182, - "step": 3109 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019894811607928026, - "loss": 1.8294, - "step": 3110 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989477778529392, - "loss": 1.8483, - "step": 3111 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989474396265981, - "loss": 1.6819, - "step": 3112 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019894710140025705, - "loss": 1.7171, - "step": 3113 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198946763173916, - "loss": 1.7693, - "step": 3114 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989464249475749, - "loss": 1.8183, - "step": 3115 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019894608672123385, - "loss": 1.6875, - "step": 3116 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989457484948928, - "loss": 1.7331, - "step": 3117 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019894541026855173, - "loss": 1.7518, - "step": 3118 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019894507204221067, - "loss": 1.8147, - "step": 3119 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989447338158696, - "loss": 1.9393, - "step": 3120 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019894439558952853, - "loss": 1.6457, - "step": 3121 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019894405736318747, - "loss": 1.7571, - "step": 3122 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989437191368464, - "loss": 1.8577, - "step": 3123 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019894338091050532, - "loss": 1.7469, - "step": 3124 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019894304268416426, - "loss": 1.8546, - "step": 3125 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989427044578232, - "loss": 1.7999, - "step": 3126 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019894236623148212, - "loss": 1.7914, - "step": 3127 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019894202800514106, - "loss": 1.8285, - "step": 3128 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989416897788, - "loss": 1.8517, - "step": 3129 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019894135155245892, - "loss": 1.8366, - "step": 3130 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019894101332611786, - "loss": 1.7537, - "step": 3131 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989406750997768, - "loss": 1.8094, - "step": 3132 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989403368734357, - "loss": 1.8656, - "step": 3133 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019893999864709465, - "loss": 1.8049, - "step": 3134 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989396604207536, - "loss": 1.7436, - "step": 3135 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989393221944125, - "loss": 1.7999, - "step": 3136 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019893898396807145, - "loss": 1.5875, - "step": 3137 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989386457417304, - "loss": 1.9028, - "step": 3138 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989383075153893, - "loss": 1.6621, - "step": 3139 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019893796928904825, - "loss": 1.8212, - "step": 3140 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019893763106270719, - "loss": 1.6142, - "step": 3141 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989372928363661, - "loss": 1.8196, - "step": 3142 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019893695461002504, - "loss": 1.7341, - "step": 3143 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019893661638368398, - "loss": 1.9539, - "step": 3144 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989362781573429, - "loss": 1.8187, - "step": 3145 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019893593993100184, - "loss": 1.7094, - "step": 3146 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019893560170466078, - "loss": 1.8928, - "step": 3147 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989352634783197, - "loss": 1.8278, - "step": 3148 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019893492525197863, - "loss": 1.8771, - "step": 3149 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019893458702563758, - "loss": 1.7827, - "step": 3150 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989342487992965, - "loss": 1.7621, - "step": 3151 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019893391057295543, - "loss": 1.8064, - "step": 3152 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019893357234661437, - "loss": 1.8841, - "step": 3153 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019893323412027329, - "loss": 1.7604, - "step": 3154 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019893289589393223, - "loss": 1.856, - "step": 3155 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019893255766759117, - "loss": 1.6497, - "step": 3156 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019893221944125008, - "loss": 1.715, - "step": 3157 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019893188121490902, - "loss": 1.9138, - "step": 3158 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019893154298856796, - "loss": 1.7895, - "step": 3159 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019893120476222688, - "loss": 1.8398, - "step": 3160 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019893086653588582, - "loss": 1.6808, - "step": 3161 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019893052830954476, - "loss": 1.7241, - "step": 3162 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019893019008320367, - "loss": 1.6492, - "step": 3163 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019892985185686262, - "loss": 1.7338, - "step": 3164 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019892951363052156, - "loss": 1.7635, - "step": 3165 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019892917540418047, - "loss": 1.7803, - "step": 3166 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989288371778394, - "loss": 1.854, - "step": 3167 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019892849895149835, - "loss": 1.93, - "step": 3168 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019892816072515727, - "loss": 1.904, - "step": 3169 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989278224988162, - "loss": 1.7247, - "step": 3170 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019892748427247515, - "loss": 1.7998, - "step": 3171 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019892714604613406, - "loss": 1.6297, - "step": 3172 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198926807819793, - "loss": 1.7713, - "step": 3173 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019892646959345195, - "loss": 1.8926, - "step": 3174 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989261313671109, - "loss": 1.8801, - "step": 3175 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019892579314076983, - "loss": 1.717, - "step": 3176 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019892545491442874, - "loss": 1.8558, - "step": 3177 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019892511668808768, - "loss": 1.7945, - "step": 3178 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019892477846174662, - "loss": 1.7156, - "step": 3179 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019892444023540554, - "loss": 1.8345, - "step": 3180 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019892410200906448, - "loss": 1.7486, - "step": 3181 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019892376378272342, - "loss": 1.7084, - "step": 3182 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019892342555638233, - "loss": 1.9667, - "step": 3183 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019892308733004128, - "loss": 1.7778, - "step": 3184 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019892274910370022, - "loss": 1.8363, - "step": 3185 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019892241087735913, - "loss": 1.7543, - "step": 3186 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019892207265101807, - "loss": 1.5885, - "step": 3187 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198921734424677, - "loss": 1.852, - "step": 3188 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019892139619833593, - "loss": 1.6886, - "step": 3189 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019892105797199487, - "loss": 1.7952, - "step": 3190 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989207197456538, - "loss": 1.6834, - "step": 3191 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019892038151931275, - "loss": 1.8324, - "step": 3192 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019892004329297166, - "loss": 1.7397, - "step": 3193 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989197050666306, - "loss": 1.79, - "step": 3194 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019891936684028955, - "loss": 1.5528, - "step": 3195 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019891902861394846, - "loss": 1.7523, - "step": 3196 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989186903876074, - "loss": 1.6652, - "step": 3197 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019891835216126634, - "loss": 1.8797, - "step": 3198 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019891801393492526, - "loss": 1.6882, - "step": 3199 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989176757085842, - "loss": 1.8096, - "step": 3200 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019891733748224314, - "loss": 1.8321, - "step": 3201 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019891699925590205, - "loss": 1.8131, - "step": 3202 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198916661029561, - "loss": 1.7048, - "step": 3203 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019891632280321994, - "loss": 1.8111, - "step": 3204 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019891598457687885, - "loss": 1.8343, - "step": 3205 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989156463505378, - "loss": 1.6693, - "step": 3206 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019891530812419673, - "loss": 1.6952, - "step": 3207 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019891496989785565, - "loss": 1.8374, - "step": 3208 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989146316715146, - "loss": 1.715, - "step": 3209 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019891429344517353, - "loss": 1.6922, - "step": 3210 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019891395521883244, - "loss": 1.594, - "step": 3211 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019891361699249138, - "loss": 1.7844, - "step": 3212 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019891327876615033, - "loss": 1.8779, - "step": 3213 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019891294053980924, - "loss": 1.763, - "step": 3214 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019891260231346818, - "loss": 1.7531, - "step": 3215 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019891226408712712, - "loss": 1.8879, - "step": 3216 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019891192586078604, - "loss": 1.7072, - "step": 3217 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019891158763444498, - "loss": 1.832, - "step": 3218 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019891124940810392, - "loss": 1.8056, - "step": 3219 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019891091118176283, - "loss": 1.711, - "step": 3220 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019891057295542177, - "loss": 1.7596, - "step": 3221 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019891023472908071, - "loss": 1.7837, - "step": 3222 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019890989650273963, - "loss": 1.8279, - "step": 3223 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019890955827639857, - "loss": 1.8395, - "step": 3224 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989092200500575, - "loss": 1.8245, - "step": 3225 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019890888182371642, - "loss": 1.5113, - "step": 3226 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019890854359737537, - "loss": 1.8053, - "step": 3227 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989082053710343, - "loss": 1.7515, - "step": 3228 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019890786714469322, - "loss": 1.8128, - "step": 3229 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989075289183522, - "loss": 1.8207, - "step": 3230 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989071906920111, - "loss": 1.7221, - "step": 3231 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019890685246567004, - "loss": 1.7246, - "step": 3232 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019890651423932899, - "loss": 1.6511, - "step": 3233 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989061760129879, - "loss": 1.8309, - "step": 3234 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019890583778664684, - "loss": 1.653, - "step": 3235 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019890549956030578, - "loss": 1.8217, - "step": 3236 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989051613339647, - "loss": 1.7502, - "step": 3237 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019890482310762364, - "loss": 1.9159, - "step": 3238 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019890448488128258, - "loss": 1.6581, - "step": 3239 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989041466549415, - "loss": 1.7608, - "step": 3240 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019890380842860043, - "loss": 1.7192, - "step": 3241 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019890347020225937, - "loss": 1.7675, - "step": 3242 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001989031319759183, - "loss": 1.8253, - "step": 3243 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019890279374957723, - "loss": 1.9583, - "step": 3244 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019890245552323617, - "loss": 1.8078, - "step": 3245 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019890211729689508, - "loss": 1.856, - "step": 3246 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019890177907055403, - "loss": 1.7485, - "step": 3247 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019890144084421297, - "loss": 1.8625, - "step": 3248 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019890110261787188, - "loss": 1.7945, - "step": 3249 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019890076439153082, - "loss": 1.771, - "step": 3250 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019890042616518976, - "loss": 1.661, - "step": 3251 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019890008793884868, - "loss": 1.8474, - "step": 3252 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019889974971250762, - "loss": 1.6972, - "step": 3253 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019889941148616656, - "loss": 1.7838, - "step": 3254 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019889907325982547, - "loss": 1.7582, - "step": 3255 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019889873503348441, - "loss": 1.7347, - "step": 3256 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019889839680714336, - "loss": 1.8582, - "step": 3257 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019889805858080227, - "loss": 1.8367, - "step": 3258 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988977203544612, - "loss": 1.7684, - "step": 3259 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019889738212812015, - "loss": 1.8189, - "step": 3260 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019889704390177907, - "loss": 1.6988, - "step": 3261 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198896705675438, - "loss": 1.8908, - "step": 3262 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019889636744909695, - "loss": 1.8748, - "step": 3263 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019889602922275586, - "loss": 1.8047, - "step": 3264 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988956909964148, - "loss": 1.8478, - "step": 3265 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019889535277007374, - "loss": 1.8952, - "step": 3266 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019889501454373266, - "loss": 1.8651, - "step": 3267 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988946763173916, - "loss": 1.681, - "step": 3268 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019889433809105054, - "loss": 1.5901, - "step": 3269 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019889399986470946, - "loss": 1.7853, - "step": 3270 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988936616383684, - "loss": 1.702, - "step": 3271 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019889332341202734, - "loss": 1.7398, - "step": 3272 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019889298518568625, - "loss": 1.6584, - "step": 3273 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988926469593452, - "loss": 1.8321, - "step": 3274 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019889230873300413, - "loss": 1.8199, - "step": 3275 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019889197050666307, - "loss": 1.653, - "step": 3276 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198891632280322, - "loss": 1.7156, - "step": 3277 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019889129405398093, - "loss": 1.7796, - "step": 3278 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019889095582763987, - "loss": 1.7733, - "step": 3279 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019889061760129879, - "loss": 1.7773, - "step": 3280 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019889027937495773, - "loss": 1.6921, - "step": 3281 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019888994114861667, - "loss": 1.7034, - "step": 3282 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019888960292227558, - "loss": 1.8103, - "step": 3283 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019888926469593452, - "loss": 1.7422, - "step": 3284 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019888892646959346, - "loss": 1.7649, - "step": 3285 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988885882432524, - "loss": 1.8576, - "step": 3286 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019888825001691135, - "loss": 1.793, - "step": 3287 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019888791179057026, - "loss": 1.918, - "step": 3288 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988875735642292, - "loss": 1.7928, - "step": 3289 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019888723533788814, - "loss": 1.8085, - "step": 3290 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019888689711154706, - "loss": 1.7427, - "step": 3291 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198886558885206, - "loss": 1.8473, - "step": 3292 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019888622065886494, - "loss": 1.7503, - "step": 3293 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019888588243252385, - "loss": 1.849, - "step": 3294 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988855442061828, - "loss": 1.9079, - "step": 3295 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019888520597984173, - "loss": 1.7322, - "step": 3296 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019888486775350065, - "loss": 1.9896, - "step": 3297 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988845295271596, - "loss": 1.6903, - "step": 3298 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019888419130081853, - "loss": 1.6984, - "step": 3299 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019888385307447745, - "loss": 1.7733, - "step": 3300 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019888351484813639, - "loss": 1.8348, - "step": 3301 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019888317662179533, - "loss": 1.8502, - "step": 3302 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019888283839545424, - "loss": 1.7989, - "step": 3303 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019888250016911318, - "loss": 1.5902, - "step": 3304 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019888216194277212, - "loss": 1.6144, - "step": 3305 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019888182371643104, - "loss": 1.7939, - "step": 3306 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019888148549008998, - "loss": 1.8371, - "step": 3307 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019888114726374892, - "loss": 1.7879, - "step": 3308 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019888080903740783, - "loss": 1.725, - "step": 3309 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019888047081106678, - "loss": 1.8511, - "step": 3310 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019888013258472572, - "loss": 1.8741, - "step": 3311 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019887979435838463, - "loss": 1.7689, - "step": 3312 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019887945613204357, - "loss": 1.7144, - "step": 3313 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988791179057025, - "loss": 1.7845, - "step": 3314 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019887877967936143, - "loss": 1.7249, - "step": 3315 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019887844145302037, - "loss": 1.9458, - "step": 3316 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988781032266793, - "loss": 1.8123, - "step": 3317 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019887776500033822, - "loss": 1.7152, - "step": 3318 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019887742677399716, - "loss": 1.7945, - "step": 3319 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988770885476561, - "loss": 1.8423, - "step": 3320 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019887675032131502, - "loss": 1.911, - "step": 3321 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019887641209497396, - "loss": 1.6149, - "step": 3322 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988760738686329, - "loss": 1.7652, - "step": 3323 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019887573564229182, - "loss": 1.7292, - "step": 3324 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019887539741595076, - "loss": 1.6645, - "step": 3325 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988750591896097, - "loss": 1.7703, - "step": 3326 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988747209632686, - "loss": 1.6092, - "step": 3327 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019887438273692755, - "loss": 1.7072, - "step": 3328 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988740445105865, - "loss": 1.8431, - "step": 3329 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988737062842454, - "loss": 1.7349, - "step": 3330 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019887336805790435, - "loss": 1.6598, - "step": 3331 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988730298315633, - "loss": 1.9014, - "step": 3332 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988726916052222, - "loss": 1.8341, - "step": 3333 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019887235337888115, - "loss": 1.8761, - "step": 3334 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988720151525401, - "loss": 1.7843, - "step": 3335 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198871676926199, - "loss": 1.8063, - "step": 3336 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019887133869985794, - "loss": 1.7431, - "step": 3337 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019887100047351688, - "loss": 1.9259, - "step": 3338 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988706622471758, - "loss": 1.6525, - "step": 3339 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019887032402083474, - "loss": 1.7155, - "step": 3340 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019886998579449368, - "loss": 1.8307, - "step": 3341 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019886964756815262, - "loss": 1.8731, - "step": 3342 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019886930934181156, - "loss": 1.8617, - "step": 3343 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988689711154705, - "loss": 1.8744, - "step": 3344 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019886863288912942, - "loss": 1.8366, - "step": 3345 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019886829466278836, - "loss": 1.7511, - "step": 3346 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988679564364473, - "loss": 1.6902, - "step": 3347 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988676182101062, - "loss": 1.7975, - "step": 3348 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019886727998376515, - "loss": 1.7917, - "step": 3349 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988669417574241, - "loss": 1.8667, - "step": 3350 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198866603531083, - "loss": 1.8128, - "step": 3351 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019886626530474195, - "loss": 1.8028, - "step": 3352 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988659270784009, - "loss": 1.7296, - "step": 3353 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988655888520598, - "loss": 1.8974, - "step": 3354 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019886525062571875, - "loss": 1.7856, - "step": 3355 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988649123993777, - "loss": 1.6941, - "step": 3356 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988645741730366, - "loss": 1.7055, - "step": 3357 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019886423594669554, - "loss": 1.8105, - "step": 3358 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019886389772035448, - "loss": 1.7735, - "step": 3359 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988635594940134, - "loss": 1.7274, - "step": 3360 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019886322126767234, - "loss": 1.6873, - "step": 3361 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019886288304133128, - "loss": 1.867, - "step": 3362 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988625448149902, - "loss": 1.9474, - "step": 3363 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019886220658864914, - "loss": 1.7802, - "step": 3364 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019886186836230808, - "loss": 1.8376, - "step": 3365 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198861530135967, - "loss": 1.809, - "step": 3366 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019886119190962593, - "loss": 1.6789, - "step": 3367 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019886085368328487, - "loss": 1.8567, - "step": 3368 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988605154569438, - "loss": 1.7249, - "step": 3369 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019886017723060273, - "loss": 1.8295, - "step": 3370 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019885983900426167, - "loss": 1.6805, - "step": 3371 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019885950077792058, - "loss": 1.6176, - "step": 3372 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019885916255157953, - "loss": 1.8081, - "step": 3373 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019885882432523847, - "loss": 1.6984, - "step": 3374 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019885848609889738, - "loss": 1.9175, - "step": 3375 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019885814787255632, - "loss": 1.8281, - "step": 3376 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019885780964621526, - "loss": 1.7618, - "step": 3377 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019885747141987418, - "loss": 1.7948, - "step": 3378 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019885713319353312, - "loss": 1.8506, - "step": 3379 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019885679496719206, - "loss": 1.8006, - "step": 3380 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019885645674085097, - "loss": 1.7863, - "step": 3381 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019885611851450991, - "loss": 1.6772, - "step": 3382 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019885578028816886, - "loss": 1.8783, - "step": 3383 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019885544206182777, - "loss": 1.6906, - "step": 3384 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988551038354867, - "loss": 1.7889, - "step": 3385 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019885476560914565, - "loss": 1.7483, - "step": 3386 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019885442738280457, - "loss": 1.9794, - "step": 3387 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988540891564635, - "loss": 1.9072, - "step": 3388 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019885375093012245, - "loss": 1.6572, - "step": 3389 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019885341270378136, - "loss": 1.5352, - "step": 3390 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988530744774403, - "loss": 1.8328, - "step": 3391 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019885273625109924, - "loss": 1.6098, - "step": 3392 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019885239802475816, - "loss": 1.8447, - "step": 3393 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988520597984171, - "loss": 1.7642, - "step": 3394 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019885172157207604, - "loss": 1.6772, - "step": 3395 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019885138334573495, - "loss": 1.8639, - "step": 3396 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988510451193939, - "loss": 1.7612, - "step": 3397 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019885070689305284, - "loss": 1.8119, - "step": 3398 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019885036866671178, - "loss": 1.8769, - "step": 3399 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019885003044037072, - "loss": 1.6761, - "step": 3400 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019884969221402966, - "loss": 1.853, - "step": 3401 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019884935398768857, - "loss": 1.769, - "step": 3402 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019884901576134752, - "loss": 1.6784, - "step": 3403 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019884867753500646, - "loss": 1.8343, - "step": 3404 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019884833930866537, - "loss": 1.5937, - "step": 3405 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988480010823243, - "loss": 1.69, - "step": 3406 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019884766285598325, - "loss": 1.8235, - "step": 3407 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019884732462964217, - "loss": 1.8338, - "step": 3408 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988469864033011, - "loss": 1.7587, - "step": 3409 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019884664817696005, - "loss": 1.6451, - "step": 3410 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019884630995061896, - "loss": 1.779, - "step": 3411 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988459717242779, - "loss": 1.7439, - "step": 3412 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019884563349793685, - "loss": 1.8377, - "step": 3413 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019884529527159576, - "loss": 1.7138, - "step": 3414 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988449570452547, - "loss": 1.7428, - "step": 3415 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019884461881891364, - "loss": 1.8334, - "step": 3416 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019884428059257256, - "loss": 1.4723, - "step": 3417 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988439423662315, - "loss": 1.764, - "step": 3418 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019884360413989044, - "loss": 1.9135, - "step": 3419 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019884326591354935, - "loss": 1.7364, - "step": 3420 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988429276872083, - "loss": 1.6219, - "step": 3421 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019884258946086723, - "loss": 1.7112, - "step": 3422 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019884225123452615, - "loss": 1.8449, - "step": 3423 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988419130081851, - "loss": 1.8262, - "step": 3424 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019884157478184403, - "loss": 1.7794, - "step": 3425 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019884123655550294, - "loss": 1.6687, - "step": 3426 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019884089832916189, - "loss": 1.8855, - "step": 3427 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019884056010282083, - "loss": 1.8156, - "step": 3428 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019884022187647974, - "loss": 1.7952, - "step": 3429 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019883988365013868, - "loss": 1.7816, - "step": 3430 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019883954542379762, - "loss": 1.6359, - "step": 3431 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019883920719745654, - "loss": 1.7687, - "step": 3432 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019883886897111548, - "loss": 1.7139, - "step": 3433 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019883853074477442, - "loss": 1.8224, - "step": 3434 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019883819251843333, - "loss": 1.7086, - "step": 3435 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019883785429209227, - "loss": 1.8485, - "step": 3436 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019883751606575122, - "loss": 1.6973, - "step": 3437 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019883717783941013, - "loss": 1.784, - "step": 3438 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019883683961306907, - "loss": 1.7628, - "step": 3439 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198836501386728, - "loss": 1.7765, - "step": 3440 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019883616316038693, - "loss": 1.7565, - "step": 3441 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019883582493404587, - "loss": 1.7377, - "step": 3442 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988354867077048, - "loss": 1.8762, - "step": 3443 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019883514848136372, - "loss": 1.7295, - "step": 3444 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019883481025502266, - "loss": 1.8313, - "step": 3445 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988344720286816, - "loss": 1.8716, - "step": 3446 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019883413380234052, - "loss": 1.6536, - "step": 3447 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019883379557599946, - "loss": 1.6871, - "step": 3448 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988334573496584, - "loss": 1.8348, - "step": 3449 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019883311912331732, - "loss": 1.7269, - "step": 3450 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019883278089697626, - "loss": 1.7798, - "step": 3451 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988324426706352, - "loss": 1.7484, - "step": 3452 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988321044442941, - "loss": 1.7228, - "step": 3453 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019883176621795305, - "loss": 1.8977, - "step": 3454 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198831427991612, - "loss": 1.7702, - "step": 3455 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019883108976527093, - "loss": 1.8742, - "step": 3456 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019883075153892988, - "loss": 1.7852, - "step": 3457 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988304133125888, - "loss": 1.7419, - "step": 3458 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019883007508624773, - "loss": 1.8368, - "step": 3459 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019882973685990667, - "loss": 1.7409, - "step": 3460 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988293986335656, - "loss": 1.8085, - "step": 3461 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019882906040722453, - "loss": 1.8683, - "step": 3462 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019882872218088347, - "loss": 1.8313, - "step": 3463 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988283839545424, - "loss": 1.8328, - "step": 3464 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019882804572820132, - "loss": 1.8329, - "step": 3465 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019882770750186026, - "loss": 1.9185, - "step": 3466 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988273692755192, - "loss": 1.8019, - "step": 3467 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019882703104917812, - "loss": 1.8883, - "step": 3468 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019882669282283706, - "loss": 1.8132, - "step": 3469 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198826354596496, - "loss": 1.6767, - "step": 3470 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019882601637015492, - "loss": 1.6957, - "step": 3471 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019882567814381386, - "loss": 1.7678, - "step": 3472 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988253399174728, - "loss": 1.934, - "step": 3473 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988250016911317, - "loss": 1.8916, - "step": 3474 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019882466346479065, - "loss": 1.8453, - "step": 3475 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988243252384496, - "loss": 1.6921, - "step": 3476 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988239870121085, - "loss": 1.9193, - "step": 3477 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019882364878576745, - "loss": 1.8405, - "step": 3478 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988233105594264, - "loss": 1.785, - "step": 3479 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988229723330853, - "loss": 1.6398, - "step": 3480 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019882263410674425, - "loss": 1.6573, - "step": 3481 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988222958804032, - "loss": 1.8035, - "step": 3482 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988219576540621, - "loss": 1.8647, - "step": 3483 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019882161942772104, - "loss": 1.7298, - "step": 3484 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019882128120137998, - "loss": 1.7205, - "step": 3485 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988209429750389, - "loss": 1.8496, - "step": 3486 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019882060474869784, - "loss": 1.8755, - "step": 3487 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019882026652235678, - "loss": 1.6807, - "step": 3488 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988199282960157, - "loss": 1.9146, - "step": 3489 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019881959006967464, - "loss": 1.6955, - "step": 3490 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019881925184333358, - "loss": 1.8457, - "step": 3491 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988189136169925, - "loss": 1.7173, - "step": 3492 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019881857539065143, - "loss": 1.7431, - "step": 3493 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019881823716431037, - "loss": 1.5905, - "step": 3494 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988178989379693, - "loss": 1.8009, - "step": 3495 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019881756071162823, - "loss": 1.8908, - "step": 3496 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019881722248528717, - "loss": 1.6518, - "step": 3497 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019881688425894608, - "loss": 1.8235, - "step": 3498 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019881654603260502, - "loss": 1.722, - "step": 3499 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019881620780626397, - "loss": 1.7723, - "step": 3500 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019881586957992288, - "loss": 1.8816, - "step": 3501 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019881553135358182, - "loss": 1.8265, - "step": 3502 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019881519312724076, - "loss": 1.8289, - "step": 3503 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019881485490089968, - "loss": 1.6792, - "step": 3504 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019881451667455862, - "loss": 1.692, - "step": 3505 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019881417844821756, - "loss": 1.7932, - "step": 3506 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019881384022187647, - "loss": 1.7442, - "step": 3507 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988135019955354, - "loss": 1.8263, - "step": 3508 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019881316376919435, - "loss": 1.8759, - "step": 3509 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019881282554285327, - "loss": 1.5383, - "step": 3510 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019881248731651224, - "loss": 1.8038, - "step": 3511 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019881214909017115, - "loss": 1.7361, - "step": 3512 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988118108638301, - "loss": 1.8876, - "step": 3513 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019881147263748903, - "loss": 1.7727, - "step": 3514 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019881113441114795, - "loss": 1.6697, - "step": 3515 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988107961848069, - "loss": 1.6961, - "step": 3516 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019881045795846583, - "loss": 1.7892, - "step": 3517 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019881011973212474, - "loss": 1.6943, - "step": 3518 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019880978150578368, - "loss": 1.8657, - "step": 3519 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019880944327944263, - "loss": 1.9639, - "step": 3520 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019880910505310154, - "loss": 1.7599, - "step": 3521 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019880876682676048, - "loss": 1.7306, - "step": 3522 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019880842860041942, - "loss": 1.7851, - "step": 3523 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019880809037407834, - "loss": 1.6254, - "step": 3524 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019880775214773728, - "loss": 1.8499, - "step": 3525 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019880741392139622, - "loss": 1.8794, - "step": 3526 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019880707569505513, - "loss": 1.848, - "step": 3527 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019880673746871407, - "loss": 1.8877, - "step": 3528 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019880639924237301, - "loss": 1.7706, - "step": 3529 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019880606101603193, - "loss": 1.7507, - "step": 3530 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019880572278969087, - "loss": 1.7481, - "step": 3531 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988053845633498, - "loss": 1.8197, - "step": 3532 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019880504633700872, - "loss": 1.742, - "step": 3533 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019880470811066767, - "loss": 1.8692, - "step": 3534 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988043698843266, - "loss": 1.726, - "step": 3535 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019880403165798552, - "loss": 1.8018, - "step": 3536 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019880369343164446, - "loss": 1.6466, - "step": 3537 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988033552053034, - "loss": 1.6442, - "step": 3538 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019880301697896232, - "loss": 1.854, - "step": 3539 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019880267875262126, - "loss": 1.8427, - "step": 3540 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988023405262802, - "loss": 1.5638, - "step": 3541 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019880200229993911, - "loss": 1.7302, - "step": 3542 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019880166407359806, - "loss": 1.7324, - "step": 3543 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198801325847257, - "loss": 1.7961, - "step": 3544 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019880098762091594, - "loss": 1.7705, - "step": 3545 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019880064939457485, - "loss": 1.7727, - "step": 3546 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001988003111682338, - "loss": 1.7909, - "step": 3547 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019879997294189273, - "loss": 1.6861, - "step": 3548 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019879963471555165, - "loss": 1.8286, - "step": 3549 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987992964892106, - "loss": 1.7809, - "step": 3550 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019879895826286953, - "loss": 1.9089, - "step": 3551 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019879862003652844, - "loss": 1.7758, - "step": 3552 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019879828181018739, - "loss": 1.7625, - "step": 3553 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019879794358384633, - "loss": 1.7406, - "step": 3554 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019879760535750524, - "loss": 1.7447, - "step": 3555 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019879726713116418, - "loss": 1.8133, - "step": 3556 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019879692890482312, - "loss": 1.6972, - "step": 3557 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019879659067848204, - "loss": 1.9362, - "step": 3558 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019879625245214098, - "loss": 1.7062, - "step": 3559 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019879591422579992, - "loss": 1.6939, - "step": 3560 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019879557599945883, - "loss": 1.6747, - "step": 3561 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019879523777311777, - "loss": 1.7264, - "step": 3562 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019879489954677672, - "loss": 1.8197, - "step": 3563 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019879456132043563, - "loss": 1.7205, - "step": 3564 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019879422309409457, - "loss": 1.8142, - "step": 3565 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987938848677535, - "loss": 1.9255, - "step": 3566 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019879354664141245, - "loss": 1.7335, - "step": 3567 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987932084150714, - "loss": 1.6386, - "step": 3568 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987928701887303, - "loss": 1.7415, - "step": 3569 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019879253196238925, - "loss": 1.8183, - "step": 3570 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987921937360482, - "loss": 1.6007, - "step": 3571 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987918555097071, - "loss": 1.7161, - "step": 3572 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019879151728336605, - "loss": 1.7286, - "step": 3573 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019879117905702499, - "loss": 1.8248, - "step": 3574 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987908408306839, - "loss": 1.7577, - "step": 3575 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019879050260434284, - "loss": 1.8452, - "step": 3576 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019879016437800178, - "loss": 1.7033, - "step": 3577 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987898261516607, - "loss": 1.8292, - "step": 3578 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019878948792531964, - "loss": 1.7821, - "step": 3579 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019878914969897858, - "loss": 1.6696, - "step": 3580 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987888114726375, - "loss": 1.8381, - "step": 3581 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019878847324629643, - "loss": 1.7874, - "step": 3582 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019878813501995538, - "loss": 1.7935, - "step": 3583 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987877967936143, - "loss": 1.6762, - "step": 3584 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019878745856727323, - "loss": 1.906, - "step": 3585 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019878712034093217, - "loss": 1.8369, - "step": 3586 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019878678211459109, - "loss": 1.7622, - "step": 3587 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019878644388825003, - "loss": 1.7669, - "step": 3588 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019878610566190897, - "loss": 1.7688, - "step": 3589 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019878576743556788, - "loss": 1.8523, - "step": 3590 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019878542920922682, - "loss": 1.7473, - "step": 3591 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019878509098288576, - "loss": 1.8717, - "step": 3592 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019878475275654468, - "loss": 1.743, - "step": 3593 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019878441453020362, - "loss": 1.8021, - "step": 3594 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019878407630386256, - "loss": 1.7254, - "step": 3595 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019878373807752147, - "loss": 1.6173, - "step": 3596 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019878339985118042, - "loss": 1.7546, - "step": 3597 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019878306162483936, - "loss": 1.6072, - "step": 3598 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019878272339849827, - "loss": 1.706, - "step": 3599 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987823851721572, - "loss": 1.9557, - "step": 3600 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019878204694581615, - "loss": 1.7055, - "step": 3601 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019878170871947507, - "loss": 1.7064, - "step": 3602 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198781370493134, - "loss": 1.7784, - "step": 3603 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019878103226679295, - "loss": 1.6851, - "step": 3604 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019878069404045186, - "loss": 1.6782, - "step": 3605 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987803558141108, - "loss": 1.7944, - "step": 3606 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019878001758776975, - "loss": 1.8765, - "step": 3607 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019877967936142866, - "loss": 1.9212, - "step": 3608 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987793411350876, - "loss": 1.8341, - "step": 3609 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019877900290874654, - "loss": 1.8694, - "step": 3610 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019877866468240546, - "loss": 1.7769, - "step": 3611 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987783264560644, - "loss": 1.8431, - "step": 3612 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019877798822972334, - "loss": 1.7217, - "step": 3613 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019877765000338225, - "loss": 1.6689, - "step": 3614 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987773117770412, - "loss": 1.7386, - "step": 3615 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019877697355070013, - "loss": 1.8418, - "step": 3616 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019877663532435905, - "loss": 1.8579, - "step": 3617 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198776297098018, - "loss": 1.8184, - "step": 3618 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019877595887167693, - "loss": 1.7646, - "step": 3619 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019877562064533585, - "loss": 1.8131, - "step": 3620 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019877528241899479, - "loss": 1.8095, - "step": 3621 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019877494419265373, - "loss": 1.6966, - "step": 3622 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019877460596631267, - "loss": 1.7235, - "step": 3623 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987742677399716, - "loss": 1.8224, - "step": 3624 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019877392951363055, - "loss": 1.757, - "step": 3625 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019877359128728946, - "loss": 1.8681, - "step": 3626 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987732530609484, - "loss": 1.6939, - "step": 3627 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019877291483460735, - "loss": 1.7145, - "step": 3628 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019877257660826626, - "loss": 1.8203, - "step": 3629 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987722383819252, - "loss": 1.688, - "step": 3630 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019877190015558414, - "loss": 1.7243, - "step": 3631 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019877156192924306, - "loss": 1.8735, - "step": 3632 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198771223702902, - "loss": 1.6506, - "step": 3633 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019877088547656094, - "loss": 1.6678, - "step": 3634 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019877054725021985, - "loss": 1.8444, - "step": 3635 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987702090238788, - "loss": 1.7664, - "step": 3636 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019876987079753774, - "loss": 1.7964, - "step": 3637 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019876953257119665, - "loss": 1.7443, - "step": 3638 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987691943448556, - "loss": 1.7796, - "step": 3639 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019876885611851453, - "loss": 1.8284, - "step": 3640 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019876851789217345, - "loss": 1.817, - "step": 3641 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987681796658324, - "loss": 1.682, - "step": 3642 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019876784143949133, - "loss": 1.7558, - "step": 3643 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019876750321315024, - "loss": 1.7651, - "step": 3644 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019876716498680918, - "loss": 1.7545, - "step": 3645 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019876682676046813, - "loss": 1.8299, - "step": 3646 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019876648853412704, - "loss": 1.7845, - "step": 3647 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019876615030778598, - "loss": 1.708, - "step": 3648 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019876581208144492, - "loss": 1.7271, - "step": 3649 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019876547385510384, - "loss": 1.8047, - "step": 3650 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019876513562876278, - "loss": 1.8643, - "step": 3651 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019876479740242172, - "loss": 1.7196, - "step": 3652 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019876445917608063, - "loss": 1.8647, - "step": 3653 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019876412094973957, - "loss": 1.8636, - "step": 3654 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019876378272339851, - "loss": 1.923, - "step": 3655 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019876344449705743, - "loss": 1.6916, - "step": 3656 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019876310627071637, - "loss": 1.7676, - "step": 3657 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987627680443753, - "loss": 1.8062, - "step": 3658 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019876242981803422, - "loss": 1.8342, - "step": 3659 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019876209159169317, - "loss": 1.8594, - "step": 3660 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987617533653521, - "loss": 1.7069, - "step": 3661 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019876141513901102, - "loss": 1.9344, - "step": 3662 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019876107691266996, - "loss": 1.7755, - "step": 3663 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987607386863289, - "loss": 1.5779, - "step": 3664 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019876040045998782, - "loss": 1.7253, - "step": 3665 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019876006223364676, - "loss": 1.7672, - "step": 3666 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987597240073057, - "loss": 1.6038, - "step": 3667 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987593857809646, - "loss": 1.6562, - "step": 3668 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019875904755462355, - "loss": 1.7432, - "step": 3669 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987587093282825, - "loss": 1.7363, - "step": 3670 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987583711019414, - "loss": 1.6299, - "step": 3671 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019875803287560035, - "loss": 1.8173, - "step": 3672 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987576946492593, - "loss": 1.6957, - "step": 3673 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987573564229182, - "loss": 1.7262, - "step": 3674 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019875701819657715, - "loss": 1.9072, - "step": 3675 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987566799702361, - "loss": 1.8611, - "step": 3676 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198756341743895, - "loss": 1.8223, - "step": 3677 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019875600351755394, - "loss": 1.6156, - "step": 3678 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987556652912129, - "loss": 1.7345, - "step": 3679 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019875532706487183, - "loss": 1.7568, - "step": 3680 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019875498883853077, - "loss": 1.8484, - "step": 3681 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987546506121897, - "loss": 1.7242, - "step": 3682 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019875431238584862, - "loss": 1.6758, - "step": 3683 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019875397415950756, - "loss": 1.7741, - "step": 3684 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987536359331665, - "loss": 1.9517, - "step": 3685 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019875329770682542, - "loss": 1.7425, - "step": 3686 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019875295948048436, - "loss": 1.7959, - "step": 3687 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987526212541433, - "loss": 1.8497, - "step": 3688 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019875228302780221, - "loss": 1.797, - "step": 3689 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019875194480146116, - "loss": 1.8311, - "step": 3690 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987516065751201, - "loss": 1.8026, - "step": 3691 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198751268348779, - "loss": 1.7178, - "step": 3692 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019875093012243795, - "loss": 1.7372, - "step": 3693 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987505918960969, - "loss": 1.8652, - "step": 3694 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987502536697558, - "loss": 1.7642, - "step": 3695 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019874991544341475, - "loss": 1.7091, - "step": 3696 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987495772170737, - "loss": 1.8301, - "step": 3697 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987492389907326, - "loss": 1.6733, - "step": 3698 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019874890076439154, - "loss": 1.9138, - "step": 3699 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019874856253805049, - "loss": 1.6153, - "step": 3700 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987482243117094, - "loss": 1.6241, - "step": 3701 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019874788608536834, - "loss": 1.8211, - "step": 3702 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019874754785902728, - "loss": 1.7034, - "step": 3703 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987472096326862, - "loss": 1.8676, - "step": 3704 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019874687140634514, - "loss": 1.6285, - "step": 3705 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019874653318000408, - "loss": 1.9137, - "step": 3706 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198746194953663, - "loss": 1.8592, - "step": 3707 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019874585672732193, - "loss": 1.747, - "step": 3708 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019874551850098087, - "loss": 1.7528, - "step": 3709 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987451802746398, - "loss": 1.8395, - "step": 3710 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019874484204829873, - "loss": 1.7311, - "step": 3711 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019874450382195767, - "loss": 1.7955, - "step": 3712 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019874416559561659, - "loss": 1.7171, - "step": 3713 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019874382736927553, - "loss": 1.7935, - "step": 3714 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019874348914293447, - "loss": 1.6877, - "step": 3715 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019874315091659338, - "loss": 1.8287, - "step": 3716 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019874281269025232, - "loss": 1.6991, - "step": 3717 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019874247446391126, - "loss": 1.7085, - "step": 3718 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019874213623757018, - "loss": 1.8276, - "step": 3719 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019874179801122912, - "loss": 1.7113, - "step": 3720 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019874145978488806, - "loss": 1.8041, - "step": 3721 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019874112155854697, - "loss": 1.715, - "step": 3722 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019874078333220592, - "loss": 1.8491, - "step": 3723 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019874044510586486, - "loss": 1.7437, - "step": 3724 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019874010687952377, - "loss": 1.7537, - "step": 3725 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987397686531827, - "loss": 1.8486, - "step": 3726 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019873943042684165, - "loss": 1.7826, - "step": 3727 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019873909220050057, - "loss": 1.8255, - "step": 3728 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987387539741595, - "loss": 1.8003, - "step": 3729 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019873841574781845, - "loss": 1.7822, - "step": 3730 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019873807752147736, - "loss": 1.7778, - "step": 3731 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987377392951363, - "loss": 1.721, - "step": 3732 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019873740106879525, - "loss": 1.6757, - "step": 3733 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019873706284245416, - "loss": 1.722, - "step": 3734 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987367246161131, - "loss": 1.9891, - "step": 3735 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019873638638977207, - "loss": 1.615, - "step": 3736 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019873604816343098, - "loss": 1.833, - "step": 3737 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019873570993708992, - "loss": 1.6913, - "step": 3738 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019873537171074887, - "loss": 1.8046, - "step": 3739 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019873503348440778, - "loss": 1.8141, - "step": 3740 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019873469525806672, - "loss": 1.6749, - "step": 3741 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019873435703172566, - "loss": 1.7809, - "step": 3742 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019873401880538458, - "loss": 1.7403, - "step": 3743 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019873368057904352, - "loss": 1.7311, - "step": 3744 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019873334235270246, - "loss": 1.652, - "step": 3745 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019873300412636137, - "loss": 1.7864, - "step": 3746 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987326659000203, - "loss": 1.6485, - "step": 3747 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019873232767367925, - "loss": 1.9437, - "step": 3748 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019873198944733817, - "loss": 1.8951, - "step": 3749 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987316512209971, - "loss": 1.6777, - "step": 3750 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019873131299465605, - "loss": 1.7265, - "step": 3751 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019873097476831496, - "loss": 1.8795, - "step": 3752 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987306365419739, - "loss": 1.7049, - "step": 3753 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019873029831563285, - "loss": 1.8251, - "step": 3754 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019872996008929176, - "loss": 1.7768, - "step": 3755 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987296218629507, - "loss": 1.7258, - "step": 3756 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019872928363660964, - "loss": 1.6109, - "step": 3757 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019872894541026856, - "loss": 1.8156, - "step": 3758 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987286071839275, - "loss": 1.7596, - "step": 3759 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019872826895758644, - "loss": 1.6055, - "step": 3760 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019872793073124535, - "loss": 1.8298, - "step": 3761 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987275925049043, - "loss": 1.8372, - "step": 3762 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019872725427856324, - "loss": 1.8123, - "step": 3763 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019872691605222215, - "loss": 1.8066, - "step": 3764 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987265778258811, - "loss": 1.7358, - "step": 3765 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019872623959954003, - "loss": 1.7071, - "step": 3766 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019872590137319895, - "loss": 1.6979, - "step": 3767 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987255631468579, - "loss": 1.7066, - "step": 3768 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019872522492051683, - "loss": 1.8338, - "step": 3769 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019872488669417574, - "loss": 1.565, - "step": 3770 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019872454846783468, - "loss": 1.8344, - "step": 3771 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019872421024149362, - "loss": 1.8058, - "step": 3772 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019872387201515254, - "loss": 1.6099, - "step": 3773 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019872353378881148, - "loss": 1.7843, - "step": 3774 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019872319556247042, - "loss": 1.7218, - "step": 3775 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019872285733612933, - "loss": 1.7856, - "step": 3776 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019872251910978828, - "loss": 1.7157, - "step": 3777 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019872218088344722, - "loss": 1.7029, - "step": 3778 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019872184265710613, - "loss": 1.8076, - "step": 3779 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019872150443076507, - "loss": 1.6737, - "step": 3780 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198721166204424, - "loss": 1.6767, - "step": 3781 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019872082797808293, - "loss": 1.6787, - "step": 3782 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019872048975174187, - "loss": 1.8106, - "step": 3783 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987201515254008, - "loss": 1.6487, - "step": 3784 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019871981329905972, - "loss": 1.765, - "step": 3785 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019871947507271866, - "loss": 1.6999, - "step": 3786 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987191368463776, - "loss": 1.806, - "step": 3787 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019871879862003652, - "loss": 1.8061, - "step": 3788 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019871846039369546, - "loss": 1.7693, - "step": 3789 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987181221673544, - "loss": 1.8358, - "step": 3790 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019871778394101332, - "loss": 1.8443, - "step": 3791 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019871744571467228, - "loss": 1.8085, - "step": 3792 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987171074883312, - "loss": 1.6821, - "step": 3793 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019871676926199014, - "loss": 1.6677, - "step": 3794 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019871643103564908, - "loss": 1.8957, - "step": 3795 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198716092809308, - "loss": 1.7492, - "step": 3796 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019871575458296694, - "loss": 1.821, - "step": 3797 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019871541635662588, - "loss": 1.7363, - "step": 3798 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987150781302848, - "loss": 1.6887, - "step": 3799 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019871473990394373, - "loss": 1.7339, - "step": 3800 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019871440167760267, - "loss": 1.7251, - "step": 3801 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987140634512616, - "loss": 1.7336, - "step": 3802 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019871372522492053, - "loss": 1.6364, - "step": 3803 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019871338699857947, - "loss": 1.8641, - "step": 3804 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019871304877223838, - "loss": 1.7091, - "step": 3805 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019871271054589733, - "loss": 1.6371, - "step": 3806 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019871237231955627, - "loss": 1.9013, - "step": 3807 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019871203409321518, - "loss": 1.8707, - "step": 3808 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019871169586687412, - "loss": 1.595, - "step": 3809 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019871135764053306, - "loss": 1.7794, - "step": 3810 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198711019414192, - "loss": 1.7988, - "step": 3811 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019871068118785092, - "loss": 1.7946, - "step": 3812 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019871034296150986, - "loss": 1.6371, - "step": 3813 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987100047351688, - "loss": 1.7975, - "step": 3814 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019870966650882771, - "loss": 1.8724, - "step": 3815 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019870932828248666, - "loss": 1.7072, - "step": 3816 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987089900561456, - "loss": 1.76, - "step": 3817 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987086518298045, - "loss": 1.8301, - "step": 3818 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019870831360346345, - "loss": 1.6717, - "step": 3819 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987079753771224, - "loss": 1.6723, - "step": 3820 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987076371507813, - "loss": 1.8924, - "step": 3821 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019870729892444025, - "loss": 1.8265, - "step": 3822 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987069606980992, - "loss": 1.722, - "step": 3823 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987066224717581, - "loss": 1.717, - "step": 3824 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019870628424541704, - "loss": 1.9437, - "step": 3825 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019870594601907599, - "loss": 1.8353, - "step": 3826 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987056077927349, - "loss": 1.8858, - "step": 3827 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019870526956639384, - "loss": 1.7589, - "step": 3828 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019870493134005278, - "loss": 1.7588, - "step": 3829 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987045931137117, - "loss": 1.7212, - "step": 3830 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019870425488737064, - "loss": 1.6028, - "step": 3831 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019870391666102958, - "loss": 1.7211, - "step": 3832 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987035784346885, - "loss": 1.751, - "step": 3833 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019870324020834743, - "loss": 1.6938, - "step": 3834 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019870290198200637, - "loss": 1.6937, - "step": 3835 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001987025637556653, - "loss": 1.6847, - "step": 3836 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019870222552932423, - "loss": 1.6674, - "step": 3837 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019870188730298317, - "loss": 1.9292, - "step": 3838 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019870154907664208, - "loss": 1.7098, - "step": 3839 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019870121085030103, - "loss": 1.6357, - "step": 3840 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019870087262395997, - "loss": 1.8738, - "step": 3841 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019870053439761888, - "loss": 1.9045, - "step": 3842 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019870019617127782, - "loss": 1.7388, - "step": 3843 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019869985794493676, - "loss": 1.6924, - "step": 3844 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019869951971859568, - "loss": 1.6286, - "step": 3845 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019869918149225462, - "loss": 1.8045, - "step": 3846 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019869884326591356, - "loss": 1.8417, - "step": 3847 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986985050395725, - "loss": 1.6801, - "step": 3848 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019869816681323144, - "loss": 1.7158, - "step": 3849 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019869782858689036, - "loss": 1.5725, - "step": 3850 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986974903605493, - "loss": 1.869, - "step": 3851 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019869715213420824, - "loss": 1.7383, - "step": 3852 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019869681390786715, - "loss": 1.7388, - "step": 3853 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986964756815261, - "loss": 1.7882, - "step": 3854 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019869613745518503, - "loss": 1.7191, - "step": 3855 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019869579922884395, - "loss": 1.7109, - "step": 3856 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986954610025029, - "loss": 1.7151, - "step": 3857 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019869512277616183, - "loss": 1.6198, - "step": 3858 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019869478454982074, - "loss": 1.7699, - "step": 3859 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019869444632347969, - "loss": 1.6445, - "step": 3860 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019869410809713863, - "loss": 1.8315, - "step": 3861 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019869376987079754, - "loss": 1.7642, - "step": 3862 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019869343164445648, - "loss": 1.724, - "step": 3863 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019869309341811542, - "loss": 1.7898, - "step": 3864 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019869275519177434, - "loss": 1.6996, - "step": 3865 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019869241696543328, - "loss": 1.8069, - "step": 3866 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019869207873909222, - "loss": 1.7884, - "step": 3867 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019869174051275113, - "loss": 1.7171, - "step": 3868 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019869140228641007, - "loss": 1.8515, - "step": 3869 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019869106406006902, - "loss": 1.6815, - "step": 3870 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019869072583372793, - "loss": 1.5674, - "step": 3871 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019869038760738687, - "loss": 1.8523, - "step": 3872 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986900493810458, - "loss": 1.7747, - "step": 3873 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019868971115470473, - "loss": 1.7468, - "step": 3874 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019868937292836367, - "loss": 1.7403, - "step": 3875 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986890347020226, - "loss": 1.7074, - "step": 3876 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019868869647568152, - "loss": 1.6841, - "step": 3877 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019868835824934046, - "loss": 1.7727, - "step": 3878 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986880200229994, - "loss": 1.6726, - "step": 3879 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019868768179665832, - "loss": 1.7049, - "step": 3880 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019868734357031726, - "loss": 1.7384, - "step": 3881 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986870053439762, - "loss": 1.6398, - "step": 3882 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019868666711763512, - "loss": 1.7167, - "step": 3883 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019868632889129406, - "loss": 1.7093, - "step": 3884 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198685990664953, - "loss": 1.8104, - "step": 3885 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986856524386119, - "loss": 1.9136, - "step": 3886 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019868531421227085, - "loss": 1.7656, - "step": 3887 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986849759859298, - "loss": 1.8594, - "step": 3888 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986846377595887, - "loss": 1.7916, - "step": 3889 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019868429953324765, - "loss": 1.7886, - "step": 3890 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986839613069066, - "loss": 1.8948, - "step": 3891 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986836230805655, - "loss": 1.838, - "step": 3892 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019868328485422445, - "loss": 1.6862, - "step": 3893 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019868294662788339, - "loss": 1.7527, - "step": 3894 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986826084015423, - "loss": 1.8177, - "step": 3895 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019868227017520124, - "loss": 1.8084, - "step": 3896 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019868193194886018, - "loss": 1.9119, - "step": 3897 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019868159372251912, - "loss": 1.7579, - "step": 3898 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019868125549617804, - "loss": 1.6465, - "step": 3899 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019868091726983698, - "loss": 1.8036, - "step": 3900 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019868057904349592, - "loss": 1.7756, - "step": 3901 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019868024081715483, - "loss": 1.6953, - "step": 3902 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019867990259081378, - "loss": 1.7515, - "step": 3903 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019867956436447272, - "loss": 1.8245, - "step": 3904 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019867922613813166, - "loss": 1.8801, - "step": 3905 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986788879117906, - "loss": 1.6279, - "step": 3906 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986785496854495, - "loss": 1.8348, - "step": 3907 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019867821145910845, - "loss": 1.6609, - "step": 3908 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986778732327674, - "loss": 1.6116, - "step": 3909 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986775350064263, - "loss": 1.7537, - "step": 3910 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019867719678008525, - "loss": 1.9332, - "step": 3911 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986768585537442, - "loss": 1.6635, - "step": 3912 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986765203274031, - "loss": 1.724, - "step": 3913 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019867618210106205, - "loss": 1.691, - "step": 3914 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198675843874721, - "loss": 1.7376, - "step": 3915 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986755056483799, - "loss": 1.8797, - "step": 3916 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019867516742203884, - "loss": 1.6997, - "step": 3917 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019867482919569778, - "loss": 1.8722, - "step": 3918 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986744909693567, - "loss": 1.7622, - "step": 3919 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019867415274301564, - "loss": 1.8079, - "step": 3920 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019867381451667458, - "loss": 1.8019, - "step": 3921 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986734762903335, - "loss": 1.7769, - "step": 3922 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019867313806399244, - "loss": 1.8454, - "step": 3923 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019867279983765138, - "loss": 1.7538, - "step": 3924 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986724616113103, - "loss": 1.6819, - "step": 3925 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019867212338496923, - "loss": 1.6768, - "step": 3926 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019867178515862817, - "loss": 1.7511, - "step": 3927 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986714469322871, - "loss": 1.7781, - "step": 3928 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019867110870594603, - "loss": 1.6393, - "step": 3929 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019867077047960497, - "loss": 1.9045, - "step": 3930 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019867043225326388, - "loss": 1.8375, - "step": 3931 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019867009402692282, - "loss": 1.6205, - "step": 3932 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019866975580058177, - "loss": 1.7671, - "step": 3933 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019866941757424068, - "loss": 1.7143, - "step": 3934 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019866907934789962, - "loss": 1.7251, - "step": 3935 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019866874112155856, - "loss": 1.7808, - "step": 3936 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019866840289521748, - "loss": 1.8064, - "step": 3937 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019866806466887642, - "loss": 1.7402, - "step": 3938 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019866772644253536, - "loss": 1.909, - "step": 3939 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019866738821619427, - "loss": 1.7379, - "step": 3940 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986670499898532, - "loss": 1.7618, - "step": 3941 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019866671176351215, - "loss": 1.6766, - "step": 3942 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019866637353717107, - "loss": 1.7876, - "step": 3943 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019866603531083, - "loss": 1.8524, - "step": 3944 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019866569708448895, - "loss": 1.8244, - "step": 3945 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019866535885814786, - "loss": 1.6973, - "step": 3946 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986650206318068, - "loss": 1.7111, - "step": 3947 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019866468240546575, - "loss": 1.6252, - "step": 3948 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019866434417912466, - "loss": 1.7859, - "step": 3949 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986640059527836, - "loss": 1.6787, - "step": 3950 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019866366772644254, - "loss": 1.7611, - "step": 3951 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019866332950010146, - "loss": 1.8034, - "step": 3952 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986629912737604, - "loss": 1.8221, - "step": 3953 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019866265304741934, - "loss": 1.6428, - "step": 3954 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019866231482107825, - "loss": 1.6096, - "step": 3955 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986619765947372, - "loss": 1.9333, - "step": 3956 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019866163836839614, - "loss": 1.849, - "step": 3957 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019866130014205505, - "loss": 1.7137, - "step": 3958 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198660961915714, - "loss": 1.7644, - "step": 3959 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019866062368937296, - "loss": 1.7859, - "step": 3960 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019866028546303187, - "loss": 1.796, - "step": 3961 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019865994723669081, - "loss": 1.642, - "step": 3962 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019865960901034976, - "loss": 1.7201, - "step": 3963 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019865927078400867, - "loss": 1.7575, - "step": 3964 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986589325576676, - "loss": 1.8246, - "step": 3965 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019865859433132655, - "loss": 1.7701, - "step": 3966 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019865825610498547, - "loss": 1.5778, - "step": 3967 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986579178786444, - "loss": 1.6045, - "step": 3968 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019865757965230335, - "loss": 1.7981, - "step": 3969 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019865724142596226, - "loss": 1.7168, - "step": 3970 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986569031996212, - "loss": 1.6739, - "step": 3971 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019865656497328014, - "loss": 1.8261, - "step": 3972 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019865622674693906, - "loss": 1.759, - "step": 3973 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198655888520598, - "loss": 1.7729, - "step": 3974 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019865555029425694, - "loss": 1.9308, - "step": 3975 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019865521206791586, - "loss": 1.7626, - "step": 3976 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986548738415748, - "loss": 1.6903, - "step": 3977 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019865453561523374, - "loss": 1.8083, - "step": 3978 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019865419738889265, - "loss": 1.7504, - "step": 3979 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986538591625516, - "loss": 1.8034, - "step": 3980 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019865352093621053, - "loss": 1.7082, - "step": 3981 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019865318270986945, - "loss": 1.7833, - "step": 3982 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986528444835284, - "loss": 1.7116, - "step": 3983 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019865250625718733, - "loss": 1.7086, - "step": 3984 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019865216803084624, - "loss": 1.7996, - "step": 3985 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019865182980450519, - "loss": 1.736, - "step": 3986 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019865149157816413, - "loss": 1.8268, - "step": 3987 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019865115335182304, - "loss": 1.7868, - "step": 3988 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019865081512548198, - "loss": 1.7204, - "step": 3989 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019865047689914092, - "loss": 1.7616, - "step": 3990 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019865013867279984, - "loss": 1.717, - "step": 3991 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019864980044645878, - "loss": 1.7504, - "step": 3992 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019864946222011772, - "loss": 1.8273, - "step": 3993 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019864912399377663, - "loss": 1.7535, - "step": 3994 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019864878576743557, - "loss": 1.6151, - "step": 3995 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019864844754109452, - "loss": 1.7086, - "step": 3996 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019864810931475343, - "loss": 1.6178, - "step": 3997 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019864777108841237, - "loss": 1.8237, - "step": 3998 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986474328620713, - "loss": 1.7024, - "step": 3999 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019864709463573023, - "loss": 1.8497, - "step": 4000 - }, - { - "epoch": 0.01, - "eval_loss": 1.2634465144230769, - "mrc_em": 10.0, - "mrc_f1": 22.119047619047628, - "nli_acc": 0.32, - "step": 4000, - "sts_f1": 0.32835820895522383, - "ynat_f1": 0.5265671997379315 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019864675640938917, - "loss": 1.6955, - "step": 4001 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986464181830481, - "loss": 1.7234, - "step": 4002 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019864607995670702, - "loss": 1.9025, - "step": 4003 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019864574173036596, - "loss": 1.7468, - "step": 4004 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986454035040249, - "loss": 1.7528, - "step": 4005 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019864506527768382, - "loss": 1.8594, - "step": 4006 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019864472705134276, - "loss": 1.8098, - "step": 4007 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986443888250017, - "loss": 1.864, - "step": 4008 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019864405059866061, - "loss": 1.8558, - "step": 4009 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019864371237231956, - "loss": 1.7493, - "step": 4010 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986433741459785, - "loss": 1.6153, - "step": 4011 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986430359196374, - "loss": 1.9971, - "step": 4012 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019864269769329635, - "loss": 1.8413, - "step": 4013 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986423594669553, - "loss": 1.754, - "step": 4014 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986420212406142, - "loss": 1.7103, - "step": 4015 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019864168301427318, - "loss": 1.5689, - "step": 4016 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019864134478793212, - "loss": 1.6786, - "step": 4017 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019864100656159103, - "loss": 1.7419, - "step": 4018 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019864066833524997, - "loss": 1.7343, - "step": 4019 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986403301089089, - "loss": 1.7056, - "step": 4020 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019863999188256783, - "loss": 1.8544, - "step": 4021 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019863965365622677, - "loss": 1.9103, - "step": 4022 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986393154298857, - "loss": 1.6589, - "step": 4023 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019863897720354462, - "loss": 1.7544, - "step": 4024 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019863863897720356, - "loss": 1.7653, - "step": 4025 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986383007508625, - "loss": 1.7728, - "step": 4026 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019863796252452142, - "loss": 1.8757, - "step": 4027 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019863762429818036, - "loss": 1.9518, - "step": 4028 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986372860718393, - "loss": 1.7172, - "step": 4029 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019863694784549822, - "loss": 1.781, - "step": 4030 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019863660961915716, - "loss": 1.9303, - "step": 4031 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986362713928161, - "loss": 1.674, - "step": 4032 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198635933166475, - "loss": 1.6971, - "step": 4033 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019863559494013395, - "loss": 1.6425, - "step": 4034 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986352567137929, - "loss": 1.6552, - "step": 4035 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986349184874518, - "loss": 1.7338, - "step": 4036 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019863458026111075, - "loss": 1.6774, - "step": 4037 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986342420347697, - "loss": 1.7944, - "step": 4038 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986339038084286, - "loss": 1.586, - "step": 4039 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019863356558208755, - "loss": 1.7384, - "step": 4040 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986332273557465, - "loss": 1.7663, - "step": 4041 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986328891294054, - "loss": 1.7737, - "step": 4042 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019863255090306434, - "loss": 1.9443, - "step": 4043 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019863221267672328, - "loss": 1.787, - "step": 4044 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986318744503822, - "loss": 1.7339, - "step": 4045 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019863153622404114, - "loss": 1.867, - "step": 4046 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019863119799770008, - "loss": 1.8462, - "step": 4047 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198630859771359, - "loss": 1.6239, - "step": 4048 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019863052154501793, - "loss": 1.801, - "step": 4049 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019863018331867688, - "loss": 1.6466, - "step": 4050 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986298450923358, - "loss": 1.8367, - "step": 4051 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019862950686599473, - "loss": 1.7126, - "step": 4052 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019862916863965367, - "loss": 1.8662, - "step": 4053 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019862883041331259, - "loss": 1.6278, - "step": 4054 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019862849218697153, - "loss": 1.8655, - "step": 4055 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019862815396063047, - "loss": 1.7476, - "step": 4056 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019862781573428938, - "loss": 1.779, - "step": 4057 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019862747750794832, - "loss": 1.9363, - "step": 4058 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019862713928160726, - "loss": 1.7342, - "step": 4059 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019862680105526618, - "loss": 1.8723, - "step": 4060 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019862646282892512, - "loss": 1.8298, - "step": 4061 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019862612460258406, - "loss": 1.834, - "step": 4062 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019862578637624298, - "loss": 1.9082, - "step": 4063 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019862544814990192, - "loss": 1.6865, - "step": 4064 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019862510992356086, - "loss": 1.6745, - "step": 4065 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019862477169721977, - "loss": 1.7947, - "step": 4066 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986244334708787, - "loss": 1.8338, - "step": 4067 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019862409524453765, - "loss": 1.7437, - "step": 4068 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019862375701819657, - "loss": 1.7496, - "step": 4069 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986234187918555, - "loss": 1.6819, - "step": 4070 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019862308056551445, - "loss": 1.6841, - "step": 4071 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019862274233917336, - "loss": 1.7953, - "step": 4072 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019862240411283233, - "loss": 1.6676, - "step": 4073 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019862206588649125, - "loss": 1.7739, - "step": 4074 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986217276601502, - "loss": 1.7821, - "step": 4075 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019862138943380913, - "loss": 1.7, - "step": 4076 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019862105120746804, - "loss": 1.8008, - "step": 4077 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019862071298112698, - "loss": 1.805, - "step": 4078 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019862037475478593, - "loss": 1.7789, - "step": 4079 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019862003652844487, - "loss": 1.7043, - "step": 4080 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019861969830210378, - "loss": 1.7986, - "step": 4081 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019861936007576272, - "loss": 1.8724, - "step": 4082 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019861902184942166, - "loss": 1.6496, - "step": 4083 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019861868362308058, - "loss": 1.7667, - "step": 4084 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019861834539673952, - "loss": 1.7851, - "step": 4085 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019861800717039846, - "loss": 1.6151, - "step": 4086 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019861766894405737, - "loss": 1.7639, - "step": 4087 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019861733071771631, - "loss": 1.7795, - "step": 4088 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019861699249137526, - "loss": 1.8584, - "step": 4089 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019861665426503417, - "loss": 1.6975, - "step": 4090 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986163160386931, - "loss": 1.7498, - "step": 4091 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019861597781235205, - "loss": 1.7349, - "step": 4092 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019861563958601097, - "loss": 1.7484, - "step": 4093 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986153013596699, - "loss": 1.7433, - "step": 4094 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019861496313332885, - "loss": 1.6936, - "step": 4095 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019861462490698776, - "loss": 1.7168, - "step": 4096 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986142866806467, - "loss": 1.7668, - "step": 4097 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019861394845430564, - "loss": 1.7351, - "step": 4098 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019861361022796456, - "loss": 1.6573, - "step": 4099 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986132720016235, - "loss": 1.7872, - "step": 4100 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019861293377528244, - "loss": 1.8818, - "step": 4101 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019861259554894135, - "loss": 1.7098, - "step": 4102 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986122573226003, - "loss": 1.6843, - "step": 4103 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019861191909625924, - "loss": 1.7934, - "step": 4104 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019861158086991815, - "loss": 1.6157, - "step": 4105 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986112426435771, - "loss": 1.8378, - "step": 4106 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019861090441723603, - "loss": 1.8027, - "step": 4107 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019861056619089495, - "loss": 1.6905, - "step": 4108 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986102279645539, - "loss": 1.8126, - "step": 4109 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019860988973821283, - "loss": 1.7298, - "step": 4110 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019860955151187174, - "loss": 1.7174, - "step": 4111 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019860921328553068, - "loss": 1.7006, - "step": 4112 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019860887505918963, - "loss": 1.6865, - "step": 4113 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019860853683284854, - "loss": 1.7556, - "step": 4114 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019860819860650748, - "loss": 1.7643, - "step": 4115 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019860786038016642, - "loss": 1.7092, - "step": 4116 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019860752215382534, - "loss": 1.746, - "step": 4117 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019860718392748428, - "loss": 1.7181, - "step": 4118 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019860684570114322, - "loss": 1.7926, - "step": 4119 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019860650747480213, - "loss": 1.7428, - "step": 4120 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019860616924846107, - "loss": 1.7446, - "step": 4121 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019860583102212001, - "loss": 1.8463, - "step": 4122 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019860549279577893, - "loss": 1.7505, - "step": 4123 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019860515456943787, - "loss": 1.8605, - "step": 4124 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986048163430968, - "loss": 1.6778, - "step": 4125 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019860447811675572, - "loss": 1.7448, - "step": 4126 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019860413989041467, - "loss": 1.8614, - "step": 4127 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986038016640736, - "loss": 1.7926, - "step": 4128 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019860346343773255, - "loss": 1.6775, - "step": 4129 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986031252113915, - "loss": 1.7281, - "step": 4130 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986027869850504, - "loss": 1.7426, - "step": 4131 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019860244875870934, - "loss": 1.7149, - "step": 4132 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019860211053236829, - "loss": 1.7413, - "step": 4133 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001986017723060272, - "loss": 1.7549, - "step": 4134 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019860143407968614, - "loss": 1.8463, - "step": 4135 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019860109585334508, - "loss": 1.7869, - "step": 4136 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198600757627004, - "loss": 1.9089, - "step": 4137 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019860041940066294, - "loss": 1.7522, - "step": 4138 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019860008117432188, - "loss": 1.674, - "step": 4139 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985997429479808, - "loss": 1.6088, - "step": 4140 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019859940472163973, - "loss": 1.7255, - "step": 4141 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019859906649529867, - "loss": 1.5932, - "step": 4142 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985987282689576, - "loss": 1.7265, - "step": 4143 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019859839004261653, - "loss": 1.7045, - "step": 4144 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019859805181627547, - "loss": 1.7554, - "step": 4145 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019859771358993439, - "loss": 1.7289, - "step": 4146 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019859737536359333, - "loss": 1.7809, - "step": 4147 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019859703713725227, - "loss": 1.7619, - "step": 4148 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019859669891091118, - "loss": 1.8159, - "step": 4149 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019859636068457012, - "loss": 1.9616, - "step": 4150 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019859602245822906, - "loss": 1.7093, - "step": 4151 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019859568423188798, - "loss": 1.6942, - "step": 4152 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019859534600554692, - "loss": 1.8297, - "step": 4153 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019859500777920586, - "loss": 1.6439, - "step": 4154 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019859466955286477, - "loss": 1.9061, - "step": 4155 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019859433132652372, - "loss": 1.7253, - "step": 4156 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019859399310018266, - "loss": 1.8247, - "step": 4157 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019859365487384157, - "loss": 1.5466, - "step": 4158 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985933166475005, - "loss": 1.8323, - "step": 4159 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019859297842115945, - "loss": 1.7782, - "step": 4160 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019859264019481837, - "loss": 1.7201, - "step": 4161 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985923019684773, - "loss": 1.767, - "step": 4162 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019859196374213625, - "loss": 1.763, - "step": 4163 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985916255157952, - "loss": 1.7567, - "step": 4164 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985912872894541, - "loss": 1.6661, - "step": 4165 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019859094906311305, - "loss": 1.7481, - "step": 4166 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019859061083677199, - "loss": 1.6628, - "step": 4167 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985902726104309, - "loss": 1.8643, - "step": 4168 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019858993438408984, - "loss": 1.6503, - "step": 4169 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019858959615774878, - "loss": 1.7785, - "step": 4170 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985892579314077, - "loss": 1.7211, - "step": 4171 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019858891970506664, - "loss": 1.7494, - "step": 4172 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019858858147872558, - "loss": 1.6608, - "step": 4173 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985882432523845, - "loss": 1.7153, - "step": 4174 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019858790502604343, - "loss": 1.8149, - "step": 4175 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019858756679970238, - "loss": 1.644, - "step": 4176 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985872285733613, - "loss": 1.775, - "step": 4177 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019858689034702023, - "loss": 1.8306, - "step": 4178 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019858655212067917, - "loss": 1.7906, - "step": 4179 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019858621389433809, - "loss": 1.7291, - "step": 4180 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019858587566799703, - "loss": 1.7223, - "step": 4181 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019858553744165597, - "loss": 1.7329, - "step": 4182 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019858519921531488, - "loss": 1.8223, - "step": 4183 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019858486098897382, - "loss": 1.9126, - "step": 4184 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019858452276263276, - "loss": 1.8339, - "step": 4185 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985841845362917, - "loss": 1.5758, - "step": 4186 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019858384630995065, - "loss": 1.7434, - "step": 4187 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019858350808360956, - "loss": 1.7147, - "step": 4188 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985831698572685, - "loss": 1.864, - "step": 4189 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019858283163092744, - "loss": 1.9, - "step": 4190 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019858249340458636, - "loss": 1.7397, - "step": 4191 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985821551782453, - "loss": 1.6742, - "step": 4192 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019858181695190424, - "loss": 1.6705, - "step": 4193 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019858147872556315, - "loss": 1.6324, - "step": 4194 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985811404992221, - "loss": 1.9029, - "step": 4195 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019858080227288104, - "loss": 1.8747, - "step": 4196 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019858046404653995, - "loss": 1.6514, - "step": 4197 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985801258201989, - "loss": 1.8448, - "step": 4198 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019857978759385783, - "loss": 1.7715, - "step": 4199 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019857944936751675, - "loss": 1.6375, - "step": 4200 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985791111411757, - "loss": 1.9789, - "step": 4201 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019857877291483463, - "loss": 1.6952, - "step": 4202 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019857843468849354, - "loss": 1.6823, - "step": 4203 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019857809646215248, - "loss": 1.6453, - "step": 4204 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019857775823581142, - "loss": 1.5396, - "step": 4205 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019857742000947034, - "loss": 1.7958, - "step": 4206 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019857708178312928, - "loss": 1.776, - "step": 4207 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019857674355678822, - "loss": 1.7391, - "step": 4208 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019857640533044713, - "loss": 1.586, - "step": 4209 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019857606710410608, - "loss": 1.8368, - "step": 4210 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019857572887776502, - "loss": 1.7701, - "step": 4211 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019857539065142393, - "loss": 1.7769, - "step": 4212 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019857505242508287, - "loss": 1.8113, - "step": 4213 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985747141987418, - "loss": 1.8101, - "step": 4214 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019857437597240073, - "loss": 1.6899, - "step": 4215 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019857403774605967, - "loss": 1.5717, - "step": 4216 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985736995197186, - "loss": 1.7187, - "step": 4217 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019857336129337752, - "loss": 1.6565, - "step": 4218 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019857302306703646, - "loss": 1.7558, - "step": 4219 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985726848406954, - "loss": 1.7249, - "step": 4220 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019857234661435432, - "loss": 1.7436, - "step": 4221 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019857200838801326, - "loss": 1.9001, - "step": 4222 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985716701616722, - "loss": 1.726, - "step": 4223 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019857133193533112, - "loss": 1.8789, - "step": 4224 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019857099370899006, - "loss": 1.8475, - "step": 4225 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198570655482649, - "loss": 1.8551, - "step": 4226 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985703172563079, - "loss": 1.8885, - "step": 4227 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019856997902996685, - "loss": 1.6342, - "step": 4228 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985696408036258, - "loss": 1.7867, - "step": 4229 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985693025772847, - "loss": 1.722, - "step": 4230 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019856896435094365, - "loss": 1.7434, - "step": 4231 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985686261246026, - "loss": 1.7905, - "step": 4232 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985682878982615, - "loss": 1.734, - "step": 4233 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019856794967192045, - "loss": 1.8439, - "step": 4234 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985676114455794, - "loss": 1.6354, - "step": 4235 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985672732192383, - "loss": 1.869, - "step": 4236 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019856693499289724, - "loss": 1.6787, - "step": 4237 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019856659676655618, - "loss": 1.7242, - "step": 4238 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985662585402151, - "loss": 1.7872, - "step": 4239 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019856592031387404, - "loss": 1.7002, - "step": 4240 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198565582087533, - "loss": 1.636, - "step": 4241 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019856524386119192, - "loss": 1.4596, - "step": 4242 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019856490563485086, - "loss": 1.6977, - "step": 4243 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985645674085098, - "loss": 1.6873, - "step": 4244 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019856422918216872, - "loss": 1.6717, - "step": 4245 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019856389095582766, - "loss": 1.7375, - "step": 4246 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985635527294866, - "loss": 1.678, - "step": 4247 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019856321450314551, - "loss": 1.7079, - "step": 4248 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019856287627680446, - "loss": 1.8005, - "step": 4249 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985625380504634, - "loss": 1.8166, - "step": 4250 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985621998241223, - "loss": 1.6343, - "step": 4251 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019856186159778125, - "loss": 1.7185, - "step": 4252 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985615233714402, - "loss": 1.6503, - "step": 4253 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985611851450991, - "loss": 1.6522, - "step": 4254 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019856084691875805, - "loss": 1.653, - "step": 4255 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198560508692417, - "loss": 1.8363, - "step": 4256 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985601704660759, - "loss": 1.842, - "step": 4257 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019855983223973484, - "loss": 1.9081, - "step": 4258 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019855949401339379, - "loss": 1.7871, - "step": 4259 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985591557870527, - "loss": 1.762, - "step": 4260 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019855881756071164, - "loss": 1.7996, - "step": 4261 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019855847933437058, - "loss": 1.8728, - "step": 4262 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985581411080295, - "loss": 1.8065, - "step": 4263 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019855780288168844, - "loss": 1.8229, - "step": 4264 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019855746465534738, - "loss": 1.8256, - "step": 4265 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985571264290063, - "loss": 1.6877, - "step": 4266 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019855678820266523, - "loss": 1.7095, - "step": 4267 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019855644997632417, - "loss": 1.741, - "step": 4268 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985561117499831, - "loss": 1.6882, - "step": 4269 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019855577352364203, - "loss": 1.7911, - "step": 4270 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019855543529730097, - "loss": 1.6891, - "step": 4271 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019855509707095988, - "loss": 1.8813, - "step": 4272 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019855475884461883, - "loss": 1.7148, - "step": 4273 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019855442061827777, - "loss": 1.5422, - "step": 4274 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019855408239193668, - "loss": 1.6576, - "step": 4275 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019855374416559562, - "loss": 1.7147, - "step": 4276 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019855340593925456, - "loss": 1.6734, - "step": 4277 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019855306771291348, - "loss": 1.8512, - "step": 4278 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019855272948657242, - "loss": 1.7347, - "step": 4279 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019855239126023136, - "loss": 1.6753, - "step": 4280 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019855205303389027, - "loss": 1.8136, - "step": 4281 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019855171480754921, - "loss": 1.7056, - "step": 4282 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019855137658120816, - "loss": 1.6541, - "step": 4283 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019855103835486707, - "loss": 1.6504, - "step": 4284 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198550700128526, - "loss": 1.6811, - "step": 4285 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019855036190218495, - "loss": 1.651, - "step": 4286 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019855002367584387, - "loss": 1.7662, - "step": 4287 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985496854495028, - "loss": 1.7829, - "step": 4288 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019854934722316175, - "loss": 1.7178, - "step": 4289 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019854900899682066, - "loss": 1.7219, - "step": 4290 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985486707704796, - "loss": 1.8021, - "step": 4291 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019854833254413854, - "loss": 1.6485, - "step": 4292 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019854799431779746, - "loss": 1.8355, - "step": 4293 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985476560914564, - "loss": 1.6963, - "step": 4294 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019854731786511534, - "loss": 1.7401, - "step": 4295 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019854697963877426, - "loss": 1.7455, - "step": 4296 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019854664141243322, - "loss": 1.6617, - "step": 4297 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019854630318609216, - "loss": 1.8371, - "step": 4298 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019854596495975108, - "loss": 1.8386, - "step": 4299 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019854562673341002, - "loss": 1.8149, - "step": 4300 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019854528850706896, - "loss": 1.7254, - "step": 4301 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019854495028072787, - "loss": 1.7934, - "step": 4302 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019854461205438682, - "loss": 1.6985, - "step": 4303 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019854427382804576, - "loss": 1.7488, - "step": 4304 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019854393560170467, - "loss": 1.7655, - "step": 4305 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985435973753636, - "loss": 1.647, - "step": 4306 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019854325914902255, - "loss": 1.6397, - "step": 4307 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019854292092268147, - "loss": 1.8598, - "step": 4308 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985425826963404, - "loss": 1.8169, - "step": 4309 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019854224446999935, - "loss": 1.7684, - "step": 4310 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019854190624365826, - "loss": 1.8171, - "step": 4311 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985415680173172, - "loss": 1.88, - "step": 4312 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019854122979097615, - "loss": 1.7818, - "step": 4313 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019854089156463506, - "loss": 1.6542, - "step": 4314 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198540553338294, - "loss": 1.8376, - "step": 4315 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019854021511195294, - "loss": 1.7849, - "step": 4316 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019853987688561186, - "loss": 1.6631, - "step": 4317 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985395386592708, - "loss": 1.5311, - "step": 4318 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019853920043292974, - "loss": 1.7979, - "step": 4319 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019853886220658865, - "loss": 1.6902, - "step": 4320 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985385239802476, - "loss": 1.8593, - "step": 4321 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019853818575390653, - "loss": 1.7809, - "step": 4322 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019853784752756545, - "loss": 1.802, - "step": 4323 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985375093012244, - "loss": 1.8674, - "step": 4324 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019853717107488333, - "loss": 1.7242, - "step": 4325 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019853683284854225, - "loss": 1.6576, - "step": 4326 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019853649462220119, - "loss": 1.728, - "step": 4327 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019853615639586013, - "loss": 1.8496, - "step": 4328 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019853581816951904, - "loss": 1.7497, - "step": 4329 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019853547994317798, - "loss": 1.7113, - "step": 4330 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019853514171683692, - "loss": 1.8731, - "step": 4331 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019853480349049584, - "loss": 1.7103, - "step": 4332 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019853446526415478, - "loss": 1.6554, - "step": 4333 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019853412703781372, - "loss": 1.9447, - "step": 4334 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019853378881147263, - "loss": 1.8282, - "step": 4335 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019853345058513158, - "loss": 1.8429, - "step": 4336 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019853311235879052, - "loss": 1.7569, - "step": 4337 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019853277413244943, - "loss": 1.8186, - "step": 4338 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019853243590610837, - "loss": 1.6623, - "step": 4339 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985320976797673, - "loss": 1.6529, - "step": 4340 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019853175945342623, - "loss": 1.6806, - "step": 4341 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019853142122708517, - "loss": 1.8431, - "step": 4342 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985310830007441, - "loss": 1.8056, - "step": 4343 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019853074477440302, - "loss": 1.7067, - "step": 4344 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019853040654806196, - "loss": 1.8882, - "step": 4345 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985300683217209, - "loss": 1.7491, - "step": 4346 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019852973009537982, - "loss": 1.6568, - "step": 4347 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019852939186903876, - "loss": 1.8484, - "step": 4348 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985290536426977, - "loss": 1.7909, - "step": 4349 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019852871541635662, - "loss": 1.5744, - "step": 4350 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019852837719001556, - "loss": 1.7044, - "step": 4351 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985280389636745, - "loss": 1.775, - "step": 4352 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019852770073733344, - "loss": 1.9052, - "step": 4353 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019852736251099238, - "loss": 1.6423, - "step": 4354 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019852702428465132, - "loss": 1.8207, - "step": 4355 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019852668605831024, - "loss": 1.7213, - "step": 4356 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019852634783196918, - "loss": 1.8674, - "step": 4357 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019852600960562812, - "loss": 1.864, - "step": 4358 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019852567137928703, - "loss": 1.6613, - "step": 4359 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019852533315294597, - "loss": 1.7653, - "step": 4360 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019852499492660491, - "loss": 1.7736, - "step": 4361 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019852465670026383, - "loss": 1.7959, - "step": 4362 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019852431847392277, - "loss": 1.6996, - "step": 4363 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985239802475817, - "loss": 1.8013, - "step": 4364 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019852364202124062, - "loss": 1.6606, - "step": 4365 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019852330379489957, - "loss": 1.5632, - "step": 4366 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985229655685585, - "loss": 1.8032, - "step": 4367 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019852262734221742, - "loss": 1.6608, - "step": 4368 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019852228911587636, - "loss": 1.6594, - "step": 4369 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985219508895353, - "loss": 1.8285, - "step": 4370 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019852161266319422, - "loss": 1.6935, - "step": 4371 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019852127443685316, - "loss": 1.6873, - "step": 4372 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985209362105121, - "loss": 1.6427, - "step": 4373 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198520597984171, - "loss": 1.7123, - "step": 4374 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019852025975782995, - "loss": 1.68, - "step": 4375 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985199215314889, - "loss": 1.6118, - "step": 4376 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985195833051478, - "loss": 1.765, - "step": 4377 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019851924507880675, - "loss": 1.7262, - "step": 4378 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985189068524657, - "loss": 1.7863, - "step": 4379 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985185686261246, - "loss": 1.9364, - "step": 4380 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019851823039978355, - "loss": 1.8681, - "step": 4381 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985178921734425, - "loss": 1.6546, - "step": 4382 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985175539471014, - "loss": 1.6666, - "step": 4383 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019851721572076034, - "loss": 1.7423, - "step": 4384 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019851687749441928, - "loss": 1.7787, - "step": 4385 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985165392680782, - "loss": 1.7741, - "step": 4386 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019851620104173714, - "loss": 1.7168, - "step": 4387 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019851586281539608, - "loss": 1.7119, - "step": 4388 - }, - { - "epoch": 0.01, - "learning_rate": 0.000198515524589055, - "loss": 1.686, - "step": 4389 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019851518636271394, - "loss": 1.7454, - "step": 4390 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019851484813637288, - "loss": 1.8238, - "step": 4391 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985145099100318, - "loss": 1.7751, - "step": 4392 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019851417168369073, - "loss": 1.7026, - "step": 4393 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019851383345734967, - "loss": 1.7675, - "step": 4394 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985134952310086, - "loss": 1.7154, - "step": 4395 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019851315700466753, - "loss": 1.6557, - "step": 4396 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019851281877832647, - "loss": 1.6346, - "step": 4397 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019851248055198538, - "loss": 1.6017, - "step": 4398 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019851214232564433, - "loss": 1.7259, - "step": 4399 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019851180409930327, - "loss": 1.772, - "step": 4400 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019851146587296218, - "loss": 1.7104, - "step": 4401 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019851112764662112, - "loss": 1.5607, - "step": 4402 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019851078942028006, - "loss": 1.8149, - "step": 4403 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019851045119393898, - "loss": 1.5891, - "step": 4404 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019851011296759792, - "loss": 1.6782, - "step": 4405 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019850977474125686, - "loss": 1.7792, - "step": 4406 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019850943651491577, - "loss": 1.7335, - "step": 4407 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019850909828857471, - "loss": 1.7528, - "step": 4408 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019850876006223366, - "loss": 1.7302, - "step": 4409 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985084218358926, - "loss": 1.8708, - "step": 4410 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019850808360955154, - "loss": 1.629, - "step": 4411 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019850774538321045, - "loss": 1.8832, - "step": 4412 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985074071568694, - "loss": 1.903, - "step": 4413 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019850706893052833, - "loss": 1.6855, - "step": 4414 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019850673070418725, - "loss": 1.6241, - "step": 4415 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985063924778462, - "loss": 1.6049, - "step": 4416 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019850605425150513, - "loss": 1.7774, - "step": 4417 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019850571602516404, - "loss": 1.7137, - "step": 4418 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019850537779882299, - "loss": 1.6913, - "step": 4419 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019850503957248193, - "loss": 1.649, - "step": 4420 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019850470134614084, - "loss": 1.7799, - "step": 4421 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019850436311979978, - "loss": 1.7462, - "step": 4422 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019850402489345872, - "loss": 1.8334, - "step": 4423 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019850368666711764, - "loss": 1.7098, - "step": 4424 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019850334844077658, - "loss": 1.7328, - "step": 4425 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019850301021443552, - "loss": 1.75, - "step": 4426 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019850267198809443, - "loss": 1.8987, - "step": 4427 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019850233376175337, - "loss": 1.6423, - "step": 4428 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019850199553541232, - "loss": 1.7674, - "step": 4429 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019850165730907123, - "loss": 1.7289, - "step": 4430 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019850131908273017, - "loss": 1.87, - "step": 4431 - }, - { - "epoch": 0.01, - "learning_rate": 0.0001985009808563891, - "loss": 1.8138, - "step": 4432 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019850064263004805, - "loss": 1.6907, - "step": 4433 - }, - { - "epoch": 0.01, - "learning_rate": 0.00019850030440370697, - "loss": 1.7541, - "step": 4434 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984999661773659, - "loss": 1.6355, - "step": 4435 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019849962795102485, - "loss": 1.6845, - "step": 4436 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019849928972468376, - "loss": 1.8301, - "step": 4437 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984989514983427, - "loss": 1.6829, - "step": 4438 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019849861327200165, - "loss": 1.6991, - "step": 4439 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019849827504566056, - "loss": 1.7575, - "step": 4440 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984979368193195, - "loss": 1.8667, - "step": 4441 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019849759859297844, - "loss": 1.7894, - "step": 4442 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019849726036663736, - "loss": 1.7632, - "step": 4443 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984969221402963, - "loss": 1.7291, - "step": 4444 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019849658391395524, - "loss": 1.6611, - "step": 4445 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019849624568761415, - "loss": 1.7102, - "step": 4446 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984959074612731, - "loss": 1.6604, - "step": 4447 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019849556923493203, - "loss": 1.7824, - "step": 4448 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019849523100859095, - "loss": 1.7623, - "step": 4449 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984948927822499, - "loss": 1.9005, - "step": 4450 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019849455455590883, - "loss": 1.7827, - "step": 4451 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019849421632956774, - "loss": 1.7286, - "step": 4452 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019849387810322669, - "loss": 1.7743, - "step": 4453 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019849353987688563, - "loss": 1.7731, - "step": 4454 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019849320165054454, - "loss": 1.7659, - "step": 4455 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019849286342420348, - "loss": 1.7883, - "step": 4456 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019849252519786242, - "loss": 1.7455, - "step": 4457 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019849218697152134, - "loss": 1.8616, - "step": 4458 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019849184874518028, - "loss": 1.7369, - "step": 4459 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019849151051883922, - "loss": 1.704, - "step": 4460 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019849117229249813, - "loss": 1.5336, - "step": 4461 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019849083406615707, - "loss": 1.7774, - "step": 4462 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019849049583981602, - "loss": 1.6766, - "step": 4463 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019849015761347493, - "loss": 1.7279, - "step": 4464 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019848981938713387, - "loss": 1.6038, - "step": 4465 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984894811607928, - "loss": 1.8113, - "step": 4466 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019848914293445175, - "loss": 1.7346, - "step": 4467 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984888047081107, - "loss": 1.6407, - "step": 4468 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984884664817696, - "loss": 1.7392, - "step": 4469 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019848812825542855, - "loss": 1.7317, - "step": 4470 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984877900290875, - "loss": 1.6959, - "step": 4471 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984874518027464, - "loss": 1.763, - "step": 4472 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019848711357640535, - "loss": 1.7021, - "step": 4473 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984867753500643, - "loss": 1.8894, - "step": 4474 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984864371237232, - "loss": 1.8627, - "step": 4475 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019848609889738214, - "loss": 1.7722, - "step": 4476 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019848576067104108, - "loss": 1.8593, - "step": 4477 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984854224447, - "loss": 1.6769, - "step": 4478 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019848508421835894, - "loss": 1.7913, - "step": 4479 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019848474599201788, - "loss": 1.7166, - "step": 4480 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984844077656768, - "loss": 1.8133, - "step": 4481 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019848406953933573, - "loss": 1.8111, - "step": 4482 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019848373131299468, - "loss": 1.5849, - "step": 4483 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984833930866536, - "loss": 1.7622, - "step": 4484 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019848305486031253, - "loss": 1.7406, - "step": 4485 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019848271663397147, - "loss": 1.7838, - "step": 4486 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019848237840763039, - "loss": 1.738, - "step": 4487 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019848204018128933, - "loss": 1.7923, - "step": 4488 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019848170195494827, - "loss": 1.7404, - "step": 4489 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019848136372860718, - "loss": 1.6904, - "step": 4490 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019848102550226612, - "loss": 1.6116, - "step": 4491 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019848068727592506, - "loss": 1.7949, - "step": 4492 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019848034904958398, - "loss": 1.8061, - "step": 4493 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019848001082324292, - "loss": 1.6399, - "step": 4494 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019847967259690186, - "loss": 1.8428, - "step": 4495 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019847933437056078, - "loss": 1.7945, - "step": 4496 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019847899614421972, - "loss": 1.9008, - "step": 4497 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019847865791787866, - "loss": 1.7691, - "step": 4498 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019847831969153757, - "loss": 1.6438, - "step": 4499 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984779814651965, - "loss": 1.7311, - "step": 4500 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019847764323885545, - "loss": 1.7856, - "step": 4501 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019847730501251437, - "loss": 1.8053, - "step": 4502 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984769667861733, - "loss": 1.7266, - "step": 4503 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019847662855983225, - "loss": 1.9686, - "step": 4504 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019847629033349116, - "loss": 1.7515, - "step": 4505 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984759521071501, - "loss": 1.7355, - "step": 4506 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019847561388080905, - "loss": 1.628, - "step": 4507 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019847527565446796, - "loss": 1.6655, - "step": 4508 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984749374281269, - "loss": 1.7882, - "step": 4509 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019847459920178584, - "loss": 1.8373, - "step": 4510 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019847426097544476, - "loss": 1.6164, - "step": 4511 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984739227491037, - "loss": 1.8306, - "step": 4512 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019847358452276264, - "loss": 1.7057, - "step": 4513 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019847324629642155, - "loss": 1.7647, - "step": 4514 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984729080700805, - "loss": 1.6085, - "step": 4515 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019847256984373944, - "loss": 1.7926, - "step": 4516 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019847223161739838, - "loss": 1.7966, - "step": 4517 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984718933910573, - "loss": 1.5969, - "step": 4518 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019847155516471623, - "loss": 1.6956, - "step": 4519 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019847121693837517, - "loss": 1.7527, - "step": 4520 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984708787120341, - "loss": 1.8073, - "step": 4521 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019847054048569306, - "loss": 1.7875, - "step": 4522 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019847020225935197, - "loss": 1.7985, - "step": 4523 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984698640330109, - "loss": 1.816, - "step": 4524 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019846952580666985, - "loss": 1.7674, - "step": 4525 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019846918758032877, - "loss": 1.8426, - "step": 4526 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984688493539877, - "loss": 1.6984, - "step": 4527 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019846851112764665, - "loss": 1.831, - "step": 4528 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019846817290130556, - "loss": 1.6454, - "step": 4529 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984678346749645, - "loss": 2.0176, - "step": 4530 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019846749644862344, - "loss": 1.6576, - "step": 4531 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019846715822228236, - "loss": 1.664, - "step": 4532 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984668199959413, - "loss": 1.8811, - "step": 4533 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019846648176960024, - "loss": 1.6908, - "step": 4534 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019846614354325915, - "loss": 1.76, - "step": 4535 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984658053169181, - "loss": 1.8447, - "step": 4536 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019846546709057704, - "loss": 1.7598, - "step": 4537 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019846512886423595, - "loss": 1.7997, - "step": 4538 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984647906378949, - "loss": 1.7913, - "step": 4539 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019846445241155383, - "loss": 1.7874, - "step": 4540 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019846411418521275, - "loss": 1.8126, - "step": 4541 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984637759588717, - "loss": 1.6806, - "step": 4542 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019846343773253063, - "loss": 1.823, - "step": 4543 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019846309950618954, - "loss": 1.793, - "step": 4544 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019846276127984848, - "loss": 1.7079, - "step": 4545 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019846242305350743, - "loss": 1.7458, - "step": 4546 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019846208482716634, - "loss": 1.7208, - "step": 4547 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019846174660082528, - "loss": 1.8014, - "step": 4548 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019846140837448422, - "loss": 1.8468, - "step": 4549 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019846107014814314, - "loss": 1.7496, - "step": 4550 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019846073192180208, - "loss": 1.759, - "step": 4551 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019846039369546102, - "loss": 1.5129, - "step": 4552 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019846005546911993, - "loss": 1.7558, - "step": 4553 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019845971724277887, - "loss": 1.6399, - "step": 4554 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019845937901643781, - "loss": 1.7661, - "step": 4555 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019845904079009673, - "loss": 1.7059, - "step": 4556 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019845870256375567, - "loss": 1.8432, - "step": 4557 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984583643374146, - "loss": 1.7115, - "step": 4558 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019845802611107352, - "loss": 1.6659, - "step": 4559 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019845768788473247, - "loss": 1.7149, - "step": 4560 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984573496583914, - "loss": 1.7261, - "step": 4561 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019845701143205032, - "loss": 1.8449, - "step": 4562 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019845667320570926, - "loss": 1.6167, - "step": 4563 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984563349793682, - "loss": 1.6548, - "step": 4564 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019845599675302712, - "loss": 1.627, - "step": 4565 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019845565852668606, - "loss": 1.7256, - "step": 4566 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198455320300345, - "loss": 1.6881, - "step": 4567 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019845498207400391, - "loss": 1.7151, - "step": 4568 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019845464384766286, - "loss": 1.6469, - "step": 4569 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984543056213218, - "loss": 1.8468, - "step": 4570 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984539673949807, - "loss": 1.782, - "step": 4571 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019845362916863965, - "loss": 1.7529, - "step": 4572 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984532909422986, - "loss": 1.6914, - "step": 4573 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984529527159575, - "loss": 1.6741, - "step": 4574 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019845261448961645, - "loss": 1.6523, - "step": 4575 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984522762632754, - "loss": 1.6314, - "step": 4576 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984519380369343, - "loss": 1.9214, - "step": 4577 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019845159981059327, - "loss": 1.7025, - "step": 4578 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984512615842522, - "loss": 1.5777, - "step": 4579 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019845092335791113, - "loss": 1.7113, - "step": 4580 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019845058513157007, - "loss": 1.7612, - "step": 4581 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198450246905229, - "loss": 1.8986, - "step": 4582 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019844990867888792, - "loss": 1.6503, - "step": 4583 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019844957045254686, - "loss": 1.7407, - "step": 4584 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984492322262058, - "loss": 1.7033, - "step": 4585 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019844889399986472, - "loss": 1.6462, - "step": 4586 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019844855577352366, - "loss": 1.6526, - "step": 4587 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984482175471826, - "loss": 1.749, - "step": 4588 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019844787932084152, - "loss": 1.7708, - "step": 4589 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019844754109450046, - "loss": 1.7266, - "step": 4590 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984472028681594, - "loss": 1.7469, - "step": 4591 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984468646418183, - "loss": 1.7482, - "step": 4592 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019844652641547725, - "loss": 1.7087, - "step": 4593 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984461881891362, - "loss": 1.688, - "step": 4594 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984458499627951, - "loss": 1.7876, - "step": 4595 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019844551173645405, - "loss": 1.6917, - "step": 4596 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198445173510113, - "loss": 1.7199, - "step": 4597 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984448352837719, - "loss": 1.735, - "step": 4598 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019844449705743085, - "loss": 1.7271, - "step": 4599 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984441588310898, - "loss": 1.7713, - "step": 4600 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984438206047487, - "loss": 1.5984, - "step": 4601 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019844348237840764, - "loss": 1.7578, - "step": 4602 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019844314415206658, - "loss": 1.7193, - "step": 4603 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984428059257255, - "loss": 1.8282, - "step": 4604 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019844246769938444, - "loss": 1.6529, - "step": 4605 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019844212947304338, - "loss": 1.7058, - "step": 4606 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984417912467023, - "loss": 1.6895, - "step": 4607 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019844145302036123, - "loss": 1.8204, - "step": 4608 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019844111479402018, - "loss": 1.6559, - "step": 4609 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984407765676791, - "loss": 1.6728, - "step": 4610 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019844043834133803, - "loss": 1.8134, - "step": 4611 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019844010011499697, - "loss": 1.6574, - "step": 4612 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019843976188865589, - "loss": 1.7771, - "step": 4613 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019843942366231483, - "loss": 1.8089, - "step": 4614 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019843908543597377, - "loss": 1.7044, - "step": 4615 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019843874720963268, - "loss": 1.878, - "step": 4616 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019843840898329162, - "loss": 1.6627, - "step": 4617 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019843807075695056, - "loss": 1.6782, - "step": 4618 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019843773253060948, - "loss": 1.8195, - "step": 4619 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019843739430426842, - "loss": 1.6676, - "step": 4620 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019843705607792736, - "loss": 1.75, - "step": 4621 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019843671785158627, - "loss": 1.7106, - "step": 4622 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019843637962524522, - "loss": 1.745, - "step": 4623 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019843604139890416, - "loss": 1.7583, - "step": 4624 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019843570317256307, - "loss": 1.6088, - "step": 4625 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198435364946222, - "loss": 1.7402, - "step": 4626 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019843502671988095, - "loss": 1.6869, - "step": 4627 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019843468849353987, - "loss": 1.7434, - "step": 4628 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984343502671988, - "loss": 1.6537, - "step": 4629 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019843401204085775, - "loss": 1.6673, - "step": 4630 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019843367381451666, - "loss": 1.6847, - "step": 4631 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984333355881756, - "loss": 1.5993, - "step": 4632 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019843299736183455, - "loss": 1.8007, - "step": 4633 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984326591354935, - "loss": 1.6769, - "step": 4634 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019843232090915243, - "loss": 1.788, - "step": 4635 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019843198268281137, - "loss": 1.8134, - "step": 4636 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019843164445647028, - "loss": 1.6722, - "step": 4637 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019843130623012922, - "loss": 1.6373, - "step": 4638 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019843096800378817, - "loss": 1.7227, - "step": 4639 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019843062977744708, - "loss": 1.7464, - "step": 4640 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019843029155110602, - "loss": 1.6242, - "step": 4641 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019842995332476496, - "loss": 1.7416, - "step": 4642 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019842961509842388, - "loss": 1.6412, - "step": 4643 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019842927687208282, - "loss": 1.8124, - "step": 4644 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019842893864574176, - "loss": 1.6767, - "step": 4645 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019842860041940067, - "loss": 1.7015, - "step": 4646 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984282621930596, - "loss": 1.805, - "step": 4647 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019842792396671855, - "loss": 1.7897, - "step": 4648 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019842758574037747, - "loss": 1.7197, - "step": 4649 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984272475140364, - "loss": 1.6726, - "step": 4650 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019842690928769535, - "loss": 1.647, - "step": 4651 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019842657106135426, - "loss": 1.6916, - "step": 4652 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984262328350132, - "loss": 1.7375, - "step": 4653 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019842589460867215, - "loss": 1.8165, - "step": 4654 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019842555638233106, - "loss": 1.7348, - "step": 4655 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019842521815599, - "loss": 1.557, - "step": 4656 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019842487992964894, - "loss": 1.7525, - "step": 4657 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019842454170330786, - "loss": 1.7405, - "step": 4658 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984242034769668, - "loss": 1.7314, - "step": 4659 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019842386525062574, - "loss": 1.8641, - "step": 4660 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019842352702428465, - "loss": 1.5921, - "step": 4661 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984231887979436, - "loss": 1.6404, - "step": 4662 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019842285057160254, - "loss": 1.877, - "step": 4663 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019842251234526145, - "loss": 1.7964, - "step": 4664 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984221741189204, - "loss": 1.6836, - "step": 4665 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019842183589257933, - "loss": 1.7969, - "step": 4666 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019842149766623825, - "loss": 1.7068, - "step": 4667 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984211594398972, - "loss": 1.6255, - "step": 4668 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019842082121355613, - "loss": 1.8228, - "step": 4669 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019842048298721504, - "loss": 1.8047, - "step": 4670 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019842014476087398, - "loss": 1.7493, - "step": 4671 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019841980653453293, - "loss": 1.6169, - "step": 4672 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019841946830819184, - "loss": 1.778, - "step": 4673 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019841913008185078, - "loss": 1.7404, - "step": 4674 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019841879185550972, - "loss": 1.7754, - "step": 4675 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019841845362916864, - "loss": 1.7968, - "step": 4676 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019841811540282758, - "loss": 1.7774, - "step": 4677 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019841777717648652, - "loss": 1.6872, - "step": 4678 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019841743895014543, - "loss": 1.657, - "step": 4679 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019841710072380437, - "loss": 1.6972, - "step": 4680 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019841676249746331, - "loss": 1.7143, - "step": 4681 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019841642427112223, - "loss": 1.7044, - "step": 4682 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019841608604478117, - "loss": 1.6577, - "step": 4683 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984157478184401, - "loss": 1.7889, - "step": 4684 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019841540959209902, - "loss": 1.7478, - "step": 4685 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019841507136575797, - "loss": 1.6381, - "step": 4686 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984147331394169, - "loss": 1.6509, - "step": 4687 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019841439491307582, - "loss": 1.7536, - "step": 4688 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019841405668673476, - "loss": 1.7494, - "step": 4689 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984137184603937, - "loss": 1.7869, - "step": 4690 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019841338023405264, - "loss": 1.7536, - "step": 4691 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019841304200771159, - "loss": 1.7222, - "step": 4692 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984127037813705, - "loss": 1.7129, - "step": 4693 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019841236555502944, - "loss": 1.6596, - "step": 4694 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019841202732868838, - "loss": 1.9272, - "step": 4695 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984116891023473, - "loss": 1.823, - "step": 4696 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019841135087600624, - "loss": 1.8413, - "step": 4697 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019841101264966518, - "loss": 1.7131, - "step": 4698 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019841067442332412, - "loss": 1.6168, - "step": 4699 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019841033619698303, - "loss": 1.7331, - "step": 4700 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019840999797064197, - "loss": 1.7719, - "step": 4701 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019840965974430092, - "loss": 1.7578, - "step": 4702 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019840932151795983, - "loss": 1.6592, - "step": 4703 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019840898329161877, - "loss": 1.7118, - "step": 4704 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984086450652777, - "loss": 1.6876, - "step": 4705 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019840830683893663, - "loss": 1.6224, - "step": 4706 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019840796861259557, - "loss": 1.7631, - "step": 4707 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984076303862545, - "loss": 1.6229, - "step": 4708 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019840729215991342, - "loss": 1.879, - "step": 4709 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019840695393357236, - "loss": 1.7866, - "step": 4710 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984066157072313, - "loss": 1.7341, - "step": 4711 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019840627748089022, - "loss": 1.7069, - "step": 4712 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019840593925454916, - "loss": 1.6056, - "step": 4713 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984056010282081, - "loss": 1.6479, - "step": 4714 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019840526280186701, - "loss": 1.8532, - "step": 4715 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019840492457552596, - "loss": 1.6945, - "step": 4716 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984045863491849, - "loss": 1.737, - "step": 4717 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984042481228438, - "loss": 1.6377, - "step": 4718 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019840390989650275, - "loss": 1.4548, - "step": 4719 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984035716701617, - "loss": 1.8411, - "step": 4720 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984032334438206, - "loss": 1.6715, - "step": 4721 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019840289521747955, - "loss": 1.8529, - "step": 4722 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984025569911385, - "loss": 1.7712, - "step": 4723 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984022187647974, - "loss": 1.6719, - "step": 4724 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019840188053845634, - "loss": 1.6459, - "step": 4725 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019840154231211529, - "loss": 1.6754, - "step": 4726 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001984012040857742, - "loss": 1.7468, - "step": 4727 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019840086585943314, - "loss": 1.7424, - "step": 4728 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019840052763309208, - "loss": 1.7758, - "step": 4729 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198400189406751, - "loss": 1.7548, - "step": 4730 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019839985118040994, - "loss": 1.8104, - "step": 4731 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019839951295406888, - "loss": 1.5384, - "step": 4732 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983991747277278, - "loss": 1.5794, - "step": 4733 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019839883650138673, - "loss": 1.7847, - "step": 4734 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019839849827504567, - "loss": 1.6044, - "step": 4735 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983981600487046, - "loss": 1.7163, - "step": 4736 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019839782182236353, - "loss": 1.7985, - "step": 4737 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019839748359602247, - "loss": 1.534, - "step": 4738 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019839714536968139, - "loss": 1.6383, - "step": 4739 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019839680714334033, - "loss": 1.8192, - "step": 4740 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019839646891699927, - "loss": 1.6779, - "step": 4741 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019839613069065818, - "loss": 1.8033, - "step": 4742 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019839579246431712, - "loss": 1.6922, - "step": 4743 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019839545423797606, - "loss": 1.7468, - "step": 4744 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019839511601163498, - "loss": 1.7366, - "step": 4745 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019839477778529392, - "loss": 1.7538, - "step": 4746 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019839443955895286, - "loss": 1.8108, - "step": 4747 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983941013326118, - "loss": 1.7801, - "step": 4748 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019839376310627074, - "loss": 1.8159, - "step": 4749 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019839342487992966, - "loss": 1.8243, - "step": 4750 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983930866535886, - "loss": 1.5528, - "step": 4751 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019839274842724754, - "loss": 1.6859, - "step": 4752 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019839241020090645, - "loss": 1.6104, - "step": 4753 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983920719745654, - "loss": 1.7141, - "step": 4754 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019839173374822433, - "loss": 1.7283, - "step": 4755 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019839139552188325, - "loss": 1.7223, - "step": 4756 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983910572955422, - "loss": 1.821, - "step": 4757 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019839071906920113, - "loss": 1.641, - "step": 4758 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019839038084286005, - "loss": 1.7525, - "step": 4759 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019839004261651899, - "loss": 1.6828, - "step": 4760 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019838970439017793, - "loss": 1.6625, - "step": 4761 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019838936616383684, - "loss": 1.8171, - "step": 4762 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019838902793749578, - "loss": 1.6812, - "step": 4763 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019838868971115472, - "loss": 1.8106, - "step": 4764 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019838835148481364, - "loss": 1.7558, - "step": 4765 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019838801325847258, - "loss": 1.8146, - "step": 4766 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019838767503213152, - "loss": 1.6841, - "step": 4767 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019838733680579043, - "loss": 1.7618, - "step": 4768 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019838699857944938, - "loss": 1.7503, - "step": 4769 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019838666035310832, - "loss": 1.6607, - "step": 4770 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019838632212676723, - "loss": 1.5501, - "step": 4771 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019838598390042617, - "loss": 1.7221, - "step": 4772 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983856456740851, - "loss": 1.6938, - "step": 4773 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019838530744774403, - "loss": 1.9985, - "step": 4774 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019838496922140297, - "loss": 1.5972, - "step": 4775 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983846309950619, - "loss": 1.7889, - "step": 4776 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019838429276872082, - "loss": 1.8484, - "step": 4777 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019838395454237976, - "loss": 1.7608, - "step": 4778 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983836163160387, - "loss": 1.7706, - "step": 4779 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019838327808969762, - "loss": 1.8901, - "step": 4780 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019838293986335656, - "loss": 1.6275, - "step": 4781 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983826016370155, - "loss": 1.8525, - "step": 4782 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019838226341067442, - "loss": 1.7281, - "step": 4783 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019838192518433336, - "loss": 1.5787, - "step": 4784 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983815869579923, - "loss": 1.7392, - "step": 4785 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019838124873165124, - "loss": 1.7245, - "step": 4786 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019838091050531015, - "loss": 1.7274, - "step": 4787 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983805722789691, - "loss": 1.8409, - "step": 4788 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019838023405262804, - "loss": 1.6701, - "step": 4789 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019837989582628695, - "loss": 1.762, - "step": 4790 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983795575999459, - "loss": 1.6877, - "step": 4791 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019837921937360483, - "loss": 1.8068, - "step": 4792 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019837888114726375, - "loss": 1.7537, - "step": 4793 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983785429209227, - "loss": 1.8017, - "step": 4794 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019837820469458163, - "loss": 1.8571, - "step": 4795 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019837786646824054, - "loss": 1.775, - "step": 4796 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019837752824189948, - "loss": 1.8879, - "step": 4797 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019837719001555842, - "loss": 1.8433, - "step": 4798 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019837685178921734, - "loss": 1.6752, - "step": 4799 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019837651356287628, - "loss": 1.9133, - "step": 4800 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019837617533653522, - "loss": 1.5767, - "step": 4801 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019837583711019413, - "loss": 1.6821, - "step": 4802 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983754988838531, - "loss": 1.7089, - "step": 4803 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019837516065751202, - "loss": 1.7067, - "step": 4804 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019837482243117096, - "loss": 1.7559, - "step": 4805 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983744842048299, - "loss": 1.7542, - "step": 4806 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983741459784888, - "loss": 1.5942, - "step": 4807 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019837380775214775, - "loss": 1.8132, - "step": 4808 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983734695258067, - "loss": 1.8237, - "step": 4809 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983731312994656, - "loss": 1.8475, - "step": 4810 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019837279307312455, - "loss": 1.7868, - "step": 4811 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983724548467835, - "loss": 1.6237, - "step": 4812 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983721166204424, - "loss": 1.7301, - "step": 4813 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019837177839410135, - "loss": 1.6293, - "step": 4814 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983714401677603, - "loss": 1.57, - "step": 4815 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983711019414192, - "loss": 1.5704, - "step": 4816 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019837076371507814, - "loss": 1.8452, - "step": 4817 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019837042548873708, - "loss": 1.7893, - "step": 4818 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198370087262396, - "loss": 1.674, - "step": 4819 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019836974903605494, - "loss": 1.7634, - "step": 4820 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019836941080971388, - "loss": 1.623, - "step": 4821 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983690725833728, - "loss": 1.6792, - "step": 4822 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019836873435703174, - "loss": 1.7018, - "step": 4823 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019836839613069068, - "loss": 1.7289, - "step": 4824 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983680579043496, - "loss": 1.7963, - "step": 4825 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019836771967800853, - "loss": 1.7122, - "step": 4826 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019836738145166747, - "loss": 1.8353, - "step": 4827 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983670432253264, - "loss": 1.577, - "step": 4828 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019836670499898533, - "loss": 1.6483, - "step": 4829 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019836636677264427, - "loss": 1.7312, - "step": 4830 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019836602854630318, - "loss": 1.6633, - "step": 4831 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019836569031996213, - "loss": 1.9226, - "step": 4832 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019836535209362107, - "loss": 1.7398, - "step": 4833 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019836501386727998, - "loss": 1.7273, - "step": 4834 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019836467564093892, - "loss": 1.725, - "step": 4835 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019836433741459786, - "loss": 1.7062, - "step": 4836 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019836399918825678, - "loss": 1.7494, - "step": 4837 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019836366096191572, - "loss": 1.7428, - "step": 4838 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019836332273557466, - "loss": 1.6707, - "step": 4839 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019836298450923357, - "loss": 1.7401, - "step": 4840 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019836264628289251, - "loss": 1.7712, - "step": 4841 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019836230805655146, - "loss": 1.7056, - "step": 4842 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019836196983021037, - "loss": 1.7567, - "step": 4843 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983616316038693, - "loss": 1.5898, - "step": 4844 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019836129337752825, - "loss": 1.6402, - "step": 4845 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019836095515118717, - "loss": 1.6395, - "step": 4846 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983606169248461, - "loss": 1.8434, - "step": 4847 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019836027869850505, - "loss": 1.8522, - "step": 4848 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019835994047216396, - "loss": 1.8527, - "step": 4849 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983596022458229, - "loss": 1.6567, - "step": 4850 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019835926401948184, - "loss": 1.6503, - "step": 4851 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019835892579314076, - "loss": 1.642, - "step": 4852 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983585875667997, - "loss": 1.7623, - "step": 4853 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019835824934045864, - "loss": 1.6296, - "step": 4854 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019835791111411755, - "loss": 1.7044, - "step": 4855 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983575728877765, - "loss": 1.6561, - "step": 4856 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019835723466143544, - "loss": 1.6673, - "step": 4857 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019835689643509435, - "loss": 1.6735, - "step": 4858 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019835655820875332, - "loss": 1.6212, - "step": 4859 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019835621998241226, - "loss": 1.7357, - "step": 4860 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019835588175607117, - "loss": 1.7748, - "step": 4861 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019835554352973012, - "loss": 1.6725, - "step": 4862 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019835520530338906, - "loss": 1.678, - "step": 4863 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019835486707704797, - "loss": 1.7042, - "step": 4864 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983545288507069, - "loss": 1.7689, - "step": 4865 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019835419062436585, - "loss": 1.7188, - "step": 4866 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019835385239802477, - "loss": 1.7818, - "step": 4867 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983535141716837, - "loss": 1.668, - "step": 4868 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019835317594534265, - "loss": 1.7295, - "step": 4869 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019835283771900156, - "loss": 1.8313, - "step": 4870 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983524994926605, - "loss": 1.6676, - "step": 4871 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019835216126631945, - "loss": 1.6838, - "step": 4872 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019835182303997836, - "loss": 1.72, - "step": 4873 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983514848136373, - "loss": 1.7146, - "step": 4874 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019835114658729624, - "loss": 1.8879, - "step": 4875 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019835080836095516, - "loss": 1.7083, - "step": 4876 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983504701346141, - "loss": 1.78, - "step": 4877 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019835013190827304, - "loss": 1.7252, - "step": 4878 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019834979368193195, - "loss": 1.6996, - "step": 4879 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983494554555909, - "loss": 1.7564, - "step": 4880 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019834911722924983, - "loss": 1.7391, - "step": 4881 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019834877900290875, - "loss": 1.5818, - "step": 4882 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983484407765677, - "loss": 1.6953, - "step": 4883 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019834810255022663, - "loss": 1.8569, - "step": 4884 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019834776432388554, - "loss": 1.6854, - "step": 4885 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019834742609754449, - "loss": 1.8546, - "step": 4886 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019834708787120343, - "loss": 1.7342, - "step": 4887 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019834674964486234, - "loss": 1.8497, - "step": 4888 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019834641141852128, - "loss": 1.6777, - "step": 4889 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019834607319218022, - "loss": 1.7714, - "step": 4890 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019834573496583914, - "loss": 1.7177, - "step": 4891 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019834539673949808, - "loss": 1.7797, - "step": 4892 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019834505851315702, - "loss": 1.672, - "step": 4893 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019834472028681593, - "loss": 1.6759, - "step": 4894 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019834438206047487, - "loss": 1.8305, - "step": 4895 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019834404383413382, - "loss": 1.7674, - "step": 4896 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019834370560779273, - "loss": 1.8043, - "step": 4897 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019834336738145167, - "loss": 1.8517, - "step": 4898 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983430291551106, - "loss": 1.6958, - "step": 4899 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019834269092876953, - "loss": 1.8153, - "step": 4900 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019834235270242847, - "loss": 1.7207, - "step": 4901 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983420144760874, - "loss": 1.6169, - "step": 4902 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019834167624974632, - "loss": 1.7506, - "step": 4903 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019834133802340526, - "loss": 1.7541, - "step": 4904 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983409997970642, - "loss": 1.793, - "step": 4905 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019834066157072312, - "loss": 1.9197, - "step": 4906 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019834032334438206, - "loss": 1.6987, - "step": 4907 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198339985118041, - "loss": 1.6777, - "step": 4908 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019833964689169992, - "loss": 1.8047, - "step": 4909 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019833930866535886, - "loss": 1.7072, - "step": 4910 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983389704390178, - "loss": 1.6362, - "step": 4911 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983386322126767, - "loss": 1.7283, - "step": 4912 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019833829398633565, - "loss": 1.7954, - "step": 4913 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983379557599946, - "loss": 1.8006, - "step": 4914 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019833761753365353, - "loss": 1.5842, - "step": 4915 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019833727930731248, - "loss": 1.7274, - "step": 4916 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019833694108097142, - "loss": 1.6876, - "step": 4917 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019833660285463033, - "loss": 1.6517, - "step": 4918 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019833626462828927, - "loss": 1.7473, - "step": 4919 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983359264019482, - "loss": 1.6632, - "step": 4920 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019833558817560713, - "loss": 1.6681, - "step": 4921 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019833524994926607, - "loss": 1.6538, - "step": 4922 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198334911722925, - "loss": 1.6062, - "step": 4923 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019833457349658392, - "loss": 1.8058, - "step": 4924 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019833423527024286, - "loss": 1.9986, - "step": 4925 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983338970439018, - "loss": 1.7164, - "step": 4926 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019833355881756072, - "loss": 1.8313, - "step": 4927 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019833322059121966, - "loss": 1.6966, - "step": 4928 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983328823648786, - "loss": 1.9016, - "step": 4929 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019833254413853752, - "loss": 1.7828, - "step": 4930 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019833220591219646, - "loss": 1.9386, - "step": 4931 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983318676858554, - "loss": 1.644, - "step": 4932 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983315294595143, - "loss": 1.8017, - "step": 4933 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019833119123317325, - "loss": 1.7524, - "step": 4934 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983308530068322, - "loss": 1.7484, - "step": 4935 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983305147804911, - "loss": 1.6365, - "step": 4936 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019833017655415005, - "loss": 1.7044, - "step": 4937 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198329838327809, - "loss": 1.6929, - "step": 4938 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983295001014679, - "loss": 1.7093, - "step": 4939 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019832916187512685, - "loss": 1.6832, - "step": 4940 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983288236487858, - "loss": 1.7757, - "step": 4941 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983284854224447, - "loss": 1.8131, - "step": 4942 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019832814719610364, - "loss": 1.6551, - "step": 4943 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019832780896976258, - "loss": 1.6681, - "step": 4944 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983274707434215, - "loss": 1.8493, - "step": 4945 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019832713251708044, - "loss": 1.7727, - "step": 4946 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019832679429073938, - "loss": 1.7522, - "step": 4947 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983264560643983, - "loss": 1.6147, - "step": 4948 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019832611783805724, - "loss": 1.7942, - "step": 4949 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019832577961171618, - "loss": 1.6712, - "step": 4950 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983254413853751, - "loss": 1.6587, - "step": 4951 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019832510315903403, - "loss": 1.8004, - "step": 4952 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019832476493269297, - "loss": 1.6512, - "step": 4953 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983244267063519, - "loss": 1.702, - "step": 4954 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019832408848001083, - "loss": 1.7835, - "step": 4955 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019832375025366977, - "loss": 1.7136, - "step": 4956 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019832341202732868, - "loss": 1.7841, - "step": 4957 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019832307380098762, - "loss": 1.7851, - "step": 4958 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019832273557464657, - "loss": 1.8563, - "step": 4959 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019832239734830548, - "loss": 1.7061, - "step": 4960 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019832205912196442, - "loss": 1.7016, - "step": 4961 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019832172089562336, - "loss": 1.6779, - "step": 4962 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019832138266928228, - "loss": 1.7567, - "step": 4963 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019832104444294122, - "loss": 1.7897, - "step": 4964 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019832070621660016, - "loss": 1.8, - "step": 4965 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019832036799025907, - "loss": 1.6974, - "step": 4966 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198320029763918, - "loss": 1.6163, - "step": 4967 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019831969153757695, - "loss": 1.7629, - "step": 4968 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019831935331123587, - "loss": 1.5535, - "step": 4969 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983190150848948, - "loss": 1.6262, - "step": 4970 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019831867685855378, - "loss": 1.7579, - "step": 4971 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983183386322127, - "loss": 1.6495, - "step": 4972 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019831800040587163, - "loss": 1.7624, - "step": 4973 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019831766217953057, - "loss": 1.7615, - "step": 4974 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983173239531895, - "loss": 1.6806, - "step": 4975 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019831698572684843, - "loss": 1.7686, - "step": 4976 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019831664750050737, - "loss": 1.6822, - "step": 4977 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019831630927416628, - "loss": 1.7133, - "step": 4978 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019831597104782523, - "loss": 1.7163, - "step": 4979 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019831563282148417, - "loss": 1.675, - "step": 4980 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019831529459514308, - "loss": 1.6593, - "step": 4981 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019831495636880202, - "loss": 1.6218, - "step": 4982 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019831461814246096, - "loss": 1.7547, - "step": 4983 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019831427991611988, - "loss": 1.6947, - "step": 4984 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019831394168977882, - "loss": 1.6805, - "step": 4985 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019831360346343776, - "loss": 1.7945, - "step": 4986 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019831326523709667, - "loss": 1.7345, - "step": 4987 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019831292701075561, - "loss": 1.7641, - "step": 4988 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019831258878441456, - "loss": 1.9354, - "step": 4989 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019831225055807347, - "loss": 1.6761, - "step": 4990 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983119123317324, - "loss": 1.6913, - "step": 4991 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019831157410539135, - "loss": 1.6321, - "step": 4992 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019831123587905027, - "loss": 1.6237, - "step": 4993 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983108976527092, - "loss": 1.5713, - "step": 4994 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019831055942636815, - "loss": 1.6294, - "step": 4995 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019831022120002706, - "loss": 1.6756, - "step": 4996 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198309882973686, - "loss": 1.8049, - "step": 4997 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019830954474734494, - "loss": 1.6166, - "step": 4998 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019830920652100386, - "loss": 1.7488, - "step": 4999 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983088682946628, - "loss": 1.6445, - "step": 5000 - }, - { - "epoch": 0.02, - "eval_loss": 1.5860877403846154, - "mrc_em": 6.0, - "mrc_f1": 16.309523809523807, - "nli_acc": 0.24, - "step": 5000, - "sts_f1": 0.15384615384615383, - "ynat_f1": 0.47310281884014604 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019830853006832174, - "loss": 1.6895, - "step": 5001 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019830819184198066, - "loss": 1.7242, - "step": 5002 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983078536156396, - "loss": 1.6187, - "step": 5003 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019830751538929854, - "loss": 1.6239, - "step": 5004 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019830717716295745, - "loss": 1.7756, - "step": 5005 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983068389366164, - "loss": 1.7696, - "step": 5006 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019830650071027533, - "loss": 1.6486, - "step": 5007 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019830616248393425, - "loss": 1.8113, - "step": 5008 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983058242575932, - "loss": 1.7628, - "step": 5009 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019830548603125213, - "loss": 1.5976, - "step": 5010 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019830514780491104, - "loss": 1.6528, - "step": 5011 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019830480957856999, - "loss": 1.8815, - "step": 5012 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019830447135222893, - "loss": 1.6468, - "step": 5013 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019830413312588784, - "loss": 1.7485, - "step": 5014 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019830379489954678, - "loss": 1.7013, - "step": 5015 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019830345667320572, - "loss": 1.6825, - "step": 5016 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019830311844686464, - "loss": 1.7182, - "step": 5017 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019830278022052358, - "loss": 1.6901, - "step": 5018 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019830244199418252, - "loss": 1.7325, - "step": 5019 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019830210376784143, - "loss": 1.7846, - "step": 5020 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019830176554150037, - "loss": 1.6155, - "step": 5021 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019830142731515932, - "loss": 1.7751, - "step": 5022 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019830108908881823, - "loss": 1.6722, - "step": 5023 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019830075086247717, - "loss": 1.7398, - "step": 5024 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001983004126361361, - "loss": 1.6282, - "step": 5025 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019830007440979503, - "loss": 1.7395, - "step": 5026 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198299736183454, - "loss": 1.7208, - "step": 5027 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982993979571129, - "loss": 1.6228, - "step": 5028 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019829905973077185, - "loss": 1.7413, - "step": 5029 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982987215044308, - "loss": 1.8002, - "step": 5030 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982983832780897, - "loss": 1.6729, - "step": 5031 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019829804505174865, - "loss": 1.7291, - "step": 5032 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982977068254076, - "loss": 1.7344, - "step": 5033 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982973685990665, - "loss": 1.7574, - "step": 5034 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019829703037272544, - "loss": 1.7607, - "step": 5035 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019829669214638438, - "loss": 1.6953, - "step": 5036 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982963539200433, - "loss": 1.5911, - "step": 5037 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019829601569370224, - "loss": 1.7499, - "step": 5038 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019829567746736118, - "loss": 1.6511, - "step": 5039 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982953392410201, - "loss": 1.7943, - "step": 5040 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019829500101467903, - "loss": 1.7542, - "step": 5041 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019829466278833798, - "loss": 1.6053, - "step": 5042 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982943245619969, - "loss": 1.7627, - "step": 5043 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019829398633565583, - "loss": 1.7538, - "step": 5044 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019829364810931477, - "loss": 1.7556, - "step": 5045 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019829330988297369, - "loss": 1.7536, - "step": 5046 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019829297165663263, - "loss": 1.8099, - "step": 5047 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019829263343029157, - "loss": 1.8069, - "step": 5048 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019829229520395048, - "loss": 1.6744, - "step": 5049 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019829195697760942, - "loss": 1.7877, - "step": 5050 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019829161875126836, - "loss": 1.8334, - "step": 5051 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982912805249273, - "loss": 1.8658, - "step": 5052 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019829094229858622, - "loss": 1.6447, - "step": 5053 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019829060407224516, - "loss": 1.7492, - "step": 5054 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982902658459041, - "loss": 1.7703, - "step": 5055 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019828992761956302, - "loss": 1.7155, - "step": 5056 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019828958939322196, - "loss": 1.7864, - "step": 5057 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982892511668809, - "loss": 1.7239, - "step": 5058 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982889129405398, - "loss": 1.724, - "step": 5059 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019828857471419875, - "loss": 1.7823, - "step": 5060 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982882364878577, - "loss": 1.9908, - "step": 5061 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982878982615166, - "loss": 1.7405, - "step": 5062 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019828756003517555, - "loss": 1.7521, - "step": 5063 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982872218088345, - "loss": 1.5992, - "step": 5064 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982868835824934, - "loss": 1.789, - "step": 5065 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019828654535615235, - "loss": 1.7599, - "step": 5066 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982862071298113, - "loss": 1.6554, - "step": 5067 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982858689034702, - "loss": 1.5743, - "step": 5068 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019828553067712914, - "loss": 1.6782, - "step": 5069 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019828519245078808, - "loss": 1.6003, - "step": 5070 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198284854224447, - "loss": 1.7306, - "step": 5071 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019828451599810594, - "loss": 1.6797, - "step": 5072 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019828417777176488, - "loss": 1.7967, - "step": 5073 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982838395454238, - "loss": 1.8381, - "step": 5074 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019828350131908273, - "loss": 1.8365, - "step": 5075 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019828316309274168, - "loss": 1.702, - "step": 5076 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982828248664006, - "loss": 1.6558, - "step": 5077 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019828248664005953, - "loss": 1.6277, - "step": 5078 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019828214841371847, - "loss": 1.6903, - "step": 5079 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019828181018737739, - "loss": 1.6176, - "step": 5080 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019828147196103633, - "loss": 1.7076, - "step": 5081 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019828113373469527, - "loss": 1.856, - "step": 5082 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019828079550835418, - "loss": 1.7067, - "step": 5083 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019828045728201315, - "loss": 1.6659, - "step": 5084 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019828011905567206, - "loss": 1.7728, - "step": 5085 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198279780829331, - "loss": 1.8721, - "step": 5086 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019827944260298995, - "loss": 1.7568, - "step": 5087 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019827910437664886, - "loss": 1.8133, - "step": 5088 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982787661503078, - "loss": 1.6823, - "step": 5089 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019827842792396674, - "loss": 1.748, - "step": 5090 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019827808969762566, - "loss": 1.7537, - "step": 5091 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982777514712846, - "loss": 1.6712, - "step": 5092 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019827741324494354, - "loss": 1.8789, - "step": 5093 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019827707501860245, - "loss": 1.7819, - "step": 5094 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982767367922614, - "loss": 1.756, - "step": 5095 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019827639856592034, - "loss": 1.66, - "step": 5096 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019827606033957925, - "loss": 1.7578, - "step": 5097 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982757221132382, - "loss": 1.6411, - "step": 5098 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019827538388689713, - "loss": 1.7258, - "step": 5099 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019827504566055605, - "loss": 1.7496, - "step": 5100 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198274707434215, - "loss": 1.8516, - "step": 5101 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019827436920787393, - "loss": 1.7367, - "step": 5102 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019827403098153284, - "loss": 1.7247, - "step": 5103 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019827369275519178, - "loss": 1.7303, - "step": 5104 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019827335452885073, - "loss": 1.7336, - "step": 5105 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019827301630250964, - "loss": 1.7391, - "step": 5106 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019827267807616858, - "loss": 1.7005, - "step": 5107 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019827233984982752, - "loss": 1.6006, - "step": 5108 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019827200162348644, - "loss": 1.8041, - "step": 5109 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019827166339714538, - "loss": 1.7687, - "step": 5110 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019827132517080432, - "loss": 1.7199, - "step": 5111 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019827098694446323, - "loss": 1.628, - "step": 5112 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019827064871812217, - "loss": 1.7774, - "step": 5113 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019827031049178111, - "loss": 1.6504, - "step": 5114 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019826997226544003, - "loss": 1.5337, - "step": 5115 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019826963403909897, - "loss": 1.8471, - "step": 5116 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982692958127579, - "loss": 1.7478, - "step": 5117 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019826895758641682, - "loss": 1.5493, - "step": 5118 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019826861936007577, - "loss": 1.7728, - "step": 5119 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982682811337347, - "loss": 1.7919, - "step": 5120 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019826794290739362, - "loss": 1.7606, - "step": 5121 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019826760468105256, - "loss": 1.6848, - "step": 5122 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982672664547115, - "loss": 1.7606, - "step": 5123 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019826692822837042, - "loss": 1.7502, - "step": 5124 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019826659000202936, - "loss": 1.7405, - "step": 5125 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982662517756883, - "loss": 1.6997, - "step": 5126 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982659135493472, - "loss": 1.8635, - "step": 5127 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019826557532300615, - "loss": 1.776, - "step": 5128 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982652370966651, - "loss": 1.7823, - "step": 5129 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198264898870324, - "loss": 1.8057, - "step": 5130 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019826456064398295, - "loss": 1.6833, - "step": 5131 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982642224176419, - "loss": 1.6477, - "step": 5132 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982638841913008, - "loss": 1.7721, - "step": 5133 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019826354596495975, - "loss": 1.7036, - "step": 5134 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982632077386187, - "loss": 1.6952, - "step": 5135 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019826286951227763, - "loss": 1.532, - "step": 5136 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019826253128593654, - "loss": 1.5999, - "step": 5137 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019826219305959548, - "loss": 1.7134, - "step": 5138 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019826185483325443, - "loss": 1.6031, - "step": 5139 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019826151660691337, - "loss": 1.7692, - "step": 5140 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982611783805723, - "loss": 1.6835, - "step": 5141 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019826084015423122, - "loss": 1.6234, - "step": 5142 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019826050192789016, - "loss": 1.726, - "step": 5143 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982601637015491, - "loss": 1.6302, - "step": 5144 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019825982547520802, - "loss": 1.893, - "step": 5145 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019825948724886696, - "loss": 1.8884, - "step": 5146 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982591490225259, - "loss": 1.7773, - "step": 5147 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019825881079618481, - "loss": 1.6989, - "step": 5148 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019825847256984376, - "loss": 1.5993, - "step": 5149 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982581343435027, - "loss": 1.7334, - "step": 5150 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982577961171616, - "loss": 1.7486, - "step": 5151 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019825745789082055, - "loss": 1.7566, - "step": 5152 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982571196644795, - "loss": 1.7027, - "step": 5153 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982567814381384, - "loss": 1.6667, - "step": 5154 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019825644321179735, - "loss": 1.6106, - "step": 5155 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982561049854563, - "loss": 1.7694, - "step": 5156 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982557667591152, - "loss": 1.6044, - "step": 5157 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019825542853277414, - "loss": 1.8014, - "step": 5158 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019825509030643309, - "loss": 1.6996, - "step": 5159 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198254752080092, - "loss": 1.9514, - "step": 5160 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019825441385375094, - "loss": 1.8681, - "step": 5161 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019825407562740988, - "loss": 1.7192, - "step": 5162 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982537374010688, - "loss": 1.6727, - "step": 5163 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019825339917472774, - "loss": 1.6607, - "step": 5164 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019825306094838668, - "loss": 1.6474, - "step": 5165 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982527227220456, - "loss": 1.8043, - "step": 5166 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019825238449570453, - "loss": 1.8161, - "step": 5167 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019825204626936347, - "loss": 1.6542, - "step": 5168 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982517080430224, - "loss": 1.6549, - "step": 5169 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019825136981668133, - "loss": 1.8143, - "step": 5170 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019825103159034027, - "loss": 1.7026, - "step": 5171 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019825069336399919, - "loss": 1.7274, - "step": 5172 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019825035513765813, - "loss": 1.7064, - "step": 5173 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019825001691131707, - "loss": 1.6389, - "step": 5174 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019824967868497598, - "loss": 1.7366, - "step": 5175 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019824934045863492, - "loss": 1.8592, - "step": 5176 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019824900223229386, - "loss": 1.6242, - "step": 5177 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019824866400595278, - "loss": 1.6801, - "step": 5178 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019824832577961172, - "loss": 1.7167, - "step": 5179 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019824798755327066, - "loss": 1.6256, - "step": 5180 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019824764932692957, - "loss": 1.8199, - "step": 5181 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019824731110058852, - "loss": 1.6854, - "step": 5182 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019824697287424746, - "loss": 1.776, - "step": 5183 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019824663464790637, - "loss": 1.8092, - "step": 5184 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982462964215653, - "loss": 1.6566, - "step": 5185 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019824595819522425, - "loss": 1.7376, - "step": 5186 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019824561996888317, - "loss": 1.7867, - "step": 5187 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982452817425421, - "loss": 1.7642, - "step": 5188 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019824494351620105, - "loss": 1.703, - "step": 5189 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019824460528985996, - "loss": 1.7444, - "step": 5190 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982442670635189, - "loss": 1.5908, - "step": 5191 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019824392883717785, - "loss": 1.6351, - "step": 5192 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019824359061083676, - "loss": 1.7021, - "step": 5193 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982432523844957, - "loss": 1.7001, - "step": 5194 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019824291415815464, - "loss": 1.6833, - "step": 5195 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019824257593181358, - "loss": 1.7929, - "step": 5196 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019824223770547252, - "loss": 1.9649, - "step": 5197 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019824189947913147, - "loss": 1.7239, - "step": 5198 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019824156125279038, - "loss": 1.7959, - "step": 5199 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019824122302644932, - "loss": 1.6243, - "step": 5200 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019824088480010826, - "loss": 1.7593, - "step": 5201 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019824054657376718, - "loss": 1.6889, - "step": 5202 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019824020834742612, - "loss": 1.7598, - "step": 5203 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019823987012108506, - "loss": 1.6867, - "step": 5204 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019823953189474397, - "loss": 1.8083, - "step": 5205 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982391936684029, - "loss": 1.677, - "step": 5206 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019823885544206185, - "loss": 1.6567, - "step": 5207 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019823851721572077, - "loss": 1.6402, - "step": 5208 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982381789893797, - "loss": 1.7854, - "step": 5209 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019823784076303865, - "loss": 1.8384, - "step": 5210 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019823750253669756, - "loss": 1.6902, - "step": 5211 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982371643103565, - "loss": 1.6385, - "step": 5212 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019823682608401545, - "loss": 1.8381, - "step": 5213 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019823648785767436, - "loss": 1.5338, - "step": 5214 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982361496313333, - "loss": 1.7976, - "step": 5215 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019823581140499224, - "loss": 1.75, - "step": 5216 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019823547317865116, - "loss": 1.5931, - "step": 5217 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982351349523101, - "loss": 1.7147, - "step": 5218 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019823479672596904, - "loss": 1.7608, - "step": 5219 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019823445849962795, - "loss": 1.6752, - "step": 5220 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982341202732869, - "loss": 1.8823, - "step": 5221 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019823378204694584, - "loss": 1.7071, - "step": 5222 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019823344382060475, - "loss": 1.7274, - "step": 5223 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982331055942637, - "loss": 1.6271, - "step": 5224 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019823276736792263, - "loss": 1.683, - "step": 5225 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019823242914158155, - "loss": 1.793, - "step": 5226 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982320909152405, - "loss": 1.7107, - "step": 5227 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019823175268889943, - "loss": 1.5341, - "step": 5228 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019823141446255834, - "loss": 1.7931, - "step": 5229 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019823107623621728, - "loss": 1.6679, - "step": 5230 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019823073800987622, - "loss": 1.6716, - "step": 5231 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019823039978353514, - "loss": 1.8924, - "step": 5232 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019823006155719408, - "loss": 1.7353, - "step": 5233 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019822972333085302, - "loss": 1.7827, - "step": 5234 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019822938510451193, - "loss": 1.6776, - "step": 5235 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019822904687817088, - "loss": 1.7732, - "step": 5236 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019822870865182982, - "loss": 1.7608, - "step": 5237 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019822837042548873, - "loss": 1.7133, - "step": 5238 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019822803219914767, - "loss": 1.6895, - "step": 5239 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982276939728066, - "loss": 1.7609, - "step": 5240 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019822735574646553, - "loss": 1.8306, - "step": 5241 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019822701752012447, - "loss": 1.7152, - "step": 5242 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982266792937834, - "loss": 1.6907, - "step": 5243 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019822634106744232, - "loss": 1.837, - "step": 5244 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019822600284110126, - "loss": 1.7257, - "step": 5245 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982256646147602, - "loss": 1.7851, - "step": 5246 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019822532638841912, - "loss": 1.7367, - "step": 5247 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019822498816207806, - "loss": 1.6872, - "step": 5248 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198224649935737, - "loss": 1.726, - "step": 5249 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019822431170939592, - "loss": 1.5713, - "step": 5250 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019822397348305486, - "loss": 1.8429, - "step": 5251 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019822363525671383, - "loss": 1.6765, - "step": 5252 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019822329703037274, - "loss": 1.7291, - "step": 5253 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019822295880403168, - "loss": 1.5893, - "step": 5254 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019822262057769062, - "loss": 1.8258, - "step": 5255 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019822228235134954, - "loss": 1.629, - "step": 5256 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019822194412500848, - "loss": 1.7693, - "step": 5257 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019822160589866742, - "loss": 1.7506, - "step": 5258 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019822126767232633, - "loss": 1.7111, - "step": 5259 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019822092944598527, - "loss": 1.9744, - "step": 5260 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019822059121964421, - "loss": 1.6874, - "step": 5261 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019822025299330313, - "loss": 1.6422, - "step": 5262 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019821991476696207, - "loss": 1.7393, - "step": 5263 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198219576540621, - "loss": 1.7488, - "step": 5264 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019821923831427993, - "loss": 1.7935, - "step": 5265 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019821890008793887, - "loss": 1.6754, - "step": 5266 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982185618615978, - "loss": 1.6946, - "step": 5267 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019821822363525672, - "loss": 1.5551, - "step": 5268 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019821788540891566, - "loss": 1.4671, - "step": 5269 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982175471825746, - "loss": 1.7882, - "step": 5270 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019821720895623352, - "loss": 1.8118, - "step": 5271 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019821687072989246, - "loss": 1.6414, - "step": 5272 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982165325035514, - "loss": 1.6722, - "step": 5273 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019821619427721031, - "loss": 1.5655, - "step": 5274 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019821585605086926, - "loss": 1.6681, - "step": 5275 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982155178245282, - "loss": 1.7418, - "step": 5276 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982151795981871, - "loss": 1.6379, - "step": 5277 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019821484137184605, - "loss": 1.643, - "step": 5278 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198214503145505, - "loss": 1.6349, - "step": 5279 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982141649191639, - "loss": 1.7756, - "step": 5280 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019821382669282285, - "loss": 1.7367, - "step": 5281 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982134884664818, - "loss": 1.7401, - "step": 5282 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982131502401407, - "loss": 1.5689, - "step": 5283 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019821281201379964, - "loss": 1.8172, - "step": 5284 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019821247378745859, - "loss": 1.6667, - "step": 5285 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982121355611175, - "loss": 1.6315, - "step": 5286 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019821179733477644, - "loss": 1.7365, - "step": 5287 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019821145910843538, - "loss": 1.8312, - "step": 5288 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982111208820943, - "loss": 1.5733, - "step": 5289 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019821078265575324, - "loss": 1.6178, - "step": 5290 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019821044442941218, - "loss": 1.7284, - "step": 5291 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982101062030711, - "loss": 1.7161, - "step": 5292 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019820976797673003, - "loss": 1.6635, - "step": 5293 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019820942975038897, - "loss": 1.76, - "step": 5294 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982090915240479, - "loss": 1.8412, - "step": 5295 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019820875329770683, - "loss": 1.7522, - "step": 5296 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019820841507136577, - "loss": 1.5549, - "step": 5297 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019820807684502468, - "loss": 1.7693, - "step": 5298 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019820773861868363, - "loss": 1.7219, - "step": 5299 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019820740039234257, - "loss": 1.648, - "step": 5300 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019820706216600148, - "loss": 1.7011, - "step": 5301 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019820672393966042, - "loss": 1.7374, - "step": 5302 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019820638571331936, - "loss": 1.6168, - "step": 5303 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019820604748697828, - "loss": 1.7048, - "step": 5304 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019820570926063722, - "loss": 1.662, - "step": 5305 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019820537103429616, - "loss": 1.7337, - "step": 5306 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019820503280795507, - "loss": 1.7747, - "step": 5307 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019820469458161404, - "loss": 1.6788, - "step": 5308 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019820435635527296, - "loss": 1.6643, - "step": 5309 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982040181289319, - "loss": 1.571, - "step": 5310 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019820367990259084, - "loss": 1.678, - "step": 5311 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019820334167624975, - "loss": 1.7371, - "step": 5312 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982030034499087, - "loss": 1.8115, - "step": 5313 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019820266522356763, - "loss": 1.4033, - "step": 5314 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019820232699722655, - "loss": 1.7098, - "step": 5315 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001982019887708855, - "loss": 1.5779, - "step": 5316 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019820165054454443, - "loss": 1.6858, - "step": 5317 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019820131231820334, - "loss": 1.8292, - "step": 5318 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019820097409186229, - "loss": 1.7422, - "step": 5319 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019820063586552123, - "loss": 1.7423, - "step": 5320 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019820029763918017, - "loss": 1.9059, - "step": 5321 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019819995941283908, - "loss": 1.696, - "step": 5322 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019819962118649802, - "loss": 1.7288, - "step": 5323 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019819928296015696, - "loss": 1.68, - "step": 5324 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019819894473381588, - "loss": 1.6049, - "step": 5325 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019819860650747482, - "loss": 1.7361, - "step": 5326 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019819826828113376, - "loss": 1.8297, - "step": 5327 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019819793005479267, - "loss": 1.7881, - "step": 5328 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019819759182845162, - "loss": 1.4288, - "step": 5329 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019819725360211056, - "loss": 1.7146, - "step": 5330 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019819691537576947, - "loss": 1.8686, - "step": 5331 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981965771494284, - "loss": 1.5835, - "step": 5332 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019819623892308735, - "loss": 1.6663, - "step": 5333 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019819590069674627, - "loss": 1.6219, - "step": 5334 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981955624704052, - "loss": 1.7112, - "step": 5335 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019819522424406415, - "loss": 1.7244, - "step": 5336 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019819488601772306, - "loss": 1.6873, - "step": 5337 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198194547791382, - "loss": 1.628, - "step": 5338 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019819420956504095, - "loss": 1.741, - "step": 5339 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019819387133869986, - "loss": 1.783, - "step": 5340 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981935331123588, - "loss": 1.7123, - "step": 5341 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019819319488601774, - "loss": 1.7369, - "step": 5342 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019819285665967666, - "loss": 1.8622, - "step": 5343 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981925184333356, - "loss": 1.7285, - "step": 5344 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019819218020699454, - "loss": 1.6714, - "step": 5345 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019819184198065345, - "loss": 1.7255, - "step": 5346 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981915037543124, - "loss": 1.7862, - "step": 5347 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019819116552797133, - "loss": 1.8188, - "step": 5348 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019819082730163025, - "loss": 1.6498, - "step": 5349 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981904890752892, - "loss": 1.8041, - "step": 5350 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019819015084894813, - "loss": 1.6833, - "step": 5351 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019818981262260705, - "loss": 1.5302, - "step": 5352 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019818947439626599, - "loss": 1.8021, - "step": 5353 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019818913616992493, - "loss": 1.6897, - "step": 5354 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019818879794358384, - "loss": 1.7846, - "step": 5355 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019818845971724278, - "loss": 1.6737, - "step": 5356 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019818812149090172, - "loss": 1.762, - "step": 5357 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019818778326456064, - "loss": 1.6066, - "step": 5358 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019818744503821958, - "loss": 1.6462, - "step": 5359 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019818710681187852, - "loss": 1.6901, - "step": 5360 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019818676858553743, - "loss": 1.6265, - "step": 5361 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019818643035919638, - "loss": 1.5262, - "step": 5362 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019818609213285532, - "loss": 1.7645, - "step": 5363 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019818575390651426, - "loss": 1.7285, - "step": 5364 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981854156801732, - "loss": 1.849, - "step": 5365 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981850774538321, - "loss": 1.9869, - "step": 5366 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019818473922749105, - "loss": 1.7101, - "step": 5367 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019818440100115, - "loss": 1.5575, - "step": 5368 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981840627748089, - "loss": 1.8196, - "step": 5369 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019818372454846785, - "loss": 1.6231, - "step": 5370 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981833863221268, - "loss": 1.7251, - "step": 5371 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981830480957857, - "loss": 1.5675, - "step": 5372 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019818270986944465, - "loss": 1.7258, - "step": 5373 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981823716431036, - "loss": 1.6187, - "step": 5374 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981820334167625, - "loss": 1.6681, - "step": 5375 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019818169519042144, - "loss": 1.7457, - "step": 5376 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019818135696408038, - "loss": 1.747, - "step": 5377 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981810187377393, - "loss": 1.835, - "step": 5378 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019818068051139824, - "loss": 1.6552, - "step": 5379 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019818034228505718, - "loss": 1.7587, - "step": 5380 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981800040587161, - "loss": 1.5493, - "step": 5381 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019817966583237504, - "loss": 1.6952, - "step": 5382 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019817932760603398, - "loss": 1.6644, - "step": 5383 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981789893796929, - "loss": 1.5784, - "step": 5384 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019817865115335183, - "loss": 1.7326, - "step": 5385 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019817831292701077, - "loss": 1.6702, - "step": 5386 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981779747006697, - "loss": 1.6513, - "step": 5387 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019817763647432863, - "loss": 1.5679, - "step": 5388 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019817729824798757, - "loss": 1.7348, - "step": 5389 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019817696002164648, - "loss": 1.6604, - "step": 5390 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019817662179530542, - "loss": 1.7102, - "step": 5391 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019817628356896437, - "loss": 1.7084, - "step": 5392 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019817594534262328, - "loss": 1.6394, - "step": 5393 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019817560711628222, - "loss": 1.6998, - "step": 5394 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019817526888994116, - "loss": 1.7362, - "step": 5395 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019817493066360008, - "loss": 1.8806, - "step": 5396 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019817459243725902, - "loss": 1.8416, - "step": 5397 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019817425421091796, - "loss": 1.7255, - "step": 5398 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019817391598457687, - "loss": 1.8501, - "step": 5399 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981735777582358, - "loss": 1.7593, - "step": 5400 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019817323953189475, - "loss": 1.6111, - "step": 5401 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019817290130555367, - "loss": 1.5386, - "step": 5402 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981725630792126, - "loss": 1.7451, - "step": 5403 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019817222485287155, - "loss": 1.6898, - "step": 5404 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981718866265305, - "loss": 1.5976, - "step": 5405 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981715484001894, - "loss": 1.6777, - "step": 5406 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019817121017384835, - "loss": 1.8585, - "step": 5407 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981708719475073, - "loss": 1.6825, - "step": 5408 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981705337211662, - "loss": 1.823, - "step": 5409 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019817019549482514, - "loss": 1.8082, - "step": 5410 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019816985726848408, - "loss": 1.6546, - "step": 5411 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198169519042143, - "loss": 1.6234, - "step": 5412 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019816918081580194, - "loss": 1.7919, - "step": 5413 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019816884258946088, - "loss": 1.8384, - "step": 5414 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981685043631198, - "loss": 1.7072, - "step": 5415 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019816816613677874, - "loss": 1.66, - "step": 5416 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019816782791043768, - "loss": 1.7774, - "step": 5417 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981674896840966, - "loss": 1.7925, - "step": 5418 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019816715145775553, - "loss": 1.6179, - "step": 5419 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019816681323141447, - "loss": 1.9403, - "step": 5420 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019816647500507341, - "loss": 1.6282, - "step": 5421 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019816613677873236, - "loss": 1.6501, - "step": 5422 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019816579855239127, - "loss": 1.7421, - "step": 5423 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981654603260502, - "loss": 1.7196, - "step": 5424 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019816512209970915, - "loss": 1.8748, - "step": 5425 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019816478387336807, - "loss": 1.5419, - "step": 5426 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198164445647027, - "loss": 1.6944, - "step": 5427 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019816410742068595, - "loss": 1.8037, - "step": 5428 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019816376919434486, - "loss": 1.775, - "step": 5429 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981634309680038, - "loss": 1.8945, - "step": 5430 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019816309274166274, - "loss": 1.8277, - "step": 5431 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019816275451532166, - "loss": 1.5999, - "step": 5432 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981624162889806, - "loss": 1.6618, - "step": 5433 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019816207806263954, - "loss": 1.6623, - "step": 5434 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019816173983629846, - "loss": 1.5687, - "step": 5435 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981614016099574, - "loss": 1.6223, - "step": 5436 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019816106338361634, - "loss": 1.7358, - "step": 5437 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019816072515727525, - "loss": 1.7361, - "step": 5438 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981603869309342, - "loss": 1.6849, - "step": 5439 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019816004870459313, - "loss": 1.6691, - "step": 5440 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019815971047825205, - "loss": 1.6971, - "step": 5441 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198159372251911, - "loss": 1.8084, - "step": 5442 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019815903402556993, - "loss": 1.6598, - "step": 5443 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019815869579922884, - "loss": 1.7638, - "step": 5444 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019815835757288779, - "loss": 1.6267, - "step": 5445 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019815801934654673, - "loss": 1.7282, - "step": 5446 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019815768112020564, - "loss": 1.7218, - "step": 5447 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019815734289386458, - "loss": 1.7195, - "step": 5448 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019815700466752352, - "loss": 1.6807, - "step": 5449 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019815666644118244, - "loss": 1.6243, - "step": 5450 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019815632821484138, - "loss": 1.771, - "step": 5451 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019815598998850032, - "loss": 1.6791, - "step": 5452 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019815565176215923, - "loss": 1.7851, - "step": 5453 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019815531353581817, - "loss": 1.8028, - "step": 5454 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019815497530947712, - "loss": 1.605, - "step": 5455 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019815463708313603, - "loss": 1.6951, - "step": 5456 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019815429885679497, - "loss": 1.5536, - "step": 5457 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981539606304539, - "loss": 1.7712, - "step": 5458 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019815362240411283, - "loss": 1.8416, - "step": 5459 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019815328417777177, - "loss": 1.8249, - "step": 5460 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981529459514307, - "loss": 1.6816, - "step": 5461 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019815260772508962, - "loss": 1.7131, - "step": 5462 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019815226949874856, - "loss": 1.6959, - "step": 5463 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981519312724075, - "loss": 1.7686, - "step": 5464 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019815159304606642, - "loss": 1.6979, - "step": 5465 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019815125481972536, - "loss": 1.7629, - "step": 5466 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981509165933843, - "loss": 1.7302, - "step": 5467 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019815057836704321, - "loss": 1.6571, - "step": 5468 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019815024014070216, - "loss": 1.7694, - "step": 5469 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981499019143611, - "loss": 1.5769, - "step": 5470 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019814956368802, - "loss": 1.6642, - "step": 5471 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019814922546167895, - "loss": 1.7635, - "step": 5472 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981488872353379, - "loss": 1.7855, - "step": 5473 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981485490089968, - "loss": 1.7018, - "step": 5474 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019814821078265575, - "loss": 1.6936, - "step": 5475 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981478725563147, - "loss": 1.6485, - "step": 5476 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019814753432997363, - "loss": 1.5831, - "step": 5477 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019814719610363257, - "loss": 1.8442, - "step": 5478 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981468578772915, - "loss": 1.6861, - "step": 5479 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019814651965095043, - "loss": 1.7219, - "step": 5480 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019814618142460937, - "loss": 1.6354, - "step": 5481 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981458431982683, - "loss": 1.806, - "step": 5482 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019814550497192722, - "loss": 1.6722, - "step": 5483 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019814516674558616, - "loss": 1.8146, - "step": 5484 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981448285192451, - "loss": 1.672, - "step": 5485 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019814449029290402, - "loss": 1.8003, - "step": 5486 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019814415206656296, - "loss": 1.7112, - "step": 5487 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981438138402219, - "loss": 1.7824, - "step": 5488 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019814347561388082, - "loss": 1.6955, - "step": 5489 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019814313738753976, - "loss": 1.7589, - "step": 5490 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981427991611987, - "loss": 1.6124, - "step": 5491 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981424609348576, - "loss": 1.6379, - "step": 5492 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019814212270851655, - "loss": 1.7447, - "step": 5493 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981417844821755, - "loss": 1.6163, - "step": 5494 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981414462558344, - "loss": 1.8421, - "step": 5495 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019814110802949335, - "loss": 1.7979, - "step": 5496 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981407698031523, - "loss": 1.8128, - "step": 5497 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981404315768112, - "loss": 1.8621, - "step": 5498 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019814009335047015, - "loss": 1.6473, - "step": 5499 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981397551241291, - "loss": 1.6742, - "step": 5500 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198139416897788, - "loss": 1.6166, - "step": 5501 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019813907867144694, - "loss": 1.9949, - "step": 5502 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019813874044510588, - "loss": 1.7834, - "step": 5503 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981384022187648, - "loss": 1.7609, - "step": 5504 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019813806399242374, - "loss": 1.8107, - "step": 5505 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019813772576608268, - "loss": 1.7579, - "step": 5506 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981373875397416, - "loss": 1.7538, - "step": 5507 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019813704931340053, - "loss": 1.6004, - "step": 5508 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019813671108705948, - "loss": 1.7713, - "step": 5509 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981363728607184, - "loss": 1.5403, - "step": 5510 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019813603463437733, - "loss": 1.6252, - "step": 5511 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019813569640803627, - "loss": 1.7398, - "step": 5512 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019813535818169519, - "loss": 1.7625, - "step": 5513 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019813501995535413, - "loss": 1.7508, - "step": 5514 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019813468172901307, - "loss": 1.6734, - "step": 5515 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019813434350267198, - "loss": 1.7458, - "step": 5516 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019813400527633092, - "loss": 1.6324, - "step": 5517 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019813366704998986, - "loss": 1.5318, - "step": 5518 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019813332882364878, - "loss": 1.7676, - "step": 5519 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019813299059730772, - "loss": 1.7033, - "step": 5520 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019813265237096666, - "loss": 1.6579, - "step": 5521 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019813231414462558, - "loss": 1.6554, - "step": 5522 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019813197591828452, - "loss": 1.7561, - "step": 5523 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019813163769194346, - "loss": 1.6329, - "step": 5524 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019813129946560237, - "loss": 1.7985, - "step": 5525 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981309612392613, - "loss": 1.8094, - "step": 5526 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019813062301292025, - "loss": 1.8566, - "step": 5527 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019813028478657917, - "loss": 1.6862, - "step": 5528 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981299465602381, - "loss": 1.5748, - "step": 5529 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019812960833389705, - "loss": 1.8523, - "step": 5530 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019812927010755596, - "loss": 1.5732, - "step": 5531 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981289318812149, - "loss": 1.6049, - "step": 5532 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019812859365487387, - "loss": 1.5989, - "step": 5533 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981282554285328, - "loss": 1.5434, - "step": 5534 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019812791720219173, - "loss": 1.8714, - "step": 5535 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019812757897585067, - "loss": 1.7594, - "step": 5536 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019812724074950958, - "loss": 1.7232, - "step": 5537 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019812690252316853, - "loss": 1.7545, - "step": 5538 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019812656429682747, - "loss": 1.5898, - "step": 5539 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019812622607048638, - "loss": 1.7691, - "step": 5540 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019812588784414532, - "loss": 1.6816, - "step": 5541 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019812554961780426, - "loss": 1.6537, - "step": 5542 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019812521139146318, - "loss": 1.6717, - "step": 5543 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019812487316512212, - "loss": 1.6348, - "step": 5544 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019812453493878106, - "loss": 1.6506, - "step": 5545 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019812419671243997, - "loss": 1.6572, - "step": 5546 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019812385848609891, - "loss": 1.8139, - "step": 5547 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019812352025975786, - "loss": 1.8522, - "step": 5548 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019812318203341677, - "loss": 1.691, - "step": 5549 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981228438070757, - "loss": 1.7202, - "step": 5550 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019812250558073465, - "loss": 1.7605, - "step": 5551 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019812216735439357, - "loss": 1.7549, - "step": 5552 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981218291280525, - "loss": 1.7791, - "step": 5553 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019812149090171145, - "loss": 1.714, - "step": 5554 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019812115267537036, - "loss": 1.7176, - "step": 5555 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981208144490293, - "loss": 1.6774, - "step": 5556 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019812047622268824, - "loss": 1.6268, - "step": 5557 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019812013799634716, - "loss": 1.6569, - "step": 5558 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981197997700061, - "loss": 1.6282, - "step": 5559 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019811946154366504, - "loss": 1.7424, - "step": 5560 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019811912331732395, - "loss": 1.6509, - "step": 5561 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981187850909829, - "loss": 1.8267, - "step": 5562 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019811844686464184, - "loss": 1.7845, - "step": 5563 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019811810863830075, - "loss": 1.8024, - "step": 5564 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981177704119597, - "loss": 1.5334, - "step": 5565 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019811743218561863, - "loss": 1.6594, - "step": 5566 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019811709395927755, - "loss": 1.5781, - "step": 5567 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981167557329365, - "loss": 1.7201, - "step": 5568 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019811641750659543, - "loss": 1.7227, - "step": 5569 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019811607928025434, - "loss": 1.6295, - "step": 5570 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019811574105391328, - "loss": 1.6286, - "step": 5571 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019811540282757223, - "loss": 1.6871, - "step": 5572 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019811506460123114, - "loss": 1.8197, - "step": 5573 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019811472637489008, - "loss": 1.8926, - "step": 5574 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019811438814854902, - "loss": 1.6657, - "step": 5575 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019811404992220794, - "loss": 1.6726, - "step": 5576 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019811371169586688, - "loss": 1.7659, - "step": 5577 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019811337346952582, - "loss": 1.7957, - "step": 5578 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019811303524318473, - "loss": 1.8094, - "step": 5579 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019811269701684367, - "loss": 1.748, - "step": 5580 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019811235879050261, - "loss": 1.6477, - "step": 5581 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019811202056416153, - "loss": 1.7976, - "step": 5582 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019811168233782047, - "loss": 1.6941, - "step": 5583 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981113441114794, - "loss": 1.6856, - "step": 5584 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019811100588513833, - "loss": 1.7305, - "step": 5585 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019811066765879727, - "loss": 1.6402, - "step": 5586 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981103294324562, - "loss": 1.7152, - "step": 5587 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019810999120611512, - "loss": 1.6939, - "step": 5588 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981096529797741, - "loss": 1.9621, - "step": 5589 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019810931475343303, - "loss": 1.7087, - "step": 5590 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019810897652709194, - "loss": 1.7001, - "step": 5591 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019810863830075089, - "loss": 1.9243, - "step": 5592 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019810830007440983, - "loss": 1.6841, - "step": 5593 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019810796184806874, - "loss": 1.6445, - "step": 5594 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019810762362172768, - "loss": 1.6855, - "step": 5595 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019810728539538662, - "loss": 1.7235, - "step": 5596 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019810694716904554, - "loss": 1.6581, - "step": 5597 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019810660894270448, - "loss": 1.7532, - "step": 5598 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019810627071636342, - "loss": 1.6848, - "step": 5599 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019810593249002233, - "loss": 1.6381, - "step": 5600 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019810559426368127, - "loss": 1.7677, - "step": 5601 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019810525603734022, - "loss": 1.68, - "step": 5602 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019810491781099913, - "loss": 1.7084, - "step": 5603 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019810457958465807, - "loss": 1.8493, - "step": 5604 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198104241358317, - "loss": 1.7265, - "step": 5605 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019810390313197593, - "loss": 1.8868, - "step": 5606 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019810356490563487, - "loss": 1.5806, - "step": 5607 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981032266792938, - "loss": 1.8967, - "step": 5608 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019810288845295272, - "loss": 1.5425, - "step": 5609 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019810255022661166, - "loss": 1.7052, - "step": 5610 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981022120002706, - "loss": 1.7013, - "step": 5611 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019810187377392952, - "loss": 1.7538, - "step": 5612 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019810153554758846, - "loss": 1.6659, - "step": 5613 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981011973212474, - "loss": 1.6501, - "step": 5614 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019810085909490632, - "loss": 1.7176, - "step": 5615 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019810052086856526, - "loss": 1.7117, - "step": 5616 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001981001826422242, - "loss": 1.7487, - "step": 5617 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980998444158831, - "loss": 1.6588, - "step": 5618 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019809950618954205, - "loss": 1.7458, - "step": 5619 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198099167963201, - "loss": 1.528, - "step": 5620 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980988297368599, - "loss": 1.7061, - "step": 5621 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019809849151051885, - "loss": 1.6837, - "step": 5622 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980981532841778, - "loss": 1.7437, - "step": 5623 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980978150578367, - "loss": 1.6149, - "step": 5624 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019809747683149565, - "loss": 1.8007, - "step": 5625 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980971386051546, - "loss": 1.6672, - "step": 5626 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980968003788135, - "loss": 1.5975, - "step": 5627 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019809646215247244, - "loss": 1.7258, - "step": 5628 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019809612392613138, - "loss": 1.7157, - "step": 5629 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980957856997903, - "loss": 1.8387, - "step": 5630 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019809544747344924, - "loss": 1.6804, - "step": 5631 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019809510924710818, - "loss": 1.8224, - "step": 5632 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980947710207671, - "loss": 1.7169, - "step": 5633 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019809443279442603, - "loss": 1.6797, - "step": 5634 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019809409456808498, - "loss": 1.734, - "step": 5635 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980937563417439, - "loss": 1.7673, - "step": 5636 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019809341811540283, - "loss": 1.6543, - "step": 5637 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019809307988906177, - "loss": 1.6216, - "step": 5638 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019809274166272069, - "loss": 1.6535, - "step": 5639 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019809240343637963, - "loss": 1.7347, - "step": 5640 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019809206521003857, - "loss": 1.763, - "step": 5641 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019809172698369748, - "loss": 1.5893, - "step": 5642 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019809138875735642, - "loss": 1.6348, - "step": 5643 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019809105053101536, - "loss": 1.5596, - "step": 5644 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980907123046743, - "loss": 1.8073, - "step": 5645 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019809037407833325, - "loss": 1.7632, - "step": 5646 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019809003585199216, - "loss": 1.5856, - "step": 5647 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980896976256511, - "loss": 1.9497, - "step": 5648 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019808935939931004, - "loss": 1.6438, - "step": 5649 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019808902117296896, - "loss": 1.7608, - "step": 5650 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980886829466279, - "loss": 1.7366, - "step": 5651 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019808834472028684, - "loss": 1.5244, - "step": 5652 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019808800649394575, - "loss": 1.7188, - "step": 5653 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980876682676047, - "loss": 1.7082, - "step": 5654 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019808733004126364, - "loss": 1.7117, - "step": 5655 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019808699181492255, - "loss": 1.6509, - "step": 5656 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980866535885815, - "loss": 1.7514, - "step": 5657 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019808631536224043, - "loss": 1.687, - "step": 5658 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019808597713589935, - "loss": 1.8086, - "step": 5659 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980856389095583, - "loss": 1.7116, - "step": 5660 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019808530068321723, - "loss": 1.7472, - "step": 5661 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019808496245687614, - "loss": 1.7972, - "step": 5662 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019808462423053508, - "loss": 1.8122, - "step": 5663 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019808428600419402, - "loss": 1.6152, - "step": 5664 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019808394777785294, - "loss": 1.7341, - "step": 5665 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019808360955151188, - "loss": 1.772, - "step": 5666 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019808327132517082, - "loss": 1.7056, - "step": 5667 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019808293309882973, - "loss": 1.9366, - "step": 5668 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019808259487248868, - "loss": 1.8043, - "step": 5669 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019808225664614762, - "loss": 1.6468, - "step": 5670 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019808191841980656, - "loss": 1.6284, - "step": 5671 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019808158019346547, - "loss": 1.6999, - "step": 5672 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980812419671244, - "loss": 1.6777, - "step": 5673 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019808090374078335, - "loss": 1.6365, - "step": 5674 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019808056551444227, - "loss": 1.5737, - "step": 5675 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980802272881012, - "loss": 1.479, - "step": 5676 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019807988906176015, - "loss": 1.7769, - "step": 5677 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019807955083541906, - "loss": 1.6214, - "step": 5678 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198079212609078, - "loss": 1.6938, - "step": 5679 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019807887438273695, - "loss": 1.6982, - "step": 5680 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019807853615639586, - "loss": 1.6172, - "step": 5681 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980781979300548, - "loss": 1.7226, - "step": 5682 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019807785970371374, - "loss": 1.6695, - "step": 5683 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019807752147737266, - "loss": 1.6262, - "step": 5684 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980771832510316, - "loss": 1.6184, - "step": 5685 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019807684502469054, - "loss": 1.7634, - "step": 5686 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019807650679834945, - "loss": 1.7287, - "step": 5687 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980761685720084, - "loss": 1.635, - "step": 5688 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019807583034566734, - "loss": 1.7731, - "step": 5689 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019807549211932625, - "loss": 1.7026, - "step": 5690 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980751538929852, - "loss": 1.7621, - "step": 5691 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019807481566664413, - "loss": 1.8034, - "step": 5692 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019807447744030305, - "loss": 1.7218, - "step": 5693 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198074139213962, - "loss": 1.7394, - "step": 5694 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019807380098762093, - "loss": 1.7859, - "step": 5695 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019807346276127984, - "loss": 1.8872, - "step": 5696 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019807312453493878, - "loss": 1.6487, - "step": 5697 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019807278630859773, - "loss": 1.8604, - "step": 5698 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019807244808225664, - "loss": 1.825, - "step": 5699 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019807210985591558, - "loss": 1.7722, - "step": 5700 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019807177162957452, - "loss": 1.6866, - "step": 5701 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019807143340323346, - "loss": 1.6072, - "step": 5702 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980710951768924, - "loss": 1.8742, - "step": 5703 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019807075695055132, - "loss": 1.8379, - "step": 5704 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019807041872421026, - "loss": 1.7067, - "step": 5705 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980700804978692, - "loss": 1.7624, - "step": 5706 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019806974227152811, - "loss": 1.7478, - "step": 5707 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019806940404518706, - "loss": 1.6419, - "step": 5708 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198069065818846, - "loss": 1.6974, - "step": 5709 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980687275925049, - "loss": 1.6888, - "step": 5710 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019806838936616385, - "loss": 1.6313, - "step": 5711 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980680511398228, - "loss": 1.7912, - "step": 5712 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980677129134817, - "loss": 1.7182, - "step": 5713 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019806737468714065, - "loss": 1.7279, - "step": 5714 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980670364607996, - "loss": 1.7107, - "step": 5715 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980666982344585, - "loss": 1.7786, - "step": 5716 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019806636000811744, - "loss": 1.6975, - "step": 5717 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019806602178177639, - "loss": 1.6046, - "step": 5718 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980656835554353, - "loss": 1.588, - "step": 5719 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019806534532909424, - "loss": 1.631, - "step": 5720 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019806500710275318, - "loss": 1.7046, - "step": 5721 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980646688764121, - "loss": 1.837, - "step": 5722 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019806433065007104, - "loss": 1.7974, - "step": 5723 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019806399242372998, - "loss": 1.7659, - "step": 5724 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980636541973889, - "loss": 1.7414, - "step": 5725 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019806331597104783, - "loss": 1.6809, - "step": 5726 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019806297774470677, - "loss": 1.8269, - "step": 5727 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980626395183657, - "loss": 1.7376, - "step": 5728 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019806230129202463, - "loss": 1.6896, - "step": 5729 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019806196306568357, - "loss": 1.7063, - "step": 5730 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019806162483934248, - "loss": 1.817, - "step": 5731 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019806128661300143, - "loss": 1.7544, - "step": 5732 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019806094838666037, - "loss": 1.6784, - "step": 5733 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019806061016031928, - "loss": 1.7101, - "step": 5734 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019806027193397822, - "loss": 1.6552, - "step": 5735 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019805993370763716, - "loss": 1.7228, - "step": 5736 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019805959548129608, - "loss": 1.6103, - "step": 5737 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019805925725495502, - "loss": 1.5689, - "step": 5738 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019805891902861396, - "loss": 1.695, - "step": 5739 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019805858080227287, - "loss": 1.7376, - "step": 5740 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019805824257593181, - "loss": 1.7498, - "step": 5741 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019805790434959076, - "loss": 1.9384, - "step": 5742 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019805756612324967, - "loss": 1.6448, - "step": 5743 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980572278969086, - "loss": 1.7049, - "step": 5744 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019805688967056755, - "loss": 1.6396, - "step": 5745 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019805655144422647, - "loss": 1.6364, - "step": 5746 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980562132178854, - "loss": 1.6195, - "step": 5747 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019805587499154435, - "loss": 1.6663, - "step": 5748 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019805553676520326, - "loss": 1.6601, - "step": 5749 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980551985388622, - "loss": 1.685, - "step": 5750 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019805486031252114, - "loss": 1.7068, - "step": 5751 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019805452208618006, - "loss": 1.7005, - "step": 5752 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198054183859839, - "loss": 1.5531, - "step": 5753 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019805384563349794, - "loss": 1.7412, - "step": 5754 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019805350740715686, - "loss": 1.7702, - "step": 5755 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980531691808158, - "loss": 1.7142, - "step": 5756 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019805283095447474, - "loss": 1.6877, - "step": 5757 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019805249272813368, - "loss": 1.6423, - "step": 5758 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019805215450179262, - "loss": 1.6407, - "step": 5759 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019805181627545156, - "loss": 1.7941, - "step": 5760 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019805147804911047, - "loss": 1.6624, - "step": 5761 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019805113982276942, - "loss": 1.8161, - "step": 5762 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019805080159642836, - "loss": 1.6863, - "step": 5763 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019805046337008727, - "loss": 1.8226, - "step": 5764 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980501251437462, - "loss": 1.7394, - "step": 5765 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019804978691740515, - "loss": 1.7935, - "step": 5766 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019804944869106407, - "loss": 1.715, - "step": 5767 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198049110464723, - "loss": 1.6825, - "step": 5768 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019804877223838195, - "loss": 1.702, - "step": 5769 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019804843401204086, - "loss": 1.7824, - "step": 5770 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980480957856998, - "loss": 1.6921, - "step": 5771 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019804775755935875, - "loss": 1.6163, - "step": 5772 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019804741933301766, - "loss": 1.7871, - "step": 5773 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980470811066766, - "loss": 1.6933, - "step": 5774 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019804674288033554, - "loss": 1.6094, - "step": 5775 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019804640465399446, - "loss": 1.792, - "step": 5776 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980460664276534, - "loss": 1.6097, - "step": 5777 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019804572820131234, - "loss": 1.685, - "step": 5778 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019804538997497125, - "loss": 1.7785, - "step": 5779 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980450517486302, - "loss": 1.6205, - "step": 5780 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019804471352228913, - "loss": 1.7972, - "step": 5781 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019804437529594805, - "loss": 1.742, - "step": 5782 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198044037069607, - "loss": 1.7645, - "step": 5783 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019804369884326593, - "loss": 1.7262, - "step": 5784 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019804336061692485, - "loss": 1.621, - "step": 5785 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019804302239058379, - "loss": 1.6131, - "step": 5786 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019804268416424273, - "loss": 1.6621, - "step": 5787 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019804234593790164, - "loss": 1.7495, - "step": 5788 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019804200771156058, - "loss": 1.7193, - "step": 5789 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019804166948521952, - "loss": 1.6581, - "step": 5790 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019804133125887844, - "loss": 1.5526, - "step": 5791 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019804099303253738, - "loss": 1.8682, - "step": 5792 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019804065480619632, - "loss": 1.7377, - "step": 5793 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019804031657985523, - "loss": 1.7132, - "step": 5794 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019803997835351418, - "loss": 1.7311, - "step": 5795 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019803964012717312, - "loss": 1.5952, - "step": 5796 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019803930190083203, - "loss": 1.7175, - "step": 5797 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019803896367449097, - "loss": 1.6291, - "step": 5798 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980386254481499, - "loss": 1.5765, - "step": 5799 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019803828722180883, - "loss": 1.7473, - "step": 5800 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019803794899546777, - "loss": 1.5857, - "step": 5801 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980376107691267, - "loss": 1.5892, - "step": 5802 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019803727254278562, - "loss": 1.6116, - "step": 5803 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019803693431644456, - "loss": 1.8249, - "step": 5804 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980365960901035, - "loss": 1.6116, - "step": 5805 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019803625786376242, - "loss": 1.7909, - "step": 5806 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019803591963742136, - "loss": 1.6336, - "step": 5807 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980355814110803, - "loss": 1.6299, - "step": 5808 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019803524318473922, - "loss": 1.8167, - "step": 5809 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019803490495839816, - "loss": 1.6473, - "step": 5810 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980345667320571, - "loss": 1.8027, - "step": 5811 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198034228505716, - "loss": 1.711, - "step": 5812 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019803389027937495, - "loss": 1.6132, - "step": 5813 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019803355205303392, - "loss": 1.7083, - "step": 5814 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019803321382669284, - "loss": 1.5891, - "step": 5815 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019803287560035178, - "loss": 1.6312, - "step": 5816 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019803253737401072, - "loss": 1.889, - "step": 5817 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019803219914766963, - "loss": 1.7007, - "step": 5818 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019803186092132857, - "loss": 1.704, - "step": 5819 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019803152269498751, - "loss": 1.6245, - "step": 5820 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019803118446864643, - "loss": 1.6945, - "step": 5821 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019803084624230537, - "loss": 1.571, - "step": 5822 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980305080159643, - "loss": 1.6554, - "step": 5823 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019803016978962322, - "loss": 1.7868, - "step": 5824 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019802983156328217, - "loss": 1.792, - "step": 5825 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980294933369411, - "loss": 1.7009, - "step": 5826 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019802915511060002, - "loss": 1.6095, - "step": 5827 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019802881688425896, - "loss": 1.6454, - "step": 5828 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980284786579179, - "loss": 1.73, - "step": 5829 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019802814043157682, - "loss": 1.7626, - "step": 5830 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019802780220523576, - "loss": 1.7125, - "step": 5831 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980274639788947, - "loss": 1.7177, - "step": 5832 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980271257525536, - "loss": 1.5826, - "step": 5833 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019802678752621255, - "loss": 1.7295, - "step": 5834 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980264492998715, - "loss": 1.7571, - "step": 5835 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980261110735304, - "loss": 1.771, - "step": 5836 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019802577284718935, - "loss": 1.763, - "step": 5837 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980254346208483, - "loss": 1.4586, - "step": 5838 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980250963945072, - "loss": 1.7785, - "step": 5839 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019802475816816615, - "loss": 1.7809, - "step": 5840 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980244199418251, - "loss": 1.912, - "step": 5841 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198024081715484, - "loss": 1.7051, - "step": 5842 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019802374348914294, - "loss": 1.6331, - "step": 5843 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019802340526280188, - "loss": 1.6503, - "step": 5844 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980230670364608, - "loss": 1.733, - "step": 5845 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019802272881011974, - "loss": 1.7584, - "step": 5846 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019802239058377868, - "loss": 1.761, - "step": 5847 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980220523574376, - "loss": 1.7757, - "step": 5848 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019802171413109654, - "loss": 1.6386, - "step": 5849 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019802137590475548, - "loss": 1.6847, - "step": 5850 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980210376784144, - "loss": 1.8025, - "step": 5851 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019802069945207333, - "loss": 1.8631, - "step": 5852 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019802036122573227, - "loss": 1.7672, - "step": 5853 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980200229993912, - "loss": 1.6095, - "step": 5854 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019801968477305013, - "loss": 1.7816, - "step": 5855 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019801934654670907, - "loss": 1.6267, - "step": 5856 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019801900832036798, - "loss": 1.7944, - "step": 5857 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019801867009402693, - "loss": 1.7714, - "step": 5858 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019801833186768587, - "loss": 1.7689, - "step": 5859 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019801799364134478, - "loss": 1.7323, - "step": 5860 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019801765541500372, - "loss": 1.6936, - "step": 5861 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019801731718866266, - "loss": 1.6636, - "step": 5862 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019801697896232158, - "loss": 1.7383, - "step": 5863 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019801664073598052, - "loss": 1.7094, - "step": 5864 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019801630250963946, - "loss": 1.7903, - "step": 5865 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019801596428329837, - "loss": 1.6121, - "step": 5866 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019801562605695731, - "loss": 1.5932, - "step": 5867 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019801528783061626, - "loss": 1.7856, - "step": 5868 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019801494960427517, - "loss": 1.7019, - "step": 5869 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019801461137793414, - "loss": 1.6701, - "step": 5870 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019801427315159308, - "loss": 1.7097, - "step": 5871 - }, - { - "epoch": 0.02, - "learning_rate": 0.000198013934925252, - "loss": 1.5404, - "step": 5872 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019801359669891093, - "loss": 1.62, - "step": 5873 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019801325847256987, - "loss": 1.4765, - "step": 5874 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980129202462288, - "loss": 1.665, - "step": 5875 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019801258201988773, - "loss": 1.7487, - "step": 5876 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019801224379354667, - "loss": 1.7334, - "step": 5877 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019801190556720559, - "loss": 1.6844, - "step": 5878 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019801156734086453, - "loss": 1.7075, - "step": 5879 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019801122911452347, - "loss": 1.7663, - "step": 5880 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019801089088818238, - "loss": 1.5736, - "step": 5881 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019801055266184132, - "loss": 1.7563, - "step": 5882 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019801021443550026, - "loss": 1.676, - "step": 5883 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019800987620915918, - "loss": 1.7678, - "step": 5884 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019800953798281812, - "loss": 1.7663, - "step": 5885 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019800919975647706, - "loss": 1.6277, - "step": 5886 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019800886153013597, - "loss": 1.7835, - "step": 5887 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019800852330379492, - "loss": 1.6612, - "step": 5888 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019800818507745386, - "loss": 1.7382, - "step": 5889 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019800784685111277, - "loss": 1.7019, - "step": 5890 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980075086247717, - "loss": 1.6883, - "step": 5891 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019800717039843065, - "loss": 1.7897, - "step": 5892 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019800683217208957, - "loss": 1.7405, - "step": 5893 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980064939457485, - "loss": 1.7227, - "step": 5894 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019800615571940745, - "loss": 1.6654, - "step": 5895 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019800581749306636, - "loss": 1.7634, - "step": 5896 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980054792667253, - "loss": 1.8323, - "step": 5897 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019800514104038425, - "loss": 1.5795, - "step": 5898 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019800480281404316, - "loss": 1.6493, - "step": 5899 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980044645877021, - "loss": 1.7828, - "step": 5900 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019800412636136104, - "loss": 1.797, - "step": 5901 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019800378813501996, - "loss": 1.5602, - "step": 5902 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980034499086789, - "loss": 1.693, - "step": 5903 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019800311168233784, - "loss": 1.7841, - "step": 5904 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019800277345599675, - "loss": 1.6441, - "step": 5905 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980024352296557, - "loss": 1.6184, - "step": 5906 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019800209700331463, - "loss": 1.6717, - "step": 5907 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019800175877697355, - "loss": 1.7556, - "step": 5908 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001980014205506325, - "loss": 1.6151, - "step": 5909 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019800108232429143, - "loss": 1.7405, - "step": 5910 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019800074409795034, - "loss": 1.8516, - "step": 5911 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019800040587160929, - "loss": 1.6415, - "step": 5912 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019800006764526823, - "loss": 1.5348, - "step": 5913 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019799972941892714, - "loss": 1.5283, - "step": 5914 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019799939119258608, - "loss": 1.8268, - "step": 5915 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019799905296624502, - "loss": 1.6498, - "step": 5916 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019799871473990394, - "loss": 1.7017, - "step": 5917 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019799837651356288, - "loss": 1.6621, - "step": 5918 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019799803828722182, - "loss": 1.8112, - "step": 5919 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019799770006088073, - "loss": 1.6892, - "step": 5920 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019799736183453967, - "loss": 1.7986, - "step": 5921 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019799702360819862, - "loss": 1.7537, - "step": 5922 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019799668538185753, - "loss": 1.719, - "step": 5923 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019799634715551647, - "loss": 1.7685, - "step": 5924 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979960089291754, - "loss": 1.6194, - "step": 5925 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019799567070283435, - "loss": 1.6795, - "step": 5926 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979953324764933, - "loss": 1.7695, - "step": 5927 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979949942501522, - "loss": 1.5989, - "step": 5928 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019799465602381115, - "loss": 1.6893, - "step": 5929 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979943177974701, - "loss": 1.7437, - "step": 5930 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197993979571129, - "loss": 1.7935, - "step": 5931 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019799364134478795, - "loss": 1.7406, - "step": 5932 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979933031184469, - "loss": 1.7303, - "step": 5933 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979929648921058, - "loss": 1.5734, - "step": 5934 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019799262666576474, - "loss": 1.6803, - "step": 5935 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019799228843942368, - "loss": 1.8389, - "step": 5936 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979919502130826, - "loss": 1.8359, - "step": 5937 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019799161198674154, - "loss": 1.814, - "step": 5938 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019799127376040048, - "loss": 1.6995, - "step": 5939 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019799093553405942, - "loss": 1.7562, - "step": 5940 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019799059730771833, - "loss": 1.6871, - "step": 5941 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019799025908137728, - "loss": 1.9448, - "step": 5942 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019798992085503622, - "loss": 1.5763, - "step": 5943 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019798958262869513, - "loss": 1.6724, - "step": 5944 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019798924440235407, - "loss": 1.6604, - "step": 5945 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197988906176013, - "loss": 1.796, - "step": 5946 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019798856794967193, - "loss": 1.6505, - "step": 5947 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019798822972333087, - "loss": 1.6578, - "step": 5948 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979878914969898, - "loss": 1.5906, - "step": 5949 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019798755327064872, - "loss": 1.9154, - "step": 5950 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019798721504430766, - "loss": 1.7118, - "step": 5951 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979868768179666, - "loss": 1.543, - "step": 5952 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019798653859162552, - "loss": 1.7476, - "step": 5953 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019798620036528446, - "loss": 1.7395, - "step": 5954 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979858621389434, - "loss": 1.7055, - "step": 5955 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019798552391260232, - "loss": 1.6762, - "step": 5956 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019798518568626126, - "loss": 1.7958, - "step": 5957 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979848474599202, - "loss": 1.8389, - "step": 5958 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979845092335791, - "loss": 1.5328, - "step": 5959 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019798417100723805, - "loss": 1.6935, - "step": 5960 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197983832780897, - "loss": 1.8023, - "step": 5961 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979834945545559, - "loss": 1.713, - "step": 5962 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019798315632821485, - "loss": 1.6855, - "step": 5963 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979828181018738, - "loss": 1.6806, - "step": 5964 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979824798755327, - "loss": 1.6965, - "step": 5965 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019798214164919165, - "loss": 1.7481, - "step": 5966 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979818034228506, - "loss": 1.6796, - "step": 5967 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979814651965095, - "loss": 1.7406, - "step": 5968 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019798112697016844, - "loss": 1.6921, - "step": 5969 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019798078874382738, - "loss": 1.6969, - "step": 5970 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979804505174863, - "loss": 1.776, - "step": 5971 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019798011229114524, - "loss": 1.6793, - "step": 5972 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019797977406480418, - "loss": 1.743, - "step": 5973 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979794358384631, - "loss": 1.7506, - "step": 5974 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019797909761212204, - "loss": 1.7946, - "step": 5975 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019797875938578098, - "loss": 1.6165, - "step": 5976 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979784211594399, - "loss": 1.6849, - "step": 5977 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019797808293309883, - "loss": 1.686, - "step": 5978 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019797774470675777, - "loss": 1.5933, - "step": 5979 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979774064804167, - "loss": 1.6796, - "step": 5980 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019797706825407563, - "loss": 1.7009, - "step": 5981 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019797673002773457, - "loss": 1.8282, - "step": 5982 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979763918013935, - "loss": 1.6908, - "step": 5983 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019797605357505245, - "loss": 1.6298, - "step": 5984 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019797571534871137, - "loss": 1.7932, - "step": 5985 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979753771223703, - "loss": 1.7085, - "step": 5986 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019797503889602925, - "loss": 1.6904, - "step": 5987 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019797470066968816, - "loss": 1.6354, - "step": 5988 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979743624433471, - "loss": 1.5868, - "step": 5989 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019797402421700604, - "loss": 1.7012, - "step": 5990 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019797368599066496, - "loss": 1.683, - "step": 5991 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979733477643239, - "loss": 1.7702, - "step": 5992 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019797300953798284, - "loss": 1.6562, - "step": 5993 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019797267131164175, - "loss": 1.7278, - "step": 5994 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979723330853007, - "loss": 1.7047, - "step": 5995 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019797199485895964, - "loss": 1.8513, - "step": 5996 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019797165663261855, - "loss": 1.801, - "step": 5997 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979713184062775, - "loss": 1.7478, - "step": 5998 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019797098017993643, - "loss": 1.5967, - "step": 5999 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019797064195359535, - "loss": 1.7843, - "step": 6000 - }, - { - "epoch": 0.02, - "eval_loss": 1.4240534855769231, - "mrc_em": 7.0, - "mrc_f1": 19.21428571428572, - "nli_acc": 0.23, - "step": 6000, - "sts_f1": 0.4444444444444444, - "ynat_f1": 0.5705582232893157 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979703037272543, - "loss": 1.6539, - "step": 6001 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019796996550091323, - "loss": 1.8277, - "step": 6002 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019796962727457214, - "loss": 1.8868, - "step": 6003 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019796928904823108, - "loss": 1.841, - "step": 6004 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019796895082189003, - "loss": 1.573, - "step": 6005 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019796861259554894, - "loss": 1.8176, - "step": 6006 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019796827436920788, - "loss": 1.7252, - "step": 6007 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019796793614286682, - "loss": 1.7268, - "step": 6008 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019796759791652574, - "loss": 1.6694, - "step": 6009 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019796725969018468, - "loss": 1.509, - "step": 6010 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019796692146384362, - "loss": 1.6224, - "step": 6011 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019796658323750253, - "loss": 1.6492, - "step": 6012 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019796624501116147, - "loss": 1.8179, - "step": 6013 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019796590678482041, - "loss": 1.6867, - "step": 6014 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019796556855847933, - "loss": 1.8327, - "step": 6015 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019796523033213827, - "loss": 1.588, - "step": 6016 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979648921057972, - "loss": 1.5239, - "step": 6017 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019796455387945613, - "loss": 1.7598, - "step": 6018 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019796421565311507, - "loss": 1.7199, - "step": 6019 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197963877426774, - "loss": 1.8149, - "step": 6020 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019796353920043292, - "loss": 1.633, - "step": 6021 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019796320097409186, - "loss": 1.5339, - "step": 6022 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979628627477508, - "loss": 1.9053, - "step": 6023 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019796252452140974, - "loss": 1.6463, - "step": 6024 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019796218629506866, - "loss": 1.5575, - "step": 6025 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979618480687276, - "loss": 1.7041, - "step": 6026 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019796150984238654, - "loss": 1.709, - "step": 6027 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019796117161604546, - "loss": 1.7235, - "step": 6028 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979608333897044, - "loss": 1.6675, - "step": 6029 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019796049516336334, - "loss": 1.6891, - "step": 6030 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019796015693702225, - "loss": 1.6846, - "step": 6031 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979598187106812, - "loss": 1.6555, - "step": 6032 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019795948048434013, - "loss": 1.8244, - "step": 6033 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019795914225799905, - "loss": 1.804, - "step": 6034 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197958804031658, - "loss": 1.7717, - "step": 6035 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019795846580531693, - "loss": 1.8218, - "step": 6036 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019795812757897584, - "loss": 1.7072, - "step": 6037 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979577893526348, - "loss": 1.7444, - "step": 6038 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019795745112629373, - "loss": 1.8906, - "step": 6039 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019795711289995267, - "loss": 1.6039, - "step": 6040 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979567746736116, - "loss": 1.5795, - "step": 6041 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019795643644727052, - "loss": 1.6582, - "step": 6042 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019795609822092946, - "loss": 1.7084, - "step": 6043 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979557599945884, - "loss": 1.5891, - "step": 6044 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019795542176824732, - "loss": 1.5275, - "step": 6045 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019795508354190626, - "loss": 1.6872, - "step": 6046 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979547453155652, - "loss": 1.6191, - "step": 6047 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019795440708922412, - "loss": 1.7791, - "step": 6048 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019795406886288306, - "loss": 1.8667, - "step": 6049 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197953730636542, - "loss": 1.6513, - "step": 6050 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979533924102009, - "loss": 1.9015, - "step": 6051 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019795305418385985, - "loss": 1.6985, - "step": 6052 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979527159575188, - "loss": 1.7362, - "step": 6053 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979523777311777, - "loss": 1.6467, - "step": 6054 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019795203950483665, - "loss": 1.6176, - "step": 6055 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979517012784956, - "loss": 1.7548, - "step": 6056 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979513630521545, - "loss": 1.6521, - "step": 6057 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019795102482581345, - "loss": 1.9736, - "step": 6058 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979506865994724, - "loss": 1.6557, - "step": 6059 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979503483731313, - "loss": 1.6467, - "step": 6060 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019795001014679024, - "loss": 1.6315, - "step": 6061 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019794967192044918, - "loss": 1.7381, - "step": 6062 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979493336941081, - "loss": 1.682, - "step": 6063 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019794899546776704, - "loss": 1.5701, - "step": 6064 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019794865724142598, - "loss": 1.6703, - "step": 6065 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979483190150849, - "loss": 1.6467, - "step": 6066 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019794798078874383, - "loss": 1.7399, - "step": 6067 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019794764256240278, - "loss": 1.6096, - "step": 6068 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979473043360617, - "loss": 1.6415, - "step": 6069 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019794696610972063, - "loss": 1.753, - "step": 6070 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019794662788337957, - "loss": 1.6408, - "step": 6071 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019794628965703849, - "loss": 1.5982, - "step": 6072 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019794595143069743, - "loss": 1.6378, - "step": 6073 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019794561320435637, - "loss": 1.6371, - "step": 6074 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019794527497801528, - "loss": 1.7119, - "step": 6075 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019794493675167422, - "loss": 1.7388, - "step": 6076 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019794459852533316, - "loss": 1.6968, - "step": 6077 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019794426029899208, - "loss": 1.6908, - "step": 6078 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019794392207265102, - "loss": 1.5421, - "step": 6079 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019794358384630996, - "loss": 1.6286, - "step": 6080 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019794324561996887, - "loss": 1.7432, - "step": 6081 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019794290739362782, - "loss": 1.7143, - "step": 6082 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019794256916728676, - "loss": 1.7526, - "step": 6083 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019794223094094567, - "loss": 1.7311, - "step": 6084 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979418927146046, - "loss": 1.6545, - "step": 6085 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019794155448826355, - "loss": 1.6973, - "step": 6086 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019794121626192247, - "loss": 1.7368, - "step": 6087 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979408780355814, - "loss": 1.7022, - "step": 6088 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019794053980924035, - "loss": 1.6565, - "step": 6089 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019794020158289926, - "loss": 1.898, - "step": 6090 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979398633565582, - "loss": 1.6894, - "step": 6091 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019793952513021715, - "loss": 1.6617, - "step": 6092 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019793918690387606, - "loss": 1.6767, - "step": 6093 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197938848677535, - "loss": 1.6512, - "step": 6094 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019793851045119397, - "loss": 1.6905, - "step": 6095 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019793817222485288, - "loss": 1.841, - "step": 6096 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019793783399851182, - "loss": 1.6692, - "step": 6097 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019793749577217077, - "loss": 1.7372, - "step": 6098 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019793715754582968, - "loss": 1.6214, - "step": 6099 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019793681931948862, - "loss": 1.4981, - "step": 6100 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019793648109314756, - "loss": 1.7565, - "step": 6101 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019793614286680648, - "loss": 1.6483, - "step": 6102 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019793580464046542, - "loss": 1.7448, - "step": 6103 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019793546641412436, - "loss": 1.7067, - "step": 6104 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019793512818778327, - "loss": 1.7509, - "step": 6105 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979347899614422, - "loss": 1.7944, - "step": 6106 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019793445173510115, - "loss": 1.7525, - "step": 6107 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019793411350876007, - "loss": 1.6733, - "step": 6108 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197933775282419, - "loss": 1.8251, - "step": 6109 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019793343705607795, - "loss": 1.5848, - "step": 6110 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019793309882973686, - "loss": 1.7439, - "step": 6111 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979327606033958, - "loss": 1.7016, - "step": 6112 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019793242237705475, - "loss": 1.7345, - "step": 6113 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019793208415071366, - "loss": 1.5855, - "step": 6114 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979317459243726, - "loss": 1.6703, - "step": 6115 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019793140769803154, - "loss": 1.8122, - "step": 6116 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019793106947169046, - "loss": 1.6989, - "step": 6117 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979307312453494, - "loss": 1.6495, - "step": 6118 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019793039301900834, - "loss": 1.6048, - "step": 6119 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019793005479266725, - "loss": 1.6703, - "step": 6120 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979297165663262, - "loss": 1.7467, - "step": 6121 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019792937833998514, - "loss": 1.625, - "step": 6122 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019792904011364405, - "loss": 1.8648, - "step": 6123 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197928701887303, - "loss": 1.5364, - "step": 6124 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019792836366096193, - "loss": 1.6266, - "step": 6125 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019792802543462085, - "loss": 1.8159, - "step": 6126 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979276872082798, - "loss": 1.8034, - "step": 6127 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019792734898193873, - "loss": 1.8634, - "step": 6128 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019792701075559764, - "loss": 1.6416, - "step": 6129 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019792667252925658, - "loss": 1.7202, - "step": 6130 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019792633430291553, - "loss": 1.5846, - "step": 6131 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019792599607657444, - "loss": 1.4991, - "step": 6132 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019792565785023338, - "loss": 1.5822, - "step": 6133 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019792531962389232, - "loss": 1.731, - "step": 6134 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019792498139755124, - "loss": 1.5859, - "step": 6135 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019792464317121018, - "loss": 1.6811, - "step": 6136 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019792430494486912, - "loss": 1.7227, - "step": 6137 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019792396671852803, - "loss": 1.5661, - "step": 6138 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019792362849218697, - "loss": 1.649, - "step": 6139 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019792329026584591, - "loss": 1.7245, - "step": 6140 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019792295203950483, - "loss": 1.7467, - "step": 6141 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019792261381316377, - "loss": 1.5744, - "step": 6142 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979222755868227, - "loss": 1.8216, - "step": 6143 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019792193736048162, - "loss": 1.7277, - "step": 6144 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019792159913414057, - "loss": 1.5844, - "step": 6145 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979212609077995, - "loss": 1.7285, - "step": 6146 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019792092268145842, - "loss": 1.7349, - "step": 6147 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019792058445511736, - "loss": 1.5755, - "step": 6148 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979202462287763, - "loss": 1.7871, - "step": 6149 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019791990800243522, - "loss": 1.5851, - "step": 6150 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019791956977609419, - "loss": 1.8151, - "step": 6151 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019791923154975313, - "loss": 1.705, - "step": 6152 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019791889332341204, - "loss": 1.8152, - "step": 6153 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019791855509707098, - "loss": 1.58, - "step": 6154 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019791821687072992, - "loss": 1.7465, - "step": 6155 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019791787864438884, - "loss": 1.7525, - "step": 6156 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019791754041804778, - "loss": 1.586, - "step": 6157 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019791720219170672, - "loss": 1.692, - "step": 6158 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019791686396536563, - "loss": 1.6377, - "step": 6159 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019791652573902457, - "loss": 1.6604, - "step": 6160 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019791618751268352, - "loss": 1.7557, - "step": 6161 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019791584928634243, - "loss": 1.7253, - "step": 6162 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019791551106000137, - "loss": 1.6129, - "step": 6163 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979151728336603, - "loss": 1.5799, - "step": 6164 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019791483460731923, - "loss": 1.5192, - "step": 6165 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019791449638097817, - "loss": 1.8443, - "step": 6166 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979141581546371, - "loss": 1.6928, - "step": 6167 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019791381992829602, - "loss": 1.6977, - "step": 6168 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019791348170195496, - "loss": 1.7484, - "step": 6169 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979131434756139, - "loss": 1.671, - "step": 6170 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019791280524927282, - "loss": 1.4557, - "step": 6171 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019791246702293176, - "loss": 1.6652, - "step": 6172 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979121287965907, - "loss": 1.6964, - "step": 6173 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019791179057024961, - "loss": 1.6407, - "step": 6174 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019791145234390856, - "loss": 1.7701, - "step": 6175 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979111141175675, - "loss": 1.6303, - "step": 6176 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979107758912264, - "loss": 1.683, - "step": 6177 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019791043766488535, - "loss": 1.6571, - "step": 6178 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979100994385443, - "loss": 1.7858, - "step": 6179 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979097612122032, - "loss": 1.8626, - "step": 6180 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019790942298586215, - "loss": 1.7357, - "step": 6181 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979090847595211, - "loss": 1.556, - "step": 6182 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019790874653318, - "loss": 1.6136, - "step": 6183 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019790840830683894, - "loss": 1.7327, - "step": 6184 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019790807008049789, - "loss": 1.8259, - "step": 6185 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979077318541568, - "loss": 1.5903, - "step": 6186 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019790739362781574, - "loss": 1.6317, - "step": 6187 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019790705540147468, - "loss": 1.6176, - "step": 6188 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979067171751336, - "loss": 1.6795, - "step": 6189 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019790637894879254, - "loss": 1.6145, - "step": 6190 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019790604072245148, - "loss": 1.7442, - "step": 6191 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979057024961104, - "loss": 1.6818, - "step": 6192 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019790536426976933, - "loss": 1.8384, - "step": 6193 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019790502604342827, - "loss": 1.7411, - "step": 6194 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979046878170872, - "loss": 1.674, - "step": 6195 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019790434959074613, - "loss": 1.7642, - "step": 6196 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019790401136440507, - "loss": 1.6188, - "step": 6197 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019790367313806399, - "loss": 1.6859, - "step": 6198 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019790333491172293, - "loss": 1.5272, - "step": 6199 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019790299668538187, - "loss": 1.6448, - "step": 6200 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019790265845904078, - "loss": 1.6829, - "step": 6201 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019790232023269972, - "loss": 1.9127, - "step": 6202 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019790198200635866, - "loss": 1.7837, - "step": 6203 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019790164378001758, - "loss": 1.6057, - "step": 6204 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019790130555367652, - "loss": 1.7945, - "step": 6205 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019790096732733546, - "loss": 1.5734, - "step": 6206 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001979006291009944, - "loss": 1.7516, - "step": 6207 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019790029087465334, - "loss": 1.7694, - "step": 6208 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019789995264831228, - "loss": 1.7525, - "step": 6209 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978996144219712, - "loss": 1.6069, - "step": 6210 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019789927619563014, - "loss": 1.6766, - "step": 6211 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019789893796928908, - "loss": 1.9386, - "step": 6212 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197898599742948, - "loss": 1.7066, - "step": 6213 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019789826151660693, - "loss": 1.7371, - "step": 6214 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019789792329026588, - "loss": 1.5859, - "step": 6215 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978975850639248, - "loss": 1.7376, - "step": 6216 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019789724683758373, - "loss": 1.6513, - "step": 6217 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019789690861124267, - "loss": 1.7456, - "step": 6218 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978965703849016, - "loss": 1.6514, - "step": 6219 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019789623215856053, - "loss": 1.613, - "step": 6220 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019789589393221947, - "loss": 1.7938, - "step": 6221 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019789555570587838, - "loss": 1.6447, - "step": 6222 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019789521747953732, - "loss": 1.7006, - "step": 6223 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019789487925319627, - "loss": 1.737, - "step": 6224 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019789454102685518, - "loss": 1.5999, - "step": 6225 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019789420280051412, - "loss": 1.5518, - "step": 6226 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019789386457417306, - "loss": 1.6964, - "step": 6227 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019789352634783198, - "loss": 1.8633, - "step": 6228 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019789318812149092, - "loss": 1.677, - "step": 6229 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019789284989514986, - "loss": 1.7377, - "step": 6230 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019789251166880877, - "loss": 1.7142, - "step": 6231 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978921734424677, - "loss": 1.6625, - "step": 6232 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019789183521612665, - "loss": 1.6376, - "step": 6233 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019789149698978557, - "loss": 1.7649, - "step": 6234 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978911587634445, - "loss": 1.6869, - "step": 6235 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019789082053710345, - "loss": 1.8478, - "step": 6236 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019789048231076236, - "loss": 1.7809, - "step": 6237 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978901440844213, - "loss": 1.6696, - "step": 6238 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019788980585808025, - "loss": 1.6503, - "step": 6239 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019788946763173916, - "loss": 1.7405, - "step": 6240 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978891294053981, - "loss": 1.7534, - "step": 6241 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019788879117905704, - "loss": 1.7029, - "step": 6242 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019788845295271596, - "loss": 1.6581, - "step": 6243 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978881147263749, - "loss": 1.5714, - "step": 6244 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019788777650003384, - "loss": 1.7302, - "step": 6245 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019788743827369275, - "loss": 1.7227, - "step": 6246 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978871000473517, - "loss": 1.778, - "step": 6247 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019788676182101064, - "loss": 1.8568, - "step": 6248 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019788642359466955, - "loss": 1.7039, - "step": 6249 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978860853683285, - "loss": 1.6939, - "step": 6250 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019788574714198743, - "loss": 1.6727, - "step": 6251 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019788540891564635, - "loss": 1.8346, - "step": 6252 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978850706893053, - "loss": 1.7796, - "step": 6253 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019788473246296423, - "loss": 1.7721, - "step": 6254 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019788439423662314, - "loss": 1.691, - "step": 6255 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019788405601028208, - "loss": 1.6374, - "step": 6256 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019788371778394102, - "loss": 1.6396, - "step": 6257 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019788337955759994, - "loss": 1.5666, - "step": 6258 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019788304133125888, - "loss": 1.6941, - "step": 6259 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019788270310491782, - "loss": 1.7682, - "step": 6260 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019788236487857673, - "loss": 1.885, - "step": 6261 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019788202665223568, - "loss": 1.5241, - "step": 6262 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019788168842589462, - "loss": 1.6912, - "step": 6263 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019788135019955356, - "loss": 1.7115, - "step": 6264 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978810119732125, - "loss": 1.7803, - "step": 6265 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978806737468714, - "loss": 1.7601, - "step": 6266 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019788033552053035, - "loss": 1.7446, - "step": 6267 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978799972941893, - "loss": 1.6293, - "step": 6268 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978796590678482, - "loss": 1.6368, - "step": 6269 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019787932084150715, - "loss": 1.8176, - "step": 6270 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978789826151661, - "loss": 1.6743, - "step": 6271 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197878644388825, - "loss": 1.7005, - "step": 6272 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019787830616248395, - "loss": 1.6382, - "step": 6273 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978779679361429, - "loss": 1.7729, - "step": 6274 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978776297098018, - "loss": 1.5898, - "step": 6275 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019787729148346074, - "loss": 1.7583, - "step": 6276 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019787695325711968, - "loss": 1.6487, - "step": 6277 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978766150307786, - "loss": 1.6333, - "step": 6278 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019787627680443754, - "loss": 1.5896, - "step": 6279 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019787593857809648, - "loss": 1.7008, - "step": 6280 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978756003517554, - "loss": 1.7229, - "step": 6281 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019787526212541434, - "loss": 1.6135, - "step": 6282 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019787492389907328, - "loss": 1.8369, - "step": 6283 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978745856727322, - "loss": 1.7102, - "step": 6284 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019787424744639113, - "loss": 1.6787, - "step": 6285 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019787390922005007, - "loss": 1.7595, - "step": 6286 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197873570993709, - "loss": 1.6706, - "step": 6287 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019787323276736793, - "loss": 1.7068, - "step": 6288 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019787289454102687, - "loss": 1.686, - "step": 6289 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019787255631468578, - "loss": 1.7755, - "step": 6290 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019787221808834473, - "loss": 1.7049, - "step": 6291 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019787187986200367, - "loss": 1.4658, - "step": 6292 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978715416356626, - "loss": 1.5859, - "step": 6293 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019787120340932152, - "loss": 1.6739, - "step": 6294 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019787086518298046, - "loss": 1.5825, - "step": 6295 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978705269566394, - "loss": 1.6373, - "step": 6296 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019787018873029832, - "loss": 1.8245, - "step": 6297 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019786985050395726, - "loss": 1.7008, - "step": 6298 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978695122776162, - "loss": 1.7464, - "step": 6299 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019786917405127511, - "loss": 1.6606, - "step": 6300 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019786883582493406, - "loss": 1.6297, - "step": 6301 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197868497598593, - "loss": 1.698, - "step": 6302 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978681593722519, - "loss": 1.838, - "step": 6303 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019786782114591085, - "loss": 1.7522, - "step": 6304 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978674829195698, - "loss": 1.6098, - "step": 6305 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978671446932287, - "loss": 1.6557, - "step": 6306 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019786680646688765, - "loss": 1.6578, - "step": 6307 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978664682405466, - "loss": 1.8121, - "step": 6308 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978661300142055, - "loss": 1.7917, - "step": 6309 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019786579178786444, - "loss": 1.6403, - "step": 6310 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019786545356152339, - "loss": 1.702, - "step": 6311 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978651153351823, - "loss": 1.5862, - "step": 6312 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019786477710884124, - "loss": 1.6051, - "step": 6313 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019786443888250018, - "loss": 1.7612, - "step": 6314 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978641006561591, - "loss": 2.0284, - "step": 6315 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019786376242981804, - "loss": 1.6472, - "step": 6316 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019786342420347698, - "loss": 1.5835, - "step": 6317 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978630859771359, - "loss": 1.6505, - "step": 6318 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019786274775079486, - "loss": 1.6533, - "step": 6319 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019786240952445377, - "loss": 1.6116, - "step": 6320 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019786207129811272, - "loss": 1.6486, - "step": 6321 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019786173307177166, - "loss": 1.7969, - "step": 6322 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019786139484543057, - "loss": 1.7448, - "step": 6323 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978610566190895, - "loss": 1.5857, - "step": 6324 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019786071839274845, - "loss": 1.7047, - "step": 6325 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019786038016640737, - "loss": 1.6736, - "step": 6326 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978600419400663, - "loss": 1.6475, - "step": 6327 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019785970371372525, - "loss": 1.6478, - "step": 6328 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019785936548738416, - "loss": 1.8385, - "step": 6329 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978590272610431, - "loss": 1.7956, - "step": 6330 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019785868903470205, - "loss": 1.6357, - "step": 6331 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019785835080836096, - "loss": 1.6439, - "step": 6332 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978580125820199, - "loss": 1.8347, - "step": 6333 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019785767435567884, - "loss": 1.6713, - "step": 6334 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019785733612933776, - "loss": 1.7271, - "step": 6335 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978569979029967, - "loss": 1.7158, - "step": 6336 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019785665967665564, - "loss": 1.5442, - "step": 6337 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019785632145031455, - "loss": 1.5938, - "step": 6338 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978559832239735, - "loss": 1.6926, - "step": 6339 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019785564499763243, - "loss": 1.6876, - "step": 6340 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019785530677129135, - "loss": 1.6837, - "step": 6341 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978549685449503, - "loss": 1.758, - "step": 6342 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019785463031860923, - "loss": 1.6366, - "step": 6343 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019785429209226814, - "loss": 1.6977, - "step": 6344 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019785395386592709, - "loss": 1.8122, - "step": 6345 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019785361563958603, - "loss": 1.6188, - "step": 6346 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019785327741324494, - "loss": 1.6396, - "step": 6347 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019785293918690388, - "loss": 1.6529, - "step": 6348 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019785260096056282, - "loss": 1.7034, - "step": 6349 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019785226273422174, - "loss": 1.546, - "step": 6350 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019785192450788068, - "loss": 1.6117, - "step": 6351 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019785158628153962, - "loss": 1.6912, - "step": 6352 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019785124805519853, - "loss": 1.6986, - "step": 6353 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019785090982885747, - "loss": 1.5994, - "step": 6354 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019785057160251642, - "loss": 1.8057, - "step": 6355 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019785023337617533, - "loss": 1.7152, - "step": 6356 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019784989514983427, - "loss": 1.7194, - "step": 6357 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978495569234932, - "loss": 1.8006, - "step": 6358 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019784921869715213, - "loss": 1.6837, - "step": 6359 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019784888047081107, - "loss": 1.5432, - "step": 6360 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019784854224447, - "loss": 1.6101, - "step": 6361 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019784820401812892, - "loss": 1.7394, - "step": 6362 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019784786579178786, - "loss": 1.8855, - "step": 6363 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978475275654468, - "loss": 1.762, - "step": 6364 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019784718933910572, - "loss": 1.7592, - "step": 6365 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019784685111276466, - "loss": 1.7292, - "step": 6366 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978465128864236, - "loss": 1.5407, - "step": 6367 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019784617466008252, - "loss": 1.6522, - "step": 6368 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019784583643374146, - "loss": 1.799, - "step": 6369 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978454982074004, - "loss": 1.6851, - "step": 6370 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978451599810593, - "loss": 1.7496, - "step": 6371 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019784482175471825, - "loss": 1.6974, - "step": 6372 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978444835283772, - "loss": 1.5807, - "step": 6373 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978441453020361, - "loss": 1.9082, - "step": 6374 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019784380707569508, - "loss": 1.8704, - "step": 6375 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019784346884935402, - "loss": 1.6658, - "step": 6376 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019784313062301293, - "loss": 1.733, - "step": 6377 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019784279239667187, - "loss": 1.5918, - "step": 6378 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978424541703308, - "loss": 1.7653, - "step": 6379 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019784211594398973, - "loss": 1.6026, - "step": 6380 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019784177771764867, - "loss": 1.755, - "step": 6381 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978414394913076, - "loss": 1.7184, - "step": 6382 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019784110126496652, - "loss": 1.585, - "step": 6383 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019784076303862547, - "loss": 1.5853, - "step": 6384 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978404248122844, - "loss": 1.7872, - "step": 6385 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019784008658594332, - "loss": 1.6626, - "step": 6386 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019783974835960226, - "loss": 1.6656, - "step": 6387 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978394101332612, - "loss": 1.786, - "step": 6388 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019783907190692012, - "loss": 1.835, - "step": 6389 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019783873368057906, - "loss": 1.8455, - "step": 6390 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197838395454238, - "loss": 1.8048, - "step": 6391 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978380572278969, - "loss": 1.5736, - "step": 6392 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019783771900155585, - "loss": 1.7539, - "step": 6393 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978373807752148, - "loss": 1.6392, - "step": 6394 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978370425488737, - "loss": 1.8891, - "step": 6395 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019783670432253265, - "loss": 1.658, - "step": 6396 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978363660961916, - "loss": 1.7871, - "step": 6397 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978360278698505, - "loss": 1.8341, - "step": 6398 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019783568964350945, - "loss": 1.6497, - "step": 6399 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978353514171684, - "loss": 1.7582, - "step": 6400 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978350131908273, - "loss": 1.8277, - "step": 6401 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019783467496448624, - "loss": 1.7428, - "step": 6402 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019783433673814518, - "loss": 1.6963, - "step": 6403 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978339985118041, - "loss": 1.533, - "step": 6404 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019783366028546304, - "loss": 1.5851, - "step": 6405 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019783332205912198, - "loss": 1.6426, - "step": 6406 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978329838327809, - "loss": 1.6916, - "step": 6407 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019783264560643984, - "loss": 1.7126, - "step": 6408 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019783230738009878, - "loss": 1.6968, - "step": 6409 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978319691537577, - "loss": 1.7157, - "step": 6410 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019783163092741663, - "loss": 1.5958, - "step": 6411 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019783129270107557, - "loss": 1.7348, - "step": 6412 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978309544747345, - "loss": 1.6612, - "step": 6413 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019783061624839343, - "loss": 1.745, - "step": 6414 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019783027802205237, - "loss": 1.6993, - "step": 6415 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019782993979571128, - "loss": 1.6554, - "step": 6416 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019782960156937022, - "loss": 1.5346, - "step": 6417 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019782926334302917, - "loss": 1.6718, - "step": 6418 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019782892511668808, - "loss": 1.7186, - "step": 6419 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019782858689034702, - "loss": 1.7597, - "step": 6420 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019782824866400596, - "loss": 1.6113, - "step": 6421 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019782791043766488, - "loss": 1.7446, - "step": 6422 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019782757221132382, - "loss": 1.6346, - "step": 6423 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019782723398498276, - "loss": 1.7498, - "step": 6424 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019782689575864167, - "loss": 1.6933, - "step": 6425 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978265575323006, - "loss": 1.8744, - "step": 6426 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019782621930595955, - "loss": 1.8051, - "step": 6427 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019782588107961847, - "loss": 1.679, - "step": 6428 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978255428532774, - "loss": 1.8543, - "step": 6429 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019782520462693635, - "loss": 1.8717, - "step": 6430 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019782486640059526, - "loss": 1.7161, - "step": 6431 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019782452817425423, - "loss": 1.6026, - "step": 6432 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019782418994791317, - "loss": 1.5747, - "step": 6433 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978238517215721, - "loss": 1.6917, - "step": 6434 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019782351349523103, - "loss": 1.6862, - "step": 6435 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019782317526888997, - "loss": 1.6923, - "step": 6436 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019782283704254888, - "loss": 1.7738, - "step": 6437 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019782249881620783, - "loss": 1.7048, - "step": 6438 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019782216058986677, - "loss": 1.7959, - "step": 6439 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019782182236352568, - "loss": 1.5998, - "step": 6440 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019782148413718462, - "loss": 1.6686, - "step": 6441 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019782114591084356, - "loss": 1.9402, - "step": 6442 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019782080768450248, - "loss": 1.6657, - "step": 6443 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019782046945816142, - "loss": 1.7294, - "step": 6444 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019782013123182036, - "loss": 1.5912, - "step": 6445 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019781979300547927, - "loss": 1.7446, - "step": 6446 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019781945477913821, - "loss": 1.7236, - "step": 6447 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019781911655279716, - "loss": 1.5417, - "step": 6448 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019781877832645607, - "loss": 1.5794, - "step": 6449 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197818440100115, - "loss": 1.6408, - "step": 6450 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019781810187377395, - "loss": 1.6441, - "step": 6451 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019781776364743287, - "loss": 1.6143, - "step": 6452 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978174254210918, - "loss": 1.747, - "step": 6453 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019781708719475075, - "loss": 1.5821, - "step": 6454 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019781674896840966, - "loss": 1.5623, - "step": 6455 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978164107420686, - "loss": 1.6528, - "step": 6456 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019781607251572754, - "loss": 1.7821, - "step": 6457 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019781573428938646, - "loss": 1.7374, - "step": 6458 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978153960630454, - "loss": 1.586, - "step": 6459 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019781505783670434, - "loss": 1.7217, - "step": 6460 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019781471961036326, - "loss": 1.7831, - "step": 6461 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978143813840222, - "loss": 1.6224, - "step": 6462 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019781404315768114, - "loss": 1.6793, - "step": 6463 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019781370493134005, - "loss": 1.6089, - "step": 6464 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197813366704999, - "loss": 1.687, - "step": 6465 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019781302847865793, - "loss": 1.8784, - "step": 6466 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019781269025231685, - "loss": 1.8873, - "step": 6467 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978123520259758, - "loss": 1.6981, - "step": 6468 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019781201379963473, - "loss": 1.7202, - "step": 6469 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019781167557329364, - "loss": 1.6915, - "step": 6470 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019781133734695259, - "loss": 1.7258, - "step": 6471 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019781099912061153, - "loss": 1.6467, - "step": 6472 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019781066089427044, - "loss": 1.7917, - "step": 6473 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019781032266792938, - "loss": 1.7303, - "step": 6474 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019780998444158832, - "loss": 1.723, - "step": 6475 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019780964621524724, - "loss": 1.7591, - "step": 6476 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019780930798890618, - "loss": 1.7066, - "step": 6477 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019780896976256512, - "loss": 1.6625, - "step": 6478 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019780863153622403, - "loss": 1.8372, - "step": 6479 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019780829330988297, - "loss": 1.662, - "step": 6480 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019780795508354192, - "loss": 1.7245, - "step": 6481 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019780761685720083, - "loss": 1.6653, - "step": 6482 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019780727863085977, - "loss": 1.68, - "step": 6483 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978069404045187, - "loss": 1.5532, - "step": 6484 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019780660217817763, - "loss": 1.679, - "step": 6485 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019780626395183657, - "loss": 1.9261, - "step": 6486 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978059257254955, - "loss": 1.7302, - "step": 6487 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019780558749915445, - "loss": 1.6286, - "step": 6488 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978052492728134, - "loss": 1.7184, - "step": 6489 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019780491104647233, - "loss": 1.6208, - "step": 6490 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019780457282013125, - "loss": 1.6594, - "step": 6491 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001978042345937902, - "loss": 1.6587, - "step": 6492 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019780389636744913, - "loss": 1.7401, - "step": 6493 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019780355814110804, - "loss": 1.7745, - "step": 6494 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019780321991476698, - "loss": 1.7729, - "step": 6495 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019780288168842592, - "loss": 1.696, - "step": 6496 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019780254346208484, - "loss": 1.4971, - "step": 6497 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019780220523574378, - "loss": 1.7545, - "step": 6498 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019780186700940272, - "loss": 1.6531, - "step": 6499 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019780152878306163, - "loss": 1.8231, - "step": 6500 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019780119055672058, - "loss": 1.6883, - "step": 6501 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019780085233037952, - "loss": 1.8971, - "step": 6502 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019780051410403843, - "loss": 1.6762, - "step": 6503 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019780017587769737, - "loss": 1.6921, - "step": 6504 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977998376513563, - "loss": 1.6325, - "step": 6505 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019779949942501523, - "loss": 1.5097, - "step": 6506 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019779916119867417, - "loss": 1.7451, - "step": 6507 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977988229723331, - "loss": 1.6076, - "step": 6508 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019779848474599202, - "loss": 1.6918, - "step": 6509 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019779814651965096, - "loss": 1.6499, - "step": 6510 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977978082933099, - "loss": 1.7209, - "step": 6511 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019779747006696882, - "loss": 1.5245, - "step": 6512 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019779713184062776, - "loss": 1.6734, - "step": 6513 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977967936142867, - "loss": 1.4915, - "step": 6514 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019779645538794562, - "loss": 1.6564, - "step": 6515 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019779611716160456, - "loss": 1.6831, - "step": 6516 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977957789352635, - "loss": 1.4993, - "step": 6517 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977954407089224, - "loss": 1.7913, - "step": 6518 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019779510248258135, - "loss": 1.6491, - "step": 6519 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977947642562403, - "loss": 1.6918, - "step": 6520 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977944260298992, - "loss": 1.7633, - "step": 6521 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019779408780355815, - "loss": 1.6595, - "step": 6522 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977937495772171, - "loss": 1.6816, - "step": 6523 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197793411350876, - "loss": 1.7649, - "step": 6524 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019779307312453495, - "loss": 1.7242, - "step": 6525 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977927348981939, - "loss": 1.6631, - "step": 6526 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977923966718528, - "loss": 1.7068, - "step": 6527 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019779205844551174, - "loss": 1.7894, - "step": 6528 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019779172021917068, - "loss": 1.5336, - "step": 6529 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977913819928296, - "loss": 1.6429, - "step": 6530 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019779104376648854, - "loss": 1.749, - "step": 6531 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019779070554014748, - "loss": 1.6799, - "step": 6532 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977903673138064, - "loss": 1.7924, - "step": 6533 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019779002908746533, - "loss": 1.5969, - "step": 6534 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019778969086112428, - "loss": 1.6968, - "step": 6535 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977893526347832, - "loss": 1.7425, - "step": 6536 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019778901440844213, - "loss": 1.4927, - "step": 6537 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019778867618210107, - "loss": 1.7815, - "step": 6538 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019778833795575999, - "loss": 1.6728, - "step": 6539 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019778799972941893, - "loss": 1.7103, - "step": 6540 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019778766150307787, - "loss": 1.5659, - "step": 6541 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019778732327673678, - "loss": 1.7632, - "step": 6542 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019778698505039572, - "loss": 1.7063, - "step": 6543 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019778664682405466, - "loss": 1.6164, - "step": 6544 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977863085977136, - "loss": 1.6429, - "step": 6545 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019778597037137255, - "loss": 1.5122, - "step": 6546 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019778563214503146, - "loss": 1.6776, - "step": 6547 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977852939186904, - "loss": 1.6478, - "step": 6548 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019778495569234934, - "loss": 1.7517, - "step": 6549 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019778461746600826, - "loss": 1.6925, - "step": 6550 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977842792396672, - "loss": 1.6948, - "step": 6551 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019778394101332614, - "loss": 1.7665, - "step": 6552 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019778360278698505, - "loss": 1.7805, - "step": 6553 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197783264560644, - "loss": 1.7043, - "step": 6554 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019778292633430294, - "loss": 1.6762, - "step": 6555 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019778258810796185, - "loss": 1.5779, - "step": 6556 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977822498816208, - "loss": 1.6799, - "step": 6557 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019778191165527973, - "loss": 1.7525, - "step": 6558 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019778157342893867, - "loss": 1.6629, - "step": 6559 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977812352025976, - "loss": 1.6881, - "step": 6560 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019778089697625653, - "loss": 1.648, - "step": 6561 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019778055874991547, - "loss": 1.5172, - "step": 6562 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019778022052357438, - "loss": 1.5851, - "step": 6563 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019777988229723333, - "loss": 1.8652, - "step": 6564 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019777954407089227, - "loss": 1.6675, - "step": 6565 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019777920584455118, - "loss": 1.7301, - "step": 6566 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019777886761821012, - "loss": 1.6532, - "step": 6567 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019777852939186906, - "loss": 1.6508, - "step": 6568 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019777819116552798, - "loss": 1.6567, - "step": 6569 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019777785293918692, - "loss": 1.6956, - "step": 6570 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019777751471284586, - "loss": 1.7704, - "step": 6571 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019777717648650477, - "loss": 1.6187, - "step": 6572 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019777683826016371, - "loss": 1.5906, - "step": 6573 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019777650003382266, - "loss": 1.6276, - "step": 6574 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019777616180748157, - "loss": 1.7999, - "step": 6575 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977758235811405, - "loss": 1.6113, - "step": 6576 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019777548535479945, - "loss": 1.7127, - "step": 6577 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019777514712845837, - "loss": 1.6054, - "step": 6578 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977748089021173, - "loss": 1.7574, - "step": 6579 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019777447067577625, - "loss": 1.7956, - "step": 6580 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019777413244943516, - "loss": 1.6407, - "step": 6581 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977737942230941, - "loss": 1.5757, - "step": 6582 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019777345599675304, - "loss": 1.6318, - "step": 6583 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019777311777041196, - "loss": 1.7073, - "step": 6584 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977727795440709, - "loss": 1.7032, - "step": 6585 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019777244131772984, - "loss": 1.6312, - "step": 6586 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019777210309138875, - "loss": 1.8496, - "step": 6587 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977717648650477, - "loss": 1.7097, - "step": 6588 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019777142663870664, - "loss": 1.7559, - "step": 6589 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019777108841236555, - "loss": 1.736, - "step": 6590 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977707501860245, - "loss": 1.6021, - "step": 6591 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019777041195968343, - "loss": 1.6515, - "step": 6592 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019777007373334235, - "loss": 1.6595, - "step": 6593 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977697355070013, - "loss": 1.6576, - "step": 6594 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019776939728066023, - "loss": 1.6593, - "step": 6595 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019776905905431914, - "loss": 1.8084, - "step": 6596 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019776872082797808, - "loss": 1.451, - "step": 6597 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019776838260163703, - "loss": 1.5431, - "step": 6598 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019776804437529594, - "loss": 1.7585, - "step": 6599 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977677061489549, - "loss": 1.7751, - "step": 6600 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019776736792261382, - "loss": 1.6218, - "step": 6601 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019776702969627276, - "loss": 1.4858, - "step": 6602 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977666914699317, - "loss": 1.6656, - "step": 6603 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019776635324359062, - "loss": 1.8496, - "step": 6604 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019776601501724956, - "loss": 1.5586, - "step": 6605 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977656767909085, - "loss": 1.7913, - "step": 6606 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019776533856456741, - "loss": 1.6943, - "step": 6607 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019776500033822636, - "loss": 1.8024, - "step": 6608 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977646621118853, - "loss": 1.6653, - "step": 6609 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977643238855442, - "loss": 1.6679, - "step": 6610 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019776398565920315, - "loss": 1.5807, - "step": 6611 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977636474328621, - "loss": 1.8017, - "step": 6612 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197763309206521, - "loss": 1.6494, - "step": 6613 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019776297098017995, - "loss": 1.7343, - "step": 6614 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977626327538389, - "loss": 1.646, - "step": 6615 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977622945274978, - "loss": 1.6202, - "step": 6616 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019776195630115674, - "loss": 1.6713, - "step": 6617 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019776161807481569, - "loss": 1.6343, - "step": 6618 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977612798484746, - "loss": 1.6999, - "step": 6619 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019776094162213354, - "loss": 1.687, - "step": 6620 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019776060339579248, - "loss": 1.6452, - "step": 6621 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977602651694514, - "loss": 1.7978, - "step": 6622 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019775992694311034, - "loss": 1.7181, - "step": 6623 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019775958871676928, - "loss": 1.7213, - "step": 6624 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977592504904282, - "loss": 1.6643, - "step": 6625 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019775891226408713, - "loss": 1.7185, - "step": 6626 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019775857403774607, - "loss": 1.4628, - "step": 6627 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197758235811405, - "loss": 1.6189, - "step": 6628 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019775789758506393, - "loss": 1.5886, - "step": 6629 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019775755935872287, - "loss": 1.7013, - "step": 6630 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019775722113238179, - "loss": 1.5125, - "step": 6631 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019775688290604073, - "loss": 1.5587, - "step": 6632 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019775654467969967, - "loss": 1.6596, - "step": 6633 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019775620645335858, - "loss": 1.6599, - "step": 6634 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019775586822701752, - "loss": 1.536, - "step": 6635 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019775553000067646, - "loss": 1.7301, - "step": 6636 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019775519177433538, - "loss": 1.6915, - "step": 6637 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019775485354799432, - "loss": 1.7073, - "step": 6638 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019775451532165326, - "loss": 1.9412, - "step": 6639 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019775417709531217, - "loss": 1.6975, - "step": 6640 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019775383886897112, - "loss": 1.8138, - "step": 6641 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019775350064263006, - "loss": 1.8487, - "step": 6642 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197753162416289, - "loss": 1.6169, - "step": 6643 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977528241899479, - "loss": 1.695, - "step": 6644 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019775248596360685, - "loss": 1.7347, - "step": 6645 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977521477372658, - "loss": 1.6402, - "step": 6646 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977518095109247, - "loss": 1.7404, - "step": 6647 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019775147128458365, - "loss": 1.5605, - "step": 6648 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977511330582426, - "loss": 1.6321, - "step": 6649 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977507948319015, - "loss": 1.7068, - "step": 6650 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019775045660556045, - "loss": 1.7731, - "step": 6651 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977501183792194, - "loss": 1.864, - "step": 6652 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977497801528783, - "loss": 1.7182, - "step": 6653 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019774944192653724, - "loss": 1.5638, - "step": 6654 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019774910370019618, - "loss": 1.7088, - "step": 6655 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019774876547385512, - "loss": 1.7639, - "step": 6656 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019774842724751407, - "loss": 1.6955, - "step": 6657 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019774808902117298, - "loss": 1.8792, - "step": 6658 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019774775079483192, - "loss": 1.7742, - "step": 6659 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019774741256849086, - "loss": 1.886, - "step": 6660 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019774707434214978, - "loss": 1.5671, - "step": 6661 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019774673611580872, - "loss": 1.6797, - "step": 6662 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019774639788946766, - "loss": 1.6391, - "step": 6663 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019774605966312657, - "loss": 1.6494, - "step": 6664 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977457214367855, - "loss": 1.7333, - "step": 6665 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019774538321044445, - "loss": 1.8186, - "step": 6666 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019774504498410337, - "loss": 1.7579, - "step": 6667 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977447067577623, - "loss": 1.7088, - "step": 6668 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019774436853142125, - "loss": 1.8175, - "step": 6669 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019774403030508016, - "loss": 1.7952, - "step": 6670 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977436920787391, - "loss": 1.7186, - "step": 6671 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019774335385239805, - "loss": 1.5949, - "step": 6672 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019774301562605696, - "loss": 1.8093, - "step": 6673 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977426773997159, - "loss": 1.7757, - "step": 6674 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019774233917337484, - "loss": 1.6498, - "step": 6675 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019774200094703376, - "loss": 1.8714, - "step": 6676 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977416627206927, - "loss": 1.6258, - "step": 6677 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019774132449435164, - "loss": 1.6258, - "step": 6678 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019774098626801055, - "loss": 1.6546, - "step": 6679 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977406480416695, - "loss": 1.6758, - "step": 6680 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019774030981532844, - "loss": 1.6373, - "step": 6681 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019773997158898735, - "loss": 1.8452, - "step": 6682 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977396333626463, - "loss": 1.6469, - "step": 6683 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019773929513630523, - "loss": 1.6036, - "step": 6684 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019773895690996415, - "loss": 1.6418, - "step": 6685 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977386186836231, - "loss": 1.8564, - "step": 6686 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019773828045728203, - "loss": 1.6452, - "step": 6687 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019773794223094094, - "loss": 1.7054, - "step": 6688 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019773760400459988, - "loss": 1.72, - "step": 6689 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019773726577825882, - "loss": 1.6665, - "step": 6690 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019773692755191774, - "loss": 1.674, - "step": 6691 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019773658932557668, - "loss": 1.7351, - "step": 6692 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019773625109923562, - "loss": 1.6299, - "step": 6693 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019773591287289453, - "loss": 1.6283, - "step": 6694 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019773557464655348, - "loss": 1.7359, - "step": 6695 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019773523642021242, - "loss": 1.6811, - "step": 6696 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019773489819387133, - "loss": 1.7595, - "step": 6697 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019773455996753027, - "loss": 1.553, - "step": 6698 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977342217411892, - "loss": 1.7157, - "step": 6699 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019773388351484813, - "loss": 1.6957, - "step": 6700 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019773354528850707, - "loss": 1.5542, - "step": 6701 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197733207062166, - "loss": 1.7156, - "step": 6702 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019773286883582492, - "loss": 1.5328, - "step": 6703 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019773253060948386, - "loss": 1.6853, - "step": 6704 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977321923831428, - "loss": 1.508, - "step": 6705 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019773185415680172, - "loss": 1.7655, - "step": 6706 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019773151593046066, - "loss": 1.6113, - "step": 6707 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977311777041196, - "loss": 1.876, - "step": 6708 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019773083947777852, - "loss": 1.725, - "step": 6709 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019773050125143746, - "loss": 1.6629, - "step": 6710 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977301630250964, - "loss": 1.6881, - "step": 6711 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019772982479875534, - "loss": 1.6211, - "step": 6712 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019772948657241428, - "loss": 1.7069, - "step": 6713 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019772914834607322, - "loss": 1.7086, - "step": 6714 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019772881011973214, - "loss": 1.8083, - "step": 6715 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019772847189339108, - "loss": 1.7431, - "step": 6716 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019772813366705002, - "loss": 1.7464, - "step": 6717 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019772779544070893, - "loss": 1.5456, - "step": 6718 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019772745721436787, - "loss": 1.9582, - "step": 6719 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019772711898802681, - "loss": 1.6196, - "step": 6720 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019772678076168573, - "loss": 1.6167, - "step": 6721 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019772644253534467, - "loss": 1.629, - "step": 6722 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977261043090036, - "loss": 1.6226, - "step": 6723 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019772576608266253, - "loss": 1.6213, - "step": 6724 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019772542785632147, - "loss": 1.6953, - "step": 6725 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977250896299804, - "loss": 1.7641, - "step": 6726 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019772475140363932, - "loss": 1.7042, - "step": 6727 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019772441317729826, - "loss": 1.646, - "step": 6728 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977240749509572, - "loss": 1.7691, - "step": 6729 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019772373672461612, - "loss": 1.5985, - "step": 6730 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019772339849827506, - "loss": 1.7311, - "step": 6731 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197723060271934, - "loss": 1.6653, - "step": 6732 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019772272204559291, - "loss": 1.6954, - "step": 6733 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019772238381925186, - "loss": 1.7698, - "step": 6734 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977220455929108, - "loss": 1.7069, - "step": 6735 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977217073665697, - "loss": 1.5712, - "step": 6736 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019772136914022865, - "loss": 1.6812, - "step": 6737 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977210309138876, - "loss": 1.6758, - "step": 6738 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977206926875465, - "loss": 1.6047, - "step": 6739 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019772035446120545, - "loss": 1.7129, - "step": 6740 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977200162348644, - "loss": 1.7186, - "step": 6741 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977196780085233, - "loss": 1.6183, - "step": 6742 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019771933978218224, - "loss": 1.6022, - "step": 6743 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019771900155584119, - "loss": 1.676, - "step": 6744 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977186633295001, - "loss": 1.7756, - "step": 6745 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019771832510315904, - "loss": 1.6513, - "step": 6746 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019771798687681798, - "loss": 1.7755, - "step": 6747 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977176486504769, - "loss": 1.7169, - "step": 6748 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019771731042413584, - "loss": 1.806, - "step": 6749 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019771697219779478, - "loss": 1.6795, - "step": 6750 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977166339714537, - "loss": 1.5379, - "step": 6751 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019771629574511263, - "loss": 1.5943, - "step": 6752 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019771595751877157, - "loss": 1.7384, - "step": 6753 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977156192924305, - "loss": 1.7292, - "step": 6754 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019771528106608943, - "loss": 1.6766, - "step": 6755 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019771494283974837, - "loss": 1.6758, - "step": 6756 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019771460461340728, - "loss": 1.7294, - "step": 6757 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019771426638706623, - "loss": 1.7313, - "step": 6758 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019771392816072517, - "loss": 1.7082, - "step": 6759 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019771358993438408, - "loss": 1.5326, - "step": 6760 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019771325170804302, - "loss": 1.6017, - "step": 6761 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019771291348170196, - "loss": 1.6662, - "step": 6762 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019771257525536088, - "loss": 1.6701, - "step": 6763 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019771223702901982, - "loss": 1.7619, - "step": 6764 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019771189880267876, - "loss": 1.6367, - "step": 6765 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019771156057633767, - "loss": 1.7134, - "step": 6766 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019771122234999661, - "loss": 1.7103, - "step": 6767 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019771088412365556, - "loss": 1.6284, - "step": 6768 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977105458973145, - "loss": 1.625, - "step": 6769 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019771020767097344, - "loss": 1.8095, - "step": 6770 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019770986944463238, - "loss": 1.5941, - "step": 6771 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977095312182913, - "loss": 1.8046, - "step": 6772 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019770919299195023, - "loss": 1.6854, - "step": 6773 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019770885476560918, - "loss": 1.766, - "step": 6774 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977085165392681, - "loss": 1.7129, - "step": 6775 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019770817831292703, - "loss": 1.7951, - "step": 6776 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019770784008658597, - "loss": 1.6469, - "step": 6777 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019770750186024489, - "loss": 1.82, - "step": 6778 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019770716363390383, - "loss": 1.6839, - "step": 6779 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019770682540756277, - "loss": 1.752, - "step": 6780 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019770648718122168, - "loss": 1.5528, - "step": 6781 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019770614895488062, - "loss": 1.8273, - "step": 6782 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019770581072853956, - "loss": 1.4316, - "step": 6783 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019770547250219848, - "loss": 1.7979, - "step": 6784 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019770513427585742, - "loss": 1.5784, - "step": 6785 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019770479604951636, - "loss": 1.8314, - "step": 6786 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019770445782317527, - "loss": 1.5214, - "step": 6787 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019770411959683422, - "loss": 1.6821, - "step": 6788 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019770378137049316, - "loss": 1.6602, - "step": 6789 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019770344314415207, - "loss": 1.7344, - "step": 6790 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197703104917811, - "loss": 1.6507, - "step": 6791 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019770276669146995, - "loss": 1.8987, - "step": 6792 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019770242846512887, - "loss": 1.5882, - "step": 6793 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977020902387878, - "loss": 1.6174, - "step": 6794 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019770175201244675, - "loss": 1.6227, - "step": 6795 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019770141378610566, - "loss": 1.6165, - "step": 6796 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977010755597646, - "loss": 1.5642, - "step": 6797 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019770073733342355, - "loss": 1.8196, - "step": 6798 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019770039910708246, - "loss": 1.7351, - "step": 6799 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001977000608807414, - "loss": 1.67, - "step": 6800 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019769972265440034, - "loss": 1.6012, - "step": 6801 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019769938442805926, - "loss": 1.7127, - "step": 6802 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976990462017182, - "loss": 1.717, - "step": 6803 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019769870797537714, - "loss": 1.7944, - "step": 6804 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019769836974903605, - "loss": 1.883, - "step": 6805 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197698031522695, - "loss": 1.6363, - "step": 6806 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019769769329635393, - "loss": 1.7537, - "step": 6807 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019769735507001285, - "loss": 1.7373, - "step": 6808 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976970168436718, - "loss": 1.6085, - "step": 6809 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019769667861733073, - "loss": 1.6936, - "step": 6810 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019769634039098965, - "loss": 1.6875, - "step": 6811 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976960021646486, - "loss": 1.531, - "step": 6812 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019769566393830753, - "loss": 1.6543, - "step": 6813 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019769532571196644, - "loss": 1.6511, - "step": 6814 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019769498748562538, - "loss": 1.7035, - "step": 6815 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019769464925928432, - "loss": 1.6583, - "step": 6816 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019769431103294324, - "loss": 1.7023, - "step": 6817 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019769397280660218, - "loss": 1.7493, - "step": 6818 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019769363458026112, - "loss": 1.6984, - "step": 6819 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019769329635392003, - "loss": 1.7769, - "step": 6820 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019769295812757898, - "loss": 1.6678, - "step": 6821 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019769261990123792, - "loss": 1.6922, - "step": 6822 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019769228167489683, - "loss": 1.6862, - "step": 6823 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019769194344855577, - "loss": 1.6687, - "step": 6824 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976916052222147, - "loss": 1.7441, - "step": 6825 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019769126699587365, - "loss": 1.7064, - "step": 6826 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976909287695326, - "loss": 1.7145, - "step": 6827 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019769059054319154, - "loss": 1.5799, - "step": 6828 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019769025231685045, - "loss": 1.723, - "step": 6829 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976899140905094, - "loss": 1.6716, - "step": 6830 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019768957586416833, - "loss": 1.6949, - "step": 6831 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019768923763782725, - "loss": 1.6783, - "step": 6832 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976888994114862, - "loss": 1.7826, - "step": 6833 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019768856118514513, - "loss": 1.7801, - "step": 6834 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019768822295880404, - "loss": 1.6743, - "step": 6835 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019768788473246298, - "loss": 1.7512, - "step": 6836 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019768754650612193, - "loss": 1.8412, - "step": 6837 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019768720827978084, - "loss": 1.673, - "step": 6838 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019768687005343978, - "loss": 1.7412, - "step": 6839 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019768653182709872, - "loss": 1.5812, - "step": 6840 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019768619360075764, - "loss": 1.7503, - "step": 6841 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019768585537441658, - "loss": 1.6574, - "step": 6842 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019768551714807552, - "loss": 1.5585, - "step": 6843 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019768517892173443, - "loss": 1.6058, - "step": 6844 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019768484069539337, - "loss": 1.7336, - "step": 6845 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019768450246905231, - "loss": 1.6616, - "step": 6846 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019768416424271123, - "loss": 1.7265, - "step": 6847 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019768382601637017, - "loss": 1.7651, - "step": 6848 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976834877900291, - "loss": 1.7002, - "step": 6849 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019768314956368802, - "loss": 1.5609, - "step": 6850 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019768281133734697, - "loss": 1.6612, - "step": 6851 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976824731110059, - "loss": 1.5819, - "step": 6852 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019768213488466482, - "loss": 1.6445, - "step": 6853 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019768179665832376, - "loss": 1.816, - "step": 6854 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976814584319827, - "loss": 1.9059, - "step": 6855 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019768112020564162, - "loss": 1.5732, - "step": 6856 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019768078197930056, - "loss": 1.7139, - "step": 6857 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976804437529595, - "loss": 1.7313, - "step": 6858 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976801055266184, - "loss": 1.6392, - "step": 6859 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019767976730027735, - "loss": 1.6337, - "step": 6860 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976794290739363, - "loss": 1.7161, - "step": 6861 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976790908475952, - "loss": 1.6266, - "step": 6862 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019767875262125415, - "loss": 1.7678, - "step": 6863 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976784143949131, - "loss": 1.622, - "step": 6864 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197678076168572, - "loss": 1.7603, - "step": 6865 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019767773794223095, - "loss": 1.6098, - "step": 6866 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976773997158899, - "loss": 1.7569, - "step": 6867 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976770614895488, - "loss": 1.5873, - "step": 6868 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019767672326320774, - "loss": 1.6234, - "step": 6869 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019767638503686668, - "loss": 1.6332, - "step": 6870 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976760468105256, - "loss": 1.7918, - "step": 6871 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019767570858418454, - "loss": 1.58, - "step": 6872 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019767537035784348, - "loss": 1.6651, - "step": 6873 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976750321315024, - "loss": 1.5981, - "step": 6874 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019767469390516134, - "loss": 1.5699, - "step": 6875 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019767435567882028, - "loss": 1.6267, - "step": 6876 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976740174524792, - "loss": 1.5456, - "step": 6877 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019767367922613813, - "loss": 1.6397, - "step": 6878 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019767334099979707, - "loss": 1.8494, - "step": 6879 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197673002773456, - "loss": 1.5828, - "step": 6880 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019767266454711496, - "loss": 1.6805, - "step": 6881 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019767232632077387, - "loss": 1.6304, - "step": 6882 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976719880944328, - "loss": 1.5423, - "step": 6883 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019767164986809175, - "loss": 1.7162, - "step": 6884 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019767131164175067, - "loss": 1.7316, - "step": 6885 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976709734154096, - "loss": 1.7505, - "step": 6886 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019767063518906855, - "loss": 1.6999, - "step": 6887 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019767029696272746, - "loss": 1.7748, - "step": 6888 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976699587363864, - "loss": 1.812, - "step": 6889 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019766962051004534, - "loss": 1.7717, - "step": 6890 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019766928228370426, - "loss": 1.6698, - "step": 6891 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976689440573632, - "loss": 1.7784, - "step": 6892 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019766860583102214, - "loss": 1.7044, - "step": 6893 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019766826760468106, - "loss": 1.707, - "step": 6894 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019766792937834, - "loss": 1.7784, - "step": 6895 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019766759115199894, - "loss": 1.6476, - "step": 6896 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019766725292565785, - "loss": 1.7347, - "step": 6897 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976669146993168, - "loss": 1.7271, - "step": 6898 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019766657647297573, - "loss": 1.6902, - "step": 6899 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019766623824663465, - "loss": 1.6623, - "step": 6900 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976659000202936, - "loss": 1.8335, - "step": 6901 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019766556179395253, - "loss": 1.6394, - "step": 6902 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019766522356761144, - "loss": 1.7577, - "step": 6903 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019766488534127039, - "loss": 1.7069, - "step": 6904 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019766454711492933, - "loss": 1.6671, - "step": 6905 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019766420888858824, - "loss": 1.9028, - "step": 6906 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019766387066224718, - "loss": 1.6701, - "step": 6907 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019766353243590612, - "loss": 1.691, - "step": 6908 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019766319420956504, - "loss": 1.8365, - "step": 6909 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019766285598322398, - "loss": 1.5421, - "step": 6910 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019766251775688292, - "loss": 1.8207, - "step": 6911 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019766217953054186, - "loss": 1.654, - "step": 6912 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019766184130420077, - "loss": 1.6995, - "step": 6913 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019766150307785972, - "loss": 1.6949, - "step": 6914 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019766116485151866, - "loss": 1.7475, - "step": 6915 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019766082662517757, - "loss": 1.7641, - "step": 6916 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976604883988365, - "loss": 1.7572, - "step": 6917 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019766015017249545, - "loss": 1.6436, - "step": 6918 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019765981194615437, - "loss": 1.5807, - "step": 6919 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976594737198133, - "loss": 1.7443, - "step": 6920 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019765913549347225, - "loss": 1.7811, - "step": 6921 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019765879726713116, - "loss": 1.75, - "step": 6922 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976584590407901, - "loss": 1.7072, - "step": 6923 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019765812081444905, - "loss": 1.8817, - "step": 6924 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019765778258810796, - "loss": 1.8086, - "step": 6925 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976574443617669, - "loss": 1.6007, - "step": 6926 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019765710613542584, - "loss": 1.673, - "step": 6927 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019765676790908476, - "loss": 1.6578, - "step": 6928 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976564296827437, - "loss": 1.853, - "step": 6929 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019765609145640264, - "loss": 1.6844, - "step": 6930 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019765575323006155, - "loss": 1.7748, - "step": 6931 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976554150037205, - "loss": 1.7089, - "step": 6932 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019765507677737943, - "loss": 1.7048, - "step": 6933 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019765473855103835, - "loss": 1.5073, - "step": 6934 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976544003246973, - "loss": 1.714, - "step": 6935 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019765406209835623, - "loss": 1.6754, - "step": 6936 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019765372387201517, - "loss": 1.692, - "step": 6937 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976533856456741, - "loss": 1.7763, - "step": 6938 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019765304741933303, - "loss": 1.6822, - "step": 6939 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019765270919299197, - "loss": 1.592, - "step": 6940 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976523709666509, - "loss": 1.7096, - "step": 6941 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019765203274030982, - "loss": 1.6822, - "step": 6942 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019765169451396876, - "loss": 1.7061, - "step": 6943 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976513562876277, - "loss": 1.6822, - "step": 6944 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019765101806128662, - "loss": 1.5976, - "step": 6945 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019765067983494556, - "loss": 1.7795, - "step": 6946 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976503416086045, - "loss": 1.8135, - "step": 6947 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019765000338226342, - "loss": 1.8115, - "step": 6948 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019764966515592236, - "loss": 1.7549, - "step": 6949 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976493269295813, - "loss": 1.5809, - "step": 6950 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976489887032402, - "loss": 1.6139, - "step": 6951 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019764865047689915, - "loss": 1.7157, - "step": 6952 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976483122505581, - "loss": 1.63, - "step": 6953 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197647974024217, - "loss": 1.6578, - "step": 6954 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019764763579787595, - "loss": 1.6096, - "step": 6955 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976472975715349, - "loss": 1.6105, - "step": 6956 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976469593451938, - "loss": 1.5373, - "step": 6957 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019764662111885275, - "loss": 1.553, - "step": 6958 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976462828925117, - "loss": 1.5514, - "step": 6959 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976459446661706, - "loss": 1.6039, - "step": 6960 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019764560643982954, - "loss": 1.6095, - "step": 6961 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019764526821348848, - "loss": 1.9375, - "step": 6962 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976449299871474, - "loss": 1.716, - "step": 6963 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019764459176080634, - "loss": 1.6785, - "step": 6964 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019764425353446528, - "loss": 1.763, - "step": 6965 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976439153081242, - "loss": 1.7076, - "step": 6966 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019764357708178313, - "loss": 1.6072, - "step": 6967 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019764323885544208, - "loss": 1.5187, - "step": 6968 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197642900629101, - "loss": 1.6094, - "step": 6969 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019764256240275993, - "loss": 1.781, - "step": 6970 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019764222417641887, - "loss": 1.6738, - "step": 6971 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019764188595007779, - "loss": 1.6516, - "step": 6972 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019764154772373673, - "loss": 1.7753, - "step": 6973 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019764120949739567, - "loss": 1.7859, - "step": 6974 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019764087127105458, - "loss": 1.6546, - "step": 6975 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019764053304471352, - "loss": 1.6051, - "step": 6976 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019764019481837247, - "loss": 1.5919, - "step": 6977 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019763985659203138, - "loss": 1.7344, - "step": 6978 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019763951836569032, - "loss": 1.704, - "step": 6979 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019763918013934926, - "loss": 1.6845, - "step": 6980 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019763884191300818, - "loss": 1.754, - "step": 6981 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019763850368666712, - "loss": 1.5417, - "step": 6982 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019763816546032606, - "loss": 1.6608, - "step": 6983 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019763782723398497, - "loss": 1.6488, - "step": 6984 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976374890076439, - "loss": 1.7684, - "step": 6985 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019763715078130285, - "loss": 1.7203, - "step": 6986 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019763681255496177, - "loss": 1.635, - "step": 6987 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976364743286207, - "loss": 1.7633, - "step": 6988 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019763613610227965, - "loss": 1.7044, - "step": 6989 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019763579787593856, - "loss": 1.5951, - "step": 6990 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976354596495975, - "loss": 1.7606, - "step": 6991 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019763512142325645, - "loss": 1.7868, - "step": 6992 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976347831969154, - "loss": 1.6162, - "step": 6993 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019763444497057433, - "loss": 1.7571, - "step": 6994 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019763410674423327, - "loss": 1.6607, - "step": 6995 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019763376851789218, - "loss": 1.8328, - "step": 6996 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019763343029155113, - "loss": 1.5997, - "step": 6997 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019763309206521007, - "loss": 1.7538, - "step": 6998 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019763275383886898, - "loss": 1.6968, - "step": 6999 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019763241561252792, - "loss": 1.9471, - "step": 7000 - }, - { - "epoch": 0.02, - "eval_loss": 1.2708082932692308, - "mrc_em": 20.0, - "mrc_f1": 26.93506493506494, - "nli_acc": 0.3, - "step": 7000, - "sts_f1": 0.4222222222222222, - "ynat_f1": 0.5112745098039215 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019763207738618686, - "loss": 1.5519, - "step": 7001 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019763173915984578, - "loss": 1.6819, - "step": 7002 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019763140093350472, - "loss": 1.6946, - "step": 7003 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019763106270716366, - "loss": 1.5894, - "step": 7004 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019763072448082257, - "loss": 1.7344, - "step": 7005 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019763038625448151, - "loss": 1.7545, - "step": 7006 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019763004802814046, - "loss": 1.7656, - "step": 7007 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019762970980179937, - "loss": 1.6838, - "step": 7008 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976293715754583, - "loss": 1.6001, - "step": 7009 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019762903334911725, - "loss": 1.7129, - "step": 7010 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019762869512277617, - "loss": 1.5786, - "step": 7011 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976283568964351, - "loss": 1.6875, - "step": 7012 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019762801867009405, - "loss": 1.6992, - "step": 7013 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019762768044375296, - "loss": 1.747, - "step": 7014 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976273422174119, - "loss": 1.7853, - "step": 7015 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019762700399107084, - "loss": 1.7456, - "step": 7016 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019762666576472976, - "loss": 1.568, - "step": 7017 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976263275383887, - "loss": 1.5614, - "step": 7018 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019762598931204764, - "loss": 1.5817, - "step": 7019 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019762565108570655, - "loss": 1.4759, - "step": 7020 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976253128593655, - "loss": 1.7738, - "step": 7021 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019762497463302444, - "loss": 1.9146, - "step": 7022 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019762463640668335, - "loss": 1.6195, - "step": 7023 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976242981803423, - "loss": 1.5824, - "step": 7024 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019762395995400123, - "loss": 1.716, - "step": 7025 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019762362172766015, - "loss": 1.7151, - "step": 7026 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976232835013191, - "loss": 1.6133, - "step": 7027 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019762294527497803, - "loss": 1.6799, - "step": 7028 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019762260704863694, - "loss": 1.6624, - "step": 7029 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019762226882229588, - "loss": 1.5957, - "step": 7030 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019762193059595483, - "loss": 1.6531, - "step": 7031 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019762159236961374, - "loss": 1.7613, - "step": 7032 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019762125414327268, - "loss": 1.7362, - "step": 7033 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019762091591693162, - "loss": 1.6721, - "step": 7034 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019762057769059054, - "loss": 1.7165, - "step": 7035 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019762023946424948, - "loss": 1.7239, - "step": 7036 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019761990123790842, - "loss": 1.6205, - "step": 7037 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019761956301156733, - "loss": 1.6881, - "step": 7038 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019761922478522627, - "loss": 1.7468, - "step": 7039 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019761888655888521, - "loss": 1.6752, - "step": 7040 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019761854833254413, - "loss": 1.6112, - "step": 7041 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019761821010620307, - "loss": 1.8021, - "step": 7042 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197617871879862, - "loss": 1.6404, - "step": 7043 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019761753365352093, - "loss": 1.6442, - "step": 7044 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019761719542717987, - "loss": 1.746, - "step": 7045 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976168572008388, - "loss": 1.7138, - "step": 7046 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019761651897449772, - "loss": 1.6573, - "step": 7047 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019761618074815666, - "loss": 1.7322, - "step": 7048 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019761584252181563, - "loss": 1.7474, - "step": 7049 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019761550429547454, - "loss": 1.6297, - "step": 7050 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019761516606913349, - "loss": 1.7113, - "step": 7051 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019761482784279243, - "loss": 1.617, - "step": 7052 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019761448961645134, - "loss": 1.6955, - "step": 7053 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019761415139011028, - "loss": 1.6236, - "step": 7054 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019761381316376922, - "loss": 1.6546, - "step": 7055 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019761347493742814, - "loss": 1.7062, - "step": 7056 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019761313671108708, - "loss": 1.7258, - "step": 7057 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019761279848474602, - "loss": 1.6118, - "step": 7058 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019761246025840493, - "loss": 1.6298, - "step": 7059 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019761212203206387, - "loss": 1.6241, - "step": 7060 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019761178380572282, - "loss": 1.5268, - "step": 7061 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019761144557938173, - "loss": 1.5986, - "step": 7062 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019761110735304067, - "loss": 1.7243, - "step": 7063 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976107691266996, - "loss": 1.744, - "step": 7064 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019761043090035853, - "loss": 1.6192, - "step": 7065 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019761009267401747, - "loss": 1.805, - "step": 7066 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976097544476764, - "loss": 1.6554, - "step": 7067 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019760941622133532, - "loss": 1.8394, - "step": 7068 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019760907799499426, - "loss": 1.6204, - "step": 7069 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976087397686532, - "loss": 1.5615, - "step": 7070 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019760840154231212, - "loss": 1.7304, - "step": 7071 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019760806331597106, - "loss": 1.5042, - "step": 7072 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019760772508963, - "loss": 1.6977, - "step": 7073 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019760738686328892, - "loss": 1.6466, - "step": 7074 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019760704863694786, - "loss": 1.7903, - "step": 7075 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976067104106068, - "loss": 1.5701, - "step": 7076 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976063721842657, - "loss": 1.7675, - "step": 7077 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019760603395792465, - "loss": 1.7107, - "step": 7078 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976056957315836, - "loss": 1.7023, - "step": 7079 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976053575052425, - "loss": 1.8352, - "step": 7080 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019760501927890145, - "loss": 1.7186, - "step": 7081 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976046810525604, - "loss": 1.57, - "step": 7082 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976043428262193, - "loss": 1.6788, - "step": 7083 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019760400459987825, - "loss": 1.768, - "step": 7084 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976036663735372, - "loss": 1.8147, - "step": 7085 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976033281471961, - "loss": 1.6327, - "step": 7086 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019760298992085504, - "loss": 1.676, - "step": 7087 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019760265169451398, - "loss": 1.618, - "step": 7088 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976023134681729, - "loss": 1.6848, - "step": 7089 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019760197524183184, - "loss": 1.5944, - "step": 7090 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019760163701549078, - "loss": 1.7201, - "step": 7091 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976012987891497, - "loss": 1.7256, - "step": 7092 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019760096056280863, - "loss": 1.5938, - "step": 7093 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019760062233646758, - "loss": 1.6505, - "step": 7094 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001976002841101265, - "loss": 1.7288, - "step": 7095 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019759994588378543, - "loss": 1.6641, - "step": 7096 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019759960765744437, - "loss": 1.7593, - "step": 7097 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019759926943110329, - "loss": 1.6774, - "step": 7098 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019759893120476223, - "loss": 1.7413, - "step": 7099 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019759859297842117, - "loss": 1.7911, - "step": 7100 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019759825475208008, - "loss": 1.8139, - "step": 7101 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019759791652573902, - "loss": 1.6311, - "step": 7102 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019759757829939796, - "loss": 1.631, - "step": 7103 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019759724007305688, - "loss": 1.7482, - "step": 7104 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019759690184671582, - "loss": 1.5806, - "step": 7105 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975965636203748, - "loss": 1.6889, - "step": 7106 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975962253940337, - "loss": 1.5773, - "step": 7107 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019759588716769264, - "loss": 1.7004, - "step": 7108 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019759554894135158, - "loss": 1.7744, - "step": 7109 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975952107150105, - "loss": 1.6204, - "step": 7110 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019759487248866944, - "loss": 1.6289, - "step": 7111 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019759453426232838, - "loss": 1.6055, - "step": 7112 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975941960359873, - "loss": 1.6091, - "step": 7113 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019759385780964624, - "loss": 1.4741, - "step": 7114 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019759351958330518, - "loss": 1.7495, - "step": 7115 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975931813569641, - "loss": 1.6617, - "step": 7116 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019759284313062303, - "loss": 1.5347, - "step": 7117 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019759250490428197, - "loss": 1.7299, - "step": 7118 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975921666779409, - "loss": 1.5934, - "step": 7119 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019759182845159983, - "loss": 1.7254, - "step": 7120 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019759149022525877, - "loss": 1.8241, - "step": 7121 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019759115199891768, - "loss": 1.8293, - "step": 7122 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019759081377257662, - "loss": 1.7624, - "step": 7123 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019759047554623557, - "loss": 1.7239, - "step": 7124 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019759013731989448, - "loss": 1.559, - "step": 7125 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019758979909355342, - "loss": 1.6348, - "step": 7126 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019758946086721236, - "loss": 1.8148, - "step": 7127 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019758912264087128, - "loss": 1.7637, - "step": 7128 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019758878441453022, - "loss": 1.6452, - "step": 7129 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019758844618818916, - "loss": 1.6345, - "step": 7130 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019758810796184807, - "loss": 1.4759, - "step": 7131 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197587769735507, - "loss": 1.6528, - "step": 7132 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019758743150916595, - "loss": 1.7393, - "step": 7133 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019758709328282487, - "loss": 1.6231, - "step": 7134 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975867550564838, - "loss": 1.6817, - "step": 7135 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019758641683014275, - "loss": 1.879, - "step": 7136 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019758607860380166, - "loss": 1.6845, - "step": 7137 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975857403774606, - "loss": 1.5413, - "step": 7138 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019758540215111955, - "loss": 1.7487, - "step": 7139 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019758506392477846, - "loss": 1.8204, - "step": 7140 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975847256984374, - "loss": 1.8268, - "step": 7141 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019758438747209634, - "loss": 1.8318, - "step": 7142 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019758404924575526, - "loss": 1.7201, - "step": 7143 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975837110194142, - "loss": 1.774, - "step": 7144 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019758337279307314, - "loss": 1.6006, - "step": 7145 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019758303456673205, - "loss": 1.7745, - "step": 7146 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197582696340391, - "loss": 1.596, - "step": 7147 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019758235811404994, - "loss": 1.6345, - "step": 7148 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019758201988770885, - "loss": 1.7737, - "step": 7149 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975816816613678, - "loss": 1.7425, - "step": 7150 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019758134343502673, - "loss": 1.7376, - "step": 7151 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019758100520868565, - "loss": 1.837, - "step": 7152 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975806669823446, - "loss": 1.7086, - "step": 7153 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019758032875600353, - "loss": 1.6778, - "step": 7154 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019757999052966244, - "loss": 1.5481, - "step": 7155 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019757965230332138, - "loss": 1.8324, - "step": 7156 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019757931407698033, - "loss": 1.5759, - "step": 7157 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019757897585063924, - "loss": 1.5448, - "step": 7158 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019757863762429818, - "loss": 1.7033, - "step": 7159 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019757829939795712, - "loss": 1.655, - "step": 7160 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019757796117161604, - "loss": 1.7783, - "step": 7161 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197577622945275, - "loss": 1.8234, - "step": 7162 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019757728471893392, - "loss": 1.7402, - "step": 7163 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019757694649259286, - "loss": 1.6673, - "step": 7164 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975766082662518, - "loss": 1.5939, - "step": 7165 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019757627003991071, - "loss": 1.8125, - "step": 7166 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019757593181356966, - "loss": 1.8895, - "step": 7167 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975755935872286, - "loss": 1.846, - "step": 7168 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975752553608875, - "loss": 1.8733, - "step": 7169 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019757491713454645, - "loss": 1.7188, - "step": 7170 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975745789082054, - "loss": 1.7259, - "step": 7171 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975742406818643, - "loss": 1.5511, - "step": 7172 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019757390245552325, - "loss": 1.6872, - "step": 7173 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975735642291822, - "loss": 1.6056, - "step": 7174 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975732260028411, - "loss": 1.6553, - "step": 7175 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019757288777650004, - "loss": 1.6569, - "step": 7176 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019757254955015899, - "loss": 1.8255, - "step": 7177 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975722113238179, - "loss": 1.73, - "step": 7178 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019757187309747684, - "loss": 1.7203, - "step": 7179 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019757153487113578, - "loss": 1.4502, - "step": 7180 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019757119664479472, - "loss": 1.8737, - "step": 7181 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019757085841845364, - "loss": 1.718, - "step": 7182 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019757052019211258, - "loss": 1.5083, - "step": 7183 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019757018196577152, - "loss": 1.7466, - "step": 7184 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019756984373943043, - "loss": 1.6708, - "step": 7185 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019756950551308937, - "loss": 1.744, - "step": 7186 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019756916728674832, - "loss": 1.7309, - "step": 7187 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019756882906040723, - "loss": 1.6553, - "step": 7188 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019756849083406617, - "loss": 1.6666, - "step": 7189 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975681526077251, - "loss": 1.6798, - "step": 7190 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019756781438138403, - "loss": 1.6639, - "step": 7191 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019756747615504297, - "loss": 1.604, - "step": 7192 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975671379287019, - "loss": 1.6364, - "step": 7193 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019756679970236082, - "loss": 1.5876, - "step": 7194 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019756646147601976, - "loss": 1.7654, - "step": 7195 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975661232496787, - "loss": 1.7074, - "step": 7196 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019756578502333762, - "loss": 1.8101, - "step": 7197 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019756544679699656, - "loss": 1.7227, - "step": 7198 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975651085706555, - "loss": 1.6766, - "step": 7199 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019756477034431441, - "loss": 1.733, - "step": 7200 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019756443211797336, - "loss": 1.6434, - "step": 7201 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975640938916323, - "loss": 1.76, - "step": 7202 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975637556652912, - "loss": 1.6186, - "step": 7203 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019756341743895015, - "loss": 1.5348, - "step": 7204 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975630792126091, - "loss": 1.4335, - "step": 7205 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197562740986268, - "loss": 1.5112, - "step": 7206 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019756240275992695, - "loss": 1.6503, - "step": 7207 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975620645335859, - "loss": 1.6873, - "step": 7208 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975617263072448, - "loss": 1.6946, - "step": 7209 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019756138808090374, - "loss": 1.6398, - "step": 7210 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019756104985456269, - "loss": 1.7099, - "step": 7211 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975607116282216, - "loss": 1.6854, - "step": 7212 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019756037340188054, - "loss": 1.716, - "step": 7213 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019756003517553948, - "loss": 1.7299, - "step": 7214 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975596969491984, - "loss": 1.6762, - "step": 7215 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019755935872285734, - "loss": 1.607, - "step": 7216 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019755902049651628, - "loss": 1.6549, - "step": 7217 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019755868227017522, - "loss": 1.895, - "step": 7218 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019755834404383416, - "loss": 1.7881, - "step": 7219 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019755800581749307, - "loss": 1.7361, - "step": 7220 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019755766759115202, - "loss": 1.6443, - "step": 7221 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019755732936481096, - "loss": 1.7743, - "step": 7222 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019755699113846987, - "loss": 1.7138, - "step": 7223 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975566529121288, - "loss": 1.6769, - "step": 7224 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019755631468578775, - "loss": 1.6129, - "step": 7225 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019755597645944667, - "loss": 1.6597, - "step": 7226 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975556382331056, - "loss": 1.5562, - "step": 7227 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019755530000676455, - "loss": 1.5001, - "step": 7228 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019755496178042346, - "loss": 1.898, - "step": 7229 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975546235540824, - "loss": 1.7432, - "step": 7230 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019755428532774135, - "loss": 1.5895, - "step": 7231 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019755394710140026, - "loss": 1.8235, - "step": 7232 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975536088750592, - "loss": 1.6861, - "step": 7233 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019755327064871814, - "loss": 1.6537, - "step": 7234 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019755293242237706, - "loss": 1.7392, - "step": 7235 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197552594196036, - "loss": 1.7316, - "step": 7236 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019755225596969494, - "loss": 1.807, - "step": 7237 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019755191774335385, - "loss": 1.6396, - "step": 7238 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975515795170128, - "loss": 1.7624, - "step": 7239 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019755124129067173, - "loss": 1.6981, - "step": 7240 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019755090306433065, - "loss": 1.8244, - "step": 7241 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975505648379896, - "loss": 1.6489, - "step": 7242 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019755022661164853, - "loss": 1.7241, - "step": 7243 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019754988838530745, - "loss": 1.6756, - "step": 7244 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975495501589664, - "loss": 1.6078, - "step": 7245 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019754921193262533, - "loss": 1.5843, - "step": 7246 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019754887370628424, - "loss": 1.8082, - "step": 7247 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019754853547994318, - "loss": 1.7086, - "step": 7248 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019754819725360212, - "loss": 1.6682, - "step": 7249 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019754785902726104, - "loss": 1.7415, - "step": 7250 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019754752080091998, - "loss": 1.8109, - "step": 7251 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019754718257457892, - "loss": 1.6397, - "step": 7252 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019754684434823783, - "loss": 1.6387, - "step": 7253 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019754650612189678, - "loss": 1.5192, - "step": 7254 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019754616789555572, - "loss": 1.5677, - "step": 7255 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019754582966921463, - "loss": 1.6913, - "step": 7256 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019754549144287357, - "loss": 1.599, - "step": 7257 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975451532165325, - "loss": 1.7229, - "step": 7258 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019754481499019143, - "loss": 1.6792, - "step": 7259 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019754447676385037, - "loss": 1.5455, - "step": 7260 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975441385375093, - "loss": 1.5898, - "step": 7261 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019754380031116822, - "loss": 1.8438, - "step": 7262 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019754346208482716, - "loss": 1.6408, - "step": 7263 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975431238584861, - "loss": 1.7386, - "step": 7264 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019754278563214505, - "loss": 1.8699, - "step": 7265 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019754244740580396, - "loss": 1.7985, - "step": 7266 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975421091794629, - "loss": 1.6463, - "step": 7267 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019754177095312184, - "loss": 1.5912, - "step": 7268 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019754143272678076, - "loss": 1.6893, - "step": 7269 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975410945004397, - "loss": 1.6536, - "step": 7270 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019754075627409864, - "loss": 1.5613, - "step": 7271 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019754041804775755, - "loss": 1.6487, - "step": 7272 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975400798214165, - "loss": 1.7375, - "step": 7273 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019753974159507544, - "loss": 1.6898, - "step": 7274 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019753940336873438, - "loss": 1.6374, - "step": 7275 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019753906514239332, - "loss": 1.5383, - "step": 7276 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019753872691605223, - "loss": 1.6836, - "step": 7277 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019753838868971117, - "loss": 1.734, - "step": 7278 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019753805046337011, - "loss": 1.7029, - "step": 7279 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019753771223702903, - "loss": 1.6398, - "step": 7280 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019753737401068797, - "loss": 1.7207, - "step": 7281 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975370357843469, - "loss": 1.623, - "step": 7282 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019753669755800582, - "loss": 1.7865, - "step": 7283 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019753635933166477, - "loss": 1.7241, - "step": 7284 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975360211053237, - "loss": 1.7903, - "step": 7285 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019753568287898262, - "loss": 1.7505, - "step": 7286 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019753534465264156, - "loss": 1.7282, - "step": 7287 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975350064263005, - "loss": 1.6918, - "step": 7288 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019753466819995942, - "loss": 1.5857, - "step": 7289 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019753432997361836, - "loss": 1.7055, - "step": 7290 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975339917472773, - "loss": 1.7072, - "step": 7291 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975336535209362, - "loss": 1.7312, - "step": 7292 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019753331529459515, - "loss": 1.7773, - "step": 7293 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975329770682541, - "loss": 1.6261, - "step": 7294 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197532638841913, - "loss": 1.7168, - "step": 7295 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019753230061557195, - "loss": 1.7703, - "step": 7296 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975319623892309, - "loss": 1.6689, - "step": 7297 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975316241628898, - "loss": 1.7434, - "step": 7298 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019753128593654875, - "loss": 1.5938, - "step": 7299 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975309477102077, - "loss": 1.8365, - "step": 7300 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975306094838666, - "loss": 1.4923, - "step": 7301 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019753027125752554, - "loss": 1.79, - "step": 7302 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019752993303118448, - "loss": 1.6076, - "step": 7303 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975295948048434, - "loss": 1.5792, - "step": 7304 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019752925657850234, - "loss": 1.7491, - "step": 7305 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019752891835216128, - "loss": 1.5506, - "step": 7306 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975285801258202, - "loss": 1.6576, - "step": 7307 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019752824189947914, - "loss": 1.5837, - "step": 7308 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019752790367313808, - "loss": 1.713, - "step": 7309 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197527565446797, - "loss": 1.6831, - "step": 7310 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019752722722045593, - "loss": 1.7462, - "step": 7311 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019752688899411487, - "loss": 1.6181, - "step": 7312 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975265507677738, - "loss": 1.5786, - "step": 7313 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019752621254143273, - "loss": 1.6089, - "step": 7314 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019752587431509167, - "loss": 1.7781, - "step": 7315 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019752553608875058, - "loss": 1.649, - "step": 7316 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019752519786240953, - "loss": 1.6679, - "step": 7317 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019752485963606847, - "loss": 1.848, - "step": 7318 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019752452140972738, - "loss": 1.7106, - "step": 7319 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019752418318338632, - "loss": 1.8037, - "step": 7320 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019752384495704526, - "loss": 1.6825, - "step": 7321 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019752350673070418, - "loss": 1.5846, - "step": 7322 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019752316850436312, - "loss": 1.8025, - "step": 7323 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019752283027802206, - "loss": 1.6287, - "step": 7324 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019752249205168097, - "loss": 1.6601, - "step": 7325 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019752215382533991, - "loss": 1.7555, - "step": 7326 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019752181559899886, - "loss": 1.7206, - "step": 7327 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019752147737265777, - "loss": 1.6872, - "step": 7328 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975211391463167, - "loss": 1.6738, - "step": 7329 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019752080091997568, - "loss": 1.6311, - "step": 7330 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975204626936346, - "loss": 1.7316, - "step": 7331 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019752012446729353, - "loss": 1.7693, - "step": 7332 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019751978624095247, - "loss": 1.7592, - "step": 7333 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975194480146114, - "loss": 1.7646, - "step": 7334 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019751910978827033, - "loss": 1.5565, - "step": 7335 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019751877156192927, - "loss": 1.69, - "step": 7336 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019751843333558819, - "loss": 1.786, - "step": 7337 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019751809510924713, - "loss": 1.6518, - "step": 7338 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019751775688290607, - "loss": 1.6569, - "step": 7339 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019751741865656498, - "loss": 1.6956, - "step": 7340 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019751708043022392, - "loss": 1.7869, - "step": 7341 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019751674220388286, - "loss": 1.6448, - "step": 7342 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019751640397754178, - "loss": 1.5036, - "step": 7343 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019751606575120072, - "loss": 1.9248, - "step": 7344 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019751572752485966, - "loss": 1.6576, - "step": 7345 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019751538929851857, - "loss": 1.7546, - "step": 7346 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019751505107217752, - "loss": 1.7801, - "step": 7347 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019751471284583646, - "loss": 1.671, - "step": 7348 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019751437461949537, - "loss": 1.7128, - "step": 7349 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975140363931543, - "loss": 1.5944, - "step": 7350 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019751369816681325, - "loss": 1.7462, - "step": 7351 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019751335994047217, - "loss": 1.7742, - "step": 7352 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975130217141311, - "loss": 1.8263, - "step": 7353 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019751268348779005, - "loss": 1.7081, - "step": 7354 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019751234526144896, - "loss": 1.5636, - "step": 7355 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975120070351079, - "loss": 1.7552, - "step": 7356 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019751166880876685, - "loss": 1.7477, - "step": 7357 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019751133058242576, - "loss": 1.6964, - "step": 7358 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975109923560847, - "loss": 1.6056, - "step": 7359 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019751065412974364, - "loss": 1.4895, - "step": 7360 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019751031590340256, - "loss": 1.581, - "step": 7361 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975099776770615, - "loss": 1.7366, - "step": 7362 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019750963945072044, - "loss": 1.5403, - "step": 7363 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019750930122437935, - "loss": 1.6488, - "step": 7364 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975089629980383, - "loss": 1.6637, - "step": 7365 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019750862477169723, - "loss": 1.5142, - "step": 7366 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019750828654535615, - "loss": 1.7119, - "step": 7367 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975079483190151, - "loss": 1.6478, - "step": 7368 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019750761009267403, - "loss": 1.7692, - "step": 7369 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019750727186633294, - "loss": 1.7497, - "step": 7370 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019750693363999189, - "loss": 1.6044, - "step": 7371 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019750659541365083, - "loss": 1.6991, - "step": 7372 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019750625718730974, - "loss": 1.7883, - "step": 7373 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019750591896096868, - "loss": 1.63, - "step": 7374 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019750558073462762, - "loss": 1.7743, - "step": 7375 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019750524250828654, - "loss": 1.6843, - "step": 7376 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019750490428194548, - "loss": 1.7088, - "step": 7377 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019750456605560442, - "loss": 1.5841, - "step": 7378 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019750422782926333, - "loss": 1.6141, - "step": 7379 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019750388960292227, - "loss": 1.6822, - "step": 7380 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019750355137658122, - "loss": 1.6513, - "step": 7381 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019750321315024013, - "loss": 1.7355, - "step": 7382 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019750287492389907, - "loss": 1.8708, - "step": 7383 - }, - { - "epoch": 0.02, - "learning_rate": 0.000197502536697558, - "loss": 1.6065, - "step": 7384 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019750219847121693, - "loss": 1.6169, - "step": 7385 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975018602448759, - "loss": 1.5116, - "step": 7386 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019750152201853484, - "loss": 1.5828, - "step": 7387 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019750118379219375, - "loss": 1.7442, - "step": 7388 - }, - { - "epoch": 0.02, - "learning_rate": 0.0001975008455658527, - "loss": 1.7463, - "step": 7389 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019750050733951163, - "loss": 1.6543, - "step": 7390 - }, - { - "epoch": 0.02, - "learning_rate": 0.00019750016911317055, - "loss": 1.6274, - "step": 7391 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974998308868295, - "loss": 1.7145, - "step": 7392 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019749949266048843, - "loss": 1.6797, - "step": 7393 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019749915443414734, - "loss": 1.6683, - "step": 7394 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019749881620780628, - "loss": 1.6806, - "step": 7395 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019749847798146522, - "loss": 1.7302, - "step": 7396 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019749813975512414, - "loss": 1.6141, - "step": 7397 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019749780152878308, - "loss": 1.6065, - "step": 7398 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019749746330244202, - "loss": 1.8599, - "step": 7399 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019749712507610093, - "loss": 1.6632, - "step": 7400 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019749678684975988, - "loss": 1.6682, - "step": 7401 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019749644862341882, - "loss": 1.7444, - "step": 7402 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019749611039707773, - "loss": 1.7236, - "step": 7403 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019749577217073667, - "loss": 1.6737, - "step": 7404 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974954339443956, - "loss": 1.7847, - "step": 7405 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019749509571805453, - "loss": 1.6644, - "step": 7406 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019749475749171347, - "loss": 1.5306, - "step": 7407 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974944192653724, - "loss": 1.7041, - "step": 7408 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019749408103903132, - "loss": 1.5649, - "step": 7409 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019749374281269027, - "loss": 1.6026, - "step": 7410 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974934045863492, - "loss": 1.6452, - "step": 7411 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019749306636000812, - "loss": 1.7839, - "step": 7412 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019749272813366706, - "loss": 1.6075, - "step": 7413 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197492389907326, - "loss": 1.6401, - "step": 7414 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019749205168098492, - "loss": 1.6129, - "step": 7415 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019749171345464386, - "loss": 1.572, - "step": 7416 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974913752283028, - "loss": 1.5705, - "step": 7417 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974910370019617, - "loss": 1.7401, - "step": 7418 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019749069877562065, - "loss": 1.5252, - "step": 7419 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974903605492796, - "loss": 1.8146, - "step": 7420 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974900223229385, - "loss": 1.6367, - "step": 7421 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019748968409659745, - "loss": 1.8041, - "step": 7422 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974893458702564, - "loss": 1.7049, - "step": 7423 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974890076439153, - "loss": 1.7206, - "step": 7424 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019748866941757425, - "loss": 1.6347, - "step": 7425 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974883311912332, - "loss": 1.6388, - "step": 7426 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974879929648921, - "loss": 1.7375, - "step": 7427 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019748765473855104, - "loss": 1.6948, - "step": 7428 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019748731651220998, - "loss": 1.6753, - "step": 7429 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974869782858689, - "loss": 1.5236, - "step": 7430 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019748664005952784, - "loss": 1.5527, - "step": 7431 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019748630183318678, - "loss": 1.7132, - "step": 7432 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974859636068457, - "loss": 1.7481, - "step": 7433 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019748562538050464, - "loss": 1.6339, - "step": 7434 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019748528715416358, - "loss": 1.5929, - "step": 7435 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974849489278225, - "loss": 1.6718, - "step": 7436 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019748461070148143, - "loss": 1.5431, - "step": 7437 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019748427247514037, - "loss": 1.667, - "step": 7438 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974839342487993, - "loss": 1.6729, - "step": 7439 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019748359602245823, - "loss": 1.6646, - "step": 7440 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019748325779611717, - "loss": 1.6924, - "step": 7441 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019748291956977608, - "loss": 1.6366, - "step": 7442 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019748258134343505, - "loss": 1.7222, - "step": 7443 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019748224311709397, - "loss": 1.5579, - "step": 7444 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974819048907529, - "loss": 1.6947, - "step": 7445 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019748156666441185, - "loss": 1.8116, - "step": 7446 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974812284380708, - "loss": 1.6252, - "step": 7447 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974808902117297, - "loss": 1.5739, - "step": 7448 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019748055198538864, - "loss": 1.701, - "step": 7449 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019748021375904759, - "loss": 1.6153, - "step": 7450 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974798755327065, - "loss": 1.7084, - "step": 7451 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019747953730636544, - "loss": 1.726, - "step": 7452 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019747919908002438, - "loss": 1.7021, - "step": 7453 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974788608536833, - "loss": 1.615, - "step": 7454 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019747852262734224, - "loss": 1.6288, - "step": 7455 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019747818440100118, - "loss": 1.6398, - "step": 7456 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974778461746601, - "loss": 1.5738, - "step": 7457 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019747750794831903, - "loss": 1.6504, - "step": 7458 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019747716972197797, - "loss": 1.7281, - "step": 7459 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974768314956369, - "loss": 1.6519, - "step": 7460 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019747649326929583, - "loss": 1.7352, - "step": 7461 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019747615504295477, - "loss": 1.7088, - "step": 7462 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019747581681661368, - "loss": 1.6199, - "step": 7463 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019747547859027263, - "loss": 1.6416, - "step": 7464 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019747514036393157, - "loss": 1.8573, - "step": 7465 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019747480213759048, - "loss": 1.7585, - "step": 7466 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019747446391124942, - "loss": 1.7951, - "step": 7467 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019747412568490836, - "loss": 1.635, - "step": 7468 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019747378745856728, - "loss": 1.7354, - "step": 7469 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019747344923222622, - "loss": 1.787, - "step": 7470 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019747311100588516, - "loss": 1.7352, - "step": 7471 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019747277277954407, - "loss": 1.5636, - "step": 7472 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019747243455320301, - "loss": 1.5823, - "step": 7473 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019747209632686196, - "loss": 1.5648, - "step": 7474 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019747175810052087, - "loss": 1.7269, - "step": 7475 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974714198741798, - "loss": 1.6075, - "step": 7476 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019747108164783875, - "loss": 1.6715, - "step": 7477 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019747074342149767, - "loss": 1.5575, - "step": 7478 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974704051951566, - "loss": 1.6524, - "step": 7479 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019747006696881555, - "loss": 1.7364, - "step": 7480 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019746972874247446, - "loss": 1.5854, - "step": 7481 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974693905161334, - "loss": 1.5356, - "step": 7482 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019746905228979234, - "loss": 1.616, - "step": 7483 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019746871406345126, - "loss": 1.6578, - "step": 7484 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974683758371102, - "loss": 1.7226, - "step": 7485 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019746803761076914, - "loss": 1.5312, - "step": 7486 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019746769938442806, - "loss": 1.8427, - "step": 7487 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197467361158087, - "loss": 1.7707, - "step": 7488 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019746702293174594, - "loss": 1.7514, - "step": 7489 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019746668470540485, - "loss": 1.6612, - "step": 7490 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974663464790638, - "loss": 1.6296, - "step": 7491 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019746600825272273, - "loss": 1.7321, - "step": 7492 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019746567002638165, - "loss": 1.777, - "step": 7493 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974653318000406, - "loss": 1.5822, - "step": 7494 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019746499357369953, - "loss": 1.6326, - "step": 7495 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019746465534735844, - "loss": 1.6013, - "step": 7496 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019746431712101739, - "loss": 1.6355, - "step": 7497 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019746397889467633, - "loss": 1.705, - "step": 7498 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019746364066833527, - "loss": 1.6431, - "step": 7499 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974633024419942, - "loss": 1.523, - "step": 7500 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019746296421565312, - "loss": 1.7154, - "step": 7501 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019746262598931206, - "loss": 1.6708, - "step": 7502 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197462287762971, - "loss": 1.6458, - "step": 7503 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019746194953662992, - "loss": 1.6573, - "step": 7504 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019746161131028886, - "loss": 1.7533, - "step": 7505 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974612730839478, - "loss": 1.6546, - "step": 7506 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019746093485760672, - "loss": 1.5649, - "step": 7507 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019746059663126566, - "loss": 1.608, - "step": 7508 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974602584049246, - "loss": 1.68, - "step": 7509 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974599201785835, - "loss": 1.6972, - "step": 7510 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019745958195224245, - "loss": 1.6295, - "step": 7511 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974592437259014, - "loss": 1.7932, - "step": 7512 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974589054995603, - "loss": 1.739, - "step": 7513 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019745856727321925, - "loss": 1.5499, - "step": 7514 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974582290468782, - "loss": 1.5907, - "step": 7515 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974578908205371, - "loss": 1.7258, - "step": 7516 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019745755259419605, - "loss": 1.6716, - "step": 7517 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197457214367855, - "loss": 1.7176, - "step": 7518 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974568761415139, - "loss": 1.5531, - "step": 7519 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019745653791517284, - "loss": 1.5969, - "step": 7520 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019745619968883178, - "loss": 1.6155, - "step": 7521 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974558614624907, - "loss": 1.5697, - "step": 7522 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019745552323614964, - "loss": 1.4803, - "step": 7523 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019745518500980858, - "loss": 1.8116, - "step": 7524 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974548467834675, - "loss": 1.4464, - "step": 7525 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019745450855712643, - "loss": 1.6616, - "step": 7526 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019745417033078538, - "loss": 1.6815, - "step": 7527 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974538321044443, - "loss": 1.6808, - "step": 7528 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019745349387810323, - "loss": 1.6096, - "step": 7529 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019745315565176217, - "loss": 1.7787, - "step": 7530 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974528174254211, - "loss": 1.701, - "step": 7531 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019745247919908003, - "loss": 1.8497, - "step": 7532 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019745214097273897, - "loss": 1.7131, - "step": 7533 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974518027463979, - "loss": 1.7795, - "step": 7534 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019745146452005682, - "loss": 1.548, - "step": 7535 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019745112629371576, - "loss": 1.7292, - "step": 7536 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974507880673747, - "loss": 1.6002, - "step": 7537 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019745044984103362, - "loss": 1.6138, - "step": 7538 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019745011161469256, - "loss": 1.5854, - "step": 7539 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974497733883515, - "loss": 1.6962, - "step": 7540 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019744943516201042, - "loss": 1.7679, - "step": 7541 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019744909693566936, - "loss": 1.4482, - "step": 7542 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974487587093283, - "loss": 1.6624, - "step": 7543 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974484204829872, - "loss": 1.5132, - "step": 7544 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019744808225664615, - "loss": 1.7016, - "step": 7545 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974477440303051, - "loss": 1.7266, - "step": 7546 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197447405803964, - "loss": 1.5589, - "step": 7547 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019744706757762295, - "loss": 1.5737, - "step": 7548 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974467293512819, - "loss": 1.6849, - "step": 7549 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974463911249408, - "loss": 1.708, - "step": 7550 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019744605289859975, - "loss": 1.7636, - "step": 7551 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974457146722587, - "loss": 1.6353, - "step": 7552 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974453764459176, - "loss": 1.7696, - "step": 7553 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019744503821957654, - "loss": 1.83, - "step": 7554 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019744469999323548, - "loss": 1.614, - "step": 7555 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019744436176689442, - "loss": 1.5775, - "step": 7556 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019744402354055337, - "loss": 1.6212, - "step": 7557 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019744368531421228, - "loss": 1.8, - "step": 7558 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019744334708787122, - "loss": 1.7219, - "step": 7559 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019744300886153016, - "loss": 1.6148, - "step": 7560 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019744267063518908, - "loss": 1.6342, - "step": 7561 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019744233240884802, - "loss": 1.6761, - "step": 7562 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019744199418250696, - "loss": 1.7909, - "step": 7563 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019744165595616587, - "loss": 1.7827, - "step": 7564 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974413177298248, - "loss": 1.649, - "step": 7565 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019744097950348375, - "loss": 1.5803, - "step": 7566 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019744064127714267, - "loss": 1.598, - "step": 7567 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974403030508016, - "loss": 1.4461, - "step": 7568 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019743996482446055, - "loss": 1.7544, - "step": 7569 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019743962659811946, - "loss": 1.8085, - "step": 7570 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974392883717784, - "loss": 1.8355, - "step": 7571 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019743895014543735, - "loss": 1.6515, - "step": 7572 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019743861191909626, - "loss": 1.4902, - "step": 7573 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974382736927552, - "loss": 1.6528, - "step": 7574 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019743793546641414, - "loss": 1.5344, - "step": 7575 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019743759724007306, - "loss": 1.7238, - "step": 7576 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197437259013732, - "loss": 1.7932, - "step": 7577 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019743692078739094, - "loss": 1.7101, - "step": 7578 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019743658256104985, - "loss": 1.6622, - "step": 7579 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974362443347088, - "loss": 1.6987, - "step": 7580 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019743590610836774, - "loss": 1.6197, - "step": 7581 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019743556788202665, - "loss": 1.618, - "step": 7582 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974352296556856, - "loss": 1.6838, - "step": 7583 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019743489142934453, - "loss": 1.7606, - "step": 7584 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019743455320300345, - "loss": 1.7666, - "step": 7585 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974342149766624, - "loss": 1.6523, - "step": 7586 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019743387675032133, - "loss": 1.748, - "step": 7587 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019743353852398024, - "loss": 1.7374, - "step": 7588 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019743320029763918, - "loss": 1.7579, - "step": 7589 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019743286207129813, - "loss": 1.7452, - "step": 7590 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019743252384495704, - "loss": 1.7536, - "step": 7591 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019743218561861598, - "loss": 1.6199, - "step": 7592 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019743184739227492, - "loss": 1.7256, - "step": 7593 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019743150916593384, - "loss": 1.7932, - "step": 7594 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019743117093959278, - "loss": 1.7186, - "step": 7595 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019743083271325172, - "loss": 1.8103, - "step": 7596 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019743049448691063, - "loss": 1.7392, - "step": 7597 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019743015626056957, - "loss": 1.6465, - "step": 7598 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019742981803422851, - "loss": 1.5489, - "step": 7599 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019742947980788743, - "loss": 1.5891, - "step": 7600 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019742914158154637, - "loss": 1.6655, - "step": 7601 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974288033552053, - "loss": 1.6879, - "step": 7602 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019742846512886422, - "loss": 1.6529, - "step": 7603 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019742812690252317, - "loss": 1.6813, - "step": 7604 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974277886761821, - "loss": 1.5437, - "step": 7605 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019742745044984102, - "loss": 1.679, - "step": 7606 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019742711222349996, - "loss": 1.5225, - "step": 7607 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974267739971589, - "loss": 1.6046, - "step": 7608 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019742643577081782, - "loss": 1.6627, - "step": 7609 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019742609754447676, - "loss": 1.7344, - "step": 7610 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019742575931813573, - "loss": 1.6576, - "step": 7611 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019742542109179464, - "loss": 1.8362, - "step": 7612 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019742508286545358, - "loss": 1.6582, - "step": 7613 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019742474463911252, - "loss": 1.6077, - "step": 7614 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019742440641277144, - "loss": 1.6197, - "step": 7615 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019742406818643038, - "loss": 1.8477, - "step": 7616 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019742372996008932, - "loss": 1.7005, - "step": 7617 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019742339173374823, - "loss": 1.6339, - "step": 7618 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019742305350740717, - "loss": 1.8388, - "step": 7619 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019742271528106612, - "loss": 1.7321, - "step": 7620 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019742237705472503, - "loss": 1.5702, - "step": 7621 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019742203882838397, - "loss": 1.5418, - "step": 7622 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974217006020429, - "loss": 1.6475, - "step": 7623 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019742136237570183, - "loss": 1.6875, - "step": 7624 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019742102414936077, - "loss": 1.717, - "step": 7625 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974206859230197, - "loss": 1.8303, - "step": 7626 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019742034769667862, - "loss": 1.7045, - "step": 7627 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019742000947033756, - "loss": 1.8254, - "step": 7628 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974196712439965, - "loss": 1.7239, - "step": 7629 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019741933301765542, - "loss": 1.665, - "step": 7630 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019741899479131436, - "loss": 1.7095, - "step": 7631 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974186565649733, - "loss": 1.8125, - "step": 7632 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019741831833863221, - "loss": 1.6785, - "step": 7633 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019741798011229116, - "loss": 1.8039, - "step": 7634 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974176418859501, - "loss": 1.7281, - "step": 7635 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197417303659609, - "loss": 1.8816, - "step": 7636 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019741696543326795, - "loss": 1.7224, - "step": 7637 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974166272069269, - "loss": 1.5955, - "step": 7638 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974162889805858, - "loss": 1.6459, - "step": 7639 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019741595075424475, - "loss": 1.6241, - "step": 7640 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974156125279037, - "loss": 1.6546, - "step": 7641 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974152743015626, - "loss": 1.645, - "step": 7642 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019741493607522154, - "loss": 1.6723, - "step": 7643 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019741459784888049, - "loss": 1.4857, - "step": 7644 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974142596225394, - "loss": 1.812, - "step": 7645 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019741392139619834, - "loss": 1.6509, - "step": 7646 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019741358316985728, - "loss": 1.7532, - "step": 7647 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974132449435162, - "loss": 1.769, - "step": 7648 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019741290671717514, - "loss": 1.6539, - "step": 7649 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019741256849083408, - "loss": 1.5903, - "step": 7650 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197412230264493, - "loss": 1.5866, - "step": 7651 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019741189203815193, - "loss": 1.4934, - "step": 7652 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019741155381181087, - "loss": 1.7346, - "step": 7653 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974112155854698, - "loss": 1.5236, - "step": 7654 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019741087735912873, - "loss": 1.7795, - "step": 7655 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019741053913278767, - "loss": 1.7004, - "step": 7656 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019741020090644659, - "loss": 1.6518, - "step": 7657 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019740986268010553, - "loss": 1.8168, - "step": 7658 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019740952445376447, - "loss": 1.7005, - "step": 7659 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019740918622742338, - "loss": 1.7471, - "step": 7660 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019740884800108232, - "loss": 1.902, - "step": 7661 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019740850977474126, - "loss": 1.6287, - "step": 7662 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019740817154840018, - "loss": 1.6573, - "step": 7663 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019740783332205912, - "loss": 1.6785, - "step": 7664 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019740749509571806, - "loss": 1.8767, - "step": 7665 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019740715686937697, - "loss": 1.5459, - "step": 7666 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019740681864303594, - "loss": 1.6955, - "step": 7667 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019740648041669488, - "loss": 1.6422, - "step": 7668 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974061421903538, - "loss": 1.5405, - "step": 7669 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019740580396401274, - "loss": 1.7378, - "step": 7670 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019740546573767168, - "loss": 1.6116, - "step": 7671 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974051275113306, - "loss": 1.565, - "step": 7672 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019740478928498954, - "loss": 1.8174, - "step": 7673 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019740445105864848, - "loss": 1.6902, - "step": 7674 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974041128323074, - "loss": 1.6032, - "step": 7675 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019740377460596633, - "loss": 1.6969, - "step": 7676 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019740343637962527, - "loss": 1.8009, - "step": 7677 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001974030981532842, - "loss": 1.7459, - "step": 7678 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019740275992694313, - "loss": 1.8705, - "step": 7679 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019740242170060207, - "loss": 1.6529, - "step": 7680 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019740208347426098, - "loss": 1.6842, - "step": 7681 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019740174524791992, - "loss": 1.518, - "step": 7682 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019740140702157887, - "loss": 1.6833, - "step": 7683 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019740106879523778, - "loss": 1.8001, - "step": 7684 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019740073056889672, - "loss": 1.5786, - "step": 7685 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019740039234255566, - "loss": 1.7613, - "step": 7686 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019740005411621458, - "loss": 1.7399, - "step": 7687 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019739971588987352, - "loss": 1.6756, - "step": 7688 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019739937766353246, - "loss": 1.6954, - "step": 7689 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019739903943719137, - "loss": 1.7524, - "step": 7690 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973987012108503, - "loss": 1.7642, - "step": 7691 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019739836298450925, - "loss": 1.7492, - "step": 7692 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019739802475816817, - "loss": 1.7146, - "step": 7693 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973976865318271, - "loss": 1.6786, - "step": 7694 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019739734830548605, - "loss": 1.7578, - "step": 7695 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019739701007914496, - "loss": 1.7021, - "step": 7696 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973966718528039, - "loss": 1.5961, - "step": 7697 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019739633362646285, - "loss": 1.6869, - "step": 7698 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019739599540012176, - "loss": 1.7257, - "step": 7699 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973956571737807, - "loss": 1.7641, - "step": 7700 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019739531894743964, - "loss": 1.735, - "step": 7701 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019739498072109856, - "loss": 1.7297, - "step": 7702 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973946424947575, - "loss": 1.7748, - "step": 7703 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019739430426841644, - "loss": 1.6183, - "step": 7704 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019739396604207535, - "loss": 1.7028, - "step": 7705 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973936278157343, - "loss": 1.5414, - "step": 7706 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019739328958939324, - "loss": 1.6289, - "step": 7707 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019739295136305215, - "loss": 1.6398, - "step": 7708 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973926131367111, - "loss": 1.6097, - "step": 7709 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019739227491037003, - "loss": 1.5829, - "step": 7710 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019739193668402895, - "loss": 1.6952, - "step": 7711 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973915984576879, - "loss": 1.6113, - "step": 7712 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019739126023134683, - "loss": 1.7689, - "step": 7713 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019739092200500574, - "loss": 1.6401, - "step": 7714 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019739058377866468, - "loss": 1.7084, - "step": 7715 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019739024555232362, - "loss": 1.7671, - "step": 7716 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019738990732598254, - "loss": 1.6751, - "step": 7717 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019738956909964148, - "loss": 1.7032, - "step": 7718 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019738923087330042, - "loss": 1.5409, - "step": 7719 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019738889264695933, - "loss": 1.6831, - "step": 7720 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019738855442061828, - "loss": 1.8386, - "step": 7721 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019738821619427722, - "loss": 1.7066, - "step": 7722 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019738787796793616, - "loss": 1.5328, - "step": 7723 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973875397415951, - "loss": 1.7185, - "step": 7724 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019738720151525404, - "loss": 1.7638, - "step": 7725 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019738686328891295, - "loss": 1.7492, - "step": 7726 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973865250625719, - "loss": 1.7045, - "step": 7727 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019738618683623084, - "loss": 1.7605, - "step": 7728 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019738584860988975, - "loss": 1.6078, - "step": 7729 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973855103835487, - "loss": 1.6475, - "step": 7730 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019738517215720763, - "loss": 1.7175, - "step": 7731 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019738483393086655, - "loss": 1.6466, - "step": 7732 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973844957045255, - "loss": 1.7094, - "step": 7733 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019738415747818443, - "loss": 1.6644, - "step": 7734 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019738381925184334, - "loss": 1.5359, - "step": 7735 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019738348102550228, - "loss": 1.6093, - "step": 7736 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019738314279916123, - "loss": 1.5861, - "step": 7737 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019738280457282014, - "loss": 1.7213, - "step": 7738 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019738246634647908, - "loss": 1.8359, - "step": 7739 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019738212812013802, - "loss": 1.6832, - "step": 7740 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019738178989379694, - "loss": 1.5239, - "step": 7741 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019738145166745588, - "loss": 1.7247, - "step": 7742 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019738111344111482, - "loss": 1.6855, - "step": 7743 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019738077521477373, - "loss": 1.528, - "step": 7744 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019738043698843267, - "loss": 1.6475, - "step": 7745 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019738009876209161, - "loss": 1.673, - "step": 7746 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019737976053575053, - "loss": 1.6019, - "step": 7747 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019737942230940947, - "loss": 1.6851, - "step": 7748 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973790840830684, - "loss": 1.5391, - "step": 7749 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019737874585672733, - "loss": 1.7374, - "step": 7750 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019737840763038627, - "loss": 1.7464, - "step": 7751 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973780694040452, - "loss": 1.6459, - "step": 7752 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019737773117770412, - "loss": 1.5853, - "step": 7753 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019737739295136306, - "loss": 1.6136, - "step": 7754 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197377054725022, - "loss": 1.5742, - "step": 7755 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019737671649868092, - "loss": 1.5765, - "step": 7756 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019737637827233986, - "loss": 1.842, - "step": 7757 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973760400459988, - "loss": 1.6071, - "step": 7758 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019737570181965771, - "loss": 1.7935, - "step": 7759 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019737536359331666, - "loss": 1.8002, - "step": 7760 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973750253669756, - "loss": 1.5565, - "step": 7761 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973746871406345, - "loss": 1.6226, - "step": 7762 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019737434891429345, - "loss": 1.7854, - "step": 7763 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973740106879524, - "loss": 1.7147, - "step": 7764 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973736724616113, - "loss": 1.6068, - "step": 7765 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019737333423527025, - "loss": 1.5734, - "step": 7766 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973729960089292, - "loss": 1.743, - "step": 7767 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973726577825881, - "loss": 1.7174, - "step": 7768 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019737231955624704, - "loss": 1.4954, - "step": 7769 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019737198132990599, - "loss": 1.6834, - "step": 7770 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973716431035649, - "loss": 1.8495, - "step": 7771 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019737130487722384, - "loss": 1.6791, - "step": 7772 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019737096665088278, - "loss": 1.7421, - "step": 7773 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973706284245417, - "loss": 1.7019, - "step": 7774 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019737029019820064, - "loss": 1.4334, - "step": 7775 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019736995197185958, - "loss": 1.7166, - "step": 7776 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973696137455185, - "loss": 1.6642, - "step": 7777 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019736927551917743, - "loss": 1.6968, - "step": 7778 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019736893729283637, - "loss": 1.6723, - "step": 7779 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019736859906649532, - "loss": 1.477, - "step": 7780 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019736826084015426, - "loss": 1.6632, - "step": 7781 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019736792261381317, - "loss": 1.5992, - "step": 7782 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973675843874721, - "loss": 1.6683, - "step": 7783 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019736724616113105, - "loss": 1.7797, - "step": 7784 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019736690793478997, - "loss": 1.5792, - "step": 7785 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973665697084489, - "loss": 1.6844, - "step": 7786 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019736623148210785, - "loss": 1.7786, - "step": 7787 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019736589325576676, - "loss": 1.5277, - "step": 7788 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973655550294257, - "loss": 1.8547, - "step": 7789 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019736521680308465, - "loss": 1.6969, - "step": 7790 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019736487857674356, - "loss": 1.5134, - "step": 7791 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973645403504025, - "loss": 1.649, - "step": 7792 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019736420212406144, - "loss": 1.5885, - "step": 7793 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019736386389772036, - "loss": 1.4991, - "step": 7794 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973635256713793, - "loss": 1.6406, - "step": 7795 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019736318744503824, - "loss": 1.5829, - "step": 7796 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019736284921869715, - "loss": 1.6383, - "step": 7797 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973625109923561, - "loss": 1.6723, - "step": 7798 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019736217276601503, - "loss": 1.82, - "step": 7799 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019736183453967398, - "loss": 1.7496, - "step": 7800 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973614963133329, - "loss": 1.6659, - "step": 7801 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019736115808699183, - "loss": 1.7858, - "step": 7802 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019736081986065077, - "loss": 1.6637, - "step": 7803 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019736048163430969, - "loss": 1.5965, - "step": 7804 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019736014340796863, - "loss": 1.7392, - "step": 7805 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019735980518162757, - "loss": 1.7285, - "step": 7806 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019735946695528648, - "loss": 1.6931, - "step": 7807 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019735912872894542, - "loss": 1.6328, - "step": 7808 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019735879050260436, - "loss": 1.6313, - "step": 7809 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019735845227626328, - "loss": 1.687, - "step": 7810 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019735811404992222, - "loss": 1.7097, - "step": 7811 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019735777582358116, - "loss": 1.7963, - "step": 7812 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019735743759724007, - "loss": 1.753, - "step": 7813 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019735709937089902, - "loss": 1.7211, - "step": 7814 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019735676114455796, - "loss": 1.5157, - "step": 7815 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019735642291821687, - "loss": 1.6689, - "step": 7816 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973560846918758, - "loss": 1.5269, - "step": 7817 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019735574646553475, - "loss": 1.7485, - "step": 7818 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019735540823919367, - "loss": 1.656, - "step": 7819 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973550700128526, - "loss": 1.6189, - "step": 7820 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019735473178651155, - "loss": 1.6559, - "step": 7821 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019735439356017046, - "loss": 1.7101, - "step": 7822 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973540553338294, - "loss": 1.7349, - "step": 7823 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019735371710748835, - "loss": 1.6757, - "step": 7824 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019735337888114726, - "loss": 1.7668, - "step": 7825 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973530406548062, - "loss": 1.7774, - "step": 7826 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019735270242846514, - "loss": 1.557, - "step": 7827 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019735236420212406, - "loss": 1.5762, - "step": 7828 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197352025975783, - "loss": 1.5255, - "step": 7829 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019735168774944194, - "loss": 1.5441, - "step": 7830 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019735134952310085, - "loss": 1.659, - "step": 7831 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973510112967598, - "loss": 1.6986, - "step": 7832 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019735067307041873, - "loss": 1.6626, - "step": 7833 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019735033484407765, - "loss": 1.7451, - "step": 7834 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973499966177366, - "loss": 1.6193, - "step": 7835 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019734965839139553, - "loss": 1.7473, - "step": 7836 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019734932016505447, - "loss": 1.6812, - "step": 7837 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019734898193871341, - "loss": 1.7551, - "step": 7838 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019734864371237233, - "loss": 1.683, - "step": 7839 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019734830548603127, - "loss": 1.6359, - "step": 7840 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973479672596902, - "loss": 1.7289, - "step": 7841 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019734762903334912, - "loss": 1.6275, - "step": 7842 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019734729080700807, - "loss": 1.6326, - "step": 7843 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197346952580667, - "loss": 1.589, - "step": 7844 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019734661435432592, - "loss": 1.7464, - "step": 7845 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019734627612798486, - "loss": 1.5654, - "step": 7846 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973459379016438, - "loss": 1.6856, - "step": 7847 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019734559967530272, - "loss": 1.6632, - "step": 7848 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019734526144896166, - "loss": 1.7421, - "step": 7849 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973449232226206, - "loss": 1.5711, - "step": 7850 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973445849962795, - "loss": 1.7239, - "step": 7851 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019734424676993845, - "loss": 1.6582, - "step": 7852 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973439085435974, - "loss": 1.821, - "step": 7853 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973435703172563, - "loss": 1.6756, - "step": 7854 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019734323209091525, - "loss": 1.6545, - "step": 7855 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973428938645742, - "loss": 1.6617, - "step": 7856 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973425556382331, - "loss": 1.7016, - "step": 7857 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019734221741189205, - "loss": 1.4223, - "step": 7858 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197341879185551, - "loss": 1.5902, - "step": 7859 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973415409592099, - "loss": 1.698, - "step": 7860 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019734120273286884, - "loss": 1.7754, - "step": 7861 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019734086450652778, - "loss": 1.658, - "step": 7862 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973405262801867, - "loss": 1.7344, - "step": 7863 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019734018805384564, - "loss": 1.7316, - "step": 7864 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019733984982750458, - "loss": 1.7998, - "step": 7865 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973395116011635, - "loss": 1.7314, - "step": 7866 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019733917337482244, - "loss": 1.8072, - "step": 7867 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019733883514848138, - "loss": 1.6762, - "step": 7868 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973384969221403, - "loss": 1.6803, - "step": 7869 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019733815869579923, - "loss": 1.6057, - "step": 7870 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019733782046945817, - "loss": 1.5824, - "step": 7871 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973374822431171, - "loss": 1.7532, - "step": 7872 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019733714401677603, - "loss": 1.8147, - "step": 7873 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019733680579043497, - "loss": 1.7005, - "step": 7874 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019733646756409388, - "loss": 1.5986, - "step": 7875 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019733612933775282, - "loss": 1.7133, - "step": 7876 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019733579111141177, - "loss": 1.6417, - "step": 7877 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019733545288507068, - "loss": 1.7029, - "step": 7878 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019733511465872962, - "loss": 1.7011, - "step": 7879 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019733477643238856, - "loss": 1.6137, - "step": 7880 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019733443820604748, - "loss": 1.7466, - "step": 7881 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019733409997970642, - "loss": 1.5141, - "step": 7882 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019733376175336536, - "loss": 1.57, - "step": 7883 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973334235270243, - "loss": 1.5958, - "step": 7884 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973330853006832, - "loss": 1.6271, - "step": 7885 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019733274707434215, - "loss": 1.6474, - "step": 7886 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973324088480011, - "loss": 1.6303, - "step": 7887 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019733207062166, - "loss": 1.8331, - "step": 7888 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019733173239531895, - "loss": 1.6536, - "step": 7889 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973313941689779, - "loss": 1.8689, - "step": 7890 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973310559426368, - "loss": 1.6198, - "step": 7891 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019733071771629577, - "loss": 1.813, - "step": 7892 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973303794899547, - "loss": 1.6762, - "step": 7893 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019733004126361363, - "loss": 1.546, - "step": 7894 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019732970303727257, - "loss": 1.5861, - "step": 7895 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019732936481093148, - "loss": 1.6201, - "step": 7896 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019732902658459043, - "loss": 1.8122, - "step": 7897 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019732868835824937, - "loss": 1.7028, - "step": 7898 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019732835013190828, - "loss": 1.807, - "step": 7899 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019732801190556722, - "loss": 1.5234, - "step": 7900 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019732767367922616, - "loss": 1.6438, - "step": 7901 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019732733545288508, - "loss": 1.668, - "step": 7902 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019732699722654402, - "loss": 1.8079, - "step": 7903 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019732665900020296, - "loss": 1.5493, - "step": 7904 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019732632077386187, - "loss": 1.5838, - "step": 7905 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019732598254752081, - "loss": 1.6848, - "step": 7906 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019732564432117976, - "loss": 1.5653, - "step": 7907 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019732530609483867, - "loss": 1.6774, - "step": 7908 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973249678684976, - "loss": 1.7373, - "step": 7909 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019732462964215655, - "loss": 1.6646, - "step": 7910 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019732429141581547, - "loss": 1.689, - "step": 7911 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973239531894744, - "loss": 1.6742, - "step": 7912 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019732361496313335, - "loss": 1.7073, - "step": 7913 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019732327673679226, - "loss": 1.677, - "step": 7914 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973229385104512, - "loss": 1.5803, - "step": 7915 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019732260028411014, - "loss": 1.6243, - "step": 7916 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019732226205776906, - "loss": 1.7535, - "step": 7917 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197321923831428, - "loss": 1.6233, - "step": 7918 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019732158560508694, - "loss": 1.6856, - "step": 7919 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019732124737874586, - "loss": 1.7271, - "step": 7920 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973209091524048, - "loss": 1.5158, - "step": 7921 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019732057092606374, - "loss": 1.5741, - "step": 7922 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019732023269972265, - "loss": 1.6933, - "step": 7923 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973198944733816, - "loss": 1.8514, - "step": 7924 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019731955624704053, - "loss": 1.7209, - "step": 7925 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019731921802069945, - "loss": 1.6389, - "step": 7926 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973188797943584, - "loss": 1.6562, - "step": 7927 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019731854156801733, - "loss": 1.6197, - "step": 7928 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019731820334167624, - "loss": 1.5206, - "step": 7929 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019731786511533519, - "loss": 1.754, - "step": 7930 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019731752688899413, - "loss": 1.7776, - "step": 7931 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019731718866265304, - "loss": 1.6045, - "step": 7932 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019731685043631198, - "loss": 1.7017, - "step": 7933 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019731651220997092, - "loss": 1.7324, - "step": 7934 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019731617398362984, - "loss": 1.5046, - "step": 7935 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019731583575728878, - "loss": 1.7688, - "step": 7936 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019731549753094772, - "loss": 1.6659, - "step": 7937 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019731515930460663, - "loss": 1.8561, - "step": 7938 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019731482107826557, - "loss": 1.7288, - "step": 7939 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019731448285192452, - "loss": 1.6928, - "step": 7940 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019731414462558343, - "loss": 1.4034, - "step": 7941 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019731380639924237, - "loss": 1.6445, - "step": 7942 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973134681729013, - "loss": 1.6404, - "step": 7943 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019731312994656023, - "loss": 1.6288, - "step": 7944 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019731279172021917, - "loss": 1.5741, - "step": 7945 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973124534938781, - "loss": 1.5973, - "step": 7946 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019731211526753702, - "loss": 1.6499, - "step": 7947 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197311777041196, - "loss": 1.7351, - "step": 7948 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019731143881485493, - "loss": 1.5443, - "step": 7949 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019731110058851385, - "loss": 1.7187, - "step": 7950 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973107623621728, - "loss": 1.5745, - "step": 7951 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019731042413583173, - "loss": 1.91, - "step": 7952 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019731008590949064, - "loss": 1.8878, - "step": 7953 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019730974768314958, - "loss": 1.7762, - "step": 7954 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019730940945680852, - "loss": 1.7215, - "step": 7955 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019730907123046744, - "loss": 1.6996, - "step": 7956 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019730873300412638, - "loss": 1.6578, - "step": 7957 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019730839477778532, - "loss": 1.8115, - "step": 7958 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019730805655144423, - "loss": 1.7171, - "step": 7959 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019730771832510318, - "loss": 1.7575, - "step": 7960 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019730738009876212, - "loss": 1.481, - "step": 7961 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019730704187242103, - "loss": 1.5452, - "step": 7962 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019730670364607997, - "loss": 1.7048, - "step": 7963 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973063654197389, - "loss": 1.6757, - "step": 7964 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019730602719339783, - "loss": 1.6443, - "step": 7965 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019730568896705677, - "loss": 1.729, - "step": 7966 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973053507407157, - "loss": 1.6219, - "step": 7967 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019730501251437462, - "loss": 1.5697, - "step": 7968 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019730467428803356, - "loss": 1.5809, - "step": 7969 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973043360616925, - "loss": 1.5872, - "step": 7970 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019730399783535142, - "loss": 1.659, - "step": 7971 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019730365960901036, - "loss": 1.6843, - "step": 7972 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973033213826693, - "loss": 1.7176, - "step": 7973 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019730298315632822, - "loss": 1.5916, - "step": 7974 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019730264492998716, - "loss": 1.6974, - "step": 7975 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973023067036461, - "loss": 1.7279, - "step": 7976 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197301968477305, - "loss": 1.6579, - "step": 7977 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019730163025096395, - "loss": 1.7778, - "step": 7978 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973012920246229, - "loss": 1.6255, - "step": 7979 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973009537982818, - "loss": 1.768, - "step": 7980 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019730061557194075, - "loss": 1.6811, - "step": 7981 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001973002773455997, - "loss": 1.6692, - "step": 7982 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972999391192586, - "loss": 1.6751, - "step": 7983 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019729960089291755, - "loss": 1.6984, - "step": 7984 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972992626665765, - "loss": 1.6451, - "step": 7985 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972989244402354, - "loss": 1.5182, - "step": 7986 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019729858621389434, - "loss": 1.5453, - "step": 7987 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019729824798755328, - "loss": 1.6817, - "step": 7988 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972979097612122, - "loss": 1.696, - "step": 7989 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019729757153487114, - "loss": 1.5831, - "step": 7990 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019729723330853008, - "loss": 1.6045, - "step": 7991 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197296895082189, - "loss": 1.6894, - "step": 7992 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019729655685584793, - "loss": 1.6752, - "step": 7993 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019729621862950688, - "loss": 1.802, - "step": 7994 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972958804031658, - "loss": 1.741, - "step": 7995 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019729554217682473, - "loss": 1.806, - "step": 7996 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019729520395048367, - "loss": 1.6827, - "step": 7997 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019729486572414259, - "loss": 1.6631, - "step": 7998 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019729452749780153, - "loss": 1.5628, - "step": 7999 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019729418927146047, - "loss": 1.6841, - "step": 8000 - }, - { - "epoch": 0.03, - "eval_loss": 1.3000676081730769, - "mrc_em": 14.0, - "mrc_f1": 24.7857142857143, - "nli_acc": 0.22, - "step": 8000, - "sts_f1": 0.5555555555555556, - "ynat_f1": 0.6331722627643405 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019729385104511938, - "loss": 1.5002, - "step": 8001 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019729351281877832, - "loss": 1.6907, - "step": 8002 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019729317459243727, - "loss": 1.7596, - "step": 8003 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972928363660962, - "loss": 1.6599, - "step": 8004 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019729249813975515, - "loss": 1.5739, - "step": 8005 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972921599134141, - "loss": 1.6098, - "step": 8006 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197291821687073, - "loss": 1.6076, - "step": 8007 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019729148346073194, - "loss": 1.7261, - "step": 8008 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019729114523439088, - "loss": 1.6905, - "step": 8009 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972908070080498, - "loss": 1.6742, - "step": 8010 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019729046878170874, - "loss": 1.5814, - "step": 8011 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019729013055536768, - "loss": 1.7072, - "step": 8012 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972897923290266, - "loss": 1.8087, - "step": 8013 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019728945410268554, - "loss": 1.8579, - "step": 8014 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019728911587634448, - "loss": 1.7256, - "step": 8015 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972887776500034, - "loss": 1.5334, - "step": 8016 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019728843942366233, - "loss": 1.4918, - "step": 8017 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019728810119732127, - "loss": 1.5994, - "step": 8018 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972877629709802, - "loss": 1.6319, - "step": 8019 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019728742474463913, - "loss": 1.6239, - "step": 8020 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019728708651829807, - "loss": 1.7908, - "step": 8021 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019728674829195698, - "loss": 1.6874, - "step": 8022 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019728641006561593, - "loss": 1.5997, - "step": 8023 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019728607183927487, - "loss": 1.6754, - "step": 8024 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019728573361293378, - "loss": 1.7593, - "step": 8025 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019728539538659272, - "loss": 1.5713, - "step": 8026 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019728505716025166, - "loss": 1.6625, - "step": 8027 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019728471893391058, - "loss": 1.7539, - "step": 8028 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019728438070756952, - "loss": 1.7026, - "step": 8029 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019728404248122846, - "loss": 1.7774, - "step": 8030 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019728370425488737, - "loss": 1.7426, - "step": 8031 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019728336602854631, - "loss": 1.5834, - "step": 8032 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019728302780220526, - "loss": 1.7476, - "step": 8033 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019728268957586417, - "loss": 1.7571, - "step": 8034 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972823513495231, - "loss": 1.5348, - "step": 8035 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019728201312318205, - "loss": 1.6323, - "step": 8036 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019728167489684097, - "loss": 1.7461, - "step": 8037 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972813366704999, - "loss": 1.729, - "step": 8038 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019728099844415885, - "loss": 1.7247, - "step": 8039 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019728066021781776, - "loss": 1.7043, - "step": 8040 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972803219914767, - "loss": 1.6968, - "step": 8041 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019727998376513564, - "loss": 1.6008, - "step": 8042 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019727964553879456, - "loss": 1.6333, - "step": 8043 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972793073124535, - "loss": 1.6696, - "step": 8044 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019727896908611244, - "loss": 1.5781, - "step": 8045 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019727863085977135, - "loss": 1.5904, - "step": 8046 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972782926334303, - "loss": 1.7286, - "step": 8047 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019727795440708924, - "loss": 1.8364, - "step": 8048 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019727761618074815, - "loss": 1.7827, - "step": 8049 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972772779544071, - "loss": 1.6541, - "step": 8050 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019727693972806603, - "loss": 1.7168, - "step": 8051 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019727660150172495, - "loss": 1.6399, - "step": 8052 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972762632753839, - "loss": 1.6682, - "step": 8053 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019727592504904283, - "loss": 1.6126, - "step": 8054 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019727558682270174, - "loss": 1.4986, - "step": 8055 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019727524859636068, - "loss": 1.6082, - "step": 8056 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019727491037001963, - "loss": 1.7101, - "step": 8057 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019727457214367854, - "loss": 1.7915, - "step": 8058 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019727423391733748, - "loss": 1.6398, - "step": 8059 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019727389569099642, - "loss": 1.7142, - "step": 8060 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019727355746465536, - "loss": 1.7638, - "step": 8061 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972732192383143, - "loss": 1.5507, - "step": 8062 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019727288101197322, - "loss": 1.7943, - "step": 8063 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019727254278563216, - "loss": 1.6382, - "step": 8064 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972722045592911, - "loss": 1.7034, - "step": 8065 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019727186633295004, - "loss": 1.6578, - "step": 8066 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019727152810660896, - "loss": 1.7003, - "step": 8067 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972711898802679, - "loss": 1.6836, - "step": 8068 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019727085165392684, - "loss": 1.6749, - "step": 8069 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019727051342758575, - "loss": 1.6283, - "step": 8070 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972701752012447, - "loss": 1.6349, - "step": 8071 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019726983697490363, - "loss": 1.606, - "step": 8072 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019726949874856255, - "loss": 1.6608, - "step": 8073 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972691605222215, - "loss": 1.8139, - "step": 8074 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019726882229588043, - "loss": 1.6173, - "step": 8075 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019726848406953934, - "loss": 1.6411, - "step": 8076 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019726814584319829, - "loss": 1.7559, - "step": 8077 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019726780761685723, - "loss": 1.6731, - "step": 8078 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019726746939051614, - "loss": 1.696, - "step": 8079 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019726713116417508, - "loss": 1.831, - "step": 8080 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019726679293783402, - "loss": 1.8277, - "step": 8081 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019726645471149294, - "loss": 1.7277, - "step": 8082 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019726611648515188, - "loss": 1.6724, - "step": 8083 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019726577825881082, - "loss": 1.6558, - "step": 8084 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019726544003246973, - "loss": 1.6924, - "step": 8085 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019726510180612867, - "loss": 1.7453, - "step": 8086 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019726476357978762, - "loss": 1.6702, - "step": 8087 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019726442535344653, - "loss": 1.6962, - "step": 8088 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019726408712710547, - "loss": 1.8306, - "step": 8089 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972637489007644, - "loss": 1.6346, - "step": 8090 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019726341067442333, - "loss": 1.6833, - "step": 8091 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019726307244808227, - "loss": 1.7801, - "step": 8092 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972627342217412, - "loss": 1.5108, - "step": 8093 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019726239599540012, - "loss": 1.6362, - "step": 8094 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019726205776905906, - "loss": 1.6693, - "step": 8095 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197261719542718, - "loss": 1.5772, - "step": 8096 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019726138131637692, - "loss": 1.6323, - "step": 8097 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019726104309003586, - "loss": 1.6333, - "step": 8098 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972607048636948, - "loss": 1.6687, - "step": 8099 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019726036663735372, - "loss": 1.5859, - "step": 8100 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019726002841101266, - "loss": 1.7489, - "step": 8101 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972596901846716, - "loss": 1.6864, - "step": 8102 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972593519583305, - "loss": 1.6433, - "step": 8103 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019725901373198945, - "loss": 1.6887, - "step": 8104 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972586755056484, - "loss": 1.8091, - "step": 8105 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972583372793073, - "loss": 1.6813, - "step": 8106 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019725799905296625, - "loss": 1.6901, - "step": 8107 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972576608266252, - "loss": 1.6637, - "step": 8108 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972573226002841, - "loss": 1.5266, - "step": 8109 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019725698437394305, - "loss": 1.7406, - "step": 8110 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197256646147602, - "loss": 1.411, - "step": 8111 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972563079212609, - "loss": 1.665, - "step": 8112 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019725596969491984, - "loss": 1.7006, - "step": 8113 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019725563146857878, - "loss": 1.7348, - "step": 8114 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972552932422377, - "loss": 1.7577, - "step": 8115 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019725495501589664, - "loss": 1.5426, - "step": 8116 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019725461678955558, - "loss": 1.761, - "step": 8117 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019725427856321452, - "loss": 1.5649, - "step": 8118 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019725394033687346, - "loss": 1.6873, - "step": 8119 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019725360211053238, - "loss": 1.6515, - "step": 8120 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019725326388419132, - "loss": 1.6508, - "step": 8121 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019725292565785026, - "loss": 1.7829, - "step": 8122 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019725258743150917, - "loss": 1.6011, - "step": 8123 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972522492051681, - "loss": 1.6511, - "step": 8124 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019725191097882705, - "loss": 1.6733, - "step": 8125 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019725157275248597, - "loss": 1.6842, - "step": 8126 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972512345261449, - "loss": 1.7397, - "step": 8127 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019725089629980385, - "loss": 1.6443, - "step": 8128 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019725055807346276, - "loss": 1.7336, - "step": 8129 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972502198471217, - "loss": 1.7242, - "step": 8130 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019724988162078065, - "loss": 1.7554, - "step": 8131 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019724954339443956, - "loss": 1.7278, - "step": 8132 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972492051680985, - "loss": 1.586, - "step": 8133 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019724886694175744, - "loss": 1.6758, - "step": 8134 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019724852871541636, - "loss": 1.6376, - "step": 8135 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972481904890753, - "loss": 1.8146, - "step": 8136 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019724785226273424, - "loss": 1.5675, - "step": 8137 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019724751403639315, - "loss": 1.7616, - "step": 8138 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972471758100521, - "loss": 1.6423, - "step": 8139 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019724683758371104, - "loss": 1.5418, - "step": 8140 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019724649935736995, - "loss": 1.8011, - "step": 8141 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972461611310289, - "loss": 1.7085, - "step": 8142 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019724582290468783, - "loss": 1.7253, - "step": 8143 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019724548467834675, - "loss": 1.725, - "step": 8144 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972451464520057, - "loss": 1.8116, - "step": 8145 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019724480822566463, - "loss": 1.6309, - "step": 8146 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019724446999932354, - "loss": 1.6099, - "step": 8147 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019724413177298248, - "loss": 1.642, - "step": 8148 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019724379354664142, - "loss": 1.7098, - "step": 8149 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019724345532030034, - "loss": 1.6139, - "step": 8150 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019724311709395928, - "loss": 1.544, - "step": 8151 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019724277886761822, - "loss": 1.7814, - "step": 8152 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019724244064127716, - "loss": 1.7398, - "step": 8153 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019724210241493608, - "loss": 1.7432, - "step": 8154 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019724176418859502, - "loss": 1.741, - "step": 8155 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019724142596225396, - "loss": 1.7846, - "step": 8156 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019724108773591287, - "loss": 1.5987, - "step": 8157 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972407495095718, - "loss": 1.6091, - "step": 8158 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019724041128323075, - "loss": 1.6726, - "step": 8159 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019724007305688967, - "loss": 1.7103, - "step": 8160 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972397348305486, - "loss": 1.6878, - "step": 8161 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019723939660420755, - "loss": 1.6047, - "step": 8162 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019723905837786646, - "loss": 1.5873, - "step": 8163 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972387201515254, - "loss": 1.674, - "step": 8164 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019723838192518435, - "loss": 1.4986, - "step": 8165 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019723804369884326, - "loss": 1.6866, - "step": 8166 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972377054725022, - "loss": 1.7819, - "step": 8167 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019723736724616114, - "loss": 1.7627, - "step": 8168 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019723702901982006, - "loss": 1.7327, - "step": 8169 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197236690793479, - "loss": 1.6458, - "step": 8170 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019723635256713794, - "loss": 1.5261, - "step": 8171 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019723601434079685, - "loss": 1.7238, - "step": 8172 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019723567611445582, - "loss": 1.7013, - "step": 8173 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019723533788811474, - "loss": 1.5511, - "step": 8174 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019723499966177368, - "loss": 1.7234, - "step": 8175 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019723466143543262, - "loss": 1.6361, - "step": 8176 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019723432320909153, - "loss": 1.695, - "step": 8177 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019723398498275047, - "loss": 1.6767, - "step": 8178 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019723364675640941, - "loss": 1.6118, - "step": 8179 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019723330853006833, - "loss": 1.7069, - "step": 8180 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019723297030372727, - "loss": 1.7224, - "step": 8181 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972326320773862, - "loss": 1.7259, - "step": 8182 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019723229385104513, - "loss": 1.6882, - "step": 8183 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019723195562470407, - "loss": 1.7218, - "step": 8184 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197231617398363, - "loss": 1.6574, - "step": 8185 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019723127917202192, - "loss": 1.5433, - "step": 8186 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019723094094568086, - "loss": 1.7187, - "step": 8187 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972306027193398, - "loss": 1.6145, - "step": 8188 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019723026449299872, - "loss": 1.8046, - "step": 8189 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019722992626665766, - "loss": 1.765, - "step": 8190 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972295880403166, - "loss": 1.6506, - "step": 8191 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019722924981397551, - "loss": 1.5572, - "step": 8192 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019722891158763446, - "loss": 1.5906, - "step": 8193 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972285733612934, - "loss": 1.5328, - "step": 8194 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972282351349523, - "loss": 1.5947, - "step": 8195 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019722789690861125, - "loss": 1.698, - "step": 8196 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972275586822702, - "loss": 1.7293, - "step": 8197 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972272204559291, - "loss": 1.5129, - "step": 8198 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019722688222958805, - "loss": 1.6291, - "step": 8199 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197226544003247, - "loss": 1.5785, - "step": 8200 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972262057769059, - "loss": 1.6573, - "step": 8201 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019722586755056484, - "loss": 1.6744, - "step": 8202 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019722552932422379, - "loss": 1.754, - "step": 8203 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972251910978827, - "loss": 1.6048, - "step": 8204 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019722485287154164, - "loss": 1.6142, - "step": 8205 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019722451464520058, - "loss": 1.7773, - "step": 8206 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972241764188595, - "loss": 1.643, - "step": 8207 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019722383819251844, - "loss": 1.6643, - "step": 8208 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019722349996617738, - "loss": 1.727, - "step": 8209 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972231617398363, - "loss": 1.7843, - "step": 8210 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019722282351349523, - "loss": 1.756, - "step": 8211 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019722248528715417, - "loss": 1.7785, - "step": 8212 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972221470608131, - "loss": 1.8114, - "step": 8213 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019722180883447203, - "loss": 1.7858, - "step": 8214 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019722147060813097, - "loss": 1.7226, - "step": 8215 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019722113238178988, - "loss": 1.57, - "step": 8216 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019722079415544883, - "loss": 1.804, - "step": 8217 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019722045592910777, - "loss": 1.6967, - "step": 8218 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019722011770276668, - "loss": 1.7255, - "step": 8219 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019721977947642562, - "loss": 1.6738, - "step": 8220 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019721944125008456, - "loss": 1.7613, - "step": 8221 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019721910302374348, - "loss": 1.6052, - "step": 8222 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019721876479740242, - "loss": 1.6844, - "step": 8223 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019721842657106136, - "loss": 1.7533, - "step": 8224 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019721808834472027, - "loss": 1.6915, - "step": 8225 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019721775011837921, - "loss": 1.5776, - "step": 8226 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019721741189203816, - "loss": 1.6978, - "step": 8227 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019721707366569707, - "loss": 1.8212, - "step": 8228 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019721673543935604, - "loss": 1.7061, - "step": 8229 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019721639721301498, - "loss": 1.5233, - "step": 8230 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972160589866739, - "loss": 1.5646, - "step": 8231 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019721572076033283, - "loss": 1.5637, - "step": 8232 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019721538253399178, - "loss": 1.7023, - "step": 8233 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972150443076507, - "loss": 1.6111, - "step": 8234 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019721470608130963, - "loss": 1.6629, - "step": 8235 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019721436785496857, - "loss": 1.692, - "step": 8236 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019721402962862749, - "loss": 1.6687, - "step": 8237 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019721369140228643, - "loss": 1.7062, - "step": 8238 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019721335317594537, - "loss": 1.5435, - "step": 8239 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019721301494960428, - "loss": 1.6496, - "step": 8240 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019721267672326322, - "loss": 1.5817, - "step": 8241 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019721233849692216, - "loss": 1.6509, - "step": 8242 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019721200027058108, - "loss": 1.6073, - "step": 8243 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019721166204424002, - "loss": 1.6343, - "step": 8244 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019721132381789896, - "loss": 1.706, - "step": 8245 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019721098559155787, - "loss": 1.845, - "step": 8246 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019721064736521682, - "loss": 1.6106, - "step": 8247 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019721030913887576, - "loss": 1.7153, - "step": 8248 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019720997091253467, - "loss": 1.6031, - "step": 8249 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972096326861936, - "loss": 1.7183, - "step": 8250 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019720929445985255, - "loss": 1.6281, - "step": 8251 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019720895623351147, - "loss": 1.6395, - "step": 8252 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972086180071704, - "loss": 1.6714, - "step": 8253 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019720827978082935, - "loss": 1.5448, - "step": 8254 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019720794155448826, - "loss": 1.5596, - "step": 8255 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972076033281472, - "loss": 1.5921, - "step": 8256 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019720726510180615, - "loss": 1.7332, - "step": 8257 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019720692687546506, - "loss": 1.5205, - "step": 8258 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197206588649124, - "loss": 1.608, - "step": 8259 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019720625042278294, - "loss": 1.6215, - "step": 8260 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019720591219644186, - "loss": 1.7428, - "step": 8261 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972055739701008, - "loss": 1.6074, - "step": 8262 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019720523574375974, - "loss": 1.6819, - "step": 8263 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019720489751741865, - "loss": 1.7284, - "step": 8264 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972045592910776, - "loss": 1.5201, - "step": 8265 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019720422106473653, - "loss": 1.5822, - "step": 8266 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019720388283839545, - "loss": 1.6904, - "step": 8267 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972035446120544, - "loss": 1.8288, - "step": 8268 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019720320638571333, - "loss": 1.6977, - "step": 8269 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019720286815937225, - "loss": 1.4341, - "step": 8270 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001972025299330312, - "loss": 1.7997, - "step": 8271 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019720219170669013, - "loss": 1.7114, - "step": 8272 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019720185348034904, - "loss": 1.6589, - "step": 8273 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019720151525400798, - "loss": 1.6367, - "step": 8274 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019720117702766692, - "loss": 1.719, - "step": 8275 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019720083880132584, - "loss": 1.608, - "step": 8276 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019720050057498478, - "loss": 1.6793, - "step": 8277 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019720016234864372, - "loss": 1.5915, - "step": 8278 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019719982412230263, - "loss": 1.5058, - "step": 8279 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019719948589596158, - "loss": 1.7621, - "step": 8280 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019719914766962052, - "loss": 1.6435, - "step": 8281 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019719880944327943, - "loss": 1.5086, - "step": 8282 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019719847121693837, - "loss": 1.5375, - "step": 8283 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971981329905973, - "loss": 1.6058, - "step": 8284 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019719779476425625, - "loss": 1.7954, - "step": 8285 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971974565379152, - "loss": 1.6545, - "step": 8286 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019719711831157414, - "loss": 1.5727, - "step": 8287 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019719678008523305, - "loss": 1.5689, - "step": 8288 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197196441858892, - "loss": 1.6513, - "step": 8289 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019719610363255093, - "loss": 1.7066, - "step": 8290 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019719576540620985, - "loss": 1.6545, - "step": 8291 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971954271798688, - "loss": 1.8097, - "step": 8292 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019719508895352773, - "loss": 1.6764, - "step": 8293 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019719475072718664, - "loss": 1.6307, - "step": 8294 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019719441250084558, - "loss": 1.5685, - "step": 8295 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019719407427450453, - "loss": 1.5794, - "step": 8296 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019719373604816344, - "loss": 1.7057, - "step": 8297 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019719339782182238, - "loss": 1.7509, - "step": 8298 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019719305959548132, - "loss": 1.5753, - "step": 8299 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019719272136914024, - "loss": 1.5481, - "step": 8300 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019719238314279918, - "loss": 1.6652, - "step": 8301 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019719204491645812, - "loss": 1.7851, - "step": 8302 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019719170669011703, - "loss": 1.7075, - "step": 8303 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019719136846377597, - "loss": 1.7932, - "step": 8304 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019719103023743491, - "loss": 1.5589, - "step": 8305 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019719069201109383, - "loss": 1.6799, - "step": 8306 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019719035378475277, - "loss": 1.5753, - "step": 8307 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971900155584117, - "loss": 1.7593, - "step": 8308 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019718967733207062, - "loss": 1.6624, - "step": 8309 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019718933910572957, - "loss": 1.6261, - "step": 8310 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971890008793885, - "loss": 1.5927, - "step": 8311 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019718866265304742, - "loss": 1.8872, - "step": 8312 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019718832442670636, - "loss": 1.6138, - "step": 8313 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971879862003653, - "loss": 1.8001, - "step": 8314 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019718764797402422, - "loss": 1.8049, - "step": 8315 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019718730974768316, - "loss": 1.752, - "step": 8316 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971869715213421, - "loss": 1.7814, - "step": 8317 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197186633295001, - "loss": 1.6179, - "step": 8318 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019718629506865995, - "loss": 1.696, - "step": 8319 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971859568423189, - "loss": 1.5652, - "step": 8320 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971856186159778, - "loss": 1.4339, - "step": 8321 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019718528038963675, - "loss": 1.6417, - "step": 8322 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971849421632957, - "loss": 1.6668, - "step": 8323 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971846039369546, - "loss": 1.5729, - "step": 8324 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019718426571061355, - "loss": 1.5389, - "step": 8325 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971839274842725, - "loss": 1.7186, - "step": 8326 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971835892579314, - "loss": 1.5335, - "step": 8327 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019718325103159034, - "loss": 1.8446, - "step": 8328 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019718291280524928, - "loss": 1.715, - "step": 8329 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971825745789082, - "loss": 1.6097, - "step": 8330 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019718223635256714, - "loss": 1.7022, - "step": 8331 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019718189812622608, - "loss": 1.6502, - "step": 8332 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197181559899885, - "loss": 1.5656, - "step": 8333 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019718122167354394, - "loss": 1.6249, - "step": 8334 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019718088344720288, - "loss": 1.7182, - "step": 8335 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971805452208618, - "loss": 1.5947, - "step": 8336 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019718020699452073, - "loss": 1.7175, - "step": 8337 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019717986876817967, - "loss": 1.5886, - "step": 8338 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971795305418386, - "loss": 1.6146, - "step": 8339 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019717919231549753, - "loss": 1.6334, - "step": 8340 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971788540891565, - "loss": 1.6159, - "step": 8341 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971785158628154, - "loss": 1.7658, - "step": 8342 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019717817763647435, - "loss": 1.653, - "step": 8343 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971778394101333, - "loss": 1.6863, - "step": 8344 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971775011837922, - "loss": 1.6291, - "step": 8345 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019717716295745115, - "loss": 1.5819, - "step": 8346 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971768247311101, - "loss": 1.747, - "step": 8347 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197176486504769, - "loss": 1.7092, - "step": 8348 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019717614827842794, - "loss": 1.8599, - "step": 8349 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019717581005208689, - "loss": 1.6612, - "step": 8350 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971754718257458, - "loss": 1.798, - "step": 8351 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019717513359940474, - "loss": 1.6365, - "step": 8352 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019717479537306368, - "loss": 1.6395, - "step": 8353 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971744571467226, - "loss": 1.5905, - "step": 8354 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019717411892038154, - "loss": 1.5978, - "step": 8355 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019717378069404048, - "loss": 1.5625, - "step": 8356 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971734424676994, - "loss": 1.6162, - "step": 8357 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019717310424135833, - "loss": 1.6155, - "step": 8358 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019717276601501727, - "loss": 1.7093, - "step": 8359 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971724277886762, - "loss": 1.5122, - "step": 8360 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019717208956233513, - "loss": 1.824, - "step": 8361 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019717175133599407, - "loss": 1.697, - "step": 8362 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019717141310965299, - "loss": 1.6133, - "step": 8363 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019717107488331193, - "loss": 1.8198, - "step": 8364 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019717073665697087, - "loss": 1.708, - "step": 8365 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019717039843062978, - "loss": 1.8287, - "step": 8366 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019717006020428872, - "loss": 1.6397, - "step": 8367 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019716972197794766, - "loss": 1.5813, - "step": 8368 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019716938375160658, - "loss": 1.744, - "step": 8369 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019716904552526552, - "loss": 1.7335, - "step": 8370 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019716870729892446, - "loss": 1.7081, - "step": 8371 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019716836907258337, - "loss": 1.6598, - "step": 8372 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019716803084624232, - "loss": 1.6272, - "step": 8373 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019716769261990126, - "loss": 1.7036, - "step": 8374 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019716735439356017, - "loss": 1.5848, - "step": 8375 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971670161672191, - "loss": 1.5612, - "step": 8376 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019716667794087805, - "loss": 1.6139, - "step": 8377 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019716633971453697, - "loss": 1.7136, - "step": 8378 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971660014881959, - "loss": 1.7845, - "step": 8379 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019716566326185485, - "loss": 1.4993, - "step": 8380 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019716532503551376, - "loss": 1.5584, - "step": 8381 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971649868091727, - "loss": 1.6381, - "step": 8382 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019716464858283165, - "loss": 1.748, - "step": 8383 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019716431035649056, - "loss": 1.6998, - "step": 8384 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971639721301495, - "loss": 1.6833, - "step": 8385 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019716363390380844, - "loss": 1.6464, - "step": 8386 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019716329567746736, - "loss": 1.6605, - "step": 8387 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971629574511263, - "loss": 1.5887, - "step": 8388 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019716261922478524, - "loss": 1.7566, - "step": 8389 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019716228099844415, - "loss": 1.6128, - "step": 8390 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971619427721031, - "loss": 1.7847, - "step": 8391 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019716160454576203, - "loss": 1.5172, - "step": 8392 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019716126631942095, - "loss": 1.7684, - "step": 8393 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971609280930799, - "loss": 1.826, - "step": 8394 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019716058986673883, - "loss": 1.7619, - "step": 8395 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019716025164039774, - "loss": 1.7225, - "step": 8396 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971599134140567, - "loss": 1.7241, - "step": 8397 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019715957518771563, - "loss": 1.7032, - "step": 8398 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019715923696137457, - "loss": 1.6735, - "step": 8399 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971588987350335, - "loss": 1.6849, - "step": 8400 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019715856050869242, - "loss": 1.6138, - "step": 8401 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019715822228235136, - "loss": 1.8137, - "step": 8402 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971578840560103, - "loss": 1.73, - "step": 8403 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019715754582966922, - "loss": 1.7052, - "step": 8404 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019715720760332816, - "loss": 1.4987, - "step": 8405 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971568693769871, - "loss": 1.452, - "step": 8406 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019715653115064602, - "loss": 1.7023, - "step": 8407 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019715619292430496, - "loss": 1.6346, - "step": 8408 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971558546979639, - "loss": 1.547, - "step": 8409 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971555164716228, - "loss": 1.6842, - "step": 8410 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019715517824528175, - "loss": 1.4089, - "step": 8411 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971548400189407, - "loss": 1.5646, - "step": 8412 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971545017925996, - "loss": 1.6768, - "step": 8413 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019715416356625855, - "loss": 1.5775, - "step": 8414 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971538253399175, - "loss": 1.7845, - "step": 8415 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971534871135764, - "loss": 1.5918, - "step": 8416 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019715314888723535, - "loss": 1.7416, - "step": 8417 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971528106608943, - "loss": 1.5589, - "step": 8418 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019715247243455323, - "loss": 1.6306, - "step": 8419 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019715213420821214, - "loss": 1.5212, - "step": 8420 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019715179598187108, - "loss": 1.6194, - "step": 8421 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019715145775553002, - "loss": 1.6447, - "step": 8422 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019715111952918894, - "loss": 1.731, - "step": 8423 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019715078130284788, - "loss": 1.7085, - "step": 8424 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019715044307650682, - "loss": 1.648, - "step": 8425 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019715010485016573, - "loss": 1.6255, - "step": 8426 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019714976662382468, - "loss": 1.636, - "step": 8427 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019714942839748362, - "loss": 1.7304, - "step": 8428 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019714909017114253, - "loss": 1.716, - "step": 8429 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019714875194480147, - "loss": 1.6345, - "step": 8430 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019714841371846041, - "loss": 1.6059, - "step": 8431 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019714807549211933, - "loss": 1.6492, - "step": 8432 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019714773726577827, - "loss": 1.5854, - "step": 8433 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971473990394372, - "loss": 1.698, - "step": 8434 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019714706081309612, - "loss": 1.5011, - "step": 8435 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019714672258675507, - "loss": 1.7704, - "step": 8436 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197146384360414, - "loss": 1.5584, - "step": 8437 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019714604613407292, - "loss": 1.9316, - "step": 8438 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019714570790773186, - "loss": 1.6796, - "step": 8439 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971453696813908, - "loss": 1.6946, - "step": 8440 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019714503145504972, - "loss": 1.5834, - "step": 8441 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019714469322870866, - "loss": 1.8686, - "step": 8442 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971443550023676, - "loss": 1.6208, - "step": 8443 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971440167760265, - "loss": 1.5938, - "step": 8444 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019714367854968545, - "loss": 1.8573, - "step": 8445 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971433403233444, - "loss": 1.7285, - "step": 8446 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971430020970033, - "loss": 1.6125, - "step": 8447 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019714266387066225, - "loss": 1.7983, - "step": 8448 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971423256443212, - "loss": 1.6918, - "step": 8449 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971419874179801, - "loss": 1.7364, - "step": 8450 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019714164919163905, - "loss": 1.6526, - "step": 8451 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197141310965298, - "loss": 1.57, - "step": 8452 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971409727389569, - "loss": 1.6135, - "step": 8453 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019714063451261587, - "loss": 1.5796, - "step": 8454 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019714029628627478, - "loss": 1.7968, - "step": 8455 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019713995805993373, - "loss": 1.6133, - "step": 8456 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019713961983359267, - "loss": 1.7193, - "step": 8457 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019713928160725158, - "loss": 1.5212, - "step": 8458 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019713894338091052, - "loss": 1.7261, - "step": 8459 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019713860515456946, - "loss": 1.5709, - "step": 8460 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019713826692822838, - "loss": 1.6576, - "step": 8461 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019713792870188732, - "loss": 1.5771, - "step": 8462 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019713759047554626, - "loss": 1.823, - "step": 8463 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019713725224920517, - "loss": 1.6223, - "step": 8464 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019713691402286411, - "loss": 1.6667, - "step": 8465 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019713657579652306, - "loss": 1.5422, - "step": 8466 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019713623757018197, - "loss": 1.7882, - "step": 8467 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971358993438409, - "loss": 1.6093, - "step": 8468 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019713556111749985, - "loss": 1.5301, - "step": 8469 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019713522289115877, - "loss": 1.5701, - "step": 8470 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971348846648177, - "loss": 1.8128, - "step": 8471 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019713454643847665, - "loss": 1.59, - "step": 8472 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019713420821213556, - "loss": 1.7141, - "step": 8473 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971338699857945, - "loss": 1.7692, - "step": 8474 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019713353175945344, - "loss": 1.6096, - "step": 8475 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019713319353311236, - "loss": 1.7344, - "step": 8476 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971328553067713, - "loss": 1.6258, - "step": 8477 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019713251708043024, - "loss": 1.7248, - "step": 8478 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019713217885408915, - "loss": 1.7255, - "step": 8479 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971318406277481, - "loss": 1.7984, - "step": 8480 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019713150240140704, - "loss": 1.5889, - "step": 8481 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019713116417506595, - "loss": 1.6185, - "step": 8482 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971308259487249, - "loss": 1.6439, - "step": 8483 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019713048772238383, - "loss": 1.838, - "step": 8484 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019713014949604275, - "loss": 1.5294, - "step": 8485 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971298112697017, - "loss": 1.5849, - "step": 8486 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019712947304336063, - "loss": 1.7713, - "step": 8487 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019712913481701954, - "loss": 1.5973, - "step": 8488 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019712879659067848, - "loss": 1.5279, - "step": 8489 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019712845836433743, - "loss": 1.5298, - "step": 8490 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019712812013799634, - "loss": 1.5794, - "step": 8491 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019712778191165528, - "loss": 1.774, - "step": 8492 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019712744368531422, - "loss": 1.697, - "step": 8493 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019712710545897314, - "loss": 1.8314, - "step": 8494 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019712676723263208, - "loss": 1.6442, - "step": 8495 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019712642900629102, - "loss": 1.7315, - "step": 8496 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019712609077994993, - "loss": 1.692, - "step": 8497 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019712575255360887, - "loss": 1.5166, - "step": 8498 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019712541432726781, - "loss": 1.6712, - "step": 8499 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019712507610092673, - "loss": 1.5975, - "step": 8500 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019712473787458567, - "loss": 1.6072, - "step": 8501 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971243996482446, - "loss": 1.6096, - "step": 8502 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019712406142190355, - "loss": 1.6701, - "step": 8503 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019712372319556247, - "loss": 1.7429, - "step": 8504 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971233849692214, - "loss": 1.6579, - "step": 8505 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019712304674288035, - "loss": 1.7226, - "step": 8506 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019712270851653926, - "loss": 1.7631, - "step": 8507 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971223702901982, - "loss": 1.7011, - "step": 8508 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019712203206385714, - "loss": 1.7273, - "step": 8509 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019712169383751609, - "loss": 1.7003, - "step": 8510 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019712135561117503, - "loss": 1.6629, - "step": 8511 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019712101738483394, - "loss": 1.621, - "step": 8512 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019712067915849288, - "loss": 1.6531, - "step": 8513 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019712034093215182, - "loss": 1.6331, - "step": 8514 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019712000270581074, - "loss": 1.6338, - "step": 8515 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019711966447946968, - "loss": 1.6485, - "step": 8516 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019711932625312862, - "loss": 1.583, - "step": 8517 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019711898802678753, - "loss": 1.7243, - "step": 8518 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019711864980044647, - "loss": 1.8201, - "step": 8519 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019711831157410542, - "loss": 1.694, - "step": 8520 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019711797334776433, - "loss": 1.7325, - "step": 8521 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019711763512142327, - "loss": 1.67, - "step": 8522 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971172968950822, - "loss": 1.8101, - "step": 8523 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019711695866874113, - "loss": 1.7244, - "step": 8524 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019711662044240007, - "loss": 1.5702, - "step": 8525 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197116282216059, - "loss": 1.598, - "step": 8526 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019711594398971792, - "loss": 1.6502, - "step": 8527 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019711560576337686, - "loss": 1.6968, - "step": 8528 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971152675370358, - "loss": 1.7357, - "step": 8529 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019711492931069472, - "loss": 1.5237, - "step": 8530 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019711459108435366, - "loss": 1.7161, - "step": 8531 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971142528580126, - "loss": 1.7186, - "step": 8532 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019711391463167152, - "loss": 1.705, - "step": 8533 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019711357640533046, - "loss": 1.6577, - "step": 8534 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971132381789894, - "loss": 1.6864, - "step": 8535 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971128999526483, - "loss": 1.5059, - "step": 8536 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019711256172630725, - "loss": 1.5902, - "step": 8537 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971122234999662, - "loss": 1.7128, - "step": 8538 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971118852736251, - "loss": 1.6188, - "step": 8539 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019711154704728405, - "loss": 1.7699, - "step": 8540 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197111208820943, - "loss": 1.6833, - "step": 8541 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971108705946019, - "loss": 1.6199, - "step": 8542 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019711053236826085, - "loss": 1.637, - "step": 8543 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971101941419198, - "loss": 1.579, - "step": 8544 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971098559155787, - "loss": 1.6524, - "step": 8545 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019710951768923764, - "loss": 1.8208, - "step": 8546 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019710917946289658, - "loss": 1.7694, - "step": 8547 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971088412365555, - "loss": 1.6181, - "step": 8548 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019710850301021444, - "loss": 1.6061, - "step": 8549 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019710816478387338, - "loss": 1.6658, - "step": 8550 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971078265575323, - "loss": 1.7264, - "step": 8551 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019710748833119123, - "loss": 1.7522, - "step": 8552 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019710715010485018, - "loss": 1.5413, - "step": 8553 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971068118785091, - "loss": 1.7042, - "step": 8554 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019710647365216803, - "loss": 1.604, - "step": 8555 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019710613542582697, - "loss": 1.6677, - "step": 8556 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019710579719948589, - "loss": 1.6604, - "step": 8557 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019710545897314483, - "loss": 1.6176, - "step": 8558 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019710512074680377, - "loss": 1.6671, - "step": 8559 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019710478252046268, - "loss": 1.556, - "step": 8560 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019710444429412162, - "loss": 1.5527, - "step": 8561 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019710410606778056, - "loss": 1.7579, - "step": 8562 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019710376784143948, - "loss": 1.6916, - "step": 8563 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019710342961509842, - "loss": 1.7049, - "step": 8564 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019710309138875736, - "loss": 1.6946, - "step": 8565 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971027531624163, - "loss": 1.7868, - "step": 8566 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019710241493607524, - "loss": 1.6494, - "step": 8567 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019710207670973418, - "loss": 1.7976, - "step": 8568 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971017384833931, - "loss": 1.6174, - "step": 8569 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019710140025705204, - "loss": 1.5378, - "step": 8570 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019710106203071098, - "loss": 1.7338, - "step": 8571 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001971007238043699, - "loss": 1.5956, - "step": 8572 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019710038557802884, - "loss": 1.6952, - "step": 8573 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019710004735168778, - "loss": 1.6053, - "step": 8574 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970997091253467, - "loss": 1.5529, - "step": 8575 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019709937089900563, - "loss": 1.6061, - "step": 8576 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019709903267266457, - "loss": 1.4953, - "step": 8577 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970986944463235, - "loss": 1.4491, - "step": 8578 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019709835621998243, - "loss": 1.7187, - "step": 8579 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019709801799364137, - "loss": 1.6397, - "step": 8580 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019709767976730028, - "loss": 1.6628, - "step": 8581 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019709734154095922, - "loss": 1.8762, - "step": 8582 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019709700331461817, - "loss": 1.7639, - "step": 8583 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019709666508827708, - "loss": 1.6156, - "step": 8584 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019709632686193602, - "loss": 1.562, - "step": 8585 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019709598863559496, - "loss": 1.7796, - "step": 8586 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019709565040925388, - "loss": 1.8325, - "step": 8587 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019709531218291282, - "loss": 1.5901, - "step": 8588 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019709497395657176, - "loss": 1.5667, - "step": 8589 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019709463573023067, - "loss": 1.6083, - "step": 8590 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970942975038896, - "loss": 1.686, - "step": 8591 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019709395927754855, - "loss": 1.6667, - "step": 8592 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019709362105120747, - "loss": 1.5911, - "step": 8593 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970932828248664, - "loss": 1.6129, - "step": 8594 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019709294459852535, - "loss": 1.7175, - "step": 8595 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019709260637218427, - "loss": 1.8482, - "step": 8596 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970922681458432, - "loss": 1.5698, - "step": 8597 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019709192991950215, - "loss": 1.6109, - "step": 8598 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019709159169316106, - "loss": 1.7568, - "step": 8599 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019709125346682, - "loss": 1.7889, - "step": 8600 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019709091524047894, - "loss": 1.7292, - "step": 8601 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019709057701413786, - "loss": 1.6074, - "step": 8602 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970902387877968, - "loss": 1.6917, - "step": 8603 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019708990056145574, - "loss": 1.6193, - "step": 8604 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019708956233511465, - "loss": 1.5938, - "step": 8605 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970892241087736, - "loss": 1.5782, - "step": 8606 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019708888588243254, - "loss": 1.6464, - "step": 8607 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019708854765609145, - "loss": 1.6673, - "step": 8608 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970882094297504, - "loss": 1.6005, - "step": 8609 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019708787120340933, - "loss": 1.5736, - "step": 8610 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019708753297706825, - "loss": 1.6176, - "step": 8611 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970871947507272, - "loss": 1.6155, - "step": 8612 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019708685652438613, - "loss": 1.6721, - "step": 8613 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019708651829804504, - "loss": 1.5892, - "step": 8614 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019708618007170398, - "loss": 1.7462, - "step": 8615 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019708584184536293, - "loss": 1.6845, - "step": 8616 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019708550361902184, - "loss": 1.7235, - "step": 8617 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019708516539268078, - "loss": 1.6261, - "step": 8618 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019708482716633972, - "loss": 1.6938, - "step": 8619 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019708448893999864, - "loss": 1.9628, - "step": 8620 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019708415071365758, - "loss": 1.7774, - "step": 8621 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019708381248731654, - "loss": 1.6593, - "step": 8622 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019708347426097546, - "loss": 1.6223, - "step": 8623 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970831360346344, - "loss": 1.5781, - "step": 8624 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019708279780829334, - "loss": 1.7062, - "step": 8625 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019708245958195226, - "loss": 1.6834, - "step": 8626 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970821213556112, - "loss": 1.7999, - "step": 8627 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019708178312927014, - "loss": 1.6818, - "step": 8628 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019708144490292905, - "loss": 1.7237, - "step": 8629 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197081106676588, - "loss": 1.5949, - "step": 8630 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019708076845024693, - "loss": 1.6888, - "step": 8631 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019708043022390585, - "loss": 1.5666, - "step": 8632 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970800919975648, - "loss": 1.651, - "step": 8633 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019707975377122373, - "loss": 1.6028, - "step": 8634 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019707941554488264, - "loss": 1.6543, - "step": 8635 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019707907731854159, - "loss": 1.9668, - "step": 8636 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019707873909220053, - "loss": 1.6316, - "step": 8637 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019707840086585944, - "loss": 1.6226, - "step": 8638 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019707806263951838, - "loss": 1.6088, - "step": 8639 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019707772441317732, - "loss": 1.9043, - "step": 8640 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019707738618683624, - "loss": 1.7212, - "step": 8641 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019707704796049518, - "loss": 1.5988, - "step": 8642 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019707670973415412, - "loss": 1.6283, - "step": 8643 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019707637150781303, - "loss": 1.7012, - "step": 8644 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019707603328147197, - "loss": 1.7259, - "step": 8645 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019707569505513092, - "loss": 1.6071, - "step": 8646 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019707535682878983, - "loss": 1.5843, - "step": 8647 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019707501860244877, - "loss": 1.6444, - "step": 8648 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970746803761077, - "loss": 1.5735, - "step": 8649 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019707434214976663, - "loss": 1.6425, - "step": 8650 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019707400392342557, - "loss": 1.6687, - "step": 8651 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970736656970845, - "loss": 1.6562, - "step": 8652 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019707332747074342, - "loss": 1.6266, - "step": 8653 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019707298924440236, - "loss": 1.6281, - "step": 8654 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970726510180613, - "loss": 1.6318, - "step": 8655 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019707231279172022, - "loss": 1.7767, - "step": 8656 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019707197456537916, - "loss": 1.621, - "step": 8657 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970716363390381, - "loss": 1.6582, - "step": 8658 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019707129811269701, - "loss": 1.553, - "step": 8659 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019707095988635596, - "loss": 1.5898, - "step": 8660 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970706216600149, - "loss": 1.7578, - "step": 8661 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970702834336738, - "loss": 1.5491, - "step": 8662 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019706994520733275, - "loss": 1.6774, - "step": 8663 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970696069809917, - "loss": 1.699, - "step": 8664 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970692687546506, - "loss": 1.7137, - "step": 8665 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019706893052830955, - "loss": 1.6363, - "step": 8666 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970685923019685, - "loss": 1.6192, - "step": 8667 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970682540756274, - "loss": 1.6232, - "step": 8668 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019706791584928634, - "loss": 1.626, - "step": 8669 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019706757762294529, - "loss": 1.5644, - "step": 8670 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970672393966042, - "loss": 1.6994, - "step": 8671 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019706690117026314, - "loss": 1.6979, - "step": 8672 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019706656294392208, - "loss": 1.6376, - "step": 8673 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197066224717581, - "loss": 1.685, - "step": 8674 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019706588649123994, - "loss": 1.7634, - "step": 8675 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019706554826489888, - "loss": 1.6982, - "step": 8676 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970652100385578, - "loss": 1.7903, - "step": 8677 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019706487181221676, - "loss": 1.5224, - "step": 8678 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019706453358587567, - "loss": 1.5888, - "step": 8679 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019706419535953462, - "loss": 1.5861, - "step": 8680 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019706385713319356, - "loss": 1.8048, - "step": 8681 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019706351890685247, - "loss": 1.7292, - "step": 8682 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970631806805114, - "loss": 1.6123, - "step": 8683 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019706284245417035, - "loss": 1.4518, - "step": 8684 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019706250422782927, - "loss": 1.7489, - "step": 8685 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970621660014882, - "loss": 1.6861, - "step": 8686 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019706182777514715, - "loss": 1.6308, - "step": 8687 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970614895488061, - "loss": 1.6291, - "step": 8688 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197061151322465, - "loss": 1.6508, - "step": 8689 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019706081309612395, - "loss": 1.5757, - "step": 8690 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970604748697829, - "loss": 1.6244, - "step": 8691 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970601366434418, - "loss": 1.7872, - "step": 8692 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019705979841710074, - "loss": 1.658, - "step": 8693 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019705946019075968, - "loss": 1.7457, - "step": 8694 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970591219644186, - "loss": 1.5213, - "step": 8695 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019705878373807754, - "loss": 1.7069, - "step": 8696 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019705844551173648, - "loss": 1.6928, - "step": 8697 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970581072853954, - "loss": 1.5399, - "step": 8698 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019705776905905434, - "loss": 1.5402, - "step": 8699 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019705743083271328, - "loss": 1.713, - "step": 8700 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970570926063722, - "loss": 1.6919, - "step": 8701 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019705675438003113, - "loss": 1.7366, - "step": 8702 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019705641615369007, - "loss": 1.6496, - "step": 8703 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197056077927349, - "loss": 1.7058, - "step": 8704 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019705573970100793, - "loss": 1.6904, - "step": 8705 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019705540147466687, - "loss": 1.6017, - "step": 8706 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019705506324832578, - "loss": 1.6225, - "step": 8707 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019705472502198472, - "loss": 1.7241, - "step": 8708 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019705438679564367, - "loss": 1.6621, - "step": 8709 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019705404856930258, - "loss": 1.7325, - "step": 8710 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019705371034296152, - "loss": 1.801, - "step": 8711 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019705337211662046, - "loss": 1.5191, - "step": 8712 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019705303389027938, - "loss": 1.4618, - "step": 8713 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019705269566393832, - "loss": 1.7985, - "step": 8714 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019705235743759726, - "loss": 1.806, - "step": 8715 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019705201921125617, - "loss": 1.7231, - "step": 8716 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970516809849151, - "loss": 1.7358, - "step": 8717 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019705134275857405, - "loss": 1.5974, - "step": 8718 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019705100453223297, - "loss": 1.5965, - "step": 8719 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970506663058919, - "loss": 1.643, - "step": 8720 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019705032807955085, - "loss": 1.6447, - "step": 8721 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019704998985320976, - "loss": 1.6086, - "step": 8722 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970496516268687, - "loss": 1.6397, - "step": 8723 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019704931340052765, - "loss": 1.515, - "step": 8724 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019704897517418656, - "loss": 1.6109, - "step": 8725 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970486369478455, - "loss": 1.6879, - "step": 8726 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019704829872150444, - "loss": 1.7709, - "step": 8727 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019704796049516336, - "loss": 1.6368, - "step": 8728 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970476222688223, - "loss": 1.5989, - "step": 8729 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019704728404248124, - "loss": 1.7817, - "step": 8730 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019704694581614015, - "loss": 1.6335, - "step": 8731 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970466075897991, - "loss": 1.5421, - "step": 8732 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019704626936345804, - "loss": 1.7378, - "step": 8733 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019704593113711698, - "loss": 1.7075, - "step": 8734 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019704559291077592, - "loss": 1.6876, - "step": 8735 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019704525468443483, - "loss": 1.6001, - "step": 8736 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019704491645809377, - "loss": 1.6977, - "step": 8737 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019704457823175271, - "loss": 1.7139, - "step": 8738 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019704424000541163, - "loss": 1.7514, - "step": 8739 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019704390177907057, - "loss": 1.6276, - "step": 8740 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970435635527295, - "loss": 1.746, - "step": 8741 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019704322532638842, - "loss": 1.6881, - "step": 8742 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019704288710004737, - "loss": 1.5929, - "step": 8743 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970425488737063, - "loss": 1.5873, - "step": 8744 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019704221064736522, - "loss": 1.6547, - "step": 8745 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019704187242102416, - "loss": 1.6939, - "step": 8746 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970415341946831, - "loss": 1.7615, - "step": 8747 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019704119596834202, - "loss": 1.7813, - "step": 8748 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019704085774200096, - "loss": 1.8366, - "step": 8749 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970405195156599, - "loss": 1.4194, - "step": 8750 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970401812893188, - "loss": 1.6875, - "step": 8751 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019703984306297775, - "loss": 1.5777, - "step": 8752 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970395048366367, - "loss": 1.749, - "step": 8753 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970391666102956, - "loss": 1.7055, - "step": 8754 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019703882838395455, - "loss": 1.7341, - "step": 8755 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970384901576135, - "loss": 1.5407, - "step": 8756 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970381519312724, - "loss": 1.6973, - "step": 8757 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019703781370493135, - "loss": 1.7698, - "step": 8758 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970374754785903, - "loss": 1.5924, - "step": 8759 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970371372522492, - "loss": 1.6596, - "step": 8760 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019703679902590814, - "loss": 1.7084, - "step": 8761 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019703646079956708, - "loss": 1.7044, - "step": 8762 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197036122573226, - "loss": 1.6042, - "step": 8763 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019703578434688494, - "loss": 1.7495, - "step": 8764 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019703544612054388, - "loss": 1.647, - "step": 8765 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970351078942028, - "loss": 1.6319, - "step": 8766 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019703476966786174, - "loss": 1.7986, - "step": 8767 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019703443144152068, - "loss": 1.901, - "step": 8768 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970340932151796, - "loss": 1.6367, - "step": 8769 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019703375498883853, - "loss": 1.7045, - "step": 8770 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019703341676249747, - "loss": 1.5942, - "step": 8771 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019703307853615641, - "loss": 1.7292, - "step": 8772 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019703274030981533, - "loss": 1.685, - "step": 8773 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019703240208347427, - "loss": 1.5322, - "step": 8774 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970320638571332, - "loss": 1.6486, - "step": 8775 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019703172563079213, - "loss": 1.5508, - "step": 8776 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019703138740445107, - "loss": 1.7759, - "step": 8777 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019703104917811, - "loss": 1.5644, - "step": 8778 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019703071095176892, - "loss": 1.6919, - "step": 8779 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019703037272542786, - "loss": 1.4893, - "step": 8780 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970300344990868, - "loss": 1.5683, - "step": 8781 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019702969627274572, - "loss": 1.7394, - "step": 8782 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019702935804640466, - "loss": 1.6774, - "step": 8783 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970290198200636, - "loss": 1.7838, - "step": 8784 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019702868159372251, - "loss": 1.6675, - "step": 8785 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019702834336738146, - "loss": 1.5047, - "step": 8786 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970280051410404, - "loss": 1.7338, - "step": 8787 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970276669146993, - "loss": 1.5379, - "step": 8788 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019702732868835825, - "loss": 1.6735, - "step": 8789 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970269904620172, - "loss": 1.7101, - "step": 8790 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019702665223567613, - "loss": 1.6213, - "step": 8791 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019702631400933507, - "loss": 1.6372, - "step": 8792 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197025975782994, - "loss": 1.6288, - "step": 8793 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019702563755665293, - "loss": 1.736, - "step": 8794 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019702529933031187, - "loss": 1.564, - "step": 8795 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019702496110397079, - "loss": 1.7225, - "step": 8796 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019702462287762973, - "loss": 1.5856, - "step": 8797 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019702428465128867, - "loss": 1.5864, - "step": 8798 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019702394642494758, - "loss": 1.5582, - "step": 8799 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019702360819860652, - "loss": 1.7341, - "step": 8800 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019702326997226546, - "loss": 1.6673, - "step": 8801 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019702293174592438, - "loss": 1.702, - "step": 8802 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019702259351958332, - "loss": 1.6442, - "step": 8803 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019702225529324226, - "loss": 1.6688, - "step": 8804 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019702191706690117, - "loss": 1.6334, - "step": 8805 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019702157884056012, - "loss": 1.6572, - "step": 8806 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019702124061421906, - "loss": 1.7183, - "step": 8807 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019702090238787797, - "loss": 1.6313, - "step": 8808 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970205641615369, - "loss": 1.7158, - "step": 8809 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019702022593519585, - "loss": 1.6904, - "step": 8810 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019701988770885477, - "loss": 1.5591, - "step": 8811 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970195494825137, - "loss": 1.797, - "step": 8812 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019701921125617265, - "loss": 1.6961, - "step": 8813 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019701887302983156, - "loss": 1.5652, - "step": 8814 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970185348034905, - "loss": 1.8621, - "step": 8815 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019701819657714945, - "loss": 1.6852, - "step": 8816 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019701785835080836, - "loss": 1.6125, - "step": 8817 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970175201244673, - "loss": 1.559, - "step": 8818 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019701718189812624, - "loss": 1.658, - "step": 8819 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019701684367178516, - "loss": 1.6053, - "step": 8820 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970165054454441, - "loss": 1.6939, - "step": 8821 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019701616721910304, - "loss": 1.8251, - "step": 8822 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019701582899276195, - "loss": 1.5735, - "step": 8823 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970154907664209, - "loss": 1.7823, - "step": 8824 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019701515254007983, - "loss": 1.6807, - "step": 8825 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019701481431373875, - "loss": 1.6347, - "step": 8826 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970144760873977, - "loss": 1.7053, - "step": 8827 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019701413786105663, - "loss": 1.7855, - "step": 8828 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019701379963471554, - "loss": 1.7798, - "step": 8829 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019701346140837449, - "loss": 1.7321, - "step": 8830 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019701312318203343, - "loss": 1.5957, - "step": 8831 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019701278495569234, - "loss": 1.4259, - "step": 8832 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019701244672935128, - "loss": 1.6125, - "step": 8833 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019701210850301022, - "loss": 1.5968, - "step": 8834 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019701177027666914, - "loss": 1.5156, - "step": 8835 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019701143205032808, - "loss": 1.7421, - "step": 8836 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019701109382398702, - "loss": 1.749, - "step": 8837 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019701075559764593, - "loss": 1.5512, - "step": 8838 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019701041737130487, - "loss": 1.6773, - "step": 8839 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019701007914496382, - "loss": 1.6151, - "step": 8840 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019700974091862273, - "loss": 1.5864, - "step": 8841 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019700940269228167, - "loss": 1.7689, - "step": 8842 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970090644659406, - "loss": 1.6352, - "step": 8843 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019700872623959953, - "loss": 1.6395, - "step": 8844 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019700838801325847, - "loss": 1.6622, - "step": 8845 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970080497869174, - "loss": 1.6913, - "step": 8846 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019700771156057635, - "loss": 1.668, - "step": 8847 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970073733342353, - "loss": 1.5143, - "step": 8848 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019700703510789423, - "loss": 1.717, - "step": 8849 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019700669688155315, - "loss": 1.656, - "step": 8850 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001970063586552121, - "loss": 1.6939, - "step": 8851 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019700602042887103, - "loss": 1.5899, - "step": 8852 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019700568220252994, - "loss": 1.6052, - "step": 8853 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019700534397618888, - "loss": 1.5674, - "step": 8854 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019700500574984782, - "loss": 1.6209, - "step": 8855 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019700466752350674, - "loss": 1.7481, - "step": 8856 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019700432929716568, - "loss": 1.6998, - "step": 8857 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019700399107082462, - "loss": 1.6098, - "step": 8858 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019700365284448353, - "loss": 1.6709, - "step": 8859 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019700331461814248, - "loss": 1.6438, - "step": 8860 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019700297639180142, - "loss": 1.6875, - "step": 8861 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019700263816546033, - "loss": 1.7273, - "step": 8862 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019700229993911927, - "loss": 1.6294, - "step": 8863 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019700196171277821, - "loss": 1.5992, - "step": 8864 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019700162348643713, - "loss": 1.6272, - "step": 8865 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019700128526009607, - "loss": 1.7746, - "step": 8866 - }, - { - "epoch": 0.03, - "learning_rate": 0.000197000947033755, - "loss": 1.6341, - "step": 8867 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019700060880741392, - "loss": 1.4712, - "step": 8868 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019700027058107287, - "loss": 1.6779, - "step": 8869 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969999323547318, - "loss": 1.5415, - "step": 8870 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019699959412839072, - "loss": 1.6607, - "step": 8871 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019699925590204966, - "loss": 1.7696, - "step": 8872 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969989176757086, - "loss": 1.7511, - "step": 8873 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019699857944936752, - "loss": 1.6225, - "step": 8874 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019699824122302646, - "loss": 1.5879, - "step": 8875 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969979029966854, - "loss": 1.6589, - "step": 8876 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969975647703443, - "loss": 1.5843, - "step": 8877 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019699722654400325, - "loss": 1.5463, - "step": 8878 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969968883176622, - "loss": 1.7027, - "step": 8879 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969965500913211, - "loss": 1.6204, - "step": 8880 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019699621186498005, - "loss": 1.6232, - "step": 8881 - }, - { - "epoch": 0.03, - "learning_rate": 0.000196995873638639, - "loss": 1.4706, - "step": 8882 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969955354122979, - "loss": 1.8149, - "step": 8883 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019699519718595685, - "loss": 1.677, - "step": 8884 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969948589596158, - "loss": 1.6458, - "step": 8885 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969945207332747, - "loss": 1.7716, - "step": 8886 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019699418250693364, - "loss": 1.6979, - "step": 8887 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019699384428059258, - "loss": 1.7275, - "step": 8888 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969935060542515, - "loss": 1.8393, - "step": 8889 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019699316782791044, - "loss": 1.6706, - "step": 8890 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019699282960156938, - "loss": 1.5109, - "step": 8891 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969924913752283, - "loss": 1.7032, - "step": 8892 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019699215314888724, - "loss": 1.4735, - "step": 8893 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019699181492254618, - "loss": 1.7397, - "step": 8894 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969914766962051, - "loss": 1.6712, - "step": 8895 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019699113846986403, - "loss": 1.6027, - "step": 8896 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019699080024352297, - "loss": 1.6762, - "step": 8897 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969904620171819, - "loss": 1.6696, - "step": 8898 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019699012379084083, - "loss": 1.4777, - "step": 8899 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019698978556449977, - "loss": 1.7068, - "step": 8900 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019698944733815868, - "loss": 1.5994, - "step": 8901 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019698910911181762, - "loss": 1.7742, - "step": 8902 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969887708854766, - "loss": 1.6744, - "step": 8903 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969884326591355, - "loss": 1.6704, - "step": 8904 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019698809443279445, - "loss": 1.775, - "step": 8905 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969877562064534, - "loss": 1.5817, - "step": 8906 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969874179801123, - "loss": 1.5841, - "step": 8907 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019698707975377124, - "loss": 1.6653, - "step": 8908 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019698674152743019, - "loss": 1.764, - "step": 8909 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969864033010891, - "loss": 1.6929, - "step": 8910 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019698606507474804, - "loss": 1.705, - "step": 8911 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019698572684840698, - "loss": 1.5688, - "step": 8912 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969853886220659, - "loss": 1.4896, - "step": 8913 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019698505039572484, - "loss": 1.6646, - "step": 8914 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019698471216938378, - "loss": 1.5258, - "step": 8915 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969843739430427, - "loss": 1.8163, - "step": 8916 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019698403571670163, - "loss": 1.6966, - "step": 8917 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019698369749036057, - "loss": 1.7088, - "step": 8918 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969833592640195, - "loss": 1.6077, - "step": 8919 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019698302103767843, - "loss": 1.7612, - "step": 8920 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019698268281133737, - "loss": 1.8559, - "step": 8921 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019698234458499628, - "loss": 1.6518, - "step": 8922 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019698200635865523, - "loss": 1.6773, - "step": 8923 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019698166813231417, - "loss": 1.6782, - "step": 8924 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019698132990597308, - "loss": 1.4781, - "step": 8925 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019698099167963202, - "loss": 1.7107, - "step": 8926 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019698065345329096, - "loss": 1.5894, - "step": 8927 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019698031522694988, - "loss": 1.6824, - "step": 8928 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019697997700060882, - "loss": 1.7129, - "step": 8929 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019697963877426776, - "loss": 1.7599, - "step": 8930 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019697930054792667, - "loss": 1.7198, - "step": 8931 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019697896232158561, - "loss": 1.5956, - "step": 8932 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019697862409524456, - "loss": 1.7388, - "step": 8933 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019697828586890347, - "loss": 1.6755, - "step": 8934 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969779476425624, - "loss": 1.6556, - "step": 8935 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019697760941622135, - "loss": 1.5916, - "step": 8936 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019697727118988027, - "loss": 1.6217, - "step": 8937 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969769329635392, - "loss": 1.7102, - "step": 8938 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019697659473719815, - "loss": 1.6153, - "step": 8939 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019697625651085706, - "loss": 1.6625, - "step": 8940 - }, - { - "epoch": 0.03, - "learning_rate": 0.000196975918284516, - "loss": 1.7132, - "step": 8941 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019697558005817494, - "loss": 1.7162, - "step": 8942 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019697524183183386, - "loss": 1.6331, - "step": 8943 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969749036054928, - "loss": 1.6624, - "step": 8944 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019697456537915174, - "loss": 1.8478, - "step": 8945 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019697422715281066, - "loss": 1.6675, - "step": 8946 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969738889264696, - "loss": 1.6418, - "step": 8947 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019697355070012854, - "loss": 1.5447, - "step": 8948 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019697321247378745, - "loss": 1.7641, - "step": 8949 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969728742474464, - "loss": 1.6329, - "step": 8950 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019697253602110533, - "loss": 1.6011, - "step": 8951 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019697219779476425, - "loss": 1.5729, - "step": 8952 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969718595684232, - "loss": 1.7058, - "step": 8953 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019697152134208213, - "loss": 1.6868, - "step": 8954 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019697118311574104, - "loss": 1.683, - "step": 8955 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019697084488939999, - "loss": 1.5086, - "step": 8956 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019697050666305893, - "loss": 1.6203, - "step": 8957 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019697016843671784, - "loss": 1.5655, - "step": 8958 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969698302103768, - "loss": 1.6535, - "step": 8959 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019696949198403575, - "loss": 1.5222, - "step": 8960 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019696915375769466, - "loss": 1.6639, - "step": 8961 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969688155313536, - "loss": 1.7609, - "step": 8962 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019696847730501255, - "loss": 1.5787, - "step": 8963 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019696813907867146, - "loss": 1.7596, - "step": 8964 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969678008523304, - "loss": 1.4805, - "step": 8965 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019696746262598934, - "loss": 1.6046, - "step": 8966 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019696712439964826, - "loss": 1.6377, - "step": 8967 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969667861733072, - "loss": 1.6133, - "step": 8968 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019696644794696614, - "loss": 1.8069, - "step": 8969 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019696610972062505, - "loss": 1.5614, - "step": 8970 - }, - { - "epoch": 0.03, - "learning_rate": 0.000196965771494284, - "loss": 1.5715, - "step": 8971 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019696543326794294, - "loss": 1.5788, - "step": 8972 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019696509504160185, - "loss": 1.5163, - "step": 8973 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969647568152608, - "loss": 1.6036, - "step": 8974 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019696441858891973, - "loss": 1.76, - "step": 8975 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019696408036257865, - "loss": 1.6243, - "step": 8976 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969637421362376, - "loss": 1.755, - "step": 8977 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019696340390989653, - "loss": 1.6724, - "step": 8978 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019696306568355544, - "loss": 1.8973, - "step": 8979 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019696272745721438, - "loss": 1.7206, - "step": 8980 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019696238923087332, - "loss": 1.6613, - "step": 8981 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019696205100453224, - "loss": 1.5389, - "step": 8982 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019696171277819118, - "loss": 1.5428, - "step": 8983 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019696137455185012, - "loss": 1.7077, - "step": 8984 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019696103632550903, - "loss": 1.6788, - "step": 8985 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019696069809916798, - "loss": 1.741, - "step": 8986 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019696035987282692, - "loss": 1.6487, - "step": 8987 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019696002164648583, - "loss": 1.84, - "step": 8988 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019695968342014477, - "loss": 1.652, - "step": 8989 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969593451938037, - "loss": 1.6268, - "step": 8990 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019695900696746263, - "loss": 1.8038, - "step": 8991 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019695866874112157, - "loss": 1.705, - "step": 8992 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969583305147805, - "loss": 1.7277, - "step": 8993 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019695799228843942, - "loss": 1.5407, - "step": 8994 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019695765406209836, - "loss": 1.7122, - "step": 8995 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969573158357573, - "loss": 1.5841, - "step": 8996 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019695697760941622, - "loss": 1.642, - "step": 8997 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019695663938307516, - "loss": 1.7159, - "step": 8998 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969563011567341, - "loss": 1.7995, - "step": 8999 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019695596293039302, - "loss": 1.7269, - "step": 9000 - }, - { - "epoch": 0.03, - "eval_loss": 1.3394869290865385, - "mrc_em": 19.0, - "mrc_f1": 26.60952380952381, - "nli_acc": 0.22, - "step": 9000, - "sts_f1": 0.4999999999999999, - "ynat_f1": 0.590829217062319 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019695562470405196, - "loss": 1.8735, - "step": 9001 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969552864777109, - "loss": 1.5599, - "step": 9002 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969549482513698, - "loss": 1.7842, - "step": 9003 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019695461002502875, - "loss": 1.5781, - "step": 9004 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969542717986877, - "loss": 1.6739, - "step": 9005 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969539335723466, - "loss": 1.6564, - "step": 9006 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019695359534600555, - "loss": 1.7137, - "step": 9007 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969532571196645, - "loss": 1.6366, - "step": 9008 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969529188933234, - "loss": 1.7503, - "step": 9009 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019695258066698235, - "loss": 1.6939, - "step": 9010 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969522424406413, - "loss": 1.6367, - "step": 9011 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969519042143002, - "loss": 1.7034, - "step": 9012 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019695156598795914, - "loss": 1.6824, - "step": 9013 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019695122776161808, - "loss": 1.6711, - "step": 9014 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019695088953527702, - "loss": 1.7744, - "step": 9015 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019695055130893597, - "loss": 1.8277, - "step": 9016 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019695021308259488, - "loss": 1.5278, - "step": 9017 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019694987485625382, - "loss": 1.5295, - "step": 9018 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019694953662991276, - "loss": 1.7146, - "step": 9019 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019694919840357168, - "loss": 1.6724, - "step": 9020 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019694886017723062, - "loss": 1.6836, - "step": 9021 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019694852195088956, - "loss": 1.6866, - "step": 9022 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019694818372454847, - "loss": 1.6845, - "step": 9023 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969478454982074, - "loss": 1.5969, - "step": 9024 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019694750727186635, - "loss": 1.5637, - "step": 9025 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019694716904552527, - "loss": 1.6988, - "step": 9026 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969468308191842, - "loss": 1.5462, - "step": 9027 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019694649259284315, - "loss": 1.6225, - "step": 9028 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019694615436650207, - "loss": 1.6267, - "step": 9029 - }, - { - "epoch": 0.03, - "learning_rate": 0.000196945816140161, - "loss": 1.7366, - "step": 9030 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019694547791381995, - "loss": 1.638, - "step": 9031 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019694513968747886, - "loss": 1.5913, - "step": 9032 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969448014611378, - "loss": 1.8596, - "step": 9033 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019694446323479674, - "loss": 1.7284, - "step": 9034 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019694412500845566, - "loss": 1.6061, - "step": 9035 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969437867821146, - "loss": 1.6141, - "step": 9036 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019694344855577354, - "loss": 1.6123, - "step": 9037 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019694311032943248, - "loss": 1.7077, - "step": 9038 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969427721030914, - "loss": 1.7782, - "step": 9039 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019694243387675034, - "loss": 1.7031, - "step": 9040 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019694209565040928, - "loss": 1.7386, - "step": 9041 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969417574240682, - "loss": 1.7431, - "step": 9042 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019694141919772713, - "loss": 1.5326, - "step": 9043 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019694108097138607, - "loss": 1.6611, - "step": 9044 - }, - { - "epoch": 0.03, - "learning_rate": 0.000196940742745045, - "loss": 1.7799, - "step": 9045 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019694040451870393, - "loss": 1.6265, - "step": 9046 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019694006629236287, - "loss": 1.6132, - "step": 9047 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019693972806602178, - "loss": 1.5282, - "step": 9048 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019693938983968073, - "loss": 1.7233, - "step": 9049 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019693905161333967, - "loss": 1.6552, - "step": 9050 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019693871338699858, - "loss": 1.7476, - "step": 9051 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019693837516065752, - "loss": 1.6049, - "step": 9052 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019693803693431646, - "loss": 1.6958, - "step": 9053 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019693769870797538, - "loss": 1.6715, - "step": 9054 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019693736048163432, - "loss": 1.5695, - "step": 9055 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019693702225529326, - "loss": 1.7518, - "step": 9056 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019693668402895217, - "loss": 1.5868, - "step": 9057 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019693634580261111, - "loss": 1.6309, - "step": 9058 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019693600757627006, - "loss": 1.6892, - "step": 9059 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019693566934992897, - "loss": 1.7244, - "step": 9060 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969353311235879, - "loss": 1.5644, - "step": 9061 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019693499289724685, - "loss": 1.5529, - "step": 9062 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019693465467090577, - "loss": 1.4366, - "step": 9063 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969343164445647, - "loss": 1.6383, - "step": 9064 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019693397821822365, - "loss": 1.5569, - "step": 9065 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019693363999188256, - "loss": 1.5735, - "step": 9066 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969333017655415, - "loss": 1.7702, - "step": 9067 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019693296353920044, - "loss": 1.6789, - "step": 9068 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019693262531285936, - "loss": 1.8017, - "step": 9069 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969322870865183, - "loss": 1.5789, - "step": 9070 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019693194886017724, - "loss": 1.7313, - "step": 9071 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019693161063383618, - "loss": 1.6224, - "step": 9072 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019693127240749512, - "loss": 1.6666, - "step": 9073 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019693093418115404, - "loss": 1.5795, - "step": 9074 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019693059595481298, - "loss": 1.5791, - "step": 9075 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019693025772847192, - "loss": 1.6364, - "step": 9076 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019692991950213083, - "loss": 1.7194, - "step": 9077 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019692958127578977, - "loss": 1.6516, - "step": 9078 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019692924304944872, - "loss": 1.5195, - "step": 9079 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019692890482310763, - "loss": 1.6455, - "step": 9080 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019692856659676657, - "loss": 1.5765, - "step": 9081 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969282283704255, - "loss": 1.8231, - "step": 9082 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019692789014408443, - "loss": 1.5171, - "step": 9083 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019692755191774337, - "loss": 1.6815, - "step": 9084 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969272136914023, - "loss": 1.5436, - "step": 9085 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019692687546506122, - "loss": 1.5725, - "step": 9086 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019692653723872016, - "loss": 1.6717, - "step": 9087 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969261990123791, - "loss": 1.7597, - "step": 9088 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019692586078603802, - "loss": 1.7478, - "step": 9089 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019692552255969696, - "loss": 1.6196, - "step": 9090 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969251843333559, - "loss": 1.5382, - "step": 9091 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019692484610701481, - "loss": 1.6823, - "step": 9092 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019692450788067376, - "loss": 1.7893, - "step": 9093 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969241696543327, - "loss": 1.6236, - "step": 9094 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969238314279916, - "loss": 1.7691, - "step": 9095 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019692349320165055, - "loss": 1.5972, - "step": 9096 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969231549753095, - "loss": 1.6729, - "step": 9097 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969228167489684, - "loss": 1.7466, - "step": 9098 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019692247852262735, - "loss": 1.6299, - "step": 9099 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969221402962863, - "loss": 1.6342, - "step": 9100 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969218020699452, - "loss": 1.6503, - "step": 9101 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019692146384360414, - "loss": 1.7075, - "step": 9102 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019692112561726309, - "loss": 1.6782, - "step": 9103 - }, - { - "epoch": 0.03, - "learning_rate": 0.000196920787390922, - "loss": 1.603, - "step": 9104 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019692044916458094, - "loss": 1.6529, - "step": 9105 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019692011093823988, - "loss": 1.4636, - "step": 9106 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969197727118988, - "loss": 1.7402, - "step": 9107 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019691943448555774, - "loss": 1.7273, - "step": 9108 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019691909625921668, - "loss": 1.5442, - "step": 9109 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969187580328756, - "loss": 1.6784, - "step": 9110 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019691841980653453, - "loss": 1.7047, - "step": 9111 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019691808158019347, - "loss": 1.6198, - "step": 9112 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969177433538524, - "loss": 1.7314, - "step": 9113 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019691740512751133, - "loss": 1.6924, - "step": 9114 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019691706690117027, - "loss": 1.7658, - "step": 9115 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019691672867482919, - "loss": 1.7028, - "step": 9116 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019691639044848813, - "loss": 1.6513, - "step": 9117 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019691605222214707, - "loss": 1.8916, - "step": 9118 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019691571399580598, - "loss": 1.6114, - "step": 9119 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019691537576946492, - "loss": 1.6583, - "step": 9120 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019691503754312386, - "loss": 1.6461, - "step": 9121 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019691469931678278, - "loss": 1.6473, - "step": 9122 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019691436109044172, - "loss": 1.6142, - "step": 9123 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019691402286410066, - "loss": 1.6186, - "step": 9124 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969136846377596, - "loss": 1.4952, - "step": 9125 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019691334641141852, - "loss": 1.6393, - "step": 9126 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019691300818507746, - "loss": 1.6974, - "step": 9127 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969126699587364, - "loss": 1.6357, - "step": 9128 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019691233173239534, - "loss": 1.7135, - "step": 9129 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019691199350605428, - "loss": 1.799, - "step": 9130 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969116552797132, - "loss": 1.5879, - "step": 9131 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019691131705337214, - "loss": 1.7336, - "step": 9132 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019691097882703108, - "loss": 1.9116, - "step": 9133 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019691064060069, - "loss": 1.6047, - "step": 9134 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019691030237434893, - "loss": 1.6342, - "step": 9135 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019690996414800787, - "loss": 1.7609, - "step": 9136 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969096259216668, - "loss": 1.6556, - "step": 9137 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019690928769532573, - "loss": 1.5601, - "step": 9138 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019690894946898467, - "loss": 1.6524, - "step": 9139 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019690861124264358, - "loss": 1.615, - "step": 9140 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019690827301630252, - "loss": 1.7566, - "step": 9141 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019690793478996147, - "loss": 1.6048, - "step": 9142 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019690759656362038, - "loss": 1.5678, - "step": 9143 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019690725833727932, - "loss": 1.7467, - "step": 9144 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019690692011093826, - "loss": 1.7893, - "step": 9145 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019690658188459718, - "loss": 1.7018, - "step": 9146 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019690624365825612, - "loss": 1.7091, - "step": 9147 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019690590543191506, - "loss": 1.7907, - "step": 9148 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019690556720557397, - "loss": 1.7106, - "step": 9149 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969052289792329, - "loss": 1.6355, - "step": 9150 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019690489075289185, - "loss": 1.7319, - "step": 9151 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019690455252655077, - "loss": 1.5459, - "step": 9152 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969042143002097, - "loss": 1.5998, - "step": 9153 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019690387607386865, - "loss": 1.6049, - "step": 9154 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019690353784752756, - "loss": 1.6693, - "step": 9155 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969031996211865, - "loss": 1.5923, - "step": 9156 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019690286139484545, - "loss": 1.607, - "step": 9157 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019690252316850436, - "loss": 1.7846, - "step": 9158 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969021849421633, - "loss": 1.505, - "step": 9159 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019690184671582224, - "loss": 1.7123, - "step": 9160 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019690150848948116, - "loss": 1.8188, - "step": 9161 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969011702631401, - "loss": 1.6787, - "step": 9162 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019690083203679904, - "loss": 1.5405, - "step": 9163 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019690049381045795, - "loss": 1.6192, - "step": 9164 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001969001555841169, - "loss": 1.6185, - "step": 9165 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019689981735777584, - "loss": 1.7353, - "step": 9166 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019689947913143475, - "loss": 1.7469, - "step": 9167 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968991409050937, - "loss": 1.8446, - "step": 9168 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019689880267875263, - "loss": 1.6301, - "step": 9169 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019689846445241155, - "loss": 1.6177, - "step": 9170 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968981262260705, - "loss": 1.5849, - "step": 9171 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019689778799972943, - "loss": 1.6684, - "step": 9172 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019689744977338834, - "loss": 1.624, - "step": 9173 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019689711154704728, - "loss": 1.4792, - "step": 9174 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019689677332070622, - "loss": 1.6188, - "step": 9175 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019689643509436514, - "loss": 1.5866, - "step": 9176 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019689609686802408, - "loss": 1.5349, - "step": 9177 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019689575864168302, - "loss": 1.608, - "step": 9178 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019689542041534193, - "loss": 1.7518, - "step": 9179 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019689508218900088, - "loss": 1.5916, - "step": 9180 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019689474396265982, - "loss": 1.6013, - "step": 9181 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019689440573631873, - "loss": 1.5862, - "step": 9182 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019689406750997767, - "loss": 1.5221, - "step": 9183 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019689372928363664, - "loss": 1.507, - "step": 9184 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019689339105729555, - "loss": 1.7171, - "step": 9185 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968930528309545, - "loss": 1.5881, - "step": 9186 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019689271460461344, - "loss": 1.6018, - "step": 9187 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019689237637827235, - "loss": 1.7558, - "step": 9188 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968920381519313, - "loss": 1.6616, - "step": 9189 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019689169992559023, - "loss": 1.6216, - "step": 9190 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019689136169924915, - "loss": 1.6076, - "step": 9191 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968910234729081, - "loss": 1.5711, - "step": 9192 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019689068524656703, - "loss": 1.5836, - "step": 9193 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019689034702022594, - "loss": 1.539, - "step": 9194 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019689000879388488, - "loss": 1.647, - "step": 9195 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019688967056754383, - "loss": 1.5794, - "step": 9196 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019688933234120274, - "loss": 1.6889, - "step": 9197 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019688899411486168, - "loss": 1.5184, - "step": 9198 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019688865588852062, - "loss": 1.4916, - "step": 9199 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019688831766217954, - "loss": 1.5689, - "step": 9200 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019688797943583848, - "loss": 1.545, - "step": 9201 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019688764120949742, - "loss": 1.5367, - "step": 9202 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019688730298315633, - "loss": 1.5893, - "step": 9203 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019688696475681527, - "loss": 1.7192, - "step": 9204 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019688662653047421, - "loss": 1.5757, - "step": 9205 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019688628830413313, - "loss": 1.6865, - "step": 9206 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019688595007779207, - "loss": 1.7867, - "step": 9207 - }, - { - "epoch": 0.03, - "learning_rate": 0.000196885611851451, - "loss": 1.682, - "step": 9208 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019688527362510993, - "loss": 1.7003, - "step": 9209 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019688493539876887, - "loss": 1.4955, - "step": 9210 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968845971724278, - "loss": 1.6283, - "step": 9211 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019688425894608672, - "loss": 1.6011, - "step": 9212 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019688392071974566, - "loss": 1.6907, - "step": 9213 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968835824934046, - "loss": 1.7151, - "step": 9214 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019688324426706352, - "loss": 1.5531, - "step": 9215 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019688290604072246, - "loss": 1.5503, - "step": 9216 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968825678143814, - "loss": 1.685, - "step": 9217 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019688222958804031, - "loss": 1.5495, - "step": 9218 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019688189136169926, - "loss": 1.6988, - "step": 9219 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968815531353582, - "loss": 1.5391, - "step": 9220 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968812149090171, - "loss": 1.7237, - "step": 9221 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019688087668267605, - "loss": 1.5944, - "step": 9222 - }, - { - "epoch": 0.03, - "learning_rate": 0.000196880538456335, - "loss": 1.5591, - "step": 9223 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968802002299939, - "loss": 1.5645, - "step": 9224 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019687986200365285, - "loss": 1.6077, - "step": 9225 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968795237773118, - "loss": 1.788, - "step": 9226 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968791855509707, - "loss": 1.6086, - "step": 9227 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019687884732462964, - "loss": 1.6293, - "step": 9228 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019687850909828859, - "loss": 1.6258, - "step": 9229 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968781708719475, - "loss": 1.5895, - "step": 9230 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019687783264560644, - "loss": 1.7888, - "step": 9231 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019687749441926538, - "loss": 1.7807, - "step": 9232 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968771561929243, - "loss": 1.5728, - "step": 9233 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019687681796658324, - "loss": 1.7056, - "step": 9234 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019687647974024218, - "loss": 1.6839, - "step": 9235 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968761415139011, - "loss": 1.5183, - "step": 9236 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019687580328756003, - "loss": 1.6334, - "step": 9237 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019687546506121897, - "loss": 1.6626, - "step": 9238 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968751268348779, - "loss": 1.7788, - "step": 9239 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019687478860853686, - "loss": 1.6408, - "step": 9240 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968744503821958, - "loss": 1.6146, - "step": 9241 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968741121558547, - "loss": 1.6391, - "step": 9242 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019687377392951365, - "loss": 1.6094, - "step": 9243 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968734357031726, - "loss": 1.7434, - "step": 9244 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968730974768315, - "loss": 1.7764, - "step": 9245 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019687275925049045, - "loss": 1.7356, - "step": 9246 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968724210241494, - "loss": 1.7138, - "step": 9247 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968720827978083, - "loss": 1.6467, - "step": 9248 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019687174457146725, - "loss": 1.5543, - "step": 9249 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968714063451262, - "loss": 1.6586, - "step": 9250 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968710681187851, - "loss": 1.6614, - "step": 9251 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019687072989244404, - "loss": 1.6764, - "step": 9252 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019687039166610298, - "loss": 1.5142, - "step": 9253 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968700534397619, - "loss": 1.7651, - "step": 9254 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019686971521342084, - "loss": 1.5613, - "step": 9255 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019686937698707978, - "loss": 1.6798, - "step": 9256 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968690387607387, - "loss": 1.7645, - "step": 9257 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019686870053439763, - "loss": 1.5926, - "step": 9258 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019686836230805658, - "loss": 1.7004, - "step": 9259 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968680240817155, - "loss": 1.5547, - "step": 9260 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019686768585537443, - "loss": 1.5961, - "step": 9261 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019686734762903337, - "loss": 1.6116, - "step": 9262 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019686700940269229, - "loss": 1.7189, - "step": 9263 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019686667117635123, - "loss": 1.6228, - "step": 9264 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019686633295001017, - "loss": 1.7306, - "step": 9265 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019686599472366908, - "loss": 1.5723, - "step": 9266 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019686565649732802, - "loss": 1.8741, - "step": 9267 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019686531827098696, - "loss": 1.7027, - "step": 9268 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019686498004464588, - "loss": 1.6047, - "step": 9269 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019686464181830482, - "loss": 1.8159, - "step": 9270 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019686430359196376, - "loss": 1.6473, - "step": 9271 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019686396536562267, - "loss": 1.772, - "step": 9272 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019686362713928162, - "loss": 1.6982, - "step": 9273 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019686328891294056, - "loss": 1.74, - "step": 9274 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019686295068659947, - "loss": 1.7669, - "step": 9275 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968626124602584, - "loss": 1.6492, - "step": 9276 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019686227423391735, - "loss": 1.7398, - "step": 9277 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019686193600757627, - "loss": 1.5832, - "step": 9278 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968615977812352, - "loss": 1.6354, - "step": 9279 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019686125955489415, - "loss": 1.5328, - "step": 9280 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019686092132855306, - "loss": 1.439, - "step": 9281 - }, - { - "epoch": 0.03, - "learning_rate": 0.000196860583102212, - "loss": 1.6629, - "step": 9282 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019686024487587095, - "loss": 1.7268, - "step": 9283 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019685990664952986, - "loss": 1.6477, - "step": 9284 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968595684231888, - "loss": 1.6345, - "step": 9285 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019685923019684774, - "loss": 1.7254, - "step": 9286 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019685889197050666, - "loss": 1.7383, - "step": 9287 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968585537441656, - "loss": 1.664, - "step": 9288 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019685821551782454, - "loss": 1.5987, - "step": 9289 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019685787729148345, - "loss": 1.6598, - "step": 9290 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968575390651424, - "loss": 1.6794, - "step": 9291 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019685720083880134, - "loss": 1.561, - "step": 9292 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019685686261246025, - "loss": 1.6104, - "step": 9293 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968565243861192, - "loss": 1.6985, - "step": 9294 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019685618615977813, - "loss": 1.6651, - "step": 9295 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019685584793343707, - "loss": 1.6284, - "step": 9296 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019685550970709601, - "loss": 1.689, - "step": 9297 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019685517148075493, - "loss": 1.5253, - "step": 9298 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019685483325441387, - "loss": 1.5734, - "step": 9299 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968544950280728, - "loss": 1.5073, - "step": 9300 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019685415680173172, - "loss": 1.6721, - "step": 9301 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019685381857539067, - "loss": 1.5168, - "step": 9302 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968534803490496, - "loss": 1.8495, - "step": 9303 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019685314212270852, - "loss": 1.6716, - "step": 9304 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019685280389636746, - "loss": 1.5553, - "step": 9305 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968524656700264, - "loss": 1.7601, - "step": 9306 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019685212744368534, - "loss": 1.6779, - "step": 9307 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019685178921734426, - "loss": 1.631, - "step": 9308 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968514509910032, - "loss": 1.6335, - "step": 9309 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019685111276466214, - "loss": 1.6975, - "step": 9310 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019685077453832105, - "loss": 1.6821, - "step": 9311 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019685043631198, - "loss": 1.6267, - "step": 9312 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019685009808563894, - "loss": 1.5309, - "step": 9313 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019684975985929785, - "loss": 1.7409, - "step": 9314 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968494216329568, - "loss": 1.8237, - "step": 9315 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019684908340661573, - "loss": 1.6496, - "step": 9316 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019684874518027465, - "loss": 1.6802, - "step": 9317 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968484069539336, - "loss": 1.7181, - "step": 9318 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019684806872759253, - "loss": 1.6264, - "step": 9319 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019684773050125144, - "loss": 1.5772, - "step": 9320 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019684739227491038, - "loss": 1.6524, - "step": 9321 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019684705404856933, - "loss": 1.6348, - "step": 9322 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019684671582222824, - "loss": 1.8217, - "step": 9323 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019684637759588718, - "loss": 1.5939, - "step": 9324 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019684603936954612, - "loss": 1.5074, - "step": 9325 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019684570114320504, - "loss": 1.814, - "step": 9326 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019684536291686398, - "loss": 1.5635, - "step": 9327 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019684502469052292, - "loss": 1.6234, - "step": 9328 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019684468646418183, - "loss": 1.6055, - "step": 9329 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019684434823784077, - "loss": 1.5598, - "step": 9330 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019684401001149971, - "loss": 1.71, - "step": 9331 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019684367178515863, - "loss": 1.6724, - "step": 9332 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019684333355881757, - "loss": 1.869, - "step": 9333 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968429953324765, - "loss": 1.7317, - "step": 9334 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019684265710613542, - "loss": 1.6928, - "step": 9335 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019684231887979437, - "loss": 1.6804, - "step": 9336 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968419806534533, - "loss": 1.6154, - "step": 9337 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019684164242711222, - "loss": 1.5968, - "step": 9338 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019684130420077116, - "loss": 1.5534, - "step": 9339 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968409659744301, - "loss": 1.6295, - "step": 9340 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019684062774808902, - "loss": 1.6339, - "step": 9341 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019684028952174796, - "loss": 1.581, - "step": 9342 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968399512954069, - "loss": 1.5716, - "step": 9343 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968396130690658, - "loss": 1.8029, - "step": 9344 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019683927484272475, - "loss": 1.7219, - "step": 9345 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968389366163837, - "loss": 1.7457, - "step": 9346 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968385983900426, - "loss": 1.5589, - "step": 9347 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019683826016370155, - "loss": 1.6014, - "step": 9348 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968379219373605, - "loss": 1.6713, - "step": 9349 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968375837110194, - "loss": 1.5272, - "step": 9350 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019683724548467835, - "loss": 1.6968, - "step": 9351 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968369072583373, - "loss": 1.7181, - "step": 9352 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019683656903199623, - "loss": 1.5816, - "step": 9353 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019683623080565517, - "loss": 1.6024, - "step": 9354 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019683589257931408, - "loss": 1.6639, - "step": 9355 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019683555435297303, - "loss": 1.6298, - "step": 9356 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019683521612663197, - "loss": 1.7239, - "step": 9357 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019683487790029088, - "loss": 1.6425, - "step": 9358 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019683453967394982, - "loss": 1.5283, - "step": 9359 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019683420144760876, - "loss": 1.7295, - "step": 9360 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019683386322126768, - "loss": 1.6417, - "step": 9361 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019683352499492662, - "loss": 1.7108, - "step": 9362 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019683318676858556, - "loss": 1.5107, - "step": 9363 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019683284854224447, - "loss": 1.5302, - "step": 9364 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019683251031590341, - "loss": 1.6499, - "step": 9365 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019683217208956236, - "loss": 1.6151, - "step": 9366 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019683183386322127, - "loss": 1.5715, - "step": 9367 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968314956368802, - "loss": 1.6929, - "step": 9368 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019683115741053915, - "loss": 1.7135, - "step": 9369 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019683081918419807, - "loss": 1.6731, - "step": 9370 - }, - { - "epoch": 0.03, - "learning_rate": 0.000196830480957857, - "loss": 1.6303, - "step": 9371 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019683014273151595, - "loss": 1.6786, - "step": 9372 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019682980450517486, - "loss": 1.6737, - "step": 9373 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968294662788338, - "loss": 1.6534, - "step": 9374 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019682912805249274, - "loss": 1.7272, - "step": 9375 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019682878982615166, - "loss": 1.609, - "step": 9376 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968284515998106, - "loss": 1.6855, - "step": 9377 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019682811337346954, - "loss": 1.5552, - "step": 9378 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019682777514712846, - "loss": 1.6721, - "step": 9379 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968274369207874, - "loss": 1.705, - "step": 9380 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019682709869444634, - "loss": 1.7987, - "step": 9381 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019682676046810525, - "loss": 1.594, - "step": 9382 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968264222417642, - "loss": 1.6244, - "step": 9383 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019682608401542313, - "loss": 1.609, - "step": 9384 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019682574578908205, - "loss": 1.6215, - "step": 9385 - }, - { - "epoch": 0.03, - "learning_rate": 0.000196825407562741, - "loss": 1.4568, - "step": 9386 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019682506933639993, - "loss": 1.643, - "step": 9387 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019682473111005884, - "loss": 1.4538, - "step": 9388 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019682439288371779, - "loss": 1.6194, - "step": 9389 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019682405465737673, - "loss": 1.602, - "step": 9390 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019682371643103567, - "loss": 1.7288, - "step": 9391 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019682337820469458, - "loss": 1.6338, - "step": 9392 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019682303997835352, - "loss": 1.6515, - "step": 9393 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019682270175201246, - "loss": 1.6599, - "step": 9394 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019682236352567138, - "loss": 1.7513, - "step": 9395 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019682202529933032, - "loss": 1.705, - "step": 9396 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019682168707298926, - "loss": 1.4693, - "step": 9397 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019682134884664817, - "loss": 1.7784, - "step": 9398 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019682101062030712, - "loss": 1.609, - "step": 9399 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019682067239396606, - "loss": 1.6, - "step": 9400 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019682033416762497, - "loss": 1.7835, - "step": 9401 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968199959412839, - "loss": 1.7148, - "step": 9402 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019681965771494285, - "loss": 1.5334, - "step": 9403 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019681931948860177, - "loss": 1.639, - "step": 9404 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968189812622607, - "loss": 1.5498, - "step": 9405 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019681864303591965, - "loss": 1.7356, - "step": 9406 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019681830480957856, - "loss": 1.819, - "step": 9407 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019681796658323753, - "loss": 1.6836, - "step": 9408 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019681762835689645, - "loss": 1.7217, - "step": 9409 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968172901305554, - "loss": 1.5944, - "step": 9410 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019681695190421433, - "loss": 1.5312, - "step": 9411 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019681661367787324, - "loss": 1.5642, - "step": 9412 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019681627545153218, - "loss": 1.5638, - "step": 9413 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019681593722519112, - "loss": 1.6189, - "step": 9414 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019681559899885004, - "loss": 1.6746, - "step": 9415 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019681526077250898, - "loss": 1.7707, - "step": 9416 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019681492254616792, - "loss": 1.6711, - "step": 9417 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019681458431982683, - "loss": 1.7438, - "step": 9418 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019681424609348578, - "loss": 1.5824, - "step": 9419 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019681390786714472, - "loss": 1.5866, - "step": 9420 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019681356964080363, - "loss": 1.61, - "step": 9421 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019681323141446257, - "loss": 1.6212, - "step": 9422 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968128931881215, - "loss": 1.5557, - "step": 9423 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019681255496178043, - "loss": 1.5853, - "step": 9424 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019681221673543937, - "loss": 1.7473, - "step": 9425 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968118785090983, - "loss": 1.7517, - "step": 9426 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019681154028275722, - "loss": 1.6305, - "step": 9427 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019681120205641616, - "loss": 1.659, - "step": 9428 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968108638300751, - "loss": 1.5433, - "step": 9429 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019681052560373402, - "loss": 1.6872, - "step": 9430 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019681018737739296, - "loss": 1.7232, - "step": 9431 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968098491510519, - "loss": 1.824, - "step": 9432 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019680951092471082, - "loss": 1.7122, - "step": 9433 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019680917269836976, - "loss": 1.5264, - "step": 9434 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968088344720287, - "loss": 1.7583, - "step": 9435 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968084962456876, - "loss": 1.5623, - "step": 9436 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019680815801934655, - "loss": 1.5601, - "step": 9437 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968078197930055, - "loss": 1.6559, - "step": 9438 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968074815666644, - "loss": 1.6218, - "step": 9439 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019680714334032335, - "loss": 1.5587, - "step": 9440 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968068051139823, - "loss": 1.7203, - "step": 9441 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968064668876412, - "loss": 1.6408, - "step": 9442 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019680612866130015, - "loss": 1.5887, - "step": 9443 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968057904349591, - "loss": 1.5429, - "step": 9444 - }, - { - "epoch": 0.03, - "learning_rate": 0.000196805452208618, - "loss": 1.7355, - "step": 9445 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019680511398227694, - "loss": 1.6262, - "step": 9446 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019680477575593588, - "loss": 1.7451, - "step": 9447 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968044375295948, - "loss": 1.6748, - "step": 9448 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019680409930325374, - "loss": 1.5921, - "step": 9449 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019680376107691268, - "loss": 1.6555, - "step": 9450 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968034228505716, - "loss": 1.6559, - "step": 9451 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019680308462423053, - "loss": 1.5476, - "step": 9452 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019680274639788948, - "loss": 1.5199, - "step": 9453 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968024081715484, - "loss": 1.6291, - "step": 9454 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019680206994520733, - "loss": 1.7427, - "step": 9455 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019680173171886627, - "loss": 1.6769, - "step": 9456 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001968013934925252, - "loss": 1.6786, - "step": 9457 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019680105526618413, - "loss": 1.7939, - "step": 9458 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019680071703984307, - "loss": 1.6868, - "step": 9459 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019680037881350198, - "loss": 1.7042, - "step": 9460 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019680004058716092, - "loss": 1.6183, - "step": 9461 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019679970236081987, - "loss": 1.75, - "step": 9462 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019679936413447878, - "loss": 1.6342, - "step": 9463 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019679902590813772, - "loss": 1.721, - "step": 9464 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967986876817967, - "loss": 1.5083, - "step": 9465 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967983494554556, - "loss": 1.6344, - "step": 9466 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019679801122911454, - "loss": 1.6012, - "step": 9467 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019679767300277348, - "loss": 1.6634, - "step": 9468 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967973347764324, - "loss": 1.7755, - "step": 9469 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019679699655009134, - "loss": 1.6455, - "step": 9470 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019679665832375028, - "loss": 1.6399, - "step": 9471 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967963200974092, - "loss": 1.678, - "step": 9472 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019679598187106814, - "loss": 1.6867, - "step": 9473 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019679564364472708, - "loss": 1.6113, - "step": 9474 - }, - { - "epoch": 0.03, - "learning_rate": 0.000196795305418386, - "loss": 1.7263, - "step": 9475 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019679496719204493, - "loss": 1.8082, - "step": 9476 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019679462896570387, - "loss": 1.7472, - "step": 9477 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967942907393628, - "loss": 1.6629, - "step": 9478 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019679395251302173, - "loss": 1.6214, - "step": 9479 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019679361428668067, - "loss": 1.6925, - "step": 9480 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019679327606033958, - "loss": 1.4945, - "step": 9481 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019679293783399853, - "loss": 1.8226, - "step": 9482 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019679259960765747, - "loss": 1.5223, - "step": 9483 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019679226138131638, - "loss": 1.7863, - "step": 9484 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019679192315497532, - "loss": 1.7851, - "step": 9485 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019679158492863426, - "loss": 1.6693, - "step": 9486 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019679124670229318, - "loss": 1.6677, - "step": 9487 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019679090847595212, - "loss": 1.736, - "step": 9488 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019679057024961106, - "loss": 1.6012, - "step": 9489 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019679023202326997, - "loss": 1.886, - "step": 9490 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019678989379692891, - "loss": 1.4867, - "step": 9491 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019678955557058786, - "loss": 1.7242, - "step": 9492 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019678921734424677, - "loss": 1.7899, - "step": 9493 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967888791179057, - "loss": 1.6892, - "step": 9494 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019678854089156465, - "loss": 1.4758, - "step": 9495 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019678820266522357, - "loss": 1.6658, - "step": 9496 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967878644388825, - "loss": 1.5888, - "step": 9497 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019678752621254145, - "loss": 1.634, - "step": 9498 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019678718798620036, - "loss": 1.6579, - "step": 9499 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967868497598593, - "loss": 1.5562, - "step": 9500 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019678651153351824, - "loss": 1.5662, - "step": 9501 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019678617330717716, - "loss": 1.6659, - "step": 9502 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967858350808361, - "loss": 1.6263, - "step": 9503 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019678549685449504, - "loss": 1.4569, - "step": 9504 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019678515862815395, - "loss": 1.6285, - "step": 9505 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967848204018129, - "loss": 1.576, - "step": 9506 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019678448217547184, - "loss": 1.7502, - "step": 9507 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019678414394913075, - "loss": 1.493, - "step": 9508 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967838057227897, - "loss": 1.6352, - "step": 9509 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019678346749644863, - "loss": 1.417, - "step": 9510 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019678312927010755, - "loss": 1.5899, - "step": 9511 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967827910437665, - "loss": 1.733, - "step": 9512 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019678245281742543, - "loss": 1.4808, - "step": 9513 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019678211459108434, - "loss": 1.7269, - "step": 9514 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019678177636474328, - "loss": 1.5923, - "step": 9515 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019678143813840223, - "loss": 1.4783, - "step": 9516 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019678109991206114, - "loss": 1.5506, - "step": 9517 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019678076168572008, - "loss": 1.6621, - "step": 9518 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019678042345937902, - "loss": 1.6683, - "step": 9519 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019678008523303794, - "loss": 1.8926, - "step": 9520 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967797470066969, - "loss": 1.7, - "step": 9521 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019677940878035585, - "loss": 1.6304, - "step": 9522 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019677907055401476, - "loss": 1.7599, - "step": 9523 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967787323276737, - "loss": 1.5409, - "step": 9524 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019677839410133264, - "loss": 1.6918, - "step": 9525 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019677805587499156, - "loss": 1.7392, - "step": 9526 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967777176486505, - "loss": 1.6492, - "step": 9527 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019677737942230944, - "loss": 1.5348, - "step": 9528 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019677704119596835, - "loss": 1.5305, - "step": 9529 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967767029696273, - "loss": 1.7636, - "step": 9530 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019677636474328623, - "loss": 1.659, - "step": 9531 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019677602651694515, - "loss": 1.6849, - "step": 9532 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967756882906041, - "loss": 1.7702, - "step": 9533 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019677535006426303, - "loss": 1.5598, - "step": 9534 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019677501183792194, - "loss": 1.5814, - "step": 9535 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019677467361158089, - "loss": 1.7801, - "step": 9536 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019677433538523983, - "loss": 1.7011, - "step": 9537 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019677399715889874, - "loss": 1.7524, - "step": 9538 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019677365893255768, - "loss": 1.6958, - "step": 9539 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019677332070621662, - "loss": 1.7426, - "step": 9540 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019677298247987554, - "loss": 1.521, - "step": 9541 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019677264425353448, - "loss": 1.6671, - "step": 9542 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019677230602719342, - "loss": 1.7273, - "step": 9543 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019677196780085233, - "loss": 1.6167, - "step": 9544 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019677162957451127, - "loss": 1.6648, - "step": 9545 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019677129134817022, - "loss": 1.8005, - "step": 9546 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019677095312182913, - "loss": 1.6454, - "step": 9547 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019677061489548807, - "loss": 1.7279, - "step": 9548 - }, - { - "epoch": 0.03, - "learning_rate": 0.000196770276669147, - "loss": 1.7908, - "step": 9549 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019676993844280593, - "loss": 1.6767, - "step": 9550 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019676960021646487, - "loss": 1.7445, - "step": 9551 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967692619901238, - "loss": 1.6164, - "step": 9552 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019676892376378272, - "loss": 1.7662, - "step": 9553 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019676858553744166, - "loss": 1.6791, - "step": 9554 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967682473111006, - "loss": 1.6583, - "step": 9555 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019676790908475952, - "loss": 1.6689, - "step": 9556 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019676757085841846, - "loss": 1.7034, - "step": 9557 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967672326320774, - "loss": 1.7321, - "step": 9558 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019676689440573632, - "loss": 1.6966, - "step": 9559 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019676655617939526, - "loss": 1.6727, - "step": 9560 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967662179530542, - "loss": 1.6012, - "step": 9561 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967658797267131, - "loss": 1.7223, - "step": 9562 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019676554150037205, - "loss": 1.4909, - "step": 9563 - }, - { - "epoch": 0.03, - "learning_rate": 0.000196765203274031, - "loss": 1.6232, - "step": 9564 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967648650476899, - "loss": 1.6392, - "step": 9565 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019676452682134885, - "loss": 1.7548, - "step": 9566 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967641885950078, - "loss": 1.5428, - "step": 9567 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967638503686667, - "loss": 1.6668, - "step": 9568 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019676351214232565, - "loss": 1.6144, - "step": 9569 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967631739159846, - "loss": 1.6568, - "step": 9570 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967628356896435, - "loss": 1.5865, - "step": 9571 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019676249746330244, - "loss": 1.6696, - "step": 9572 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019676215923696138, - "loss": 1.68, - "step": 9573 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967618210106203, - "loss": 1.7315, - "step": 9574 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019676148278427924, - "loss": 1.7698, - "step": 9575 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019676114455793818, - "loss": 1.7224, - "step": 9576 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019676080633159712, - "loss": 1.6848, - "step": 9577 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019676046810525606, - "loss": 1.7018, - "step": 9578 - }, - { - "epoch": 0.03, - "learning_rate": 0.000196760129878915, - "loss": 1.8076, - "step": 9579 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019675979165257392, - "loss": 1.6573, - "step": 9580 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019675945342623286, - "loss": 1.5572, - "step": 9581 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967591151998918, - "loss": 1.482, - "step": 9582 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967587769735507, - "loss": 1.5387, - "step": 9583 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019675843874720965, - "loss": 1.4775, - "step": 9584 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967581005208686, - "loss": 1.5284, - "step": 9585 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967577622945275, - "loss": 1.7888, - "step": 9586 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019675742406818645, - "loss": 1.6387, - "step": 9587 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967570858418454, - "loss": 1.725, - "step": 9588 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967567476155043, - "loss": 1.6882, - "step": 9589 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019675640938916325, - "loss": 1.6951, - "step": 9590 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967560711628222, - "loss": 1.6233, - "step": 9591 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967557329364811, - "loss": 1.5922, - "step": 9592 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019675539471014004, - "loss": 1.6374, - "step": 9593 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019675505648379898, - "loss": 1.6786, - "step": 9594 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967547182574579, - "loss": 1.5778, - "step": 9595 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019675438003111684, - "loss": 1.5832, - "step": 9596 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019675404180477578, - "loss": 1.6776, - "step": 9597 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967537035784347, - "loss": 1.6057, - "step": 9598 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019675336535209364, - "loss": 1.56, - "step": 9599 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019675302712575258, - "loss": 1.7271, - "step": 9600 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967526888994115, - "loss": 1.6476, - "step": 9601 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019675235067307043, - "loss": 1.7287, - "step": 9602 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019675201244672937, - "loss": 1.6091, - "step": 9603 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967516742203883, - "loss": 1.7494, - "step": 9604 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019675133599404723, - "loss": 1.5962, - "step": 9605 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019675099776770617, - "loss": 1.678, - "step": 9606 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019675065954136508, - "loss": 1.7252, - "step": 9607 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019675032131502402, - "loss": 1.811, - "step": 9608 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019674998308868297, - "loss": 1.5312, - "step": 9609 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019674964486234188, - "loss": 1.8952, - "step": 9610 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019674930663600082, - "loss": 1.6467, - "step": 9611 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019674896840965976, - "loss": 1.6414, - "step": 9612 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019674863018331868, - "loss": 1.5278, - "step": 9613 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019674829195697762, - "loss": 1.7293, - "step": 9614 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019674795373063656, - "loss": 1.731, - "step": 9615 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019674761550429547, - "loss": 1.5865, - "step": 9616 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967472772779544, - "loss": 1.6114, - "step": 9617 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019674693905161335, - "loss": 1.5909, - "step": 9618 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019674660082527227, - "loss": 1.5486, - "step": 9619 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967462625989312, - "loss": 1.6149, - "step": 9620 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019674592437259015, - "loss": 1.6084, - "step": 9621 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019674558614624907, - "loss": 1.6954, - "step": 9622 - }, - { - "epoch": 0.03, - "learning_rate": 0.000196745247919908, - "loss": 1.6432, - "step": 9623 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019674490969356695, - "loss": 1.6781, - "step": 9624 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019674457146722586, - "loss": 1.5065, - "step": 9625 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967442332408848, - "loss": 1.6871, - "step": 9626 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019674389501454374, - "loss": 1.6151, - "step": 9627 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019674355678820266, - "loss": 1.649, - "step": 9628 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967432185618616, - "loss": 1.4709, - "step": 9629 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019674288033552054, - "loss": 1.6274, - "step": 9630 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019674254210917945, - "loss": 1.611, - "step": 9631 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967422038828384, - "loss": 1.5586, - "step": 9632 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019674186565649734, - "loss": 1.7793, - "step": 9633 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019674152743015628, - "loss": 1.6242, - "step": 9634 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019674118920381522, - "loss": 1.5852, - "step": 9635 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019674085097747413, - "loss": 1.6703, - "step": 9636 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019674051275113307, - "loss": 1.7664, - "step": 9637 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019674017452479201, - "loss": 1.6394, - "step": 9638 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019673983629845093, - "loss": 1.5679, - "step": 9639 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019673949807210987, - "loss": 1.6172, - "step": 9640 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967391598457688, - "loss": 1.8477, - "step": 9641 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019673882161942773, - "loss": 1.6435, - "step": 9642 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019673848339308667, - "loss": 1.702, - "step": 9643 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967381451667456, - "loss": 1.6678, - "step": 9644 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019673780694040452, - "loss": 1.4713, - "step": 9645 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019673746871406346, - "loss": 1.6948, - "step": 9646 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967371304877224, - "loss": 1.6411, - "step": 9647 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019673679226138132, - "loss": 1.5585, - "step": 9648 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019673645403504026, - "loss": 1.6162, - "step": 9649 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967361158086992, - "loss": 1.8494, - "step": 9650 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019673577758235811, - "loss": 1.5901, - "step": 9651 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019673543935601706, - "loss": 1.6794, - "step": 9652 - }, - { - "epoch": 0.03, - "learning_rate": 0.000196735101129676, - "loss": 1.6254, - "step": 9653 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967347629033349, - "loss": 1.6534, - "step": 9654 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019673442467699385, - "loss": 1.8116, - "step": 9655 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967340864506528, - "loss": 1.7502, - "step": 9656 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967337482243117, - "loss": 1.62, - "step": 9657 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019673340999797065, - "loss": 1.739, - "step": 9658 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967330717716296, - "loss": 1.6506, - "step": 9659 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019673273354528853, - "loss": 1.5125, - "step": 9660 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019673239531894744, - "loss": 1.6511, - "step": 9661 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019673205709260639, - "loss": 1.5519, - "step": 9662 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019673171886626533, - "loss": 1.6217, - "step": 9663 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019673138063992424, - "loss": 1.519, - "step": 9664 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019673104241358318, - "loss": 1.5615, - "step": 9665 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019673070418724212, - "loss": 1.6625, - "step": 9666 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019673036596090104, - "loss": 1.6311, - "step": 9667 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019673002773455998, - "loss": 1.655, - "step": 9668 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019672968950821892, - "loss": 1.7463, - "step": 9669 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019672935128187783, - "loss": 1.6996, - "step": 9670 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019672901305553677, - "loss": 1.6954, - "step": 9671 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019672867482919572, - "loss": 1.6599, - "step": 9672 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019672833660285463, - "loss": 1.7008, - "step": 9673 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019672799837651357, - "loss": 1.7902, - "step": 9674 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967276601501725, - "loss": 1.5442, - "step": 9675 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019672732192383143, - "loss": 1.671, - "step": 9676 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019672698369749037, - "loss": 1.5981, - "step": 9677 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967266454711493, - "loss": 1.6291, - "step": 9678 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019672630724480822, - "loss": 1.6065, - "step": 9679 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019672596901846716, - "loss": 1.7762, - "step": 9680 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967256307921261, - "loss": 1.6459, - "step": 9681 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019672529256578502, - "loss": 1.7083, - "step": 9682 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019672495433944396, - "loss": 1.6184, - "step": 9683 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967246161131029, - "loss": 1.7478, - "step": 9684 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019672427788676181, - "loss": 1.7058, - "step": 9685 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019672393966042076, - "loss": 1.797, - "step": 9686 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967236014340797, - "loss": 1.6905, - "step": 9687 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967232632077386, - "loss": 1.641, - "step": 9688 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019672292498139758, - "loss": 1.5812, - "step": 9689 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967225867550565, - "loss": 1.6074, - "step": 9690 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019672224852871543, - "loss": 1.6915, - "step": 9691 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019672191030237438, - "loss": 1.6623, - "step": 9692 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967215720760333, - "loss": 1.7157, - "step": 9693 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019672123384969223, - "loss": 1.658, - "step": 9694 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019672089562335117, - "loss": 1.6412, - "step": 9695 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019672055739701009, - "loss": 1.6621, - "step": 9696 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019672021917066903, - "loss": 1.5706, - "step": 9697 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019671988094432797, - "loss": 1.6706, - "step": 9698 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019671954271798688, - "loss": 1.7067, - "step": 9699 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019671920449164582, - "loss": 1.687, - "step": 9700 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019671886626530476, - "loss": 1.6097, - "step": 9701 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019671852803896368, - "loss": 1.7099, - "step": 9702 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019671818981262262, - "loss": 1.4945, - "step": 9703 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019671785158628156, - "loss": 1.7038, - "step": 9704 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019671751335994047, - "loss": 1.6439, - "step": 9705 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019671717513359942, - "loss": 1.7069, - "step": 9706 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019671683690725836, - "loss": 1.6176, - "step": 9707 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019671649868091727, - "loss": 1.4904, - "step": 9708 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967161604545762, - "loss": 1.6546, - "step": 9709 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019671582222823515, - "loss": 1.5646, - "step": 9710 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019671548400189407, - "loss": 1.7121, - "step": 9711 - }, - { - "epoch": 0.03, - "learning_rate": 0.000196715145775553, - "loss": 1.5857, - "step": 9712 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019671480754921195, - "loss": 1.5051, - "step": 9713 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019671446932287086, - "loss": 1.7175, - "step": 9714 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967141310965298, - "loss": 1.6648, - "step": 9715 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019671379287018875, - "loss": 1.7387, - "step": 9716 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019671345464384766, - "loss": 1.642, - "step": 9717 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967131164175066, - "loss": 1.7562, - "step": 9718 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019671277819116554, - "loss": 1.5785, - "step": 9719 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019671243996482446, - "loss": 1.677, - "step": 9720 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967121017384834, - "loss": 1.5906, - "step": 9721 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019671176351214234, - "loss": 1.7532, - "step": 9722 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019671142528580125, - "loss": 1.6665, - "step": 9723 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967110870594602, - "loss": 1.6968, - "step": 9724 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019671074883311914, - "loss": 1.7503, - "step": 9725 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019671041060677805, - "loss": 1.6286, - "step": 9726 - }, - { - "epoch": 0.03, - "learning_rate": 0.000196710072380437, - "loss": 1.6748, - "step": 9727 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019670973415409593, - "loss": 1.6666, - "step": 9728 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019670939592775485, - "loss": 1.5718, - "step": 9729 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967090577014138, - "loss": 1.6097, - "step": 9730 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019670871947507273, - "loss": 1.6117, - "step": 9731 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019670838124873164, - "loss": 1.5906, - "step": 9732 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019670804302239058, - "loss": 1.6305, - "step": 9733 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019670770479604952, - "loss": 1.5522, - "step": 9734 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019670736656970844, - "loss": 1.6337, - "step": 9735 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019670702834336738, - "loss": 1.7644, - "step": 9736 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019670669011702632, - "loss": 1.7461, - "step": 9737 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019670635189068523, - "loss": 1.6545, - "step": 9738 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019670601366434418, - "loss": 1.7102, - "step": 9739 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019670567543800312, - "loss": 1.7284, - "step": 9740 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019670533721166203, - "loss": 1.6573, - "step": 9741 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019670499898532097, - "loss": 1.5023, - "step": 9742 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967046607589799, - "loss": 1.5602, - "step": 9743 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019670432253263885, - "loss": 1.6644, - "step": 9744 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967039843062978, - "loss": 1.7484, - "step": 9745 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019670364607995674, - "loss": 1.7134, - "step": 9746 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019670330785361565, - "loss": 1.6457, - "step": 9747 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967029696272746, - "loss": 1.5748, - "step": 9748 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019670263140093353, - "loss": 1.9016, - "step": 9749 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019670229317459245, - "loss": 1.5017, - "step": 9750 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001967019549482514, - "loss": 1.6928, - "step": 9751 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019670161672191033, - "loss": 1.769, - "step": 9752 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019670127849556924, - "loss": 1.6969, - "step": 9753 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019670094026922818, - "loss": 1.7814, - "step": 9754 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019670060204288713, - "loss": 1.555, - "step": 9755 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019670026381654604, - "loss": 1.5499, - "step": 9756 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019669992559020498, - "loss": 1.5773, - "step": 9757 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019669958736386392, - "loss": 1.5618, - "step": 9758 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019669924913752284, - "loss": 1.661, - "step": 9759 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019669891091118178, - "loss": 1.6329, - "step": 9760 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019669857268484072, - "loss": 1.4686, - "step": 9761 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019669823445849963, - "loss": 1.6155, - "step": 9762 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019669789623215857, - "loss": 1.5378, - "step": 9763 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019669755800581751, - "loss": 1.7108, - "step": 9764 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019669721977947643, - "loss": 1.6969, - "step": 9765 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019669688155313537, - "loss": 1.5811, - "step": 9766 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966965433267943, - "loss": 1.6621, - "step": 9767 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019669620510045322, - "loss": 1.6333, - "step": 9768 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019669586687411217, - "loss": 1.7363, - "step": 9769 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966955286477711, - "loss": 1.6966, - "step": 9770 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019669519042143002, - "loss": 1.5426, - "step": 9771 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019669485219508896, - "loss": 1.7612, - "step": 9772 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966945139687479, - "loss": 1.6706, - "step": 9773 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019669417574240682, - "loss": 1.6508, - "step": 9774 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019669383751606576, - "loss": 1.8078, - "step": 9775 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966934992897247, - "loss": 1.7716, - "step": 9776 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966931610633836, - "loss": 1.7464, - "step": 9777 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019669282283704255, - "loss": 1.7023, - "step": 9778 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966924846107015, - "loss": 1.7, - "step": 9779 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966921463843604, - "loss": 1.8114, - "step": 9780 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019669180815801935, - "loss": 1.6129, - "step": 9781 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966914699316783, - "loss": 1.6488, - "step": 9782 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966911317053372, - "loss": 1.4567, - "step": 9783 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019669079347899615, - "loss": 1.7432, - "step": 9784 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966904552526551, - "loss": 1.6094, - "step": 9785 - }, - { - "epoch": 0.03, - "learning_rate": 0.000196690117026314, - "loss": 1.7249, - "step": 9786 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019668977879997294, - "loss": 1.543, - "step": 9787 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019668944057363188, - "loss": 1.5782, - "step": 9788 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966891023472908, - "loss": 1.642, - "step": 9789 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019668876412094974, - "loss": 1.561, - "step": 9790 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019668842589460868, - "loss": 1.5103, - "step": 9791 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966880876682676, - "loss": 1.7607, - "step": 9792 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019668774944192654, - "loss": 1.4803, - "step": 9793 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019668741121558548, - "loss": 1.5415, - "step": 9794 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966870729892444, - "loss": 1.6702, - "step": 9795 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019668673476290333, - "loss": 1.6792, - "step": 9796 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019668639653656227, - "loss": 1.8218, - "step": 9797 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966860583102212, - "loss": 1.6201, - "step": 9798 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019668572008388013, - "loss": 1.5155, - "step": 9799 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019668538185753907, - "loss": 1.6237, - "step": 9800 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019668504363119798, - "loss": 1.565, - "step": 9801 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019668470540485695, - "loss": 1.6212, - "step": 9802 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966843671785159, - "loss": 1.6554, - "step": 9803 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966840289521748, - "loss": 1.7097, - "step": 9804 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019668369072583375, - "loss": 1.6189, - "step": 9805 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966833524994927, - "loss": 1.7685, - "step": 9806 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966830142731516, - "loss": 1.6459, - "step": 9807 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019668267604681054, - "loss": 1.6613, - "step": 9808 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019668233782046949, - "loss": 1.5478, - "step": 9809 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966819995941284, - "loss": 1.575, - "step": 9810 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019668166136778734, - "loss": 1.6232, - "step": 9811 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019668132314144628, - "loss": 1.5821, - "step": 9812 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966809849151052, - "loss": 1.4703, - "step": 9813 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019668064668876414, - "loss": 1.6036, - "step": 9814 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019668030846242308, - "loss": 1.8316, - "step": 9815 - }, - { - "epoch": 0.03, - "learning_rate": 0.000196679970236082, - "loss": 1.6564, - "step": 9816 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019667963200974093, - "loss": 1.7223, - "step": 9817 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019667929378339987, - "loss": 1.5848, - "step": 9818 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966789555570588, - "loss": 1.6204, - "step": 9819 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019667861733071773, - "loss": 1.7524, - "step": 9820 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019667827910437667, - "loss": 1.7114, - "step": 9821 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019667794087803559, - "loss": 1.7577, - "step": 9822 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019667760265169453, - "loss": 1.738, - "step": 9823 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019667726442535347, - "loss": 1.6003, - "step": 9824 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019667692619901238, - "loss": 1.6787, - "step": 9825 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019667658797267132, - "loss": 1.7291, - "step": 9826 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019667624974633026, - "loss": 1.7386, - "step": 9827 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019667591151998918, - "loss": 1.5887, - "step": 9828 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019667557329364812, - "loss": 1.6953, - "step": 9829 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019667523506730706, - "loss": 1.5487, - "step": 9830 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019667489684096597, - "loss": 1.7369, - "step": 9831 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019667455861462492, - "loss": 1.6315, - "step": 9832 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019667422038828386, - "loss": 1.5858, - "step": 9833 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019667388216194277, - "loss": 1.5927, - "step": 9834 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966735439356017, - "loss": 1.7045, - "step": 9835 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019667320570926065, - "loss": 1.7058, - "step": 9836 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019667286748291957, - "loss": 1.6916, - "step": 9837 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966725292565785, - "loss": 1.7377, - "step": 9838 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019667219103023745, - "loss": 1.6242, - "step": 9839 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019667185280389636, - "loss": 1.7043, - "step": 9840 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966715145775553, - "loss": 1.8126, - "step": 9841 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019667117635121425, - "loss": 1.4927, - "step": 9842 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019667083812487316, - "loss": 1.6908, - "step": 9843 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966704998985321, - "loss": 1.6855, - "step": 9844 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019667016167219104, - "loss": 1.6699, - "step": 9845 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019666982344584996, - "loss": 1.631, - "step": 9846 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966694852195089, - "loss": 1.6457, - "step": 9847 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019666914699316784, - "loss": 1.4867, - "step": 9848 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019666880876682675, - "loss": 1.4728, - "step": 9849 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966684705404857, - "loss": 1.6854, - "step": 9850 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019666813231414463, - "loss": 1.6301, - "step": 9851 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019666779408780355, - "loss": 1.6245, - "step": 9852 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966674558614625, - "loss": 1.5911, - "step": 9853 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019666711763512143, - "loss": 1.5461, - "step": 9854 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019666677940878034, - "loss": 1.7009, - "step": 9855 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019666644118243929, - "loss": 1.6918, - "step": 9856 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019666610295609823, - "loss": 1.6682, - "step": 9857 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019666576472975717, - "loss": 1.7645, - "step": 9858 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966654265034161, - "loss": 1.6015, - "step": 9859 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019666508827707505, - "loss": 1.657, - "step": 9860 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019666475005073396, - "loss": 1.7262, - "step": 9861 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966644118243929, - "loss": 1.6442, - "step": 9862 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019666407359805185, - "loss": 1.7236, - "step": 9863 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019666373537171076, - "loss": 1.6493, - "step": 9864 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966633971453697, - "loss": 1.6438, - "step": 9865 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019666305891902864, - "loss": 1.6694, - "step": 9866 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019666272069268756, - "loss": 1.5975, - "step": 9867 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966623824663465, - "loss": 1.6037, - "step": 9868 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019666204424000544, - "loss": 1.7833, - "step": 9869 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019666170601366435, - "loss": 1.7487, - "step": 9870 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966613677873233, - "loss": 1.4814, - "step": 9871 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019666102956098224, - "loss": 1.6475, - "step": 9872 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019666069133464115, - "loss": 1.5875, - "step": 9873 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966603531083001, - "loss": 1.6889, - "step": 9874 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019666001488195903, - "loss": 1.5971, - "step": 9875 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019665967665561795, - "loss": 1.7025, - "step": 9876 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966593384292769, - "loss": 1.5632, - "step": 9877 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019665900020293583, - "loss": 1.6022, - "step": 9878 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019665866197659474, - "loss": 1.6191, - "step": 9879 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019665832375025368, - "loss": 1.6775, - "step": 9880 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019665798552391262, - "loss": 1.7509, - "step": 9881 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019665764729757154, - "loss": 1.6241, - "step": 9882 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019665730907123048, - "loss": 1.6508, - "step": 9883 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019665697084488942, - "loss": 1.5418, - "step": 9884 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019665663261854834, - "loss": 1.6003, - "step": 9885 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019665629439220728, - "loss": 1.5654, - "step": 9886 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019665595616586622, - "loss": 1.739, - "step": 9887 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019665561793952513, - "loss": 1.7553, - "step": 9888 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019665527971318407, - "loss": 1.6556, - "step": 9889 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019665494148684301, - "loss": 1.7153, - "step": 9890 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019665460326050193, - "loss": 1.6699, - "step": 9891 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019665426503416087, - "loss": 1.6085, - "step": 9892 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966539268078198, - "loss": 1.6263, - "step": 9893 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019665358858147872, - "loss": 1.5353, - "step": 9894 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019665325035513767, - "loss": 1.8069, - "step": 9895 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966529121287966, - "loss": 1.72, - "step": 9896 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019665257390245552, - "loss": 1.6093, - "step": 9897 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019665223567611446, - "loss": 1.7525, - "step": 9898 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966518974497734, - "loss": 1.5088, - "step": 9899 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019665155922343232, - "loss": 1.6062, - "step": 9900 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019665122099709126, - "loss": 1.7081, - "step": 9901 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966508827707502, - "loss": 1.707, - "step": 9902 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966505445444091, - "loss": 1.6311, - "step": 9903 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019665020631806805, - "loss": 1.6783, - "step": 9904 - }, - { - "epoch": 0.03, - "learning_rate": 0.000196649868091727, - "loss": 1.6276, - "step": 9905 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966495298653859, - "loss": 1.5658, - "step": 9906 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019664919163904485, - "loss": 1.7947, - "step": 9907 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966488534127038, - "loss": 1.6127, - "step": 9908 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966485151863627, - "loss": 1.6606, - "step": 9909 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019664817696002165, - "loss": 1.7177, - "step": 9910 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966478387336806, - "loss": 1.6171, - "step": 9911 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966475005073395, - "loss": 1.6349, - "step": 9912 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019664716228099844, - "loss": 1.6775, - "step": 9913 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019664682405465738, - "loss": 1.6758, - "step": 9914 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019664648582831633, - "loss": 1.7698, - "step": 9915 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019664614760197527, - "loss": 1.6536, - "step": 9916 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019664580937563418, - "loss": 1.5101, - "step": 9917 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019664547114929312, - "loss": 1.6315, - "step": 9918 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019664513292295206, - "loss": 1.6535, - "step": 9919 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019664479469661098, - "loss": 1.6629, - "step": 9920 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019664445647026992, - "loss": 1.7529, - "step": 9921 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019664411824392886, - "loss": 1.6768, - "step": 9922 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019664378001758777, - "loss": 1.7165, - "step": 9923 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019664344179124671, - "loss": 1.491, - "step": 9924 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019664310356490566, - "loss": 1.6584, - "step": 9925 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966427653385646, - "loss": 1.6592, - "step": 9926 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966424271122235, - "loss": 1.73, - "step": 9927 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019664208888588245, - "loss": 1.5201, - "step": 9928 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966417506595414, - "loss": 1.6744, - "step": 9929 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966414124332003, - "loss": 1.5452, - "step": 9930 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019664107420685925, - "loss": 1.7398, - "step": 9931 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966407359805182, - "loss": 1.8457, - "step": 9932 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966403977541771, - "loss": 1.6683, - "step": 9933 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019664005952783604, - "loss": 1.661, - "step": 9934 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019663972130149499, - "loss": 1.7245, - "step": 9935 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966393830751539, - "loss": 1.6583, - "step": 9936 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019663904484881284, - "loss": 1.5199, - "step": 9937 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019663870662247178, - "loss": 1.5642, - "step": 9938 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966383683961307, - "loss": 1.4947, - "step": 9939 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019663803016978964, - "loss": 1.4916, - "step": 9940 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019663769194344858, - "loss": 1.6039, - "step": 9941 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966373537171075, - "loss": 1.8335, - "step": 9942 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019663701549076643, - "loss": 1.55, - "step": 9943 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019663667726442537, - "loss": 1.7148, - "step": 9944 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966363390380843, - "loss": 1.6289, - "step": 9945 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019663600081174323, - "loss": 1.4892, - "step": 9946 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019663566258540217, - "loss": 1.5071, - "step": 9947 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019663532435906108, - "loss": 1.5877, - "step": 9948 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019663498613272003, - "loss": 1.6847, - "step": 9949 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019663464790637897, - "loss": 1.7351, - "step": 9950 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019663430968003788, - "loss": 1.772, - "step": 9951 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019663397145369682, - "loss": 1.6884, - "step": 9952 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019663363322735576, - "loss": 1.6368, - "step": 9953 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019663329500101468, - "loss": 1.4899, - "step": 9954 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019663295677467362, - "loss": 1.6935, - "step": 9955 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019663261854833256, - "loss": 1.6616, - "step": 9956 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019663228032199147, - "loss": 1.6257, - "step": 9957 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019663194209565041, - "loss": 1.7307, - "step": 9958 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019663160386930936, - "loss": 1.4819, - "step": 9959 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019663126564296827, - "loss": 1.697, - "step": 9960 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966309274166272, - "loss": 1.5469, - "step": 9961 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019663058919028615, - "loss": 1.5906, - "step": 9962 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019663025096394507, - "loss": 1.6306, - "step": 9963 - }, - { - "epoch": 0.03, - "learning_rate": 0.000196629912737604, - "loss": 1.7651, - "step": 9964 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019662957451126295, - "loss": 1.6987, - "step": 9965 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019662923628492186, - "loss": 1.694, - "step": 9966 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966288980585808, - "loss": 1.6309, - "step": 9967 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019662855983223974, - "loss": 1.5282, - "step": 9968 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019662822160589866, - "loss": 1.6289, - "step": 9969 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019662788337955763, - "loss": 1.7331, - "step": 9970 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019662754515321654, - "loss": 1.6063, - "step": 9971 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019662720692687548, - "loss": 1.6496, - "step": 9972 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019662686870053442, - "loss": 1.6667, - "step": 9973 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019662653047419334, - "loss": 1.6552, - "step": 9974 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019662619224785228, - "loss": 1.6505, - "step": 9975 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019662585402151122, - "loss": 1.7459, - "step": 9976 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019662551579517013, - "loss": 1.615, - "step": 9977 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019662517756882907, - "loss": 1.5701, - "step": 9978 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019662483934248802, - "loss": 1.7307, - "step": 9979 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019662450111614693, - "loss": 1.5655, - "step": 9980 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019662416288980587, - "loss": 1.582, - "step": 9981 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966238246634648, - "loss": 1.7668, - "step": 9982 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019662348643712373, - "loss": 1.7056, - "step": 9983 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019662314821078267, - "loss": 1.7998, - "step": 9984 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966228099844416, - "loss": 1.7016, - "step": 9985 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019662247175810052, - "loss": 1.6708, - "step": 9986 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019662213353175946, - "loss": 1.5925, - "step": 9987 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966217953054184, - "loss": 1.7546, - "step": 9988 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019662145707907732, - "loss": 1.4725, - "step": 9989 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019662111885273626, - "loss": 1.5336, - "step": 9990 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966207806263952, - "loss": 1.7171, - "step": 9991 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019662044240005412, - "loss": 1.6519, - "step": 9992 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019662010417371306, - "loss": 1.6406, - "step": 9993 - }, - { - "epoch": 0.03, - "learning_rate": 0.000196619765947372, - "loss": 1.6769, - "step": 9994 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966194277210309, - "loss": 1.5737, - "step": 9995 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019661908949468985, - "loss": 1.509, - "step": 9996 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966187512683488, - "loss": 1.6613, - "step": 9997 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966184130420077, - "loss": 1.8473, - "step": 9998 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019661807481566665, - "loss": 1.5893, - "step": 9999 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966177365893256, - "loss": 1.7005, - "step": 10000 - }, - { - "epoch": 0.03, - "eval_loss": 1.4290114182692308, - "mrc_em": 23.0, - "mrc_f1": 29.83333333333334, - "nli_acc": 0.2, - "step": 10000, - "sts_f1": 0.22580645161290325, - "ynat_f1": 0.5322751976284585 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966173983629845, - "loss": 1.696, - "step": 10001 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019661706013664345, - "loss": 1.5447, - "step": 10002 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966167219103024, - "loss": 1.6775, - "step": 10003 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966163836839613, - "loss": 1.662, - "step": 10004 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019661604545762024, - "loss": 1.6472, - "step": 10005 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019661570723127918, - "loss": 1.7098, - "step": 10006 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966153690049381, - "loss": 1.8189, - "step": 10007 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019661503077859704, - "loss": 1.5668, - "step": 10008 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019661469255225598, - "loss": 1.7054, - "step": 10009 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966143543259149, - "loss": 1.7001, - "step": 10010 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019661401609957383, - "loss": 1.6192, - "step": 10011 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019661367787323278, - "loss": 1.6377, - "step": 10012 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019661333964689172, - "loss": 1.5768, - "step": 10013 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019661300142055063, - "loss": 1.5951, - "step": 10014 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019661266319420957, - "loss": 1.6733, - "step": 10015 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966123249678685, - "loss": 1.6324, - "step": 10016 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019661198674152743, - "loss": 1.5278, - "step": 10017 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019661164851518637, - "loss": 1.715, - "step": 10018 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966113102888453, - "loss": 1.6695, - "step": 10019 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019661097206250422, - "loss": 1.6362, - "step": 10020 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019661063383616316, - "loss": 1.533, - "step": 10021 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966102956098221, - "loss": 1.6683, - "step": 10022 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019660995738348102, - "loss": 1.6795, - "step": 10023 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019660961915713996, - "loss": 1.7724, - "step": 10024 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966092809307989, - "loss": 1.5413, - "step": 10025 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019660894270445784, - "loss": 1.5328, - "step": 10026 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019660860447811678, - "loss": 1.6569, - "step": 10027 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966082662517757, - "loss": 1.4491, - "step": 10028 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019660792802543464, - "loss": 1.7585, - "step": 10029 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019660758979909358, - "loss": 1.6583, - "step": 10030 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966072515727525, - "loss": 1.6705, - "step": 10031 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019660691334641144, - "loss": 1.5413, - "step": 10032 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019660657512007038, - "loss": 1.7212, - "step": 10033 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966062368937293, - "loss": 1.5632, - "step": 10034 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019660589866738823, - "loss": 1.5963, - "step": 10035 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019660556044104717, - "loss": 1.6879, - "step": 10036 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001966052222147061, - "loss": 1.8191, - "step": 10037 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019660488398836503, - "loss": 1.6114, - "step": 10038 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019660454576202397, - "loss": 1.6714, - "step": 10039 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019660420753568288, - "loss": 1.5423, - "step": 10040 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019660386930934182, - "loss": 1.5431, - "step": 10041 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019660353108300077, - "loss": 1.6142, - "step": 10042 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019660319285665968, - "loss": 1.5003, - "step": 10043 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019660285463031862, - "loss": 1.6775, - "step": 10044 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019660251640397756, - "loss": 1.4955, - "step": 10045 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019660217817763648, - "loss": 1.6283, - "step": 10046 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019660183995129542, - "loss": 1.6062, - "step": 10047 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019660150172495436, - "loss": 1.6385, - "step": 10048 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019660116349861327, - "loss": 1.4996, - "step": 10049 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019660082527227221, - "loss": 1.5607, - "step": 10050 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019660048704593115, - "loss": 1.7638, - "step": 10051 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019660014881959007, - "loss": 1.6077, - "step": 10052 - }, - { - "epoch": 0.03, - "learning_rate": 0.000196599810593249, - "loss": 1.6465, - "step": 10053 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019659947236690795, - "loss": 1.6362, - "step": 10054 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019659913414056687, - "loss": 1.665, - "step": 10055 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965987959142258, - "loss": 1.7874, - "step": 10056 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019659845768788475, - "loss": 1.6083, - "step": 10057 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019659811946154366, - "loss": 1.7543, - "step": 10058 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965977812352026, - "loss": 1.6995, - "step": 10059 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019659744300886154, - "loss": 1.7352, - "step": 10060 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019659710478252046, - "loss": 1.6247, - "step": 10061 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965967665561794, - "loss": 1.5304, - "step": 10062 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019659642832983834, - "loss": 1.5881, - "step": 10063 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019659609010349725, - "loss": 1.6048, - "step": 10064 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965957518771562, - "loss": 1.7734, - "step": 10065 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019659541365081514, - "loss": 1.4925, - "step": 10066 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019659507542447405, - "loss": 1.6948, - "step": 10067 - }, - { - "epoch": 0.03, - "learning_rate": 0.000196594737198133, - "loss": 1.842, - "step": 10068 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019659439897179193, - "loss": 1.632, - "step": 10069 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019659406074545085, - "loss": 1.7744, - "step": 10070 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965937225191098, - "loss": 1.6531, - "step": 10071 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019659338429276873, - "loss": 1.4978, - "step": 10072 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019659304606642764, - "loss": 1.6217, - "step": 10073 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019659270784008658, - "loss": 1.5379, - "step": 10074 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019659236961374553, - "loss": 1.5253, - "step": 10075 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019659203138740444, - "loss": 1.4736, - "step": 10076 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019659169316106338, - "loss": 1.8085, - "step": 10077 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019659135493472232, - "loss": 1.5911, - "step": 10078 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019659101670838124, - "loss": 1.8211, - "step": 10079 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019659067848204018, - "loss": 1.7114, - "step": 10080 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019659034025569912, - "loss": 1.6121, - "step": 10081 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019659000202935806, - "loss": 1.6303, - "step": 10082 - }, - { - "epoch": 0.03, - "learning_rate": 0.000196589663803017, - "loss": 1.5672, - "step": 10083 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019658932557667594, - "loss": 1.6635, - "step": 10084 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019658898735033486, - "loss": 1.5701, - "step": 10085 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965886491239938, - "loss": 1.6253, - "step": 10086 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019658831089765274, - "loss": 1.6308, - "step": 10087 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019658797267131165, - "loss": 1.781, - "step": 10088 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965876344449706, - "loss": 1.7163, - "step": 10089 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019658729621862953, - "loss": 1.5722, - "step": 10090 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019658695799228845, - "loss": 1.689, - "step": 10091 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965866197659474, - "loss": 1.7011, - "step": 10092 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019658628153960633, - "loss": 1.5522, - "step": 10093 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019658594331326524, - "loss": 1.6358, - "step": 10094 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019658560508692419, - "loss": 1.5237, - "step": 10095 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019658526686058313, - "loss": 1.6443, - "step": 10096 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019658492863424204, - "loss": 1.6497, - "step": 10097 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019658459040790098, - "loss": 1.6083, - "step": 10098 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019658425218155992, - "loss": 1.6047, - "step": 10099 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019658391395521884, - "loss": 1.4514, - "step": 10100 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019658357572887778, - "loss": 1.5313, - "step": 10101 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019658323750253672, - "loss": 1.6409, - "step": 10102 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019658289927619563, - "loss": 1.7522, - "step": 10103 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019658256104985457, - "loss": 1.5135, - "step": 10104 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019658222282351352, - "loss": 1.6045, - "step": 10105 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019658188459717243, - "loss": 1.7576, - "step": 10106 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019658154637083137, - "loss": 1.6111, - "step": 10107 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965812081444903, - "loss": 1.7489, - "step": 10108 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019658086991814923, - "loss": 1.6324, - "step": 10109 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019658053169180817, - "loss": 1.6203, - "step": 10110 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965801934654671, - "loss": 1.6306, - "step": 10111 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019657985523912602, - "loss": 1.7088, - "step": 10112 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019657951701278496, - "loss": 1.6814, - "step": 10113 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965791787864439, - "loss": 1.6816, - "step": 10114 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019657884056010282, - "loss": 1.6144, - "step": 10115 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019657850233376176, - "loss": 1.5611, - "step": 10116 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965781641074207, - "loss": 1.5886, - "step": 10117 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019657782588107961, - "loss": 1.6353, - "step": 10118 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019657748765473856, - "loss": 1.6552, - "step": 10119 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965771494283975, - "loss": 1.63, - "step": 10120 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965768112020564, - "loss": 1.7127, - "step": 10121 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019657647297571535, - "loss": 1.6864, - "step": 10122 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965761347493743, - "loss": 1.7957, - "step": 10123 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965757965230332, - "loss": 1.5805, - "step": 10124 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019657545829669215, - "loss": 1.5599, - "step": 10125 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965751200703511, - "loss": 1.6003, - "step": 10126 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019657478184401, - "loss": 1.7042, - "step": 10127 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019657444361766894, - "loss": 1.6335, - "step": 10128 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019657410539132789, - "loss": 1.765, - "step": 10129 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965737671649868, - "loss": 1.5766, - "step": 10130 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019657342893864574, - "loss": 1.5583, - "step": 10131 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019657309071230468, - "loss": 1.7519, - "step": 10132 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965727524859636, - "loss": 1.5847, - "step": 10133 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019657241425962254, - "loss": 1.6031, - "step": 10134 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019657207603328148, - "loss": 1.4371, - "step": 10135 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965717378069404, - "loss": 1.6232, - "step": 10136 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019657139958059933, - "loss": 1.6595, - "step": 10137 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019657106135425827, - "loss": 1.7536, - "step": 10138 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019657072312791722, - "loss": 1.6033, - "step": 10139 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019657038490157616, - "loss": 1.4939, - "step": 10140 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965700466752351, - "loss": 1.6337, - "step": 10141 - }, - { - "epoch": 0.03, - "learning_rate": 0.000196569708448894, - "loss": 1.6592, - "step": 10142 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019656937022255295, - "loss": 1.7481, - "step": 10143 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965690319962119, - "loss": 1.6345, - "step": 10144 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965686937698708, - "loss": 1.6367, - "step": 10145 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019656835554352975, - "loss": 1.6496, - "step": 10146 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965680173171887, - "loss": 1.5754, - "step": 10147 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965676790908476, - "loss": 1.6501, - "step": 10148 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019656734086450655, - "loss": 1.6831, - "step": 10149 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965670026381655, - "loss": 1.6813, - "step": 10150 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965666644118244, - "loss": 1.6397, - "step": 10151 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019656632618548334, - "loss": 1.6911, - "step": 10152 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019656598795914228, - "loss": 1.6637, - "step": 10153 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965656497328012, - "loss": 1.6175, - "step": 10154 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019656531150646014, - "loss": 1.7078, - "step": 10155 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019656497328011908, - "loss": 1.6292, - "step": 10156 - }, - { - "epoch": 0.03, - "learning_rate": 0.000196564635053778, - "loss": 1.8042, - "step": 10157 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019656429682743694, - "loss": 1.7202, - "step": 10158 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019656395860109588, - "loss": 1.595, - "step": 10159 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965636203747548, - "loss": 1.7586, - "step": 10160 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019656328214841373, - "loss": 1.5125, - "step": 10161 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019656294392207267, - "loss": 1.667, - "step": 10162 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965626056957316, - "loss": 1.7046, - "step": 10163 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019656226746939053, - "loss": 1.6083, - "step": 10164 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019656192924304947, - "loss": 1.6233, - "step": 10165 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019656159101670838, - "loss": 1.63, - "step": 10166 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019656125279036732, - "loss": 1.5893, - "step": 10167 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019656091456402627, - "loss": 1.7349, - "step": 10168 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019656057633768518, - "loss": 1.4182, - "step": 10169 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019656023811134412, - "loss": 1.821, - "step": 10170 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019655989988500306, - "loss": 1.649, - "step": 10171 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019655956165866198, - "loss": 1.6277, - "step": 10172 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019655922343232092, - "loss": 1.6712, - "step": 10173 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019655888520597986, - "loss": 1.5779, - "step": 10174 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019655854697963877, - "loss": 1.6891, - "step": 10175 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965582087532977, - "loss": 1.5207, - "step": 10176 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019655787052695665, - "loss": 1.6587, - "step": 10177 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019655753230061557, - "loss": 1.5831, - "step": 10178 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965571940742745, - "loss": 1.6523, - "step": 10179 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019655685584793345, - "loss": 1.5473, - "step": 10180 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019655651762159236, - "loss": 1.8004, - "step": 10181 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965561793952513, - "loss": 1.7993, - "step": 10182 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019655584116891025, - "loss": 1.7661, - "step": 10183 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019655550294256916, - "loss": 1.6311, - "step": 10184 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965551647162281, - "loss": 1.5721, - "step": 10185 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019655482648988704, - "loss": 1.6477, - "step": 10186 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019655448826354596, - "loss": 1.5861, - "step": 10187 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965541500372049, - "loss": 1.6362, - "step": 10188 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019655381181086384, - "loss": 1.6471, - "step": 10189 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019655347358452275, - "loss": 1.5357, - "step": 10190 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965531353581817, - "loss": 1.5126, - "step": 10191 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019655279713184064, - "loss": 1.7525, - "step": 10192 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019655245890549955, - "loss": 1.7481, - "step": 10193 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965521206791585, - "loss": 1.7109, - "step": 10194 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019655178245281746, - "loss": 1.7636, - "step": 10195 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019655144422647637, - "loss": 1.5821, - "step": 10196 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019655110600013531, - "loss": 1.5884, - "step": 10197 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019655076777379426, - "loss": 1.5596, - "step": 10198 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019655042954745317, - "loss": 1.5957, - "step": 10199 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965500913211121, - "loss": 1.5794, - "step": 10200 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019654975309477105, - "loss": 1.5168, - "step": 10201 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019654941486842997, - "loss": 1.6415, - "step": 10202 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965490766420889, - "loss": 1.6832, - "step": 10203 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019654873841574785, - "loss": 1.4985, - "step": 10204 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019654840018940676, - "loss": 1.6739, - "step": 10205 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965480619630657, - "loss": 1.505, - "step": 10206 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019654772373672464, - "loss": 1.6678, - "step": 10207 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019654738551038356, - "loss": 1.7547, - "step": 10208 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965470472840425, - "loss": 1.5118, - "step": 10209 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019654670905770144, - "loss": 1.7264, - "step": 10210 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019654637083136035, - "loss": 1.7963, - "step": 10211 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965460326050193, - "loss": 1.6354, - "step": 10212 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019654569437867824, - "loss": 1.7378, - "step": 10213 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019654535615233715, - "loss": 1.5907, - "step": 10214 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965450179259961, - "loss": 1.6108, - "step": 10215 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019654467969965503, - "loss": 1.5693, - "step": 10216 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019654434147331395, - "loss": 1.6636, - "step": 10217 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965440032469729, - "loss": 1.4783, - "step": 10218 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019654366502063183, - "loss": 1.6279, - "step": 10219 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019654332679429074, - "loss": 1.6854, - "step": 10220 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019654298856794968, - "loss": 1.6781, - "step": 10221 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019654265034160863, - "loss": 1.5441, - "step": 10222 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019654231211526754, - "loss": 1.6436, - "step": 10223 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019654197388892648, - "loss": 1.673, - "step": 10224 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019654163566258542, - "loss": 1.5932, - "step": 10225 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019654129743624434, - "loss": 1.7884, - "step": 10226 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019654095920990328, - "loss": 1.648, - "step": 10227 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019654062098356222, - "loss": 1.5358, - "step": 10228 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019654028275722113, - "loss": 1.7133, - "step": 10229 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019653994453088007, - "loss": 1.495, - "step": 10230 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019653960630453901, - "loss": 1.699, - "step": 10231 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019653926807819793, - "loss": 1.5193, - "step": 10232 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019653892985185687, - "loss": 1.6671, - "step": 10233 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965385916255158, - "loss": 1.5562, - "step": 10234 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019653825339917473, - "loss": 1.5667, - "step": 10235 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019653791517283367, - "loss": 1.5388, - "step": 10236 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965375769464926, - "loss": 1.6564, - "step": 10237 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019653723872015152, - "loss": 1.5777, - "step": 10238 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019653690049381046, - "loss": 1.6373, - "step": 10239 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965365622674694, - "loss": 1.6278, - "step": 10240 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019653622404112832, - "loss": 1.7045, - "step": 10241 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019653588581478726, - "loss": 1.6566, - "step": 10242 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965355475884462, - "loss": 1.5883, - "step": 10243 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019653520936210511, - "loss": 1.7082, - "step": 10244 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019653487113576406, - "loss": 1.7353, - "step": 10245 - }, - { - "epoch": 0.03, - "learning_rate": 0.000196534532909423, - "loss": 1.6229, - "step": 10246 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965341946830819, - "loss": 1.8133, - "step": 10247 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019653385645674085, - "loss": 1.5714, - "step": 10248 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965335182303998, - "loss": 1.6563, - "step": 10249 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965331800040587, - "loss": 1.8282, - "step": 10250 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019653284177771767, - "loss": 1.5987, - "step": 10251 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965325035513766, - "loss": 1.6405, - "step": 10252 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019653216532503553, - "loss": 1.4905, - "step": 10253 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019653182709869447, - "loss": 1.5659, - "step": 10254 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019653148887235339, - "loss": 1.6391, - "step": 10255 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019653115064601233, - "loss": 1.6328, - "step": 10256 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019653081241967127, - "loss": 1.5591, - "step": 10257 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019653047419333018, - "loss": 1.5916, - "step": 10258 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019653013596698912, - "loss": 1.5803, - "step": 10259 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019652979774064806, - "loss": 1.604, - "step": 10260 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019652945951430698, - "loss": 1.6855, - "step": 10261 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019652912128796592, - "loss": 1.6241, - "step": 10262 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019652878306162486, - "loss": 1.8848, - "step": 10263 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019652844483528377, - "loss": 1.7208, - "step": 10264 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019652810660894272, - "loss": 1.4602, - "step": 10265 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019652776838260166, - "loss": 1.7165, - "step": 10266 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019652743015626057, - "loss": 1.6321, - "step": 10267 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965270919299195, - "loss": 1.6056, - "step": 10268 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019652675370357845, - "loss": 1.5563, - "step": 10269 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019652641547723737, - "loss": 1.8394, - "step": 10270 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965260772508963, - "loss": 1.77, - "step": 10271 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019652573902455525, - "loss": 1.6045, - "step": 10272 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019652540079821416, - "loss": 1.6031, - "step": 10273 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965250625718731, - "loss": 1.4801, - "step": 10274 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019652472434553205, - "loss": 1.5508, - "step": 10275 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019652438611919096, - "loss": 1.6155, - "step": 10276 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965240478928499, - "loss": 1.8013, - "step": 10277 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019652370966650884, - "loss": 1.8083, - "step": 10278 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019652337144016778, - "loss": 1.6533, - "step": 10279 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965230332138267, - "loss": 1.6413, - "step": 10280 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019652269498748564, - "loss": 1.5991, - "step": 10281 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019652235676114458, - "loss": 1.5639, - "step": 10282 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965220185348035, - "loss": 1.5692, - "step": 10283 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019652168030846243, - "loss": 1.6519, - "step": 10284 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019652134208212138, - "loss": 1.57, - "step": 10285 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965210038557803, - "loss": 1.5967, - "step": 10286 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019652066562943923, - "loss": 1.6056, - "step": 10287 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019652032740309817, - "loss": 1.5171, - "step": 10288 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019651998917675709, - "loss": 1.7902, - "step": 10289 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019651965095041603, - "loss": 1.6374, - "step": 10290 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019651931272407497, - "loss": 1.4855, - "step": 10291 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019651897449773388, - "loss": 1.5885, - "step": 10292 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019651863627139282, - "loss": 1.5996, - "step": 10293 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019651829804505176, - "loss": 1.5776, - "step": 10294 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019651795981871068, - "loss": 1.5084, - "step": 10295 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019651762159236962, - "loss": 1.5403, - "step": 10296 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019651728336602856, - "loss": 1.5167, - "step": 10297 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019651694513968747, - "loss": 1.5734, - "step": 10298 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019651660691334642, - "loss": 1.5877, - "step": 10299 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019651626868700536, - "loss": 1.7343, - "step": 10300 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019651593046066427, - "loss": 1.5991, - "step": 10301 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965155922343232, - "loss": 1.5242, - "step": 10302 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019651525400798215, - "loss": 1.5798, - "step": 10303 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019651491578164107, - "loss": 1.6092, - "step": 10304 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965145775553, - "loss": 1.5544, - "step": 10305 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019651423932895895, - "loss": 1.4701, - "step": 10306 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965139011026179, - "loss": 1.6858, - "step": 10307 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019651356287627683, - "loss": 1.706, - "step": 10308 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019651322464993575, - "loss": 1.816, - "step": 10309 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965128864235947, - "loss": 1.599, - "step": 10310 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019651254819725363, - "loss": 1.679, - "step": 10311 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019651220997091254, - "loss": 1.5929, - "step": 10312 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019651187174457148, - "loss": 1.7315, - "step": 10313 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019651153351823042, - "loss": 1.4446, - "step": 10314 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019651119529188934, - "loss": 1.661, - "step": 10315 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019651085706554828, - "loss": 1.6302, - "step": 10316 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019651051883920722, - "loss": 1.7223, - "step": 10317 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019651018061286614, - "loss": 1.7563, - "step": 10318 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019650984238652508, - "loss": 1.7012, - "step": 10319 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019650950416018402, - "loss": 1.6556, - "step": 10320 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019650916593384293, - "loss": 1.6622, - "step": 10321 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019650882770750187, - "loss": 1.7932, - "step": 10322 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019650848948116081, - "loss": 1.5246, - "step": 10323 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019650815125481973, - "loss": 1.7089, - "step": 10324 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019650781302847867, - "loss": 1.5879, - "step": 10325 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965074748021376, - "loss": 1.5649, - "step": 10326 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019650713657579652, - "loss": 1.5689, - "step": 10327 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019650679834945547, - "loss": 1.7405, - "step": 10328 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965064601231144, - "loss": 1.4224, - "step": 10329 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019650612189677332, - "loss": 1.568, - "step": 10330 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019650578367043226, - "loss": 1.6204, - "step": 10331 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965054454440912, - "loss": 1.6763, - "step": 10332 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019650510721775012, - "loss": 1.6242, - "step": 10333 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019650476899140906, - "loss": 1.8164, - "step": 10334 - }, - { - "epoch": 0.03, - "learning_rate": 0.000196504430765068, - "loss": 1.5506, - "step": 10335 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965040925387269, - "loss": 1.4207, - "step": 10336 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019650375431238585, - "loss": 1.7828, - "step": 10337 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965034160860448, - "loss": 1.7553, - "step": 10338 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965030778597037, - "loss": 1.6714, - "step": 10339 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019650273963336265, - "loss": 1.6116, - "step": 10340 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965024014070216, - "loss": 1.6634, - "step": 10341 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965020631806805, - "loss": 1.7023, - "step": 10342 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019650172495433945, - "loss": 1.5066, - "step": 10343 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965013867279984, - "loss": 1.812, - "step": 10344 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965010485016573, - "loss": 1.6909, - "step": 10345 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019650071027531624, - "loss": 1.6692, - "step": 10346 - }, - { - "epoch": 0.03, - "learning_rate": 0.00019650037204897518, - "loss": 1.545, - "step": 10347 - }, - { - "epoch": 0.03, - "learning_rate": 0.0001965000338226341, - "loss": 1.703, - "step": 10348 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019649969559629304, - "loss": 1.7139, - "step": 10349 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019649935736995198, - "loss": 1.6305, - "step": 10350 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964990191436109, - "loss": 1.6457, - "step": 10351 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019649868091726984, - "loss": 1.6215, - "step": 10352 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019649834269092878, - "loss": 1.3667, - "step": 10353 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964980044645877, - "loss": 1.8662, - "step": 10354 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019649766623824663, - "loss": 1.7429, - "step": 10355 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019649732801190557, - "loss": 1.5024, - "step": 10356 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964969897855645, - "loss": 1.5782, - "step": 10357 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019649665155922343, - "loss": 1.8281, - "step": 10358 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019649631333288237, - "loss": 1.6751, - "step": 10359 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019649597510654128, - "loss": 1.591, - "step": 10360 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019649563688020022, - "loss": 1.7716, - "step": 10361 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019649529865385917, - "loss": 1.6396, - "step": 10362 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964949604275181, - "loss": 1.6026, - "step": 10363 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019649462220117705, - "loss": 1.8117, - "step": 10364 - }, - { - "epoch": 0.04, - "learning_rate": 0.000196494283974836, - "loss": 1.6109, - "step": 10365 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964939457484949, - "loss": 1.638, - "step": 10366 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019649360752215384, - "loss": 1.7039, - "step": 10367 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019649326929581279, - "loss": 1.5134, - "step": 10368 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964929310694717, - "loss": 1.7227, - "step": 10369 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019649259284313064, - "loss": 1.5415, - "step": 10370 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019649225461678958, - "loss": 1.5577, - "step": 10371 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964919163904485, - "loss": 1.6668, - "step": 10372 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019649157816410744, - "loss": 1.7323, - "step": 10373 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019649123993776638, - "loss": 1.6063, - "step": 10374 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964909017114253, - "loss": 1.5312, - "step": 10375 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019649056348508423, - "loss": 1.6966, - "step": 10376 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019649022525874317, - "loss": 1.8693, - "step": 10377 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964898870324021, - "loss": 1.5799, - "step": 10378 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019648954880606103, - "loss": 1.614, - "step": 10379 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019648921057971997, - "loss": 1.6763, - "step": 10380 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019648887235337888, - "loss": 1.5256, - "step": 10381 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019648853412703783, - "loss": 1.5204, - "step": 10382 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019648819590069677, - "loss": 1.5824, - "step": 10383 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019648785767435568, - "loss": 1.665, - "step": 10384 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019648751944801462, - "loss": 1.4654, - "step": 10385 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019648718122167356, - "loss": 1.7346, - "step": 10386 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019648684299533248, - "loss": 1.6154, - "step": 10387 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019648650476899142, - "loss": 1.682, - "step": 10388 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019648616654265036, - "loss": 1.7042, - "step": 10389 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019648582831630927, - "loss": 1.7107, - "step": 10390 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019648549008996821, - "loss": 1.6037, - "step": 10391 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019648515186362716, - "loss": 1.5918, - "step": 10392 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019648481363728607, - "loss": 1.8118, - "step": 10393 - }, - { - "epoch": 0.04, - "learning_rate": 0.000196484475410945, - "loss": 1.5468, - "step": 10394 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019648413718460395, - "loss": 1.6371, - "step": 10395 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019648379895826287, - "loss": 1.5557, - "step": 10396 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964834607319218, - "loss": 1.6143, - "step": 10397 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019648312250558075, - "loss": 1.6521, - "step": 10398 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019648278427923966, - "loss": 1.6981, - "step": 10399 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964824460528986, - "loss": 1.6354, - "step": 10400 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019648210782655754, - "loss": 1.7447, - "step": 10401 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019648176960021646, - "loss": 1.6788, - "step": 10402 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964814313738754, - "loss": 1.5983, - "step": 10403 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019648109314753434, - "loss": 1.5912, - "step": 10404 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019648075492119326, - "loss": 1.7083, - "step": 10405 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964804166948522, - "loss": 1.6926, - "step": 10406 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019648007846851114, - "loss": 1.5969, - "step": 10407 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019647974024217005, - "loss": 1.6865, - "step": 10408 - }, - { - "epoch": 0.04, - "learning_rate": 0.000196479402015829, - "loss": 1.7003, - "step": 10409 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019647906378948793, - "loss": 1.7383, - "step": 10410 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019647872556314685, - "loss": 1.7386, - "step": 10411 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964783873368058, - "loss": 1.5445, - "step": 10412 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019647804911046473, - "loss": 1.8116, - "step": 10413 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019647771088412364, - "loss": 1.6865, - "step": 10414 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019647737265778259, - "loss": 1.5206, - "step": 10415 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019647703443144153, - "loss": 1.7184, - "step": 10416 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019647669620510044, - "loss": 1.5794, - "step": 10417 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019647635797875938, - "loss": 1.6127, - "step": 10418 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019647601975241835, - "loss": 1.7029, - "step": 10419 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019647568152607726, - "loss": 1.6434, - "step": 10420 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964753432997362, - "loss": 1.717, - "step": 10421 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019647500507339515, - "loss": 1.6627, - "step": 10422 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019647466684705406, - "loss": 1.5192, - "step": 10423 - }, - { - "epoch": 0.04, - "learning_rate": 0.000196474328620713, - "loss": 1.6401, - "step": 10424 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019647399039437194, - "loss": 1.6906, - "step": 10425 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019647365216803086, - "loss": 1.6466, - "step": 10426 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964733139416898, - "loss": 1.6438, - "step": 10427 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019647297571534874, - "loss": 1.5528, - "step": 10428 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019647263748900765, - "loss": 1.6427, - "step": 10429 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964722992626666, - "loss": 1.6677, - "step": 10430 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019647196103632554, - "loss": 1.6411, - "step": 10431 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019647162280998445, - "loss": 1.5511, - "step": 10432 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964712845836434, - "loss": 1.616, - "step": 10433 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019647094635730233, - "loss": 1.4973, - "step": 10434 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019647060813096125, - "loss": 1.7967, - "step": 10435 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964702699046202, - "loss": 1.7471, - "step": 10436 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019646993167827913, - "loss": 1.5032, - "step": 10437 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019646959345193804, - "loss": 1.5817, - "step": 10438 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019646925522559698, - "loss": 1.6889, - "step": 10439 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019646891699925592, - "loss": 1.711, - "step": 10440 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019646857877291484, - "loss": 1.6816, - "step": 10441 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019646824054657378, - "loss": 1.6314, - "step": 10442 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019646790232023272, - "loss": 1.5501, - "step": 10443 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019646756409389163, - "loss": 1.7422, - "step": 10444 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019646722586755058, - "loss": 1.5941, - "step": 10445 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019646688764120952, - "loss": 1.6984, - "step": 10446 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019646654941486843, - "loss": 1.7038, - "step": 10447 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019646621118852737, - "loss": 1.6104, - "step": 10448 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964658729621863, - "loss": 1.4671, - "step": 10449 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019646553473584523, - "loss": 1.5618, - "step": 10450 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019646519650950417, - "loss": 1.5634, - "step": 10451 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964648582831631, - "loss": 1.6347, - "step": 10452 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019646452005682202, - "loss": 1.5874, - "step": 10453 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019646418183048096, - "loss": 1.7682, - "step": 10454 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964638436041399, - "loss": 1.6088, - "step": 10455 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019646350537779882, - "loss": 1.7269, - "step": 10456 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019646316715145776, - "loss": 1.633, - "step": 10457 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964628289251167, - "loss": 1.4094, - "step": 10458 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019646249069877562, - "loss": 1.7758, - "step": 10459 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019646215247243456, - "loss": 1.6556, - "step": 10460 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964618142460935, - "loss": 1.5052, - "step": 10461 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964614760197524, - "loss": 1.488, - "step": 10462 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019646113779341135, - "loss": 1.6254, - "step": 10463 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964607995670703, - "loss": 1.5869, - "step": 10464 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964604613407292, - "loss": 1.6677, - "step": 10465 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019646012311438815, - "loss": 1.6879, - "step": 10466 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964597848880471, - "loss": 1.5863, - "step": 10467 - }, - { - "epoch": 0.04, - "learning_rate": 0.000196459446661706, - "loss": 1.6009, - "step": 10468 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019645910843536495, - "loss": 1.6499, - "step": 10469 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964587702090239, - "loss": 1.6504, - "step": 10470 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964584319826828, - "loss": 1.6103, - "step": 10471 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019645809375634174, - "loss": 1.548, - "step": 10472 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019645775553000068, - "loss": 1.6953, - "step": 10473 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964574173036596, - "loss": 1.624, - "step": 10474 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019645707907731854, - "loss": 1.5226, - "step": 10475 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964567408509775, - "loss": 1.6724, - "step": 10476 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019645640262463642, - "loss": 1.9414, - "step": 10477 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019645606439829536, - "loss": 1.5004, - "step": 10478 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964557261719543, - "loss": 1.7821, - "step": 10479 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019645538794561322, - "loss": 1.6657, - "step": 10480 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019645504971927216, - "loss": 1.5896, - "step": 10481 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964547114929311, - "loss": 1.5931, - "step": 10482 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019645437326659001, - "loss": 1.6305, - "step": 10483 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019645403504024895, - "loss": 1.5999, - "step": 10484 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964536968139079, - "loss": 1.5483, - "step": 10485 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964533585875668, - "loss": 1.65, - "step": 10486 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019645302036122575, - "loss": 1.676, - "step": 10487 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964526821348847, - "loss": 1.6108, - "step": 10488 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964523439085436, - "loss": 1.6843, - "step": 10489 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019645200568220255, - "loss": 1.359, - "step": 10490 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964516674558615, - "loss": 1.8721, - "step": 10491 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964513292295204, - "loss": 1.624, - "step": 10492 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019645099100317934, - "loss": 1.7371, - "step": 10493 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019645065277683828, - "loss": 1.7038, - "step": 10494 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964503145504972, - "loss": 1.6729, - "step": 10495 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019644997632415614, - "loss": 1.6059, - "step": 10496 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019644963809781508, - "loss": 1.6533, - "step": 10497 - }, - { - "epoch": 0.04, - "learning_rate": 0.000196449299871474, - "loss": 1.593, - "step": 10498 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019644896164513294, - "loss": 1.6572, - "step": 10499 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019644862341879188, - "loss": 1.3996, - "step": 10500 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964482851924508, - "loss": 1.7073, - "step": 10501 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019644794696610973, - "loss": 1.5494, - "step": 10502 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019644760873976867, - "loss": 1.7412, - "step": 10503 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964472705134276, - "loss": 1.544, - "step": 10504 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019644693228708653, - "loss": 1.7632, - "step": 10505 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019644659406074547, - "loss": 1.6748, - "step": 10506 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019644625583440438, - "loss": 1.5628, - "step": 10507 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019644591760806333, - "loss": 1.696, - "step": 10508 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019644557938172227, - "loss": 1.7236, - "step": 10509 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019644524115538118, - "loss": 1.5393, - "step": 10510 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019644490292904012, - "loss": 1.5334, - "step": 10511 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019644456470269906, - "loss": 1.5887, - "step": 10512 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019644422647635798, - "loss": 1.6842, - "step": 10513 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019644388825001692, - "loss": 1.6343, - "step": 10514 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019644355002367586, - "loss": 1.6942, - "step": 10515 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019644321179733477, - "loss": 1.7312, - "step": 10516 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019644287357099371, - "loss": 1.5002, - "step": 10517 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019644253534465266, - "loss": 1.6229, - "step": 10518 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019644219711831157, - "loss": 1.7087, - "step": 10519 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964418588919705, - "loss": 1.724, - "step": 10520 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019644152066562945, - "loss": 1.5711, - "step": 10521 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019644118243928837, - "loss": 1.4764, - "step": 10522 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964408442129473, - "loss": 1.5617, - "step": 10523 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019644050598660625, - "loss": 1.6294, - "step": 10524 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019644016776026516, - "loss": 1.6976, - "step": 10525 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964398295339241, - "loss": 1.5942, - "step": 10526 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019643949130758304, - "loss": 1.6435, - "step": 10527 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019643915308124196, - "loss": 1.588, - "step": 10528 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964388148549009, - "loss": 1.9335, - "step": 10529 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019643847662855984, - "loss": 1.7927, - "step": 10530 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019643813840221875, - "loss": 1.613, - "step": 10531 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019643780017587772, - "loss": 1.5989, - "step": 10532 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019643746194953664, - "loss": 1.6961, - "step": 10533 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019643712372319558, - "loss": 1.7762, - "step": 10534 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019643678549685452, - "loss": 1.5822, - "step": 10535 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019643644727051343, - "loss": 1.692, - "step": 10536 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019643610904417237, - "loss": 1.7614, - "step": 10537 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019643577081783132, - "loss": 1.5229, - "step": 10538 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019643543259149023, - "loss": 1.7416, - "step": 10539 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019643509436514917, - "loss": 1.7327, - "step": 10540 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964347561388081, - "loss": 1.568, - "step": 10541 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019643441791246703, - "loss": 1.6098, - "step": 10542 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019643407968612597, - "loss": 1.5944, - "step": 10543 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964337414597849, - "loss": 1.8773, - "step": 10544 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019643340323344382, - "loss": 1.5381, - "step": 10545 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019643306500710276, - "loss": 1.6131, - "step": 10546 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964327267807617, - "loss": 1.7222, - "step": 10547 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019643238855442065, - "loss": 1.6182, - "step": 10548 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019643205032807956, - "loss": 1.6508, - "step": 10549 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964317121017385, - "loss": 1.5615, - "step": 10550 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019643137387539744, - "loss": 1.6209, - "step": 10551 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019643103564905636, - "loss": 1.7875, - "step": 10552 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964306974227153, - "loss": 1.5793, - "step": 10553 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019643035919637424, - "loss": 1.7133, - "step": 10554 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019643002097003315, - "loss": 1.6527, - "step": 10555 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964296827436921, - "loss": 1.6745, - "step": 10556 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019642934451735103, - "loss": 1.6313, - "step": 10557 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019642900629100995, - "loss": 1.6018, - "step": 10558 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964286680646689, - "loss": 1.7177, - "step": 10559 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019642832983832783, - "loss": 1.6326, - "step": 10560 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019642799161198674, - "loss": 1.7005, - "step": 10561 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019642765338564569, - "loss": 1.5789, - "step": 10562 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019642731515930463, - "loss": 1.4796, - "step": 10563 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019642697693296354, - "loss": 1.5978, - "step": 10564 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019642663870662248, - "loss": 1.5906, - "step": 10565 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019642630048028142, - "loss": 1.6532, - "step": 10566 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019642596225394034, - "loss": 1.4673, - "step": 10567 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019642562402759928, - "loss": 1.5798, - "step": 10568 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019642528580125822, - "loss": 1.6099, - "step": 10569 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019642494757491713, - "loss": 1.5972, - "step": 10570 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019642460934857607, - "loss": 1.8051, - "step": 10571 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019642427112223502, - "loss": 1.4769, - "step": 10572 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019642393289589393, - "loss": 1.4646, - "step": 10573 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019642359466955287, - "loss": 1.6532, - "step": 10574 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964232564432118, - "loss": 1.641, - "step": 10575 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019642291821687073, - "loss": 1.7649, - "step": 10576 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019642257999052967, - "loss": 1.5867, - "step": 10577 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964222417641886, - "loss": 1.6275, - "step": 10578 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019642190353784752, - "loss": 1.6378, - "step": 10579 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019642156531150646, - "loss": 1.7362, - "step": 10580 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964212270851654, - "loss": 1.6155, - "step": 10581 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019642088885882432, - "loss": 1.7216, - "step": 10582 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019642055063248326, - "loss": 1.5578, - "step": 10583 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964202124061422, - "loss": 1.6943, - "step": 10584 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019641987417980112, - "loss": 1.6604, - "step": 10585 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019641953595346006, - "loss": 1.6607, - "step": 10586 - }, - { - "epoch": 0.04, - "learning_rate": 0.000196419197727119, - "loss": 1.8053, - "step": 10587 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019641885950077794, - "loss": 1.716, - "step": 10588 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019641852127443688, - "loss": 1.7606, - "step": 10589 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964181830480958, - "loss": 1.6058, - "step": 10590 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019641784482175474, - "loss": 1.6507, - "step": 10591 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019641750659541368, - "loss": 1.6134, - "step": 10592 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964171683690726, - "loss": 1.6149, - "step": 10593 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019641683014273153, - "loss": 1.7166, - "step": 10594 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019641649191639047, - "loss": 1.5871, - "step": 10595 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964161536900494, - "loss": 1.5465, - "step": 10596 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019641581546370833, - "loss": 1.6832, - "step": 10597 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019641547723736727, - "loss": 1.8111, - "step": 10598 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019641513901102618, - "loss": 1.7009, - "step": 10599 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019641480078468512, - "loss": 1.613, - "step": 10600 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019641446255834407, - "loss": 1.6148, - "step": 10601 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019641412433200298, - "loss": 1.7329, - "step": 10602 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019641378610566192, - "loss": 1.5784, - "step": 10603 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019641344787932086, - "loss": 1.6695, - "step": 10604 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019641310965297978, - "loss": 1.7471, - "step": 10605 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019641277142663872, - "loss": 1.7606, - "step": 10606 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019641243320029766, - "loss": 1.4913, - "step": 10607 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019641209497395657, - "loss": 1.6005, - "step": 10608 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964117567476155, - "loss": 1.6219, - "step": 10609 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019641141852127445, - "loss": 1.676, - "step": 10610 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019641108029493337, - "loss": 1.7163, - "step": 10611 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964107420685923, - "loss": 1.4872, - "step": 10612 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019641040384225125, - "loss": 1.5329, - "step": 10613 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019641006561591016, - "loss": 1.6875, - "step": 10614 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964097273895691, - "loss": 1.7398, - "step": 10615 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019640938916322805, - "loss": 1.5695, - "step": 10616 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019640905093688696, - "loss": 1.7832, - "step": 10617 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964087127105459, - "loss": 1.6217, - "step": 10618 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019640837448420484, - "loss": 1.7029, - "step": 10619 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019640803625786376, - "loss": 1.6362, - "step": 10620 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964076980315227, - "loss": 1.5804, - "step": 10621 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019640735980518164, - "loss": 1.5667, - "step": 10622 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019640702157884055, - "loss": 1.6428, - "step": 10623 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964066833524995, - "loss": 1.5767, - "step": 10624 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019640634512615844, - "loss": 1.8084, - "step": 10625 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019640600689981735, - "loss": 1.5473, - "step": 10626 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964056686734763, - "loss": 1.5496, - "step": 10627 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019640533044713523, - "loss": 1.7515, - "step": 10628 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019640499222079415, - "loss": 1.5643, - "step": 10629 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964046539944531, - "loss": 1.7221, - "step": 10630 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019640431576811203, - "loss": 1.6338, - "step": 10631 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019640397754177097, - "loss": 1.6783, - "step": 10632 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019640363931542988, - "loss": 1.6422, - "step": 10633 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019640330108908882, - "loss": 1.6123, - "step": 10634 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019640296286274777, - "loss": 1.6726, - "step": 10635 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019640262463640668, - "loss": 1.6823, - "step": 10636 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019640228641006562, - "loss": 1.6846, - "step": 10637 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019640194818372456, - "loss": 1.6153, - "step": 10638 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019640160995738348, - "loss": 1.5708, - "step": 10639 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019640127173104242, - "loss": 1.5555, - "step": 10640 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019640093350470136, - "loss": 1.6185, - "step": 10641 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019640059527836027, - "loss": 1.6061, - "step": 10642 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001964002570520192, - "loss": 1.5866, - "step": 10643 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019639991882567815, - "loss": 1.8574, - "step": 10644 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963995805993371, - "loss": 1.4875, - "step": 10645 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019639924237299604, - "loss": 1.5913, - "step": 10646 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019639890414665495, - "loss": 1.5857, - "step": 10647 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963985659203139, - "loss": 1.6288, - "step": 10648 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019639822769397283, - "loss": 1.709, - "step": 10649 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019639788946763175, - "loss": 1.6785, - "step": 10650 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963975512412907, - "loss": 1.6639, - "step": 10651 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019639721301494963, - "loss": 1.5308, - "step": 10652 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019639687478860854, - "loss": 1.7078, - "step": 10653 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019639653656226748, - "loss": 1.562, - "step": 10654 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019639619833592643, - "loss": 1.6342, - "step": 10655 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019639586010958534, - "loss": 1.7194, - "step": 10656 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019639552188324428, - "loss": 1.69, - "step": 10657 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019639518365690322, - "loss": 1.6438, - "step": 10658 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019639484543056214, - "loss": 1.7809, - "step": 10659 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019639450720422108, - "loss": 1.7138, - "step": 10660 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019639416897788002, - "loss": 1.6664, - "step": 10661 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019639383075153893, - "loss": 1.6216, - "step": 10662 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019639349252519787, - "loss": 1.5541, - "step": 10663 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019639315429885681, - "loss": 1.8037, - "step": 10664 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019639281607251573, - "loss": 1.6031, - "step": 10665 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019639247784617467, - "loss": 1.551, - "step": 10666 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963921396198336, - "loss": 1.6763, - "step": 10667 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019639180139349253, - "loss": 1.6181, - "step": 10668 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019639146316715147, - "loss": 1.5383, - "step": 10669 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963911249408104, - "loss": 1.5605, - "step": 10670 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019639078671446932, - "loss": 1.6822, - "step": 10671 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019639044848812826, - "loss": 1.6932, - "step": 10672 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963901102617872, - "loss": 1.6425, - "step": 10673 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019638977203544612, - "loss": 1.6214, - "step": 10674 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019638943380910506, - "loss": 1.4953, - "step": 10675 - }, - { - "epoch": 0.04, - "learning_rate": 0.000196389095582764, - "loss": 1.6484, - "step": 10676 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019638875735642291, - "loss": 1.7143, - "step": 10677 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019638841913008186, - "loss": 1.6423, - "step": 10678 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963880809037408, - "loss": 1.6907, - "step": 10679 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963877426773997, - "loss": 1.7264, - "step": 10680 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019638740445105865, - "loss": 1.6935, - "step": 10681 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963870662247176, - "loss": 1.7988, - "step": 10682 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963867279983765, - "loss": 1.8698, - "step": 10683 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019638638977203545, - "loss": 1.6134, - "step": 10684 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963860515456944, - "loss": 1.4731, - "step": 10685 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963857133193533, - "loss": 1.5423, - "step": 10686 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019638537509301224, - "loss": 1.6669, - "step": 10687 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019638503686667119, - "loss": 1.7087, - "step": 10688 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963846986403301, - "loss": 1.6065, - "step": 10689 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019638436041398904, - "loss": 1.6861, - "step": 10690 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019638402218764798, - "loss": 1.6226, - "step": 10691 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963836839613069, - "loss": 1.6413, - "step": 10692 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019638334573496584, - "loss": 1.4696, - "step": 10693 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019638300750862478, - "loss": 1.6413, - "step": 10694 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963826692822837, - "loss": 1.6347, - "step": 10695 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019638233105594263, - "loss": 1.6837, - "step": 10696 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019638199282960157, - "loss": 1.6495, - "step": 10697 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963816546032605, - "loss": 1.6219, - "step": 10698 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019638131637691943, - "loss": 1.6308, - "step": 10699 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963809781505784, - "loss": 1.7835, - "step": 10700 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963806399242373, - "loss": 1.5989, - "step": 10701 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019638030169789625, - "loss": 1.7758, - "step": 10702 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963799634715552, - "loss": 1.6968, - "step": 10703 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963796252452141, - "loss": 1.6492, - "step": 10704 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019637928701887305, - "loss": 1.6317, - "step": 10705 - }, - { - "epoch": 0.04, - "learning_rate": 0.000196378948792532, - "loss": 1.5943, - "step": 10706 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963786105661909, - "loss": 1.4986, - "step": 10707 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019637827233984985, - "loss": 1.6603, - "step": 10708 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963779341135088, - "loss": 1.7414, - "step": 10709 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963775958871677, - "loss": 1.5168, - "step": 10710 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019637725766082664, - "loss": 1.6752, - "step": 10711 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019637691943448558, - "loss": 1.6194, - "step": 10712 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963765812081445, - "loss": 1.7794, - "step": 10713 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019637624298180344, - "loss": 1.567, - "step": 10714 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019637590475546238, - "loss": 1.7218, - "step": 10715 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963755665291213, - "loss": 1.5016, - "step": 10716 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019637522830278023, - "loss": 1.5698, - "step": 10717 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019637489007643918, - "loss": 1.6376, - "step": 10718 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963745518500981, - "loss": 1.523, - "step": 10719 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019637421362375703, - "loss": 1.5544, - "step": 10720 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019637387539741597, - "loss": 1.7727, - "step": 10721 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019637353717107489, - "loss": 1.6955, - "step": 10722 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019637319894473383, - "loss": 1.5578, - "step": 10723 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019637286071839277, - "loss": 1.7186, - "step": 10724 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019637252249205168, - "loss": 1.6447, - "step": 10725 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019637218426571062, - "loss": 1.5537, - "step": 10726 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019637184603936956, - "loss": 1.5787, - "step": 10727 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019637150781302848, - "loss": 1.5181, - "step": 10728 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019637116958668742, - "loss": 1.7445, - "step": 10729 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019637083136034636, - "loss": 1.7562, - "step": 10730 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019637049313400527, - "loss": 1.7581, - "step": 10731 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019637015490766422, - "loss": 1.5884, - "step": 10732 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019636981668132316, - "loss": 1.724, - "step": 10733 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019636947845498207, - "loss": 1.5417, - "step": 10734 - }, - { - "epoch": 0.04, - "learning_rate": 0.000196369140228641, - "loss": 1.5575, - "step": 10735 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019636880200229995, - "loss": 1.7857, - "step": 10736 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019636846377595887, - "loss": 1.6248, - "step": 10737 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963681255496178, - "loss": 1.586, - "step": 10738 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019636778732327675, - "loss": 1.7361, - "step": 10739 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019636744909693566, - "loss": 1.7802, - "step": 10740 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963671108705946, - "loss": 1.7432, - "step": 10741 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019636677264425355, - "loss": 1.7038, - "step": 10742 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019636643441791246, - "loss": 1.5694, - "step": 10743 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963660961915714, - "loss": 1.6615, - "step": 10744 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019636575796523034, - "loss": 1.6151, - "step": 10745 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019636541973888926, - "loss": 1.5221, - "step": 10746 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963650815125482, - "loss": 1.8142, - "step": 10747 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019636474328620714, - "loss": 1.5483, - "step": 10748 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019636440505986605, - "loss": 1.6353, - "step": 10749 - }, - { - "epoch": 0.04, - "learning_rate": 0.000196364066833525, - "loss": 1.653, - "step": 10750 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019636372860718394, - "loss": 1.4599, - "step": 10751 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019636339038084285, - "loss": 1.5811, - "step": 10752 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963630521545018, - "loss": 1.6385, - "step": 10753 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019636271392816073, - "loss": 1.5545, - "step": 10754 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019636237570181965, - "loss": 1.6761, - "step": 10755 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019636203747547861, - "loss": 1.6647, - "step": 10756 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019636169924913755, - "loss": 1.7615, - "step": 10757 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019636136102279647, - "loss": 1.6637, - "step": 10758 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963610227964554, - "loss": 1.7534, - "step": 10759 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019636068457011435, - "loss": 1.6066, - "step": 10760 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019636034634377327, - "loss": 1.5522, - "step": 10761 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963600081174322, - "loss": 1.8166, - "step": 10762 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019635966989109115, - "loss": 1.785, - "step": 10763 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019635933166475006, - "loss": 1.6789, - "step": 10764 - }, - { - "epoch": 0.04, - "learning_rate": 0.000196358993438409, - "loss": 1.5163, - "step": 10765 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019635865521206794, - "loss": 1.657, - "step": 10766 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019635831698572686, - "loss": 1.6105, - "step": 10767 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963579787593858, - "loss": 1.623, - "step": 10768 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019635764053304474, - "loss": 1.5709, - "step": 10769 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019635730230670365, - "loss": 1.5891, - "step": 10770 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963569640803626, - "loss": 1.5979, - "step": 10771 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019635662585402154, - "loss": 1.5655, - "step": 10772 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019635628762768045, - "loss": 1.6099, - "step": 10773 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963559494013394, - "loss": 1.5324, - "step": 10774 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019635561117499833, - "loss": 1.39, - "step": 10775 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019635527294865725, - "loss": 1.5908, - "step": 10776 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963549347223162, - "loss": 1.5806, - "step": 10777 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019635459649597513, - "loss": 1.7153, - "step": 10778 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019635425826963404, - "loss": 1.6412, - "step": 10779 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019635392004329298, - "loss": 1.6285, - "step": 10780 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019635358181695193, - "loss": 1.6718, - "step": 10781 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019635324359061084, - "loss": 1.6318, - "step": 10782 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019635290536426978, - "loss": 1.5658, - "step": 10783 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019635256713792872, - "loss": 1.6253, - "step": 10784 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019635222891158764, - "loss": 1.6129, - "step": 10785 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019635189068524658, - "loss": 1.6217, - "step": 10786 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019635155245890552, - "loss": 1.7034, - "step": 10787 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019635121423256443, - "loss": 1.6967, - "step": 10788 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019635087600622337, - "loss": 1.7487, - "step": 10789 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019635053777988231, - "loss": 1.6313, - "step": 10790 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019635019955354123, - "loss": 1.5687, - "step": 10791 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019634986132720017, - "loss": 1.541, - "step": 10792 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963495231008591, - "loss": 1.4677, - "step": 10793 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019634918487451802, - "loss": 1.5605, - "step": 10794 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019634884664817697, - "loss": 1.5833, - "step": 10795 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963485084218359, - "loss": 1.6174, - "step": 10796 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019634817019549482, - "loss": 1.666, - "step": 10797 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019634783196915376, - "loss": 1.5977, - "step": 10798 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963474937428127, - "loss": 1.618, - "step": 10799 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019634715551647162, - "loss": 1.6133, - "step": 10800 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019634681729013056, - "loss": 1.6835, - "step": 10801 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963464790637895, - "loss": 1.5953, - "step": 10802 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963461408374484, - "loss": 1.6, - "step": 10803 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019634580261110735, - "loss": 1.7752, - "step": 10804 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963454643847663, - "loss": 1.6759, - "step": 10805 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963451261584252, - "loss": 1.7501, - "step": 10806 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019634478793208415, - "loss": 1.6084, - "step": 10807 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963444497057431, - "loss": 1.5914, - "step": 10808 - }, - { - "epoch": 0.04, - "learning_rate": 0.000196344111479402, - "loss": 1.5596, - "step": 10809 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019634377325306095, - "loss": 1.6133, - "step": 10810 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963434350267199, - "loss": 1.7067, - "step": 10811 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963430968003788, - "loss": 1.5935, - "step": 10812 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019634275857403777, - "loss": 1.5729, - "step": 10813 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963424203476967, - "loss": 1.6877, - "step": 10814 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019634208212135563, - "loss": 1.6851, - "step": 10815 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019634174389501457, - "loss": 1.5779, - "step": 10816 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963414056686735, - "loss": 1.5836, - "step": 10817 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019634106744233242, - "loss": 1.5905, - "step": 10818 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019634072921599136, - "loss": 1.5848, - "step": 10819 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963403909896503, - "loss": 1.6156, - "step": 10820 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019634005276330922, - "loss": 1.8144, - "step": 10821 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019633971453696816, - "loss": 1.6243, - "step": 10822 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963393763106271, - "loss": 1.5663, - "step": 10823 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019633903808428601, - "loss": 1.7254, - "step": 10824 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019633869985794496, - "loss": 1.689, - "step": 10825 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963383616316039, - "loss": 1.7738, - "step": 10826 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963380234052628, - "loss": 1.641, - "step": 10827 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019633768517892175, - "loss": 1.4016, - "step": 10828 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963373469525807, - "loss": 1.5509, - "step": 10829 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963370087262396, - "loss": 1.5874, - "step": 10830 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019633667049989855, - "loss": 1.6138, - "step": 10831 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963363322735575, - "loss": 1.5738, - "step": 10832 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963359940472164, - "loss": 1.6059, - "step": 10833 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019633565582087534, - "loss": 1.5485, - "step": 10834 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019633531759453429, - "loss": 1.5952, - "step": 10835 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963349793681932, - "loss": 1.7494, - "step": 10836 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019633464114185214, - "loss": 1.749, - "step": 10837 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019633430291551108, - "loss": 1.6575, - "step": 10838 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019633396468917, - "loss": 1.6698, - "step": 10839 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019633362646282894, - "loss": 1.7087, - "step": 10840 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019633328823648788, - "loss": 1.7065, - "step": 10841 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963329500101468, - "loss": 1.8271, - "step": 10842 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019633261178380573, - "loss": 1.5619, - "step": 10843 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019633227355746467, - "loss": 1.6891, - "step": 10844 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963319353311236, - "loss": 1.8199, - "step": 10845 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019633159710478253, - "loss": 1.71, - "step": 10846 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019633125887844147, - "loss": 1.5777, - "step": 10847 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019633092065210039, - "loss": 1.5657, - "step": 10848 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019633058242575933, - "loss": 1.5422, - "step": 10849 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019633024419941827, - "loss": 1.6645, - "step": 10850 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019632990597307718, - "loss": 1.6277, - "step": 10851 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019632956774673612, - "loss": 1.6146, - "step": 10852 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019632922952039506, - "loss": 1.7115, - "step": 10853 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019632889129405398, - "loss": 1.5822, - "step": 10854 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019632855306771292, - "loss": 1.7184, - "step": 10855 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019632821484137186, - "loss": 1.8427, - "step": 10856 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019632787661503077, - "loss": 1.6185, - "step": 10857 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019632753838868972, - "loss": 1.8114, - "step": 10858 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019632720016234866, - "loss": 1.6809, - "step": 10859 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019632686193600757, - "loss": 1.5705, - "step": 10860 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963265237096665, - "loss": 1.6517, - "step": 10861 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019632618548332545, - "loss": 1.6934, - "step": 10862 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019632584725698437, - "loss": 1.6285, - "step": 10863 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963255090306433, - "loss": 1.7518, - "step": 10864 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019632517080430225, - "loss": 1.7152, - "step": 10865 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019632483257796116, - "loss": 1.6278, - "step": 10866 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963244943516201, - "loss": 1.4958, - "step": 10867 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019632415612527905, - "loss": 1.6426, - "step": 10868 - }, - { - "epoch": 0.04, - "learning_rate": 0.000196323817898938, - "loss": 1.7823, - "step": 10869 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019632347967259693, - "loss": 1.7207, - "step": 10870 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019632314144625584, - "loss": 1.5735, - "step": 10871 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019632280321991478, - "loss": 1.5619, - "step": 10872 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019632246499357372, - "loss": 1.5928, - "step": 10873 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019632212676723264, - "loss": 1.9242, - "step": 10874 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019632178854089158, - "loss": 1.5951, - "step": 10875 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019632145031455052, - "loss": 1.5201, - "step": 10876 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019632111208820943, - "loss": 1.6594, - "step": 10877 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019632077386186838, - "loss": 1.57, - "step": 10878 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019632043563552732, - "loss": 1.6339, - "step": 10879 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019632009740918623, - "loss": 1.6763, - "step": 10880 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019631975918284517, - "loss": 1.6239, - "step": 10881 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963194209565041, - "loss": 1.7807, - "step": 10882 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019631908273016303, - "loss": 1.6304, - "step": 10883 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019631874450382197, - "loss": 1.5569, - "step": 10884 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963184062774809, - "loss": 1.5225, - "step": 10885 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019631806805113982, - "loss": 1.5689, - "step": 10886 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019631772982479876, - "loss": 1.7908, - "step": 10887 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963173915984577, - "loss": 1.5477, - "step": 10888 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019631705337211662, - "loss": 1.5386, - "step": 10889 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019631671514577556, - "loss": 1.7007, - "step": 10890 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963163769194345, - "loss": 1.6121, - "step": 10891 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019631603869309342, - "loss": 1.6952, - "step": 10892 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019631570046675236, - "loss": 1.6988, - "step": 10893 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963153622404113, - "loss": 1.5723, - "step": 10894 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963150240140702, - "loss": 1.5345, - "step": 10895 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019631468578772915, - "loss": 1.6919, - "step": 10896 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963143475613881, - "loss": 1.6072, - "step": 10897 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019631400933504704, - "loss": 1.7554, - "step": 10898 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019631367110870595, - "loss": 1.7031, - "step": 10899 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963133328823649, - "loss": 1.4833, - "step": 10900 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019631299465602383, - "loss": 1.8564, - "step": 10901 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019631265642968275, - "loss": 1.6069, - "step": 10902 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963123182033417, - "loss": 1.8186, - "step": 10903 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019631197997700063, - "loss": 1.5608, - "step": 10904 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019631164175065954, - "loss": 1.5988, - "step": 10905 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019631130352431848, - "loss": 1.5963, - "step": 10906 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019631096529797742, - "loss": 1.5395, - "step": 10907 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019631062707163634, - "loss": 1.6247, - "step": 10908 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019631028884529528, - "loss": 1.5461, - "step": 10909 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019630995061895422, - "loss": 1.5908, - "step": 10910 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019630961239261314, - "loss": 1.5948, - "step": 10911 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019630927416627208, - "loss": 1.5945, - "step": 10912 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019630893593993102, - "loss": 1.5845, - "step": 10913 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019630859771358993, - "loss": 1.5892, - "step": 10914 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019630825948724887, - "loss": 1.6606, - "step": 10915 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019630792126090781, - "loss": 1.4902, - "step": 10916 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019630758303456673, - "loss": 1.5112, - "step": 10917 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019630724480822567, - "loss": 1.6773, - "step": 10918 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963069065818846, - "loss": 1.4695, - "step": 10919 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019630656835554352, - "loss": 1.6622, - "step": 10920 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019630623012920247, - "loss": 1.6845, - "step": 10921 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963058919028614, - "loss": 1.7899, - "step": 10922 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019630555367652032, - "loss": 1.6695, - "step": 10923 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019630521545017926, - "loss": 1.6794, - "step": 10924 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963048772238382, - "loss": 1.5582, - "step": 10925 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019630453899749714, - "loss": 1.6909, - "step": 10926 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019630420077115608, - "loss": 1.6268, - "step": 10927 - }, - { - "epoch": 0.04, - "learning_rate": 0.000196303862544815, - "loss": 1.6406, - "step": 10928 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019630352431847394, - "loss": 1.6824, - "step": 10929 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019630318609213288, - "loss": 1.631, - "step": 10930 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963028478657918, - "loss": 1.5102, - "step": 10931 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019630250963945074, - "loss": 1.683, - "step": 10932 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019630217141310968, - "loss": 1.6511, - "step": 10933 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963018331867686, - "loss": 1.6397, - "step": 10934 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019630149496042753, - "loss": 1.5815, - "step": 10935 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019630115673408647, - "loss": 1.4851, - "step": 10936 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001963008185077454, - "loss": 1.7126, - "step": 10937 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019630048028140433, - "loss": 1.4418, - "step": 10938 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019630014205506327, - "loss": 1.5317, - "step": 10939 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019629980382872218, - "loss": 1.7526, - "step": 10940 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019629946560238113, - "loss": 1.727, - "step": 10941 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019629912737604007, - "loss": 1.6421, - "step": 10942 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019629878914969898, - "loss": 1.5823, - "step": 10943 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019629845092335792, - "loss": 1.7076, - "step": 10944 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019629811269701686, - "loss": 1.5948, - "step": 10945 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019629777447067578, - "loss": 1.7205, - "step": 10946 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019629743624433472, - "loss": 1.4659, - "step": 10947 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019629709801799366, - "loss": 1.5922, - "step": 10948 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019629675979165257, - "loss": 1.5277, - "step": 10949 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019629642156531151, - "loss": 1.6369, - "step": 10950 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019629608333897046, - "loss": 1.6893, - "step": 10951 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019629574511262937, - "loss": 1.4581, - "step": 10952 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962954068862883, - "loss": 1.8304, - "step": 10953 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019629506865994725, - "loss": 1.6872, - "step": 10954 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019629473043360617, - "loss": 1.6186, - "step": 10955 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962943922072651, - "loss": 1.6125, - "step": 10956 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019629405398092405, - "loss": 1.6878, - "step": 10957 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019629371575458296, - "loss": 1.7999, - "step": 10958 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962933775282419, - "loss": 1.8024, - "step": 10959 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019629303930190084, - "loss": 1.6795, - "step": 10960 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019629270107555976, - "loss": 1.6291, - "step": 10961 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962923628492187, - "loss": 1.6594, - "step": 10962 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019629202462287764, - "loss": 1.5771, - "step": 10963 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019629168639653655, - "loss": 1.7149, - "step": 10964 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962913481701955, - "loss": 1.6591, - "step": 10965 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019629100994385444, - "loss": 1.6499, - "step": 10966 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019629067171751335, - "loss": 1.5095, - "step": 10967 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962903334911723, - "loss": 1.6263, - "step": 10968 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019628999526483123, - "loss": 1.5911, - "step": 10969 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019628965703849015, - "loss": 1.6173, - "step": 10970 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962893188121491, - "loss": 1.5774, - "step": 10971 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019628898058580803, - "loss": 1.7951, - "step": 10972 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019628864235946694, - "loss": 1.6174, - "step": 10973 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019628830413312588, - "loss": 1.6617, - "step": 10974 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019628796590678483, - "loss": 1.5899, - "step": 10975 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019628762768044374, - "loss": 1.6607, - "step": 10976 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019628728945410268, - "loss": 1.7868, - "step": 10977 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019628695122776162, - "loss": 1.5452, - "step": 10978 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019628661300142054, - "loss": 1.4607, - "step": 10979 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019628627477507948, - "loss": 1.5699, - "step": 10980 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019628593654873845, - "loss": 1.7206, - "step": 10981 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019628559832239736, - "loss": 1.6698, - "step": 10982 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962852600960563, - "loss": 1.7112, - "step": 10983 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019628492186971524, - "loss": 1.854, - "step": 10984 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019628458364337416, - "loss": 1.6151, - "step": 10985 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962842454170331, - "loss": 1.5987, - "step": 10986 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019628390719069204, - "loss": 1.7359, - "step": 10987 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019628356896435095, - "loss": 1.6338, - "step": 10988 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962832307380099, - "loss": 1.5537, - "step": 10989 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019628289251166883, - "loss": 1.7303, - "step": 10990 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019628255428532775, - "loss": 1.7107, - "step": 10991 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962822160589867, - "loss": 1.6433, - "step": 10992 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019628187783264563, - "loss": 1.7073, - "step": 10993 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019628153960630454, - "loss": 1.5757, - "step": 10994 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019628120137996349, - "loss": 1.5945, - "step": 10995 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019628086315362243, - "loss": 1.7727, - "step": 10996 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019628052492728134, - "loss": 1.777, - "step": 10997 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019628018670094028, - "loss": 1.6764, - "step": 10998 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019627984847459922, - "loss": 1.6541, - "step": 10999 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019627951024825814, - "loss": 1.5781, - "step": 11000 - }, - { - "epoch": 0.04, - "eval_loss": 1.4279597355769231, - "mrc_em": 20.0, - "mrc_f1": 29.6904761904762, - "nli_acc": 0.32, - "step": 11000, - "sts_f1": 0.0, - "ynat_f1": 0.5095134644766998 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019627917202191708, - "loss": 1.7407, - "step": 11001 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019627883379557602, - "loss": 1.5535, - "step": 11002 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019627849556923493, - "loss": 1.716, - "step": 11003 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019627815734289387, - "loss": 1.5974, - "step": 11004 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019627781911655282, - "loss": 1.6289, - "step": 11005 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019627748089021173, - "loss": 1.6564, - "step": 11006 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019627714266387067, - "loss": 1.5151, - "step": 11007 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962768044375296, - "loss": 1.6348, - "step": 11008 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019627646621118853, - "loss": 1.6363, - "step": 11009 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019627612798484747, - "loss": 1.6926, - "step": 11010 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962757897585064, - "loss": 1.6362, - "step": 11011 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019627545153216532, - "loss": 1.5668, - "step": 11012 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019627511330582426, - "loss": 1.6185, - "step": 11013 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962747750794832, - "loss": 1.6162, - "step": 11014 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019627443685314212, - "loss": 1.681, - "step": 11015 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019627409862680106, - "loss": 1.6831, - "step": 11016 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019627376040046, - "loss": 1.7289, - "step": 11017 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019627342217411892, - "loss": 1.615, - "step": 11018 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019627308394777786, - "loss": 1.6436, - "step": 11019 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962727457214368, - "loss": 1.6751, - "step": 11020 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962724074950957, - "loss": 1.6289, - "step": 11021 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019627206926875465, - "loss": 1.5741, - "step": 11022 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962717310424136, - "loss": 1.5722, - "step": 11023 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962713928160725, - "loss": 1.6357, - "step": 11024 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019627105458973145, - "loss": 1.6916, - "step": 11025 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962707163633904, - "loss": 1.8145, - "step": 11026 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962703781370493, - "loss": 1.6793, - "step": 11027 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019627003991070825, - "loss": 1.6846, - "step": 11028 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962697016843672, - "loss": 1.6085, - "step": 11029 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962693634580261, - "loss": 1.6084, - "step": 11030 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019626902523168504, - "loss": 1.7741, - "step": 11031 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019626868700534398, - "loss": 1.7083, - "step": 11032 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962683487790029, - "loss": 1.7276, - "step": 11033 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019626801055266184, - "loss": 1.734, - "step": 11034 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019626767232632078, - "loss": 1.7191, - "step": 11035 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962673340999797, - "loss": 1.6199, - "step": 11036 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019626699587363866, - "loss": 1.576, - "step": 11037 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962666576472976, - "loss": 1.8305, - "step": 11038 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019626631942095652, - "loss": 1.4558, - "step": 11039 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019626598119461546, - "loss": 1.7368, - "step": 11040 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962656429682744, - "loss": 1.569, - "step": 11041 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962653047419333, - "loss": 1.5617, - "step": 11042 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019626496651559225, - "loss": 1.5791, - "step": 11043 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962646282892512, - "loss": 1.659, - "step": 11044 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962642900629101, - "loss": 1.6072, - "step": 11045 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019626395183656905, - "loss": 1.7133, - "step": 11046 - }, - { - "epoch": 0.04, - "learning_rate": 0.000196263613610228, - "loss": 1.3523, - "step": 11047 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962632753838869, - "loss": 1.6966, - "step": 11048 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019626293715754585, - "loss": 1.5875, - "step": 11049 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962625989312048, - "loss": 1.7856, - "step": 11050 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962622607048637, - "loss": 1.7808, - "step": 11051 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019626192247852264, - "loss": 1.7478, - "step": 11052 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019626158425218158, - "loss": 1.6206, - "step": 11053 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962612460258405, - "loss": 1.6981, - "step": 11054 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019626090779949944, - "loss": 1.6889, - "step": 11055 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019626056957315838, - "loss": 1.6273, - "step": 11056 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962602313468173, - "loss": 1.4109, - "step": 11057 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019625989312047624, - "loss": 1.5461, - "step": 11058 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019625955489413518, - "loss": 1.604, - "step": 11059 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962592166677941, - "loss": 1.6422, - "step": 11060 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019625887844145303, - "loss": 1.5894, - "step": 11061 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019625854021511197, - "loss": 1.6065, - "step": 11062 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962582019887709, - "loss": 1.7297, - "step": 11063 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019625786376242983, - "loss": 1.7213, - "step": 11064 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019625752553608877, - "loss": 1.5084, - "step": 11065 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019625718730974768, - "loss": 1.5877, - "step": 11066 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019625684908340662, - "loss": 1.637, - "step": 11067 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019625651085706557, - "loss": 1.6563, - "step": 11068 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019625617263072448, - "loss": 1.6978, - "step": 11069 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019625583440438342, - "loss": 1.6268, - "step": 11070 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019625549617804236, - "loss": 1.6846, - "step": 11071 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019625515795170128, - "loss": 1.6189, - "step": 11072 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019625481972536022, - "loss": 1.7578, - "step": 11073 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019625448149901916, - "loss": 1.6691, - "step": 11074 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019625414327267807, - "loss": 1.6722, - "step": 11075 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019625380504633701, - "loss": 1.5295, - "step": 11076 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019625346681999595, - "loss": 1.643, - "step": 11077 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019625312859365487, - "loss": 1.6429, - "step": 11078 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962527903673138, - "loss": 1.6161, - "step": 11079 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019625245214097275, - "loss": 1.6441, - "step": 11080 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019625211391463167, - "loss": 1.5219, - "step": 11081 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962517756882906, - "loss": 1.6054, - "step": 11082 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019625143746194955, - "loss": 1.6127, - "step": 11083 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019625109923560846, - "loss": 1.5855, - "step": 11084 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962507610092674, - "loss": 1.7641, - "step": 11085 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019625042278292634, - "loss": 1.6952, - "step": 11086 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019625008455658526, - "loss": 1.6207, - "step": 11087 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962497463302442, - "loss": 1.6729, - "step": 11088 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019624940810390314, - "loss": 1.6062, - "step": 11089 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019624906987756205, - "loss": 1.7145, - "step": 11090 - }, - { - "epoch": 0.04, - "learning_rate": 0.000196248731651221, - "loss": 1.6549, - "step": 11091 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019624839342487994, - "loss": 1.5532, - "step": 11092 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019624805519853888, - "loss": 1.6043, - "step": 11093 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019624771697219782, - "loss": 1.7515, - "step": 11094 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019624737874585676, - "loss": 1.6929, - "step": 11095 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019624704051951567, - "loss": 1.5312, - "step": 11096 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019624670229317461, - "loss": 1.554, - "step": 11097 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019624636406683356, - "loss": 1.5146, - "step": 11098 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019624602584049247, - "loss": 1.6533, - "step": 11099 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962456876141514, - "loss": 1.6521, - "step": 11100 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019624534938781035, - "loss": 1.477, - "step": 11101 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019624501116146927, - "loss": 1.5362, - "step": 11102 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962446729351282, - "loss": 1.6152, - "step": 11103 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019624433470878715, - "loss": 1.5969, - "step": 11104 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019624399648244606, - "loss": 1.7035, - "step": 11105 - }, - { - "epoch": 0.04, - "learning_rate": 0.000196243658256105, - "loss": 1.5885, - "step": 11106 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019624332002976394, - "loss": 1.6913, - "step": 11107 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019624298180342286, - "loss": 1.6949, - "step": 11108 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962426435770818, - "loss": 1.5594, - "step": 11109 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019624230535074074, - "loss": 1.6201, - "step": 11110 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019624196712439966, - "loss": 1.7411, - "step": 11111 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962416288980586, - "loss": 1.553, - "step": 11112 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019624129067171754, - "loss": 1.6949, - "step": 11113 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019624095244537645, - "loss": 1.662, - "step": 11114 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962406142190354, - "loss": 1.5353, - "step": 11115 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019624027599269433, - "loss": 1.7241, - "step": 11116 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019623993776635325, - "loss": 1.4899, - "step": 11117 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962395995400122, - "loss": 1.6783, - "step": 11118 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019623926131367113, - "loss": 1.7922, - "step": 11119 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019623892308733004, - "loss": 1.7259, - "step": 11120 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019623858486098899, - "loss": 1.4485, - "step": 11121 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019623824663464793, - "loss": 1.7403, - "step": 11122 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019623790840830684, - "loss": 1.7054, - "step": 11123 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019623757018196578, - "loss": 1.5645, - "step": 11124 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019623723195562472, - "loss": 1.5435, - "step": 11125 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019623689372928364, - "loss": 1.656, - "step": 11126 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019623655550294258, - "loss": 1.7917, - "step": 11127 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019623621727660152, - "loss": 1.6659, - "step": 11128 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019623587905026043, - "loss": 1.6267, - "step": 11129 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019623554082391937, - "loss": 1.7368, - "step": 11130 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019623520259757832, - "loss": 1.5419, - "step": 11131 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019623486437123723, - "loss": 1.7134, - "step": 11132 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019623452614489617, - "loss": 1.637, - "step": 11133 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962341879185551, - "loss": 1.5688, - "step": 11134 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019623384969221403, - "loss": 1.861, - "step": 11135 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019623351146587297, - "loss": 1.5478, - "step": 11136 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962331732395319, - "loss": 1.4875, - "step": 11137 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019623283501319082, - "loss": 1.4653, - "step": 11138 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019623249678684976, - "loss": 1.6212, - "step": 11139 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962321585605087, - "loss": 1.6936, - "step": 11140 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019623182033416762, - "loss": 1.6497, - "step": 11141 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019623148210782656, - "loss": 1.5839, - "step": 11142 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962311438814855, - "loss": 1.6395, - "step": 11143 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019623080565514441, - "loss": 1.7144, - "step": 11144 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019623046742880336, - "loss": 1.7021, - "step": 11145 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962301292024623, - "loss": 1.6943, - "step": 11146 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962297909761212, - "loss": 1.5826, - "step": 11147 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019622945274978015, - "loss": 1.4879, - "step": 11148 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962291145234391, - "loss": 1.7275, - "step": 11149 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019622877629709803, - "loss": 1.6692, - "step": 11150 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019622843807075698, - "loss": 1.6002, - "step": 11151 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962280998444159, - "loss": 1.7013, - "step": 11152 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019622776161807483, - "loss": 1.6177, - "step": 11153 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019622742339173377, - "loss": 1.6182, - "step": 11154 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019622708516539269, - "loss": 1.6047, - "step": 11155 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019622674693905163, - "loss": 1.5632, - "step": 11156 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019622640871271057, - "loss": 1.5899, - "step": 11157 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019622607048636948, - "loss": 1.5394, - "step": 11158 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019622573226002842, - "loss": 1.6267, - "step": 11159 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019622539403368736, - "loss": 1.6983, - "step": 11160 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019622505580734628, - "loss": 1.535, - "step": 11161 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019622471758100522, - "loss": 1.675, - "step": 11162 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019622437935466416, - "loss": 1.5467, - "step": 11163 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019622404112832307, - "loss": 1.5841, - "step": 11164 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019622370290198202, - "loss": 1.5664, - "step": 11165 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019622336467564096, - "loss": 1.5669, - "step": 11166 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962230264492999, - "loss": 1.5863, - "step": 11167 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962226882229588, - "loss": 1.5836, - "step": 11168 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019622234999661775, - "loss": 1.5193, - "step": 11169 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962220117702767, - "loss": 1.8351, - "step": 11170 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962216735439356, - "loss": 1.6318, - "step": 11171 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019622133531759455, - "loss": 1.6512, - "step": 11172 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962209970912535, - "loss": 1.7894, - "step": 11173 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962206588649124, - "loss": 1.7488, - "step": 11174 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019622032063857135, - "loss": 1.5122, - "step": 11175 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962199824122303, - "loss": 1.6035, - "step": 11176 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962196441858892, - "loss": 1.6314, - "step": 11177 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019621930595954814, - "loss": 1.7556, - "step": 11178 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019621896773320708, - "loss": 1.4862, - "step": 11179 - }, - { - "epoch": 0.04, - "learning_rate": 0.000196218629506866, - "loss": 1.6454, - "step": 11180 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019621829128052494, - "loss": 1.5497, - "step": 11181 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019621795305418388, - "loss": 1.7527, - "step": 11182 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962176148278428, - "loss": 1.5744, - "step": 11183 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019621727660150174, - "loss": 1.8419, - "step": 11184 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019621693837516068, - "loss": 1.7196, - "step": 11185 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962166001488196, - "loss": 1.6696, - "step": 11186 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019621626192247853, - "loss": 1.5333, - "step": 11187 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019621592369613747, - "loss": 1.7308, - "step": 11188 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962155854697964, - "loss": 1.466, - "step": 11189 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019621524724345533, - "loss": 1.6039, - "step": 11190 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019621490901711427, - "loss": 1.7458, - "step": 11191 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019621457079077318, - "loss": 1.5492, - "step": 11192 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019621423256443212, - "loss": 1.6181, - "step": 11193 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019621389433809107, - "loss": 1.6683, - "step": 11194 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019621355611174998, - "loss": 1.6704, - "step": 11195 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019621321788540892, - "loss": 1.6259, - "step": 11196 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019621287965906786, - "loss": 1.5216, - "step": 11197 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019621254143272678, - "loss": 1.7148, - "step": 11198 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019621220320638572, - "loss": 1.6189, - "step": 11199 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019621186498004466, - "loss": 1.6341, - "step": 11200 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019621152675370357, - "loss": 1.5797, - "step": 11201 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962111885273625, - "loss": 1.6669, - "step": 11202 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019621085030102145, - "loss": 1.6344, - "step": 11203 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019621051207468037, - "loss": 1.7624, - "step": 11204 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962101738483393, - "loss": 1.6003, - "step": 11205 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019620983562199825, - "loss": 1.4854, - "step": 11206 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962094973956572, - "loss": 1.5757, - "step": 11207 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019620915916931613, - "loss": 1.6797, - "step": 11208 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019620882094297505, - "loss": 1.7029, - "step": 11209 - }, - { - "epoch": 0.04, - "learning_rate": 0.000196208482716634, - "loss": 1.7611, - "step": 11210 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019620814449029293, - "loss": 1.5249, - "step": 11211 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019620780626395184, - "loss": 1.6189, - "step": 11212 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019620746803761078, - "loss": 1.5768, - "step": 11213 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019620712981126973, - "loss": 1.6186, - "step": 11214 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019620679158492864, - "loss": 1.6981, - "step": 11215 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019620645335858758, - "loss": 1.6221, - "step": 11216 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019620611513224652, - "loss": 1.6813, - "step": 11217 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019620577690590544, - "loss": 1.56, - "step": 11218 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019620543867956438, - "loss": 1.5445, - "step": 11219 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019620510045322332, - "loss": 1.6884, - "step": 11220 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019620476222688223, - "loss": 1.6736, - "step": 11221 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019620442400054117, - "loss": 1.6671, - "step": 11222 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019620408577420011, - "loss": 1.6642, - "step": 11223 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019620374754785903, - "loss": 1.6597, - "step": 11224 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019620340932151797, - "loss": 1.6582, - "step": 11225 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962030710951769, - "loss": 1.6653, - "step": 11226 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019620273286883582, - "loss": 1.6094, - "step": 11227 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019620239464249477, - "loss": 1.7367, - "step": 11228 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962020564161537, - "loss": 1.7374, - "step": 11229 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019620171818981262, - "loss": 1.6267, - "step": 11230 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019620137996347156, - "loss": 1.5976, - "step": 11231 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962010417371305, - "loss": 1.5829, - "step": 11232 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019620070351078942, - "loss": 1.7638, - "step": 11233 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019620036528444836, - "loss": 1.7502, - "step": 11234 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001962000270581073, - "loss": 1.5993, - "step": 11235 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961996888317662, - "loss": 1.7312, - "step": 11236 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019619935060542515, - "loss": 1.8142, - "step": 11237 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961990123790841, - "loss": 1.6072, - "step": 11238 - }, - { - "epoch": 0.04, - "learning_rate": 0.000196198674152743, - "loss": 1.6308, - "step": 11239 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019619833592640195, - "loss": 1.7031, - "step": 11240 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961979977000609, - "loss": 1.6155, - "step": 11241 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961976594737198, - "loss": 1.4829, - "step": 11242 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019619732124737875, - "loss": 1.5977, - "step": 11243 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961969830210377, - "loss": 1.5239, - "step": 11244 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961966447946966, - "loss": 1.602, - "step": 11245 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019619630656835554, - "loss": 1.6219, - "step": 11246 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019619596834201448, - "loss": 1.6701, - "step": 11247 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961956301156734, - "loss": 1.6714, - "step": 11248 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019619529188933234, - "loss": 1.5864, - "step": 11249 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019619495366299128, - "loss": 1.698, - "step": 11250 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019619461543665022, - "loss": 1.6643, - "step": 11251 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019619427721030914, - "loss": 1.7092, - "step": 11252 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019619393898396808, - "loss": 1.7728, - "step": 11253 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019619360075762702, - "loss": 1.6019, - "step": 11254 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019619326253128593, - "loss": 1.6148, - "step": 11255 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019619292430494487, - "loss": 1.7579, - "step": 11256 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019619258607860381, - "loss": 1.6706, - "step": 11257 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019619224785226273, - "loss": 1.7674, - "step": 11258 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019619190962592167, - "loss": 1.5219, - "step": 11259 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961915713995806, - "loss": 1.6733, - "step": 11260 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019619123317323953, - "loss": 1.5569, - "step": 11261 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961908949468985, - "loss": 1.5164, - "step": 11262 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961905567205574, - "loss": 1.514, - "step": 11263 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019619021849421635, - "loss": 1.6082, - "step": 11264 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961898802678753, - "loss": 1.7625, - "step": 11265 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961895420415342, - "loss": 1.5463, - "step": 11266 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019618920381519314, - "loss": 1.5805, - "step": 11267 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019618886558885209, - "loss": 1.7686, - "step": 11268 - }, - { - "epoch": 0.04, - "learning_rate": 0.000196188527362511, - "loss": 1.656, - "step": 11269 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019618818913616994, - "loss": 1.675, - "step": 11270 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019618785090982888, - "loss": 1.599, - "step": 11271 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961875126834878, - "loss": 1.7787, - "step": 11272 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019618717445714674, - "loss": 1.6419, - "step": 11273 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019618683623080568, - "loss": 1.8249, - "step": 11274 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961864980044646, - "loss": 1.58, - "step": 11275 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019618615977812353, - "loss": 1.5044, - "step": 11276 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019618582155178247, - "loss": 1.7406, - "step": 11277 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961854833254414, - "loss": 1.5687, - "step": 11278 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019618514509910033, - "loss": 1.5477, - "step": 11279 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019618480687275927, - "loss": 1.6167, - "step": 11280 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019618446864641819, - "loss": 1.6013, - "step": 11281 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019618413042007713, - "loss": 1.8128, - "step": 11282 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019618379219373607, - "loss": 1.6446, - "step": 11283 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019618345396739498, - "loss": 1.6822, - "step": 11284 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019618311574105392, - "loss": 1.6262, - "step": 11285 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019618277751471286, - "loss": 1.5986, - "step": 11286 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019618243928837178, - "loss": 1.6942, - "step": 11287 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019618210106203072, - "loss": 1.7292, - "step": 11288 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019618176283568966, - "loss": 1.6636, - "step": 11289 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019618142460934857, - "loss": 1.9092, - "step": 11290 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019618108638300752, - "loss": 1.55, - "step": 11291 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019618074815666646, - "loss": 1.5705, - "step": 11292 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019618040993032537, - "loss": 1.6834, - "step": 11293 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961800717039843, - "loss": 1.6533, - "step": 11294 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019617973347764325, - "loss": 1.5942, - "step": 11295 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019617939525130217, - "loss": 1.5809, - "step": 11296 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961790570249611, - "loss": 1.6052, - "step": 11297 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019617871879862005, - "loss": 1.5902, - "step": 11298 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019617838057227896, - "loss": 1.5696, - "step": 11299 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961780423459379, - "loss": 1.7795, - "step": 11300 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019617770411959685, - "loss": 1.5603, - "step": 11301 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019617736589325576, - "loss": 1.6849, - "step": 11302 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961770276669147, - "loss": 1.4708, - "step": 11303 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019617668944057364, - "loss": 1.6039, - "step": 11304 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019617635121423256, - "loss": 1.5642, - "step": 11305 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961760129878915, - "loss": 1.6659, - "step": 11306 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019617567476155044, - "loss": 1.6234, - "step": 11307 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019617533653520935, - "loss": 1.5386, - "step": 11308 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961749983088683, - "loss": 1.666, - "step": 11309 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019617466008252723, - "loss": 1.5801, - "step": 11310 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019617432185618615, - "loss": 1.5895, - "step": 11311 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961739836298451, - "loss": 1.6939, - "step": 11312 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019617364540350403, - "loss": 1.5846, - "step": 11313 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019617330717716294, - "loss": 1.6709, - "step": 11314 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019617296895082189, - "loss": 1.5462, - "step": 11315 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019617263072448083, - "loss": 1.653, - "step": 11316 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019617229249813974, - "loss": 1.6802, - "step": 11317 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961719542717987, - "loss": 1.761, - "step": 11318 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019617161604545765, - "loss": 1.593, - "step": 11319 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019617127781911656, - "loss": 1.6418, - "step": 11320 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961709395927755, - "loss": 1.398, - "step": 11321 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019617060136643445, - "loss": 1.6223, - "step": 11322 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019617026314009336, - "loss": 1.5911, - "step": 11323 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961699249137523, - "loss": 1.6637, - "step": 11324 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019616958668741124, - "loss": 1.5179, - "step": 11325 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019616924846107016, - "loss": 1.6645, - "step": 11326 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961689102347291, - "loss": 1.6686, - "step": 11327 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019616857200838804, - "loss": 1.48, - "step": 11328 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019616823378204695, - "loss": 1.7625, - "step": 11329 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961678955557059, - "loss": 1.6635, - "step": 11330 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019616755732936484, - "loss": 1.5418, - "step": 11331 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019616721910302375, - "loss": 1.6426, - "step": 11332 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961668808766827, - "loss": 1.6699, - "step": 11333 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019616654265034163, - "loss": 1.7181, - "step": 11334 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019616620442400055, - "loss": 1.654, - "step": 11335 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961658661976595, - "loss": 1.5973, - "step": 11336 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019616552797131843, - "loss": 1.689, - "step": 11337 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019616518974497734, - "loss": 1.7449, - "step": 11338 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019616485151863628, - "loss": 1.5965, - "step": 11339 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019616451329229522, - "loss": 1.5885, - "step": 11340 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019616417506595414, - "loss": 1.5287, - "step": 11341 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019616383683961308, - "loss": 1.695, - "step": 11342 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019616349861327202, - "loss": 1.6827, - "step": 11343 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019616316038693094, - "loss": 1.6702, - "step": 11344 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019616282216058988, - "loss": 1.7144, - "step": 11345 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019616248393424882, - "loss": 1.5774, - "step": 11346 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019616214570790773, - "loss": 1.6949, - "step": 11347 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019616180748156667, - "loss": 1.6469, - "step": 11348 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019616146925522561, - "loss": 1.4583, - "step": 11349 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019616113102888453, - "loss": 1.477, - "step": 11350 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019616079280254347, - "loss": 1.6393, - "step": 11351 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961604545762024, - "loss": 1.6252, - "step": 11352 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019616011634986132, - "loss": 1.6046, - "step": 11353 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019615977812352027, - "loss": 1.6958, - "step": 11354 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961594398971792, - "loss": 1.5989, - "step": 11355 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019615910167083812, - "loss": 1.4867, - "step": 11356 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019615876344449706, - "loss": 1.5716, - "step": 11357 - }, - { - "epoch": 0.04, - "learning_rate": 0.000196158425218156, - "loss": 1.4747, - "step": 11358 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019615808699181492, - "loss": 1.599, - "step": 11359 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019615774876547386, - "loss": 1.8527, - "step": 11360 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961574105391328, - "loss": 1.576, - "step": 11361 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961570723127917, - "loss": 1.5098, - "step": 11362 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019615673408645065, - "loss": 1.8598, - "step": 11363 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961563958601096, - "loss": 1.7163, - "step": 11364 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961560576337685, - "loss": 1.7103, - "step": 11365 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019615571940742745, - "loss": 1.5822, - "step": 11366 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961553811810864, - "loss": 1.6081, - "step": 11367 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961550429547453, - "loss": 1.6747, - "step": 11368 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019615470472840425, - "loss": 1.4726, - "step": 11369 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961543665020632, - "loss": 1.4809, - "step": 11370 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961540282757221, - "loss": 1.5442, - "step": 11371 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019615369004938104, - "loss": 1.5958, - "step": 11372 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019615335182303998, - "loss": 1.6574, - "step": 11373 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019615301359669893, - "loss": 1.504, - "step": 11374 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019615267537035787, - "loss": 1.6213, - "step": 11375 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961523371440168, - "loss": 1.7338, - "step": 11376 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019615199891767572, - "loss": 1.6827, - "step": 11377 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019615166069133466, - "loss": 1.551, - "step": 11378 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961513224649936, - "loss": 1.6982, - "step": 11379 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019615098423865252, - "loss": 1.5319, - "step": 11380 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019615064601231146, - "loss": 1.5897, - "step": 11381 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961503077859704, - "loss": 1.5106, - "step": 11382 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019614996955962931, - "loss": 1.6088, - "step": 11383 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019614963133328826, - "loss": 1.6078, - "step": 11384 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961492931069472, - "loss": 1.6511, - "step": 11385 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961489548806061, - "loss": 1.5787, - "step": 11386 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019614861665426505, - "loss": 1.694, - "step": 11387 - }, - { - "epoch": 0.04, - "learning_rate": 0.000196148278427924, - "loss": 1.7307, - "step": 11388 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961479402015829, - "loss": 1.7051, - "step": 11389 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019614760197524185, - "loss": 1.6014, - "step": 11390 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961472637489008, - "loss": 1.5888, - "step": 11391 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961469255225597, - "loss": 1.6256, - "step": 11392 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019614658729621864, - "loss": 1.7009, - "step": 11393 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019614624906987759, - "loss": 1.5685, - "step": 11394 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961459108435365, - "loss": 1.6528, - "step": 11395 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019614557261719544, - "loss": 1.7041, - "step": 11396 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019614523439085438, - "loss": 1.7164, - "step": 11397 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961448961645133, - "loss": 1.762, - "step": 11398 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019614455793817224, - "loss": 1.6028, - "step": 11399 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019614421971183118, - "loss": 1.5624, - "step": 11400 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961438814854901, - "loss": 1.7039, - "step": 11401 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019614354325914903, - "loss": 1.5891, - "step": 11402 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019614320503280797, - "loss": 1.6053, - "step": 11403 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961428668064669, - "loss": 1.6315, - "step": 11404 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019614252858012583, - "loss": 1.5467, - "step": 11405 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019614219035378477, - "loss": 1.6545, - "step": 11406 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019614185212744368, - "loss": 1.6504, - "step": 11407 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019614151390110263, - "loss": 1.6, - "step": 11408 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019614117567476157, - "loss": 1.6006, - "step": 11409 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019614083744842048, - "loss": 1.5376, - "step": 11410 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019614049922207942, - "loss": 1.5563, - "step": 11411 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019614016099573836, - "loss": 1.4837, - "step": 11412 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019613982276939728, - "loss": 1.6134, - "step": 11413 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019613948454305622, - "loss": 1.7083, - "step": 11414 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019613914631671516, - "loss": 1.5436, - "step": 11415 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019613880809037407, - "loss": 1.7526, - "step": 11416 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019613846986403301, - "loss": 1.6667, - "step": 11417 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019613813163769196, - "loss": 1.5809, - "step": 11418 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019613779341135087, - "loss": 1.5764, - "step": 11419 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961374551850098, - "loss": 1.7221, - "step": 11420 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019613711695866875, - "loss": 1.5872, - "step": 11421 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019613677873232767, - "loss": 1.4559, - "step": 11422 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961364405059866, - "loss": 1.4664, - "step": 11423 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019613610227964555, - "loss": 1.6588, - "step": 11424 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019613576405330446, - "loss": 1.5835, - "step": 11425 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961354258269634, - "loss": 1.6268, - "step": 11426 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019613508760062234, - "loss": 1.7722, - "step": 11427 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019613474937428126, - "loss": 1.604, - "step": 11428 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961344111479402, - "loss": 1.5305, - "step": 11429 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019613407292159914, - "loss": 1.5951, - "step": 11430 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019613373469525808, - "loss": 1.6106, - "step": 11431 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019613339646891702, - "loss": 1.7631, - "step": 11432 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019613305824257596, - "loss": 1.6821, - "step": 11433 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019613272001623488, - "loss": 1.7311, - "step": 11434 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019613238178989382, - "loss": 1.643, - "step": 11435 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019613204356355276, - "loss": 1.7355, - "step": 11436 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019613170533721167, - "loss": 1.622, - "step": 11437 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019613136711087062, - "loss": 1.6266, - "step": 11438 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019613102888452956, - "loss": 1.6707, - "step": 11439 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019613069065818847, - "loss": 1.7594, - "step": 11440 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961303524318474, - "loss": 1.6742, - "step": 11441 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019613001420550635, - "loss": 1.5047, - "step": 11442 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019612967597916527, - "loss": 1.6636, - "step": 11443 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961293377528242, - "loss": 1.7686, - "step": 11444 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019612899952648315, - "loss": 1.2244, - "step": 11445 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019612866130014206, - "loss": 1.6028, - "step": 11446 - }, - { - "epoch": 0.04, - "learning_rate": 0.000196128323073801, - "loss": 1.6286, - "step": 11447 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019612798484745995, - "loss": 1.5131, - "step": 11448 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019612764662111886, - "loss": 1.6153, - "step": 11449 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961273083947778, - "loss": 1.6546, - "step": 11450 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019612697016843674, - "loss": 1.8143, - "step": 11451 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019612663194209566, - "loss": 1.6445, - "step": 11452 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961262937157546, - "loss": 1.6858, - "step": 11453 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019612595548941354, - "loss": 1.7154, - "step": 11454 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019612561726307245, - "loss": 1.7329, - "step": 11455 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961252790367314, - "loss": 1.704, - "step": 11456 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019612494081039034, - "loss": 1.516, - "step": 11457 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019612460258404925, - "loss": 1.6453, - "step": 11458 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961242643577082, - "loss": 1.7296, - "step": 11459 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019612392613136713, - "loss": 1.7622, - "step": 11460 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019612358790502605, - "loss": 1.7668, - "step": 11461 - }, - { - "epoch": 0.04, - "learning_rate": 0.000196123249678685, - "loss": 1.5567, - "step": 11462 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019612291145234393, - "loss": 1.6844, - "step": 11463 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019612257322600284, - "loss": 1.5494, - "step": 11464 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019612223499966178, - "loss": 1.5664, - "step": 11465 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019612189677332072, - "loss": 1.6274, - "step": 11466 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019612155854697964, - "loss": 1.5422, - "step": 11467 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019612122032063858, - "loss": 1.5813, - "step": 11468 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019612088209429752, - "loss": 1.6561, - "step": 11469 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019612054386795643, - "loss": 1.6508, - "step": 11470 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019612020564161538, - "loss": 1.6436, - "step": 11471 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019611986741527432, - "loss": 1.7482, - "step": 11472 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019611952918893323, - "loss": 1.4661, - "step": 11473 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019611919096259217, - "loss": 1.655, - "step": 11474 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961188527362511, - "loss": 1.6166, - "step": 11475 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019611851450991003, - "loss": 1.6668, - "step": 11476 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019611817628356897, - "loss": 1.7939, - "step": 11477 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961178380572279, - "loss": 1.7413, - "step": 11478 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019611749983088682, - "loss": 1.5636, - "step": 11479 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019611716160454576, - "loss": 1.5944, - "step": 11480 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961168233782047, - "loss": 1.5581, - "step": 11481 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019611648515186362, - "loss": 1.6992, - "step": 11482 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019611614692552256, - "loss": 1.5654, - "step": 11483 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961158086991815, - "loss": 1.7637, - "step": 11484 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019611547047284042, - "loss": 1.692, - "step": 11485 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019611513224649936, - "loss": 1.6351, - "step": 11486 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961147940201583, - "loss": 1.6698, - "step": 11487 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019611445579381724, - "loss": 1.6118, - "step": 11488 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019611411756747618, - "loss": 1.6356, - "step": 11489 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961137793411351, - "loss": 1.5894, - "step": 11490 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019611344111479404, - "loss": 1.6866, - "step": 11491 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019611310288845298, - "loss": 1.544, - "step": 11492 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961127646621119, - "loss": 1.7567, - "step": 11493 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019611242643577083, - "loss": 1.5431, - "step": 11494 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019611208820942977, - "loss": 1.5328, - "step": 11495 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961117499830887, - "loss": 1.7583, - "step": 11496 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019611141175674763, - "loss": 1.4284, - "step": 11497 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019611107353040657, - "loss": 1.66, - "step": 11498 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019611073530406548, - "loss": 1.4656, - "step": 11499 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019611039707772442, - "loss": 1.5104, - "step": 11500 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019611005885138337, - "loss": 1.497, - "step": 11501 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019610972062504228, - "loss": 1.7238, - "step": 11502 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019610938239870122, - "loss": 1.5413, - "step": 11503 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019610904417236016, - "loss": 1.525, - "step": 11504 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019610870594601908, - "loss": 1.634, - "step": 11505 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019610836771967802, - "loss": 1.6493, - "step": 11506 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019610802949333696, - "loss": 1.5873, - "step": 11507 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019610769126699587, - "loss": 1.6897, - "step": 11508 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019610735304065481, - "loss": 1.3822, - "step": 11509 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019610701481431375, - "loss": 1.5418, - "step": 11510 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019610667658797267, - "loss": 1.5673, - "step": 11511 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961063383616316, - "loss": 1.6812, - "step": 11512 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019610600013529055, - "loss": 1.5539, - "step": 11513 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019610566190894947, - "loss": 1.6339, - "step": 11514 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961053236826084, - "loss": 1.6236, - "step": 11515 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019610498545626735, - "loss": 1.499, - "step": 11516 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019610464722992626, - "loss": 1.623, - "step": 11517 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961043090035852, - "loss": 1.5887, - "step": 11518 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019610397077724414, - "loss": 1.6594, - "step": 11519 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019610363255090308, - "loss": 1.633, - "step": 11520 - }, - { - "epoch": 0.04, - "learning_rate": 0.000196103294324562, - "loss": 1.7576, - "step": 11521 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019610295609822094, - "loss": 1.5881, - "step": 11522 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019610261787187988, - "loss": 1.5531, - "step": 11523 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961022796455388, - "loss": 1.7162, - "step": 11524 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019610194141919774, - "loss": 1.737, - "step": 11525 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019610160319285668, - "loss": 1.5467, - "step": 11526 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961012649665156, - "loss": 1.464, - "step": 11527 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019610092674017453, - "loss": 1.6986, - "step": 11528 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019610058851383347, - "loss": 1.5614, - "step": 11529 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001961002502874924, - "loss": 1.6473, - "step": 11530 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019609991206115133, - "loss": 1.6767, - "step": 11531 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019609957383481027, - "loss": 1.4683, - "step": 11532 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019609923560846918, - "loss": 1.6764, - "step": 11533 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019609889738212813, - "loss": 1.7594, - "step": 11534 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019609855915578707, - "loss": 1.5723, - "step": 11535 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019609822092944598, - "loss": 1.6417, - "step": 11536 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019609788270310492, - "loss": 1.7255, - "step": 11537 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019609754447676386, - "loss": 1.6307, - "step": 11538 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019609720625042278, - "loss": 1.5146, - "step": 11539 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019609686802408172, - "loss": 1.5286, - "step": 11540 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019609652979774066, - "loss": 1.6686, - "step": 11541 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019609619157139957, - "loss": 1.6563, - "step": 11542 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019609585334505854, - "loss": 1.5454, - "step": 11543 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019609551511871746, - "loss": 1.7031, - "step": 11544 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960951768923764, - "loss": 1.637, - "step": 11545 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019609483866603534, - "loss": 1.7032, - "step": 11546 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019609450043969425, - "loss": 1.594, - "step": 11547 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960941622133532, - "loss": 1.6793, - "step": 11548 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019609382398701213, - "loss": 1.6451, - "step": 11549 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019609348576067105, - "loss": 1.6679, - "step": 11550 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019609314753433, - "loss": 1.7257, - "step": 11551 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019609280930798893, - "loss": 1.7026, - "step": 11552 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019609247108164784, - "loss": 1.6287, - "step": 11553 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019609213285530679, - "loss": 1.6982, - "step": 11554 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019609179462896573, - "loss": 1.5117, - "step": 11555 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019609145640262464, - "loss": 1.5411, - "step": 11556 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019609111817628358, - "loss": 1.5307, - "step": 11557 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019609077994994252, - "loss": 1.7376, - "step": 11558 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019609044172360144, - "loss": 1.5117, - "step": 11559 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019609010349726038, - "loss": 1.6212, - "step": 11560 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019608976527091932, - "loss": 1.6935, - "step": 11561 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019608942704457823, - "loss": 1.5875, - "step": 11562 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019608908881823717, - "loss": 1.7771, - "step": 11563 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019608875059189612, - "loss": 1.7997, - "step": 11564 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019608841236555503, - "loss": 1.5883, - "step": 11565 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019608807413921397, - "loss": 1.5069, - "step": 11566 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960877359128729, - "loss": 1.6895, - "step": 11567 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019608739768653183, - "loss": 1.616, - "step": 11568 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019608705946019077, - "loss": 1.9067, - "step": 11569 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960867212338497, - "loss": 1.6914, - "step": 11570 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019608638300750862, - "loss": 1.6307, - "step": 11571 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019608604478116756, - "loss": 1.7705, - "step": 11572 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960857065548265, - "loss": 1.6521, - "step": 11573 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019608536832848542, - "loss": 1.8308, - "step": 11574 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019608503010214436, - "loss": 1.6593, - "step": 11575 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960846918758033, - "loss": 1.7767, - "step": 11576 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019608435364946221, - "loss": 1.839, - "step": 11577 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019608401542312116, - "loss": 1.6842, - "step": 11578 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960836771967801, - "loss": 1.6178, - "step": 11579 - }, - { - "epoch": 0.04, - "learning_rate": 0.000196083338970439, - "loss": 1.6251, - "step": 11580 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019608300074409795, - "loss": 1.6383, - "step": 11581 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960826625177569, - "loss": 1.5845, - "step": 11582 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960823242914158, - "loss": 1.535, - "step": 11583 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019608198606507475, - "loss": 1.8282, - "step": 11584 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960816478387337, - "loss": 1.5364, - "step": 11585 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960813096123926, - "loss": 1.6046, - "step": 11586 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019608097138605154, - "loss": 1.5908, - "step": 11587 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019608063315971049, - "loss": 1.6195, - "step": 11588 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960802949333694, - "loss": 1.6263, - "step": 11589 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019607995670702834, - "loss": 1.5802, - "step": 11590 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019607961848068728, - "loss": 1.5234, - "step": 11591 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960792802543462, - "loss": 1.5382, - "step": 11592 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019607894202800514, - "loss": 1.7375, - "step": 11593 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019607860380166408, - "loss": 1.5324, - "step": 11594 - }, - { - "epoch": 0.04, - "learning_rate": 0.000196078265575323, - "loss": 1.7503, - "step": 11595 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019607792734898193, - "loss": 1.5756, - "step": 11596 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019607758912264087, - "loss": 1.5756, - "step": 11597 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960772508962998, - "loss": 1.4673, - "step": 11598 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019607691266995876, - "loss": 1.6978, - "step": 11599 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960765744436177, - "loss": 1.5559, - "step": 11600 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960762362172766, - "loss": 1.6404, - "step": 11601 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019607589799093555, - "loss": 1.5251, - "step": 11602 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960755597645945, - "loss": 1.7355, - "step": 11603 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960752215382534, - "loss": 1.6904, - "step": 11604 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019607488331191235, - "loss": 1.6184, - "step": 11605 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960745450855713, - "loss": 1.5526, - "step": 11606 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960742068592302, - "loss": 1.6157, - "step": 11607 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019607386863288915, - "loss": 1.5811, - "step": 11608 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960735304065481, - "loss": 1.6433, - "step": 11609 - }, - { - "epoch": 0.04, - "learning_rate": 0.000196073192180207, - "loss": 1.6762, - "step": 11610 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019607285395386594, - "loss": 1.549, - "step": 11611 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019607251572752488, - "loss": 1.6745, - "step": 11612 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960721775011838, - "loss": 1.7987, - "step": 11613 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019607183927484274, - "loss": 1.6449, - "step": 11614 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019607150104850168, - "loss": 1.5658, - "step": 11615 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960711628221606, - "loss": 1.6426, - "step": 11616 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019607082459581954, - "loss": 1.5178, - "step": 11617 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019607048636947848, - "loss": 1.4596, - "step": 11618 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960701481431374, - "loss": 1.5618, - "step": 11619 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019606980991679633, - "loss": 1.644, - "step": 11620 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019606947169045527, - "loss": 1.5586, - "step": 11621 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960691334641142, - "loss": 1.8246, - "step": 11622 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019606879523777313, - "loss": 1.5365, - "step": 11623 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019606845701143207, - "loss": 1.7827, - "step": 11624 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019606811878509098, - "loss": 1.4981, - "step": 11625 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019606778055874992, - "loss": 1.5745, - "step": 11626 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019606744233240887, - "loss": 1.6403, - "step": 11627 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019606710410606778, - "loss": 1.4786, - "step": 11628 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019606676587972672, - "loss": 1.6714, - "step": 11629 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019606642765338566, - "loss": 1.6776, - "step": 11630 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019606608942704458, - "loss": 1.7713, - "step": 11631 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019606575120070352, - "loss": 1.7778, - "step": 11632 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019606541297436246, - "loss": 1.8782, - "step": 11633 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019606507474802137, - "loss": 1.7065, - "step": 11634 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960647365216803, - "loss": 1.6381, - "step": 11635 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019606439829533925, - "loss": 1.789, - "step": 11636 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019606406006899817, - "loss": 1.6085, - "step": 11637 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960637218426571, - "loss": 1.7267, - "step": 11638 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019606338361631605, - "loss": 1.7171, - "step": 11639 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019606304538997496, - "loss": 1.5302, - "step": 11640 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960627071636339, - "loss": 1.6315, - "step": 11641 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019606236893729285, - "loss": 1.6757, - "step": 11642 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019606203071095176, - "loss": 1.6771, - "step": 11643 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960616924846107, - "loss": 1.6694, - "step": 11644 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019606135425826964, - "loss": 1.8149, - "step": 11645 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019606101603192856, - "loss": 1.6139, - "step": 11646 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960606778055875, - "loss": 1.6308, - "step": 11647 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019606033957924644, - "loss": 1.6351, - "step": 11648 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019606000135290535, - "loss": 1.6675, - "step": 11649 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960596631265643, - "loss": 1.6548, - "step": 11650 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019605932490022324, - "loss": 1.6457, - "step": 11651 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019605898667388215, - "loss": 1.7227, - "step": 11652 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960586484475411, - "loss": 1.6197, - "step": 11653 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019605831022120003, - "loss": 1.6567, - "step": 11654 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019605797199485897, - "loss": 1.7914, - "step": 11655 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019605763376851791, - "loss": 1.5589, - "step": 11656 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019605729554217686, - "loss": 1.7617, - "step": 11657 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019605695731583577, - "loss": 1.6866, - "step": 11658 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960566190894947, - "loss": 1.7743, - "step": 11659 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019605628086315365, - "loss": 1.6694, - "step": 11660 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019605594263681257, - "loss": 1.5956, - "step": 11661 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960556044104715, - "loss": 1.6795, - "step": 11662 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019605526618413045, - "loss": 1.8568, - "step": 11663 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019605492795778936, - "loss": 1.5721, - "step": 11664 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960545897314483, - "loss": 1.8165, - "step": 11665 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019605425150510724, - "loss": 1.7141, - "step": 11666 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019605391327876616, - "loss": 1.6147, - "step": 11667 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960535750524251, - "loss": 1.5623, - "step": 11668 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019605323682608404, - "loss": 1.6472, - "step": 11669 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019605289859974295, - "loss": 1.5971, - "step": 11670 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960525603734019, - "loss": 1.687, - "step": 11671 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019605222214706084, - "loss": 1.6109, - "step": 11672 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019605188392071975, - "loss": 1.6462, - "step": 11673 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960515456943787, - "loss": 1.6503, - "step": 11674 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019605120746803763, - "loss": 1.6496, - "step": 11675 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019605086924169655, - "loss": 1.6167, - "step": 11676 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960505310153555, - "loss": 1.7372, - "step": 11677 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019605019278901443, - "loss": 1.6467, - "step": 11678 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019604985456267334, - "loss": 1.5864, - "step": 11679 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019604951633633228, - "loss": 1.6496, - "step": 11680 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019604917810999123, - "loss": 1.6813, - "step": 11681 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019604883988365014, - "loss": 1.7377, - "step": 11682 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019604850165730908, - "loss": 1.5642, - "step": 11683 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019604816343096802, - "loss": 1.6056, - "step": 11684 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019604782520462694, - "loss": 1.5624, - "step": 11685 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019604748697828588, - "loss": 1.656, - "step": 11686 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019604714875194482, - "loss": 1.4415, - "step": 11687 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019604681052560373, - "loss": 1.6233, - "step": 11688 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019604647229926267, - "loss": 1.6937, - "step": 11689 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019604613407292161, - "loss": 1.6933, - "step": 11690 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019604579584658053, - "loss": 1.6374, - "step": 11691 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019604545762023947, - "loss": 1.6494, - "step": 11692 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960451193938984, - "loss": 1.7066, - "step": 11693 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019604478116755733, - "loss": 1.445, - "step": 11694 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019604444294121627, - "loss": 1.7319, - "step": 11695 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960441047148752, - "loss": 1.5632, - "step": 11696 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019604376648853412, - "loss": 1.6068, - "step": 11697 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019604342826219306, - "loss": 1.532, - "step": 11698 - }, - { - "epoch": 0.04, - "learning_rate": 0.000196043090035852, - "loss": 1.7291, - "step": 11699 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019604275180951092, - "loss": 1.7522, - "step": 11700 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019604241358316986, - "loss": 1.815, - "step": 11701 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960420753568288, - "loss": 1.7072, - "step": 11702 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019604173713048771, - "loss": 1.7142, - "step": 11703 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019604139890414666, - "loss": 1.5571, - "step": 11704 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960410606778056, - "loss": 1.7367, - "step": 11705 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960407224514645, - "loss": 1.5116, - "step": 11706 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019604038422512345, - "loss": 1.6842, - "step": 11707 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960400459987824, - "loss": 1.7458, - "step": 11708 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960397077724413, - "loss": 1.6178, - "step": 11709 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019603936954610025, - "loss": 1.611, - "step": 11710 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019603903131975922, - "loss": 1.623, - "step": 11711 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019603869309341813, - "loss": 1.5672, - "step": 11712 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019603835486707707, - "loss": 1.516, - "step": 11713 - }, - { - "epoch": 0.04, - "learning_rate": 0.000196038016640736, - "loss": 1.6023, - "step": 11714 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019603767841439493, - "loss": 1.5757, - "step": 11715 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019603734018805387, - "loss": 1.5547, - "step": 11716 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960370019617128, - "loss": 1.6643, - "step": 11717 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019603666373537172, - "loss": 1.5279, - "step": 11718 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019603632550903066, - "loss": 1.7511, - "step": 11719 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960359872826896, - "loss": 1.741, - "step": 11720 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019603564905634852, - "loss": 1.5972, - "step": 11721 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019603531083000746, - "loss": 1.6209, - "step": 11722 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960349726036664, - "loss": 1.469, - "step": 11723 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019603463437732532, - "loss": 1.3623, - "step": 11724 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019603429615098426, - "loss": 1.6193, - "step": 11725 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960339579246432, - "loss": 1.7317, - "step": 11726 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960336196983021, - "loss": 1.6974, - "step": 11727 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019603328147196105, - "loss": 1.6497, - "step": 11728 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019603294324562, - "loss": 1.631, - "step": 11729 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960326050192789, - "loss": 1.6758, - "step": 11730 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019603226679293785, - "loss": 1.6478, - "step": 11731 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960319285665968, - "loss": 1.5285, - "step": 11732 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960315903402557, - "loss": 1.5947, - "step": 11733 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019603125211391465, - "loss": 1.6096, - "step": 11734 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960309138875736, - "loss": 1.5161, - "step": 11735 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960305756612325, - "loss": 1.7454, - "step": 11736 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019603023743489144, - "loss": 1.4811, - "step": 11737 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019602989920855038, - "loss": 1.6422, - "step": 11738 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960295609822093, - "loss": 1.6657, - "step": 11739 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019602922275586824, - "loss": 1.6094, - "step": 11740 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019602888452952718, - "loss": 1.6, - "step": 11741 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960285463031861, - "loss": 1.7354, - "step": 11742 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019602820807684503, - "loss": 1.5376, - "step": 11743 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019602786985050398, - "loss": 1.5251, - "step": 11744 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960275316241629, - "loss": 1.7699, - "step": 11745 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019602719339782183, - "loss": 1.5953, - "step": 11746 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019602685517148077, - "loss": 1.7841, - "step": 11747 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019602651694513969, - "loss": 1.5384, - "step": 11748 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019602617871879863, - "loss": 1.5498, - "step": 11749 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019602584049245757, - "loss": 1.6727, - "step": 11750 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019602550226611648, - "loss": 1.6752, - "step": 11751 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019602516403977542, - "loss": 1.7499, - "step": 11752 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019602482581343436, - "loss": 1.5947, - "step": 11753 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019602448758709328, - "loss": 1.5747, - "step": 11754 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019602414936075222, - "loss": 1.6851, - "step": 11755 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019602381113441116, - "loss": 1.5906, - "step": 11756 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019602347290807007, - "loss": 1.6583, - "step": 11757 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019602313468172902, - "loss": 1.4416, - "step": 11758 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019602279645538796, - "loss": 1.5784, - "step": 11759 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019602245822904687, - "loss": 1.7791, - "step": 11760 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960221200027058, - "loss": 1.4639, - "step": 11761 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019602178177636475, - "loss": 1.6638, - "step": 11762 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019602144355002367, - "loss": 1.5847, - "step": 11763 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960211053236826, - "loss": 1.6804, - "step": 11764 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019602076709734155, - "loss": 1.6961, - "step": 11765 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019602042887100046, - "loss": 1.6183, - "step": 11766 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019602009064465943, - "loss": 1.6511, - "step": 11767 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019601975241831835, - "loss": 1.5927, - "step": 11768 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960194141919773, - "loss": 1.5709, - "step": 11769 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019601907596563623, - "loss": 1.5488, - "step": 11770 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019601873773929514, - "loss": 1.6308, - "step": 11771 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019601839951295408, - "loss": 1.8427, - "step": 11772 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019601806128661302, - "loss": 1.558, - "step": 11773 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019601772306027194, - "loss": 1.6964, - "step": 11774 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019601738483393088, - "loss": 1.5982, - "step": 11775 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019601704660758982, - "loss": 1.6695, - "step": 11776 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019601670838124874, - "loss": 1.6298, - "step": 11777 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019601637015490768, - "loss": 1.5917, - "step": 11778 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019601603192856662, - "loss": 1.6011, - "step": 11779 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019601569370222553, - "loss": 1.5596, - "step": 11780 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019601535547588447, - "loss": 1.6378, - "step": 11781 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019601501724954341, - "loss": 1.6532, - "step": 11782 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019601467902320233, - "loss": 1.5695, - "step": 11783 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019601434079686127, - "loss": 1.6402, - "step": 11784 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960140025705202, - "loss": 1.5738, - "step": 11785 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019601366434417915, - "loss": 1.8209, - "step": 11786 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019601332611783807, - "loss": 1.6393, - "step": 11787 - }, - { - "epoch": 0.04, - "learning_rate": 0.000196012987891497, - "loss": 1.7059, - "step": 11788 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019601264966515595, - "loss": 1.7937, - "step": 11789 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019601231143881486, - "loss": 1.6731, - "step": 11790 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960119732124738, - "loss": 1.5394, - "step": 11791 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019601163498613274, - "loss": 1.6265, - "step": 11792 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019601129675979166, - "loss": 1.4819, - "step": 11793 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960109585334506, - "loss": 1.5348, - "step": 11794 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019601062030710954, - "loss": 1.5873, - "step": 11795 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019601028208076845, - "loss": 1.6486, - "step": 11796 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960099438544274, - "loss": 1.6345, - "step": 11797 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019600960562808634, - "loss": 1.5997, - "step": 11798 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019600926740174525, - "loss": 1.6403, - "step": 11799 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960089291754042, - "loss": 1.5068, - "step": 11800 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019600859094906313, - "loss": 1.5888, - "step": 11801 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019600825272272205, - "loss": 1.5112, - "step": 11802 - }, - { - "epoch": 0.04, - "learning_rate": 0.000196007914496381, - "loss": 1.6288, - "step": 11803 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019600757627003993, - "loss": 1.6497, - "step": 11804 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019600723804369884, - "loss": 1.561, - "step": 11805 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019600689981735778, - "loss": 1.6482, - "step": 11806 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019600656159101673, - "loss": 1.7264, - "step": 11807 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019600622336467564, - "loss": 1.7286, - "step": 11808 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019600588513833458, - "loss": 1.556, - "step": 11809 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019600554691199352, - "loss": 1.662, - "step": 11810 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019600520868565244, - "loss": 1.7372, - "step": 11811 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019600487045931138, - "loss": 1.645, - "step": 11812 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019600453223297032, - "loss": 1.4752, - "step": 11813 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019600419400662923, - "loss": 1.7728, - "step": 11814 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019600385578028817, - "loss": 1.5584, - "step": 11815 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019600351755394711, - "loss": 1.6113, - "step": 11816 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019600317932760603, - "loss": 1.6738, - "step": 11817 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019600284110126497, - "loss": 1.6095, - "step": 11818 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960025028749239, - "loss": 1.6954, - "step": 11819 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019600216464858282, - "loss": 1.7146, - "step": 11820 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019600182642224177, - "loss": 1.5181, - "step": 11821 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960014881959007, - "loss": 1.7261, - "step": 11822 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019600114996955962, - "loss": 1.6311, - "step": 11823 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960008117432186, - "loss": 1.7883, - "step": 11824 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001960004735168775, - "loss": 1.5922, - "step": 11825 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019600013529053644, - "loss": 1.4257, - "step": 11826 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019599979706419539, - "loss": 1.6808, - "step": 11827 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959994588378543, - "loss": 1.6398, - "step": 11828 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019599912061151324, - "loss": 1.6116, - "step": 11829 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019599878238517218, - "loss": 1.5551, - "step": 11830 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959984441588311, - "loss": 1.7096, - "step": 11831 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019599810593249004, - "loss": 1.8273, - "step": 11832 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019599776770614898, - "loss": 1.6272, - "step": 11833 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959974294798079, - "loss": 1.8711, - "step": 11834 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019599709125346683, - "loss": 1.6031, - "step": 11835 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019599675302712577, - "loss": 1.6506, - "step": 11836 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959964148007847, - "loss": 1.6476, - "step": 11837 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019599607657444363, - "loss": 1.5598, - "step": 11838 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019599573834810257, - "loss": 1.8772, - "step": 11839 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019599540012176148, - "loss": 1.5659, - "step": 11840 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019599506189542043, - "loss": 1.7398, - "step": 11841 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019599472366907937, - "loss": 1.5526, - "step": 11842 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019599438544273828, - "loss": 1.7425, - "step": 11843 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019599404721639722, - "loss": 1.6174, - "step": 11844 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019599370899005616, - "loss": 1.5899, - "step": 11845 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019599337076371508, - "loss": 1.5827, - "step": 11846 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019599303253737402, - "loss": 1.6065, - "step": 11847 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019599269431103296, - "loss": 1.6426, - "step": 11848 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019599235608469187, - "loss": 1.5689, - "step": 11849 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019599201785835081, - "loss": 1.575, - "step": 11850 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019599167963200976, - "loss": 1.6069, - "step": 11851 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019599134140566867, - "loss": 1.4761, - "step": 11852 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959910031793276, - "loss": 1.5601, - "step": 11853 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019599066495298655, - "loss": 1.6043, - "step": 11854 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019599032672664547, - "loss": 1.4364, - "step": 11855 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959899885003044, - "loss": 1.5662, - "step": 11856 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019598965027396335, - "loss": 1.9335, - "step": 11857 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019598931204762226, - "loss": 1.6071, - "step": 11858 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959889738212812, - "loss": 1.6274, - "step": 11859 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019598863559494014, - "loss": 1.6026, - "step": 11860 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019598829736859906, - "loss": 1.7348, - "step": 11861 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195987959142258, - "loss": 1.6778, - "step": 11862 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019598762091591694, - "loss": 1.5901, - "step": 11863 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019598728268957586, - "loss": 1.6056, - "step": 11864 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959869444632348, - "loss": 1.5891, - "step": 11865 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019598660623689374, - "loss": 1.6886, - "step": 11866 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019598626801055265, - "loss": 1.6234, - "step": 11867 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959859297842116, - "loss": 1.6928, - "step": 11868 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019598559155787053, - "loss": 1.4066, - "step": 11869 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019598525333152947, - "loss": 1.5179, - "step": 11870 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959849151051884, - "loss": 1.673, - "step": 11871 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019598457687884733, - "loss": 1.5686, - "step": 11872 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019598423865250627, - "loss": 1.478, - "step": 11873 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019598390042616519, - "loss": 1.8263, - "step": 11874 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019598356219982413, - "loss": 1.5738, - "step": 11875 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019598322397348307, - "loss": 1.7134, - "step": 11876 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019598288574714198, - "loss": 1.8217, - "step": 11877 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019598254752080092, - "loss": 1.6016, - "step": 11878 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019598220929445986, - "loss": 1.5672, - "step": 11879 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959818710681188, - "loss": 1.6541, - "step": 11880 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019598153284177775, - "loss": 1.6838, - "step": 11881 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019598119461543666, - "loss": 1.5971, - "step": 11882 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959808563890956, - "loss": 1.7584, - "step": 11883 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019598051816275454, - "loss": 1.6633, - "step": 11884 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019598017993641346, - "loss": 1.5124, - "step": 11885 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959798417100724, - "loss": 1.6459, - "step": 11886 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019597950348373134, - "loss": 1.5292, - "step": 11887 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019597916525739025, - "loss": 1.6594, - "step": 11888 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959788270310492, - "loss": 1.6963, - "step": 11889 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019597848880470814, - "loss": 1.591, - "step": 11890 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019597815057836705, - "loss": 1.6138, - "step": 11891 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195977812352026, - "loss": 1.7244, - "step": 11892 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019597747412568493, - "loss": 1.5228, - "step": 11893 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019597713589934385, - "loss": 1.8015, - "step": 11894 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959767976730028, - "loss": 1.6477, - "step": 11895 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019597645944666173, - "loss": 1.4209, - "step": 11896 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019597612122032064, - "loss": 1.5937, - "step": 11897 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019597578299397958, - "loss": 1.7276, - "step": 11898 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019597544476763852, - "loss": 1.7302, - "step": 11899 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019597510654129744, - "loss": 1.5293, - "step": 11900 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019597476831495638, - "loss": 1.6888, - "step": 11901 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019597443008861532, - "loss": 1.6055, - "step": 11902 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019597409186227423, - "loss": 1.5641, - "step": 11903 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019597375363593318, - "loss": 1.6342, - "step": 11904 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019597341540959212, - "loss": 1.806, - "step": 11905 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019597307718325103, - "loss": 1.5528, - "step": 11906 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019597273895690997, - "loss": 1.7376, - "step": 11907 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959724007305689, - "loss": 1.5819, - "step": 11908 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019597206250422783, - "loss": 1.6005, - "step": 11909 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019597172427788677, - "loss": 1.6503, - "step": 11910 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959713860515457, - "loss": 1.5531, - "step": 11911 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019597104782520462, - "loss": 1.7169, - "step": 11912 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019597070959886356, - "loss": 1.5307, - "step": 11913 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959703713725225, - "loss": 1.5654, - "step": 11914 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019597003314618142, - "loss": 1.5819, - "step": 11915 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019596969491984036, - "loss": 1.5604, - "step": 11916 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959693566934993, - "loss": 1.6832, - "step": 11917 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019596901846715822, - "loss": 1.6134, - "step": 11918 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019596868024081716, - "loss": 1.6209, - "step": 11919 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959683420144761, - "loss": 1.7827, - "step": 11920 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195968003788135, - "loss": 1.6143, - "step": 11921 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019596766556179395, - "loss": 1.664, - "step": 11922 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959673273354529, - "loss": 1.5661, - "step": 11923 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959669891091118, - "loss": 1.7455, - "step": 11924 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019596665088277075, - "loss": 1.6527, - "step": 11925 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959663126564297, - "loss": 1.6613, - "step": 11926 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959659744300886, - "loss": 1.8045, - "step": 11927 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019596563620374755, - "loss": 1.6879, - "step": 11928 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959652979774065, - "loss": 1.5598, - "step": 11929 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959649597510654, - "loss": 1.5374, - "step": 11930 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019596462152472434, - "loss": 1.7607, - "step": 11931 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019596428329838328, - "loss": 1.5649, - "step": 11932 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959639450720422, - "loss": 1.3974, - "step": 11933 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019596360684570114, - "loss": 1.5942, - "step": 11934 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019596326861936008, - "loss": 1.5919, - "step": 11935 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019596293039301902, - "loss": 1.5368, - "step": 11936 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019596259216667796, - "loss": 1.6779, - "step": 11937 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959622539403369, - "loss": 1.4451, - "step": 11938 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019596191571399582, - "loss": 1.6376, - "step": 11939 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019596157748765476, - "loss": 1.4165, - "step": 11940 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959612392613137, - "loss": 1.7023, - "step": 11941 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019596090103497261, - "loss": 1.6216, - "step": 11942 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019596056280863155, - "loss": 1.7008, - "step": 11943 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959602245822905, - "loss": 1.6205, - "step": 11944 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959598863559494, - "loss": 1.6446, - "step": 11945 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019595954812960835, - "loss": 1.7621, - "step": 11946 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959592099032673, - "loss": 1.9129, - "step": 11947 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959588716769262, - "loss": 1.5331, - "step": 11948 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019595853345058515, - "loss": 1.5415, - "step": 11949 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959581952242441, - "loss": 1.5988, - "step": 11950 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195957856997903, - "loss": 1.61, - "step": 11951 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019595751877156194, - "loss": 1.4497, - "step": 11952 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019595718054522088, - "loss": 1.5664, - "step": 11953 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959568423188798, - "loss": 1.7387, - "step": 11954 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019595650409253874, - "loss": 1.6987, - "step": 11955 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019595616586619768, - "loss": 1.5654, - "step": 11956 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959558276398566, - "loss": 1.5837, - "step": 11957 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019595548941351554, - "loss": 1.7804, - "step": 11958 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019595515118717448, - "loss": 1.5503, - "step": 11959 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959548129608334, - "loss": 1.6029, - "step": 11960 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019595447473449233, - "loss": 1.6825, - "step": 11961 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019595413650815127, - "loss": 1.666, - "step": 11962 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959537982818102, - "loss": 1.6104, - "step": 11963 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019595346005546913, - "loss": 1.5888, - "step": 11964 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019595312182912807, - "loss": 1.6717, - "step": 11965 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019595278360278698, - "loss": 1.5165, - "step": 11966 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019595244537644593, - "loss": 1.729, - "step": 11967 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019595210715010487, - "loss": 1.6439, - "step": 11968 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019595176892376378, - "loss": 1.5026, - "step": 11969 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019595143069742272, - "loss": 1.7503, - "step": 11970 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019595109247108166, - "loss": 1.5267, - "step": 11971 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019595075424474058, - "loss": 1.7242, - "step": 11972 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019595041601839952, - "loss": 1.4859, - "step": 11973 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019595007779205846, - "loss": 1.4333, - "step": 11974 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019594973956571737, - "loss": 1.6377, - "step": 11975 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019594940133937631, - "loss": 1.5921, - "step": 11976 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019594906311303526, - "loss": 1.8108, - "step": 11977 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019594872488669417, - "loss": 1.6995, - "step": 11978 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959483866603531, - "loss": 1.6605, - "step": 11979 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019594804843401205, - "loss": 1.7277, - "step": 11980 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019594771020767097, - "loss": 1.6446, - "step": 11981 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959473719813299, - "loss": 1.4985, - "step": 11982 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019594703375498885, - "loss": 1.6209, - "step": 11983 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019594669552864776, - "loss": 1.7971, - "step": 11984 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959463573023067, - "loss": 1.7685, - "step": 11985 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019594601907596564, - "loss": 1.686, - "step": 11986 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019594568084962456, - "loss": 1.6834, - "step": 11987 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959453426232835, - "loss": 1.6748, - "step": 11988 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019594500439694244, - "loss": 1.8685, - "step": 11989 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019594466617060135, - "loss": 1.7468, - "step": 11990 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959443279442603, - "loss": 1.8489, - "step": 11991 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019594398971791926, - "loss": 1.7509, - "step": 11992 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019594365149157818, - "loss": 1.5771, - "step": 11993 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019594331326523712, - "loss": 1.6322, - "step": 11994 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019594297503889606, - "loss": 1.7358, - "step": 11995 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019594263681255497, - "loss": 1.578, - "step": 11996 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019594229858621392, - "loss": 1.5698, - "step": 11997 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019594196035987286, - "loss": 1.5623, - "step": 11998 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019594162213353177, - "loss": 1.7086, - "step": 11999 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959412839071907, - "loss": 1.657, - "step": 12000 - }, - { - "epoch": 0.04, - "eval_loss": 1.2872971754807692, - "mrc_em": 21.0, - "mrc_f1": 29.066666666666674, - "nli_acc": 0.28, - "step": 12000, - "sts_f1": 0.3103448275862069, - "ynat_f1": 0.4891098484848485 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019594094568084965, - "loss": 1.797, - "step": 12001 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019594060745450857, - "loss": 1.6841, - "step": 12002 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959402692281675, - "loss": 1.6163, - "step": 12003 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019593993100182645, - "loss": 1.7141, - "step": 12004 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019593959277548536, - "loss": 1.5388, - "step": 12005 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959392545491443, - "loss": 1.6756, - "step": 12006 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019593891632280325, - "loss": 1.5465, - "step": 12007 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019593857809646216, - "loss": 1.6253, - "step": 12008 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959382398701211, - "loss": 1.538, - "step": 12009 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019593790164378004, - "loss": 1.69, - "step": 12010 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019593756341743896, - "loss": 1.7062, - "step": 12011 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959372251910979, - "loss": 1.6226, - "step": 12012 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019593688696475684, - "loss": 1.5991, - "step": 12013 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019593654873841575, - "loss": 1.4697, - "step": 12014 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959362105120747, - "loss": 1.6099, - "step": 12015 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019593587228573363, - "loss": 1.6798, - "step": 12016 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019593553405939255, - "loss": 1.675, - "step": 12017 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959351958330515, - "loss": 1.5792, - "step": 12018 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019593485760671043, - "loss": 1.7393, - "step": 12019 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019593451938036934, - "loss": 1.6163, - "step": 12020 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019593418115402829, - "loss": 1.5985, - "step": 12021 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019593384292768723, - "loss": 1.5658, - "step": 12022 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019593350470134614, - "loss": 1.7032, - "step": 12023 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019593316647500508, - "loss": 1.6027, - "step": 12024 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019593282824866402, - "loss": 1.8688, - "step": 12025 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019593249002232294, - "loss": 1.7251, - "step": 12026 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019593215179598188, - "loss": 1.5845, - "step": 12027 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019593181356964082, - "loss": 1.4914, - "step": 12028 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019593147534329973, - "loss": 1.5226, - "step": 12029 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019593113711695867, - "loss": 1.5388, - "step": 12030 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019593079889061762, - "loss": 1.8291, - "step": 12031 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019593046066427653, - "loss": 1.5803, - "step": 12032 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019593012243793547, - "loss": 1.8061, - "step": 12033 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959297842115944, - "loss": 1.5887, - "step": 12034 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019592944598525333, - "loss": 1.5229, - "step": 12035 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019592910775891227, - "loss": 1.6563, - "step": 12036 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959287695325712, - "loss": 1.6354, - "step": 12037 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019592843130623012, - "loss": 1.6448, - "step": 12038 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019592809307988906, - "loss": 1.5737, - "step": 12039 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195927754853548, - "loss": 1.5003, - "step": 12040 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019592741662720692, - "loss": 1.7525, - "step": 12041 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019592707840086586, - "loss": 1.5625, - "step": 12042 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959267401745248, - "loss": 1.6506, - "step": 12043 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019592640194818372, - "loss": 1.5634, - "step": 12044 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019592606372184266, - "loss": 1.6977, - "step": 12045 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959257254955016, - "loss": 1.5509, - "step": 12046 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959253872691605, - "loss": 1.6766, - "step": 12047 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019592504904281948, - "loss": 1.6966, - "step": 12048 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959247108164784, - "loss": 1.6123, - "step": 12049 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019592437259013734, - "loss": 1.6214, - "step": 12050 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019592403436379628, - "loss": 1.5718, - "step": 12051 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959236961374552, - "loss": 1.6305, - "step": 12052 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019592335791111413, - "loss": 1.5729, - "step": 12053 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019592301968477307, - "loss": 1.5773, - "step": 12054 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019592268145843201, - "loss": 1.7648, - "step": 12055 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019592234323209093, - "loss": 1.6855, - "step": 12056 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019592200500574987, - "loss": 1.5551, - "step": 12057 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959216667794088, - "loss": 1.6985, - "step": 12058 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019592132855306772, - "loss": 1.6728, - "step": 12059 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019592099032672667, - "loss": 1.6814, - "step": 12060 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959206521003856, - "loss": 1.5612, - "step": 12061 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019592031387404452, - "loss": 1.6363, - "step": 12062 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019591997564770346, - "loss": 1.6475, - "step": 12063 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959196374213624, - "loss": 1.8656, - "step": 12064 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019591929919502132, - "loss": 1.5429, - "step": 12065 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019591896096868026, - "loss": 1.5875, - "step": 12066 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959186227423392, - "loss": 1.4933, - "step": 12067 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959182845159981, - "loss": 1.5548, - "step": 12068 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019591794628965705, - "loss": 1.7885, - "step": 12069 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195917608063316, - "loss": 1.5332, - "step": 12070 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959172698369749, - "loss": 1.6457, - "step": 12071 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019591693161063385, - "loss": 1.6005, - "step": 12072 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959165933842928, - "loss": 1.5446, - "step": 12073 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959162551579517, - "loss": 1.6063, - "step": 12074 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019591591693161065, - "loss": 1.6395, - "step": 12075 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959155787052696, - "loss": 1.7865, - "step": 12076 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959152404789285, - "loss": 1.6145, - "step": 12077 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019591490225258744, - "loss": 1.7265, - "step": 12078 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019591456402624638, - "loss": 1.5745, - "step": 12079 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959142257999053, - "loss": 1.5309, - "step": 12080 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019591388757356424, - "loss": 1.6599, - "step": 12081 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019591354934722318, - "loss": 1.6145, - "step": 12082 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959132111208821, - "loss": 1.8331, - "step": 12083 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019591287289454104, - "loss": 1.6287, - "step": 12084 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019591253466819998, - "loss": 1.5934, - "step": 12085 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959121964418589, - "loss": 1.5247, - "step": 12086 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019591185821551783, - "loss": 1.5693, - "step": 12087 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019591151998917677, - "loss": 1.5992, - "step": 12088 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959111817628357, - "loss": 1.4553, - "step": 12089 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019591084353649463, - "loss": 1.6043, - "step": 12090 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019591050531015357, - "loss": 1.6556, - "step": 12091 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019591016708381248, - "loss": 1.6252, - "step": 12092 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019590982885747142, - "loss": 1.4656, - "step": 12093 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019590949063113037, - "loss": 1.8013, - "step": 12094 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019590915240478928, - "loss": 1.5579, - "step": 12095 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019590881417844822, - "loss": 1.5759, - "step": 12096 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019590847595210716, - "loss": 1.6365, - "step": 12097 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019590813772576608, - "loss": 1.7061, - "step": 12098 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019590779949942502, - "loss": 1.6708, - "step": 12099 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019590746127308396, - "loss": 1.4267, - "step": 12100 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019590712304674287, - "loss": 1.5869, - "step": 12101 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019590678482040181, - "loss": 1.5554, - "step": 12102 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019590644659406075, - "loss": 1.7312, - "step": 12103 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959061083677197, - "loss": 1.5985, - "step": 12104 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019590577014137864, - "loss": 1.6652, - "step": 12105 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019590543191503755, - "loss": 1.6793, - "step": 12106 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959050936886965, - "loss": 1.767, - "step": 12107 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019590475546235543, - "loss": 1.7039, - "step": 12108 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019590441723601435, - "loss": 1.449, - "step": 12109 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001959040790096733, - "loss": 1.5853, - "step": 12110 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019590374078333223, - "loss": 1.3336, - "step": 12111 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019590340255699114, - "loss": 1.7092, - "step": 12112 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019590306433065008, - "loss": 1.5246, - "step": 12113 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019590272610430903, - "loss": 1.6149, - "step": 12114 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019590238787796794, - "loss": 1.828, - "step": 12115 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019590204965162688, - "loss": 1.5195, - "step": 12116 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019590171142528582, - "loss": 1.6111, - "step": 12117 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019590137319894474, - "loss": 1.5986, - "step": 12118 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019590103497260368, - "loss": 1.697, - "step": 12119 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019590069674626262, - "loss": 1.7347, - "step": 12120 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019590035851992153, - "loss": 1.4604, - "step": 12121 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019590002029358047, - "loss": 1.4615, - "step": 12122 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019589968206723941, - "loss": 1.6542, - "step": 12123 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019589934384089833, - "loss": 1.7832, - "step": 12124 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019589900561455727, - "loss": 1.6132, - "step": 12125 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958986673882162, - "loss": 1.7243, - "step": 12126 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019589832916187513, - "loss": 1.6027, - "step": 12127 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019589799093553407, - "loss": 1.6563, - "step": 12128 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195897652709193, - "loss": 1.8023, - "step": 12129 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019589731448285192, - "loss": 1.6073, - "step": 12130 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019589697625651086, - "loss": 1.6518, - "step": 12131 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958966380301698, - "loss": 1.7527, - "step": 12132 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019589629980382872, - "loss": 1.5422, - "step": 12133 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019589596157748766, - "loss": 1.6863, - "step": 12134 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958956233511466, - "loss": 1.7267, - "step": 12135 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019589528512480551, - "loss": 1.7581, - "step": 12136 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019589494689846446, - "loss": 1.5197, - "step": 12137 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958946086721234, - "loss": 1.63, - "step": 12138 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019589427044578234, - "loss": 1.7085, - "step": 12139 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019589393221944125, - "loss": 1.6048, - "step": 12140 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958935939931002, - "loss": 1.6018, - "step": 12141 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019589325576675913, - "loss": 1.7823, - "step": 12142 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019589291754041805, - "loss": 1.5104, - "step": 12143 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195892579314077, - "loss": 1.7314, - "step": 12144 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019589224108773593, - "loss": 1.7076, - "step": 12145 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019589190286139484, - "loss": 1.6402, - "step": 12146 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019589156463505379, - "loss": 1.5955, - "step": 12147 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019589122640871273, - "loss": 1.8283, - "step": 12148 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019589088818237164, - "loss": 1.6275, - "step": 12149 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019589054995603058, - "loss": 1.7119, - "step": 12150 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019589021172968952, - "loss": 1.6722, - "step": 12151 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019588987350334844, - "loss": 1.6713, - "step": 12152 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019588953527700738, - "loss": 1.6936, - "step": 12153 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019588919705066632, - "loss": 1.591, - "step": 12154 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019588885882432523, - "loss": 1.6915, - "step": 12155 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019588852059798417, - "loss": 1.4644, - "step": 12156 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019588818237164312, - "loss": 1.7417, - "step": 12157 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019588784414530203, - "loss": 1.6633, - "step": 12158 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019588750591896097, - "loss": 1.7755, - "step": 12159 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958871676926199, - "loss": 1.6936, - "step": 12160 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019588682946627885, - "loss": 1.3332, - "step": 12161 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958864912399378, - "loss": 1.714, - "step": 12162 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958861530135967, - "loss": 1.5716, - "step": 12163 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019588581478725565, - "loss": 1.4865, - "step": 12164 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958854765609146, - "loss": 1.9097, - "step": 12165 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958851383345735, - "loss": 1.6886, - "step": 12166 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019588480010823245, - "loss": 1.6704, - "step": 12167 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958844618818914, - "loss": 1.7119, - "step": 12168 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958841236555503, - "loss": 1.6467, - "step": 12169 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019588378542920924, - "loss": 1.6302, - "step": 12170 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019588344720286818, - "loss": 1.6409, - "step": 12171 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958831089765271, - "loss": 1.6469, - "step": 12172 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019588277075018604, - "loss": 1.6264, - "step": 12173 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019588243252384498, - "loss": 1.7473, - "step": 12174 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958820942975039, - "loss": 1.7651, - "step": 12175 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019588175607116283, - "loss": 1.5018, - "step": 12176 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019588141784482178, - "loss": 1.6574, - "step": 12177 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958810796184807, - "loss": 1.641, - "step": 12178 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019588074139213963, - "loss": 1.611, - "step": 12179 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019588040316579857, - "loss": 1.5987, - "step": 12180 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019588006493945749, - "loss": 1.544, - "step": 12181 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019587972671311643, - "loss": 1.5145, - "step": 12182 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019587938848677537, - "loss": 1.7001, - "step": 12183 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019587905026043428, - "loss": 1.5753, - "step": 12184 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019587871203409322, - "loss": 1.7277, - "step": 12185 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019587837380775216, - "loss": 1.6931, - "step": 12186 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019587803558141108, - "loss": 1.5388, - "step": 12187 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019587769735507002, - "loss": 1.7235, - "step": 12188 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019587735912872896, - "loss": 1.4688, - "step": 12189 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019587702090238787, - "loss": 1.6164, - "step": 12190 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019587668267604682, - "loss": 1.5189, - "step": 12191 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019587634444970576, - "loss": 1.594, - "step": 12192 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019587600622336467, - "loss": 1.641, - "step": 12193 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958756679970236, - "loss": 1.5553, - "step": 12194 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019587532977068255, - "loss": 1.6067, - "step": 12195 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019587499154434147, - "loss": 1.4688, - "step": 12196 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958746533180004, - "loss": 1.6644, - "step": 12197 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019587431509165935, - "loss": 1.4893, - "step": 12198 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019587397686531826, - "loss": 1.5641, - "step": 12199 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958736386389772, - "loss": 1.6653, - "step": 12200 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019587330041263615, - "loss": 1.6867, - "step": 12201 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019587296218629506, - "loss": 1.4857, - "step": 12202 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195872623959954, - "loss": 1.5776, - "step": 12203 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019587228573361294, - "loss": 1.5499, - "step": 12204 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019587194750727186, - "loss": 1.5916, - "step": 12205 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958716092809308, - "loss": 1.5311, - "step": 12206 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019587127105458974, - "loss": 1.5948, - "step": 12207 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019587093282824865, - "loss": 1.6364, - "step": 12208 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958705946019076, - "loss": 1.6775, - "step": 12209 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019587025637556654, - "loss": 1.6553, - "step": 12210 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019586991814922545, - "loss": 1.6653, - "step": 12211 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958695799228844, - "loss": 1.552, - "step": 12212 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019586924169654333, - "loss": 1.6704, - "step": 12213 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019586890347020225, - "loss": 1.5943, - "step": 12214 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958685652438612, - "loss": 1.6472, - "step": 12215 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019586822701752013, - "loss": 1.6466, - "step": 12216 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019586788879117907, - "loss": 1.5925, - "step": 12217 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195867550564838, - "loss": 1.6959, - "step": 12218 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019586721233849695, - "loss": 1.6925, - "step": 12219 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019586687411215587, - "loss": 1.583, - "step": 12220 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958665358858148, - "loss": 1.5342, - "step": 12221 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019586619765947375, - "loss": 1.5388, - "step": 12222 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019586585943313266, - "loss": 1.6492, - "step": 12223 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958655212067916, - "loss": 1.689, - "step": 12224 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019586518298045054, - "loss": 1.4873, - "step": 12225 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019586484475410946, - "loss": 1.5881, - "step": 12226 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958645065277684, - "loss": 1.7994, - "step": 12227 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019586416830142734, - "loss": 1.661, - "step": 12228 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019586383007508625, - "loss": 1.717, - "step": 12229 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958634918487452, - "loss": 1.634, - "step": 12230 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019586315362240414, - "loss": 1.611, - "step": 12231 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019586281539606305, - "loss": 1.5575, - "step": 12232 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195862477169722, - "loss": 1.5703, - "step": 12233 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019586213894338093, - "loss": 1.4222, - "step": 12234 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019586180071703985, - "loss": 1.602, - "step": 12235 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958614624906988, - "loss": 1.5945, - "step": 12236 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019586112426435773, - "loss": 1.6905, - "step": 12237 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019586078603801664, - "loss": 1.6074, - "step": 12238 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019586044781167558, - "loss": 1.6675, - "step": 12239 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019586010958533453, - "loss": 1.5024, - "step": 12240 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019585977135899344, - "loss": 1.7527, - "step": 12241 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019585943313265238, - "loss": 1.5027, - "step": 12242 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019585909490631132, - "loss": 1.7006, - "step": 12243 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019585875667997024, - "loss": 1.7217, - "step": 12244 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019585841845362918, - "loss": 1.4911, - "step": 12245 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019585808022728812, - "loss": 1.6787, - "step": 12246 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019585774200094703, - "loss": 1.7695, - "step": 12247 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019585740377460597, - "loss": 1.4186, - "step": 12248 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019585706554826491, - "loss": 1.5389, - "step": 12249 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019585672732192383, - "loss": 1.5585, - "step": 12250 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019585638909558277, - "loss": 1.5407, - "step": 12251 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958560508692417, - "loss": 1.6993, - "step": 12252 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019585571264290062, - "loss": 1.5474, - "step": 12253 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019585537441655957, - "loss": 1.6652, - "step": 12254 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958550361902185, - "loss": 1.6634, - "step": 12255 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019585469796387742, - "loss": 1.5114, - "step": 12256 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019585435973753636, - "loss": 1.6626, - "step": 12257 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958540215111953, - "loss": 1.6292, - "step": 12258 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019585368328485422, - "loss": 1.6424, - "step": 12259 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019585334505851316, - "loss": 1.5321, - "step": 12260 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958530068321721, - "loss": 1.7342, - "step": 12261 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019585266860583101, - "loss": 1.7608, - "step": 12262 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019585233037948995, - "loss": 1.501, - "step": 12263 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958519921531489, - "loss": 1.5664, - "step": 12264 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958516539268078, - "loss": 1.4867, - "step": 12265 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019585131570046675, - "loss": 1.5358, - "step": 12266 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958509774741257, - "loss": 1.6067, - "step": 12267 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958506392477846, - "loss": 1.4285, - "step": 12268 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019585030102144355, - "loss": 1.7824, - "step": 12269 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958499627951025, - "loss": 1.6338, - "step": 12270 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958496245687614, - "loss": 1.5306, - "step": 12271 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019584928634242034, - "loss": 1.7209, - "step": 12272 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958489481160793, - "loss": 1.658, - "step": 12273 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019584860988973823, - "loss": 1.6563, - "step": 12274 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019584827166339717, - "loss": 1.6573, - "step": 12275 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958479334370561, - "loss": 1.5213, - "step": 12276 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019584759521071502, - "loss": 1.5841, - "step": 12277 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019584725698437396, - "loss": 1.6352, - "step": 12278 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958469187580329, - "loss": 1.4217, - "step": 12279 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019584658053169182, - "loss": 1.7361, - "step": 12280 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019584624230535076, - "loss": 1.6707, - "step": 12281 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958459040790097, - "loss": 1.5624, - "step": 12282 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019584556585266861, - "loss": 1.6318, - "step": 12283 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019584522762632756, - "loss": 1.7145, - "step": 12284 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958448893999865, - "loss": 1.6449, - "step": 12285 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958445511736454, - "loss": 1.5205, - "step": 12286 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019584421294730435, - "loss": 1.5214, - "step": 12287 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958438747209633, - "loss": 1.5197, - "step": 12288 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958435364946222, - "loss": 1.5407, - "step": 12289 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019584319826828115, - "loss": 1.4965, - "step": 12290 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958428600419401, - "loss": 1.4221, - "step": 12291 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195842521815599, - "loss": 1.584, - "step": 12292 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019584218358925794, - "loss": 1.7007, - "step": 12293 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019584184536291689, - "loss": 1.7615, - "step": 12294 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958415071365758, - "loss": 1.5744, - "step": 12295 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019584116891023474, - "loss": 1.7821, - "step": 12296 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019584083068389368, - "loss": 1.6294, - "step": 12297 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958404924575526, - "loss": 1.6902, - "step": 12298 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019584015423121154, - "loss": 1.6953, - "step": 12299 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019583981600487048, - "loss": 1.5168, - "step": 12300 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958394777785294, - "loss": 1.7701, - "step": 12301 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019583913955218833, - "loss": 1.4562, - "step": 12302 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019583880132584728, - "loss": 1.5427, - "step": 12303 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958384630995062, - "loss": 1.6855, - "step": 12304 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019583812487316513, - "loss": 1.6503, - "step": 12305 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019583778664682407, - "loss": 1.6667, - "step": 12306 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019583744842048299, - "loss": 1.849, - "step": 12307 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019583711019414193, - "loss": 1.5949, - "step": 12308 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019583677196780087, - "loss": 1.6051, - "step": 12309 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019583643374145978, - "loss": 1.5554, - "step": 12310 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019583609551511872, - "loss": 1.5768, - "step": 12311 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019583575728877766, - "loss": 1.4981, - "step": 12312 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019583541906243658, - "loss": 1.6609, - "step": 12313 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019583508083609552, - "loss": 1.5845, - "step": 12314 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019583474260975446, - "loss": 1.5727, - "step": 12315 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019583440438341337, - "loss": 1.6776, - "step": 12316 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019583406615707232, - "loss": 1.7356, - "step": 12317 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019583372793073126, - "loss": 1.699, - "step": 12318 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019583338970439017, - "loss": 1.7767, - "step": 12319 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958330514780491, - "loss": 1.7411, - "step": 12320 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019583271325170805, - "loss": 1.5932, - "step": 12321 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019583237502536697, - "loss": 1.6475, - "step": 12322 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958320367990259, - "loss": 1.5403, - "step": 12323 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019583169857268485, - "loss": 1.6679, - "step": 12324 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019583136034634376, - "loss": 1.7392, - "step": 12325 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958310221200027, - "loss": 1.729, - "step": 12326 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019583068389366165, - "loss": 1.512, - "step": 12327 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019583034566732056, - "loss": 1.5168, - "step": 12328 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019583000744097953, - "loss": 1.6799, - "step": 12329 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019582966921463847, - "loss": 1.5514, - "step": 12330 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019582933098829738, - "loss": 1.608, - "step": 12331 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019582899276195632, - "loss": 1.6877, - "step": 12332 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019582865453561527, - "loss": 1.4866, - "step": 12333 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019582831630927418, - "loss": 1.546, - "step": 12334 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019582797808293312, - "loss": 1.5803, - "step": 12335 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019582763985659206, - "loss": 1.6602, - "step": 12336 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019582730163025098, - "loss": 1.4715, - "step": 12337 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019582696340390992, - "loss": 1.6864, - "step": 12338 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019582662517756886, - "loss": 1.6871, - "step": 12339 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019582628695122777, - "loss": 1.637, - "step": 12340 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958259487248867, - "loss": 1.6666, - "step": 12341 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019582561049854565, - "loss": 1.6407, - "step": 12342 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019582527227220457, - "loss": 1.5782, - "step": 12343 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958249340458635, - "loss": 1.4924, - "step": 12344 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019582459581952245, - "loss": 1.5469, - "step": 12345 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019582425759318136, - "loss": 1.6618, - "step": 12346 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958239193668403, - "loss": 1.5888, - "step": 12347 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019582358114049925, - "loss": 1.8257, - "step": 12348 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019582324291415816, - "loss": 1.6033, - "step": 12349 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958229046878171, - "loss": 1.551, - "step": 12350 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019582256646147604, - "loss": 1.5584, - "step": 12351 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019582222823513496, - "loss": 1.4769, - "step": 12352 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958218900087939, - "loss": 1.6191, - "step": 12353 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019582155178245284, - "loss": 1.4322, - "step": 12354 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019582121355611175, - "loss": 1.7595, - "step": 12355 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958208753297707, - "loss": 1.7269, - "step": 12356 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019582053710342964, - "loss": 1.6541, - "step": 12357 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019582019887708855, - "loss": 1.8506, - "step": 12358 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958198606507475, - "loss": 1.5159, - "step": 12359 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019581952242440643, - "loss": 1.6599, - "step": 12360 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019581918419806535, - "loss": 1.653, - "step": 12361 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958188459717243, - "loss": 1.7207, - "step": 12362 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019581850774538323, - "loss": 1.5953, - "step": 12363 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019581816951904214, - "loss": 1.5804, - "step": 12364 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019581783129270108, - "loss": 1.5245, - "step": 12365 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019581749306636002, - "loss": 1.7261, - "step": 12366 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019581715484001894, - "loss": 1.5545, - "step": 12367 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019581681661367788, - "loss": 1.6772, - "step": 12368 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019581647838733682, - "loss": 1.5354, - "step": 12369 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019581614016099574, - "loss": 1.6446, - "step": 12370 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019581580193465468, - "loss": 1.5895, - "step": 12371 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019581546370831362, - "loss": 1.5881, - "step": 12372 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019581512548197253, - "loss": 1.6895, - "step": 12373 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019581478725563147, - "loss": 1.5893, - "step": 12374 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019581444902929041, - "loss": 1.509, - "step": 12375 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019581411080294933, - "loss": 1.5825, - "step": 12376 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019581377257660827, - "loss": 1.6908, - "step": 12377 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958134343502672, - "loss": 1.5175, - "step": 12378 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019581309612392612, - "loss": 1.6554, - "step": 12379 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019581275789758507, - "loss": 1.57, - "step": 12380 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195812419671244, - "loss": 1.6361, - "step": 12381 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019581208144490292, - "loss": 1.5438, - "step": 12382 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019581174321856186, - "loss": 1.5902, - "step": 12383 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958114049922208, - "loss": 1.6897, - "step": 12384 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019581106676587974, - "loss": 1.6112, - "step": 12385 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019581072853953868, - "loss": 1.6003, - "step": 12386 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958103903131976, - "loss": 1.6498, - "step": 12387 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019581005208685654, - "loss": 1.7154, - "step": 12388 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019580971386051548, - "loss": 1.7494, - "step": 12389 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958093756341744, - "loss": 1.491, - "step": 12390 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019580903740783334, - "loss": 1.5664, - "step": 12391 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019580869918149228, - "loss": 1.5614, - "step": 12392 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958083609551512, - "loss": 1.5278, - "step": 12393 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019580802272881013, - "loss": 1.5791, - "step": 12394 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019580768450246907, - "loss": 1.4668, - "step": 12395 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195807346276128, - "loss": 1.6522, - "step": 12396 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019580700804978693, - "loss": 1.6389, - "step": 12397 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019580666982344587, - "loss": 1.7189, - "step": 12398 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019580633159710478, - "loss": 1.6474, - "step": 12399 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019580599337076373, - "loss": 1.6432, - "step": 12400 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019580565514442267, - "loss": 1.4466, - "step": 12401 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019580531691808158, - "loss": 1.6603, - "step": 12402 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019580497869174052, - "loss": 1.7719, - "step": 12403 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019580464046539946, - "loss": 1.4777, - "step": 12404 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019580430223905838, - "loss": 1.4848, - "step": 12405 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019580396401271732, - "loss": 1.6144, - "step": 12406 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019580362578637626, - "loss": 1.6778, - "step": 12407 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958032875600352, - "loss": 1.7063, - "step": 12408 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019580294933369411, - "loss": 1.6634, - "step": 12409 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019580261110735306, - "loss": 1.4876, - "step": 12410 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195802272881012, - "loss": 1.8239, - "step": 12411 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958019346546709, - "loss": 1.6299, - "step": 12412 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019580159642832985, - "loss": 1.5009, - "step": 12413 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958012582019888, - "loss": 1.6205, - "step": 12414 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958009199756477, - "loss": 1.5842, - "step": 12415 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019580058174930665, - "loss": 1.5503, - "step": 12416 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001958002435229656, - "loss": 1.579, - "step": 12417 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957999052966245, - "loss": 1.6142, - "step": 12418 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019579956707028344, - "loss": 1.5967, - "step": 12419 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019579922884394239, - "loss": 1.7608, - "step": 12420 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957988906176013, - "loss": 1.5304, - "step": 12421 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019579855239126024, - "loss": 1.6783, - "step": 12422 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019579821416491918, - "loss": 1.7417, - "step": 12423 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957978759385781, - "loss": 1.7288, - "step": 12424 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019579753771223704, - "loss": 1.5974, - "step": 12425 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019579719948589598, - "loss": 1.5522, - "step": 12426 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957968612595549, - "loss": 1.7, - "step": 12427 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019579652303321383, - "loss": 1.5164, - "step": 12428 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019579618480687277, - "loss": 1.7864, - "step": 12429 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957958465805317, - "loss": 1.6232, - "step": 12430 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019579550835419063, - "loss": 1.5584, - "step": 12431 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019579517012784957, - "loss": 1.5992, - "step": 12432 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019579483190150848, - "loss": 1.5587, - "step": 12433 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019579449367516743, - "loss": 1.6382, - "step": 12434 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019579415544882637, - "loss": 1.6378, - "step": 12435 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019579381722248528, - "loss": 1.6152, - "step": 12436 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019579347899614422, - "loss": 1.5157, - "step": 12437 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019579314076980316, - "loss": 1.6375, - "step": 12438 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019579280254346208, - "loss": 1.2819, - "step": 12439 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019579246431712102, - "loss": 1.4965, - "step": 12440 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019579212609077996, - "loss": 1.6273, - "step": 12441 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957917878644389, - "loss": 1.6099, - "step": 12442 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019579144963809784, - "loss": 1.7, - "step": 12443 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019579111141175676, - "loss": 1.6372, - "step": 12444 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957907731854157, - "loss": 1.5753, - "step": 12445 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019579043495907464, - "loss": 1.5959, - "step": 12446 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019579009673273355, - "loss": 1.6137, - "step": 12447 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957897585063925, - "loss": 1.6396, - "step": 12448 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019578942028005143, - "loss": 1.771, - "step": 12449 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019578908205371035, - "loss": 1.6176, - "step": 12450 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957887438273693, - "loss": 1.7777, - "step": 12451 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019578840560102823, - "loss": 1.4648, - "step": 12452 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019578806737468714, - "loss": 1.493, - "step": 12453 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019578772914834609, - "loss": 1.5524, - "step": 12454 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019578739092200503, - "loss": 1.741, - "step": 12455 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019578705269566394, - "loss": 1.6731, - "step": 12456 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019578671446932288, - "loss": 1.4871, - "step": 12457 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019578637624298182, - "loss": 1.5884, - "step": 12458 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019578603801664074, - "loss": 1.6657, - "step": 12459 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019578569979029968, - "loss": 1.5273, - "step": 12460 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019578536156395862, - "loss": 1.6335, - "step": 12461 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019578502333761753, - "loss": 1.6768, - "step": 12462 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019578468511127647, - "loss": 1.6625, - "step": 12463 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019578434688493542, - "loss": 1.4327, - "step": 12464 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019578400865859433, - "loss": 1.4615, - "step": 12465 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019578367043225327, - "loss": 1.5555, - "step": 12466 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957833322059122, - "loss": 1.5432, - "step": 12467 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019578299397957113, - "loss": 1.5907, - "step": 12468 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019578265575323007, - "loss": 1.5939, - "step": 12469 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195782317526889, - "loss": 1.5806, - "step": 12470 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019578197930054792, - "loss": 1.4212, - "step": 12471 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019578164107420686, - "loss": 1.4555, - "step": 12472 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957813028478658, - "loss": 1.4978, - "step": 12473 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019578096462152472, - "loss": 1.6808, - "step": 12474 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019578062639518366, - "loss": 1.4938, - "step": 12475 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957802881688426, - "loss": 1.7543, - "step": 12476 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019577994994250152, - "loss": 1.5333, - "step": 12477 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019577961171616046, - "loss": 1.6184, - "step": 12478 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957792734898194, - "loss": 1.703, - "step": 12479 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957789352634783, - "loss": 1.4466, - "step": 12480 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019577859703713725, - "loss": 1.6811, - "step": 12481 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957782588107962, - "loss": 1.6762, - "step": 12482 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957779205844551, - "loss": 1.6157, - "step": 12483 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019577758235811405, - "loss": 1.4944, - "step": 12484 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195777244131773, - "loss": 1.7213, - "step": 12485 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957769059054319, - "loss": 1.5492, - "step": 12486 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019577656767909085, - "loss": 1.6869, - "step": 12487 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957762294527498, - "loss": 1.6685, - "step": 12488 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957758912264087, - "loss": 1.7118, - "step": 12489 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019577555300006764, - "loss": 1.7153, - "step": 12490 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019577521477372658, - "loss": 1.7955, - "step": 12491 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019577487654738552, - "loss": 1.3933, - "step": 12492 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019577453832104444, - "loss": 1.5925, - "step": 12493 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019577420009470338, - "loss": 1.5985, - "step": 12494 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019577386186836232, - "loss": 1.615, - "step": 12495 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019577352364202123, - "loss": 1.5576, - "step": 12496 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019577318541568018, - "loss": 1.5222, - "step": 12497 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019577284718933912, - "loss": 1.6962, - "step": 12498 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019577250896299806, - "loss": 1.416, - "step": 12499 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195772170736657, - "loss": 1.6295, - "step": 12500 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957718325103159, - "loss": 1.7091, - "step": 12501 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019577149428397485, - "loss": 1.6536, - "step": 12502 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957711560576338, - "loss": 1.8137, - "step": 12503 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957708178312927, - "loss": 1.7671, - "step": 12504 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019577047960495165, - "loss": 1.6246, - "step": 12505 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957701413786106, - "loss": 1.6257, - "step": 12506 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957698031522695, - "loss": 1.5709, - "step": 12507 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019576946492592845, - "loss": 1.6041, - "step": 12508 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957691266995874, - "loss": 1.5032, - "step": 12509 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957687884732463, - "loss": 1.6281, - "step": 12510 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019576845024690524, - "loss": 1.461, - "step": 12511 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019576811202056418, - "loss": 1.4774, - "step": 12512 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957677737942231, - "loss": 1.6515, - "step": 12513 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019576743556788204, - "loss": 1.5441, - "step": 12514 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019576709734154098, - "loss": 1.6355, - "step": 12515 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957667591151999, - "loss": 1.6279, - "step": 12516 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019576642088885884, - "loss": 1.63, - "step": 12517 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019576608266251778, - "loss": 1.6698, - "step": 12518 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957657444361767, - "loss": 1.4513, - "step": 12519 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019576540620983563, - "loss": 1.5701, - "step": 12520 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019576506798349457, - "loss": 1.663, - "step": 12521 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957647297571535, - "loss": 1.7074, - "step": 12522 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019576439153081243, - "loss": 1.6067, - "step": 12523 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019576405330447137, - "loss": 1.5947, - "step": 12524 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019576371507813028, - "loss": 1.5902, - "step": 12525 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019576337685178922, - "loss": 1.6778, - "step": 12526 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019576303862544817, - "loss": 1.7424, - "step": 12527 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019576270039910708, - "loss": 1.6521, - "step": 12528 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019576236217276602, - "loss": 1.7214, - "step": 12529 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019576202394642496, - "loss": 1.6644, - "step": 12530 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019576168572008388, - "loss": 1.6856, - "step": 12531 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019576134749374282, - "loss": 1.5251, - "step": 12532 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019576100926740176, - "loss": 1.5256, - "step": 12533 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019576067104106067, - "loss": 1.5058, - "step": 12534 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019576033281471961, - "loss": 1.6557, - "step": 12535 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019575999458837855, - "loss": 1.6044, - "step": 12536 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019575965636203747, - "loss": 1.5287, - "step": 12537 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957593181356964, - "loss": 1.4226, - "step": 12538 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019575897990935535, - "loss": 1.6107, - "step": 12539 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019575864168301427, - "loss": 1.6888, - "step": 12540 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957583034566732, - "loss": 1.7135, - "step": 12541 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019575796523033215, - "loss": 1.6241, - "step": 12542 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019575762700399106, - "loss": 1.4752, - "step": 12543 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019575728877765, - "loss": 1.4964, - "step": 12544 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019575695055130894, - "loss": 1.493, - "step": 12545 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019575661232496786, - "loss": 1.5622, - "step": 12546 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957562740986268, - "loss": 1.5701, - "step": 12547 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019575593587228574, - "loss": 1.7016, - "step": 12548 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019575559764594465, - "loss": 1.8044, - "step": 12549 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957552594196036, - "loss": 1.5893, - "step": 12550 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019575492119326254, - "loss": 1.5518, - "step": 12551 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019575458296692145, - "loss": 1.4704, - "step": 12552 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957542447405804, - "loss": 1.6609, - "step": 12553 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019575390651423936, - "loss": 1.7148, - "step": 12554 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019575356828789827, - "loss": 1.5728, - "step": 12555 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019575323006155721, - "loss": 1.7032, - "step": 12556 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019575289183521616, - "loss": 1.4604, - "step": 12557 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019575255360887507, - "loss": 1.614, - "step": 12558 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195752215382534, - "loss": 1.6218, - "step": 12559 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019575187715619295, - "loss": 1.5466, - "step": 12560 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019575153892985187, - "loss": 1.6441, - "step": 12561 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957512007035108, - "loss": 1.5614, - "step": 12562 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019575086247716975, - "loss": 1.7122, - "step": 12563 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019575052425082866, - "loss": 1.5894, - "step": 12564 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957501860244876, - "loss": 1.4569, - "step": 12565 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019574984779814654, - "loss": 1.5754, - "step": 12566 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019574950957180546, - "loss": 1.759, - "step": 12567 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957491713454644, - "loss": 1.5637, - "step": 12568 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019574883311912334, - "loss": 1.5643, - "step": 12569 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019574849489278226, - "loss": 1.719, - "step": 12570 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957481566664412, - "loss": 1.545, - "step": 12571 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019574781844010014, - "loss": 1.7488, - "step": 12572 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019574748021375905, - "loss": 1.6313, - "step": 12573 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195747141987418, - "loss": 1.8118, - "step": 12574 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019574680376107693, - "loss": 1.5206, - "step": 12575 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019574646553473585, - "loss": 1.5079, - "step": 12576 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957461273083948, - "loss": 1.4145, - "step": 12577 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019574578908205373, - "loss": 1.7744, - "step": 12578 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019574545085571264, - "loss": 1.5879, - "step": 12579 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019574511262937159, - "loss": 1.6422, - "step": 12580 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019574477440303053, - "loss": 1.5715, - "step": 12581 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019574443617668944, - "loss": 1.603, - "step": 12582 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019574409795034838, - "loss": 1.6436, - "step": 12583 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019574375972400732, - "loss": 1.7768, - "step": 12584 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019574342149766624, - "loss": 1.5828, - "step": 12585 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019574308327132518, - "loss": 1.6104, - "step": 12586 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019574274504498412, - "loss": 1.6826, - "step": 12587 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019574240681864303, - "loss": 1.6412, - "step": 12588 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019574206859230197, - "loss": 1.5986, - "step": 12589 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019574173036596092, - "loss": 1.7198, - "step": 12590 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019574139213961983, - "loss": 1.7482, - "step": 12591 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019574105391327877, - "loss": 1.7119, - "step": 12592 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957407156869377, - "loss": 1.7416, - "step": 12593 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019574037746059663, - "loss": 1.7933, - "step": 12594 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019574003923425557, - "loss": 1.6228, - "step": 12595 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957397010079145, - "loss": 1.695, - "step": 12596 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019573936278157342, - "loss": 1.5335, - "step": 12597 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019573902455523236, - "loss": 1.5878, - "step": 12598 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957386863288913, - "loss": 1.4646, - "step": 12599 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019573834810255022, - "loss": 1.587, - "step": 12600 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019573800987620916, - "loss": 1.5492, - "step": 12601 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957376716498681, - "loss": 1.6163, - "step": 12602 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019573733342352701, - "loss": 1.5962, - "step": 12603 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019573699519718596, - "loss": 1.602, - "step": 12604 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957366569708449, - "loss": 1.6463, - "step": 12605 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957363187445038, - "loss": 1.6275, - "step": 12606 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019573598051816275, - "loss": 1.6668, - "step": 12607 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957356422918217, - "loss": 1.5467, - "step": 12608 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957353040654806, - "loss": 1.6991, - "step": 12609 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019573496583913958, - "loss": 1.5369, - "step": 12610 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019573462761279852, - "loss": 1.7507, - "step": 12611 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019573428938645743, - "loss": 1.7333, - "step": 12612 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019573395116011637, - "loss": 1.5871, - "step": 12613 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957336129337753, - "loss": 1.5735, - "step": 12614 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019573327470743423, - "loss": 1.6567, - "step": 12615 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019573293648109317, - "loss": 1.5251, - "step": 12616 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957325982547521, - "loss": 1.565, - "step": 12617 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019573226002841102, - "loss": 1.5901, - "step": 12618 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019573192180206996, - "loss": 1.6138, - "step": 12619 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957315835757289, - "loss": 1.63, - "step": 12620 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019573124534938782, - "loss": 1.5641, - "step": 12621 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019573090712304676, - "loss": 1.631, - "step": 12622 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957305688967057, - "loss": 1.6407, - "step": 12623 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019573023067036462, - "loss": 1.7207, - "step": 12624 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019572989244402356, - "loss": 1.661, - "step": 12625 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957295542176825, - "loss": 1.7419, - "step": 12626 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957292159913414, - "loss": 1.4514, - "step": 12627 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019572887776500035, - "loss": 1.5819, - "step": 12628 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957285395386593, - "loss": 1.6867, - "step": 12629 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957282013123182, - "loss": 1.6039, - "step": 12630 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019572786308597715, - "loss": 1.6561, - "step": 12631 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957275248596361, - "loss": 1.5689, - "step": 12632 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195727186633295, - "loss": 1.5624, - "step": 12633 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019572684840695395, - "loss": 1.6881, - "step": 12634 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957265101806129, - "loss": 1.6192, - "step": 12635 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957261719542718, - "loss": 1.6712, - "step": 12636 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019572583372793074, - "loss": 1.5943, - "step": 12637 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019572549550158968, - "loss": 1.721, - "step": 12638 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957251572752486, - "loss": 1.6595, - "step": 12639 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019572481904890754, - "loss": 1.6171, - "step": 12640 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019572448082256648, - "loss": 1.6198, - "step": 12641 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957241425962254, - "loss": 1.6976, - "step": 12642 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019572380436988434, - "loss": 1.565, - "step": 12643 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019572346614354328, - "loss": 1.6122, - "step": 12644 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957231279172022, - "loss": 1.6131, - "step": 12645 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019572278969086113, - "loss": 1.416, - "step": 12646 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019572245146452007, - "loss": 1.6907, - "step": 12647 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195722113238179, - "loss": 1.7376, - "step": 12648 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019572177501183793, - "loss": 1.7432, - "step": 12649 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019572143678549687, - "loss": 1.6929, - "step": 12650 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019572109855915578, - "loss": 1.6615, - "step": 12651 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019572076033281472, - "loss": 1.6724, - "step": 12652 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019572042210647367, - "loss": 1.4259, - "step": 12653 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019572008388013258, - "loss": 1.3867, - "step": 12654 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019571974565379152, - "loss": 1.4371, - "step": 12655 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019571940742745046, - "loss": 1.6116, - "step": 12656 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019571906920110938, - "loss": 1.5182, - "step": 12657 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019571873097476832, - "loss": 1.667, - "step": 12658 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019571839274842726, - "loss": 1.8272, - "step": 12659 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019571805452208617, - "loss": 1.6479, - "step": 12660 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957177162957451, - "loss": 1.6991, - "step": 12661 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019571737806940405, - "loss": 1.567, - "step": 12662 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019571703984306297, - "loss": 1.5508, - "step": 12663 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957167016167219, - "loss": 1.6609, - "step": 12664 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019571636339038085, - "loss": 1.6317, - "step": 12665 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957160251640398, - "loss": 1.6204, - "step": 12666 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019571568693769873, - "loss": 1.5883, - "step": 12667 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019571534871135765, - "loss": 1.6916, - "step": 12668 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957150104850166, - "loss": 1.4856, - "step": 12669 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019571467225867553, - "loss": 1.6438, - "step": 12670 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019571433403233444, - "loss": 1.6788, - "step": 12671 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019571399580599338, - "loss": 1.6029, - "step": 12672 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019571365757965233, - "loss": 1.4971, - "step": 12673 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019571331935331127, - "loss": 1.7598, - "step": 12674 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019571298112697018, - "loss": 1.664, - "step": 12675 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019571264290062912, - "loss": 1.7855, - "step": 12676 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019571230467428806, - "loss": 1.5576, - "step": 12677 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019571196644794698, - "loss": 1.5644, - "step": 12678 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019571162822160592, - "loss": 1.5826, - "step": 12679 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019571128999526486, - "loss": 1.6508, - "step": 12680 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019571095176892377, - "loss": 1.7672, - "step": 12681 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019571061354258271, - "loss": 1.6232, - "step": 12682 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019571027531624166, - "loss": 1.5348, - "step": 12683 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019570993708990057, - "loss": 1.8437, - "step": 12684 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957095988635595, - "loss": 1.61, - "step": 12685 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019570926063721845, - "loss": 1.6493, - "step": 12686 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019570892241087737, - "loss": 1.521, - "step": 12687 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957085841845363, - "loss": 1.791, - "step": 12688 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019570824595819525, - "loss": 1.5367, - "step": 12689 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019570790773185416, - "loss": 1.6055, - "step": 12690 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957075695055131, - "loss": 1.6706, - "step": 12691 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019570723127917204, - "loss": 1.609, - "step": 12692 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019570689305283096, - "loss": 1.5388, - "step": 12693 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957065548264899, - "loss": 1.6958, - "step": 12694 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019570621660014884, - "loss": 1.5526, - "step": 12695 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019570587837380775, - "loss": 1.6728, - "step": 12696 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957055401474667, - "loss": 1.5301, - "step": 12697 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019570520192112564, - "loss": 1.6713, - "step": 12698 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019570486369478455, - "loss": 1.6665, - "step": 12699 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957045254684435, - "loss": 1.5861, - "step": 12700 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019570418724210243, - "loss": 1.5832, - "step": 12701 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019570384901576135, - "loss": 1.5394, - "step": 12702 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001957035107894203, - "loss": 1.5571, - "step": 12703 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019570317256307923, - "loss": 1.7181, - "step": 12704 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019570283433673814, - "loss": 1.6632, - "step": 12705 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019570249611039708, - "loss": 1.6865, - "step": 12706 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019570215788405603, - "loss": 1.518, - "step": 12707 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019570181965771494, - "loss": 1.5015, - "step": 12708 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019570148143137388, - "loss": 1.5032, - "step": 12709 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019570114320503282, - "loss": 1.5982, - "step": 12710 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019570080497869174, - "loss": 1.7819, - "step": 12711 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019570046675235068, - "loss": 1.6139, - "step": 12712 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019570012852600962, - "loss": 1.5651, - "step": 12713 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019569979029966853, - "loss": 1.7322, - "step": 12714 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019569945207332747, - "loss": 1.7336, - "step": 12715 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019569911384698641, - "loss": 1.7806, - "step": 12716 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019569877562064533, - "loss": 1.5994, - "step": 12717 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019569843739430427, - "loss": 1.5069, - "step": 12718 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956980991679632, - "loss": 1.6235, - "step": 12719 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019569776094162213, - "loss": 1.7035, - "step": 12720 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019569742271528107, - "loss": 1.6805, - "step": 12721 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019569708448894, - "loss": 1.6273, - "step": 12722 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019569674626259895, - "loss": 1.6356, - "step": 12723 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956964080362579, - "loss": 1.7339, - "step": 12724 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956960698099168, - "loss": 1.6684, - "step": 12725 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019569573158357574, - "loss": 1.7415, - "step": 12726 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019569539335723469, - "loss": 1.5377, - "step": 12727 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956950551308936, - "loss": 1.6792, - "step": 12728 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019569471690455254, - "loss": 1.5037, - "step": 12729 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019569437867821148, - "loss": 1.7613, - "step": 12730 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956940404518704, - "loss": 1.5852, - "step": 12731 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019569370222552934, - "loss": 1.6227, - "step": 12732 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019569336399918828, - "loss": 1.5434, - "step": 12733 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956930257728472, - "loss": 1.6537, - "step": 12734 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019569268754650613, - "loss": 1.5553, - "step": 12735 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019569234932016508, - "loss": 1.7007, - "step": 12736 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195692011093824, - "loss": 1.6121, - "step": 12737 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019569167286748293, - "loss": 1.613, - "step": 12738 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019569133464114187, - "loss": 1.6559, - "step": 12739 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019569099641480079, - "loss": 1.7681, - "step": 12740 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019569065818845973, - "loss": 1.6031, - "step": 12741 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019569031996211867, - "loss": 1.5954, - "step": 12742 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019568998173577758, - "loss": 1.7032, - "step": 12743 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019568964350943652, - "loss": 1.7376, - "step": 12744 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019568930528309546, - "loss": 1.4771, - "step": 12745 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019568896705675438, - "loss": 1.7632, - "step": 12746 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019568862883041332, - "loss": 1.5794, - "step": 12747 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019568829060407226, - "loss": 1.5319, - "step": 12748 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019568795237773117, - "loss": 1.5925, - "step": 12749 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019568761415139012, - "loss": 1.6135, - "step": 12750 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019568727592504906, - "loss": 1.7413, - "step": 12751 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019568693769870797, - "loss": 1.6554, - "step": 12752 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956865994723669, - "loss": 1.6381, - "step": 12753 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019568626124602585, - "loss": 1.6731, - "step": 12754 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019568592301968477, - "loss": 1.5534, - "step": 12755 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956855847933437, - "loss": 1.7608, - "step": 12756 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019568524656700265, - "loss": 1.442, - "step": 12757 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956849083406616, - "loss": 1.6773, - "step": 12758 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956845701143205, - "loss": 1.6638, - "step": 12759 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019568423188797945, - "loss": 1.586, - "step": 12760 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956838936616384, - "loss": 1.7175, - "step": 12761 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956835554352973, - "loss": 1.6101, - "step": 12762 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019568321720895624, - "loss": 1.5597, - "step": 12763 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019568287898261518, - "loss": 1.6351, - "step": 12764 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956825407562741, - "loss": 1.7863, - "step": 12765 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019568220252993304, - "loss": 1.5754, - "step": 12766 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019568186430359198, - "loss": 1.6193, - "step": 12767 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956815260772509, - "loss": 1.6236, - "step": 12768 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019568118785090983, - "loss": 1.5666, - "step": 12769 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019568084962456878, - "loss": 1.6306, - "step": 12770 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956805113982277, - "loss": 1.6997, - "step": 12771 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019568017317188663, - "loss": 1.6366, - "step": 12772 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019567983494554557, - "loss": 1.52, - "step": 12773 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019567949671920449, - "loss": 1.5575, - "step": 12774 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019567915849286343, - "loss": 1.631, - "step": 12775 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019567882026652237, - "loss": 1.5575, - "step": 12776 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019567848204018128, - "loss": 1.6994, - "step": 12777 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019567814381384025, - "loss": 1.526, - "step": 12778 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019567780558749916, - "loss": 1.6742, - "step": 12779 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956774673611581, - "loss": 1.7323, - "step": 12780 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019567712913481705, - "loss": 1.5759, - "step": 12781 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019567679090847596, - "loss": 1.4816, - "step": 12782 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956764526821349, - "loss": 1.6567, - "step": 12783 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019567611445579384, - "loss": 1.5618, - "step": 12784 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019567577622945276, - "loss": 1.6018, - "step": 12785 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956754380031117, - "loss": 1.6029, - "step": 12786 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019567509977677064, - "loss": 1.599, - "step": 12787 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019567476155042955, - "loss": 1.5157, - "step": 12788 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956744233240885, - "loss": 1.5859, - "step": 12789 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019567408509774744, - "loss": 1.5523, - "step": 12790 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019567374687140635, - "loss": 1.5246, - "step": 12791 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956734086450653, - "loss": 1.7157, - "step": 12792 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019567307041872423, - "loss": 1.6244, - "step": 12793 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019567273219238315, - "loss": 1.5265, - "step": 12794 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956723939660421, - "loss": 1.6897, - "step": 12795 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019567205573970103, - "loss": 1.8879, - "step": 12796 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019567171751335994, - "loss": 1.5329, - "step": 12797 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019567137928701888, - "loss": 1.5121, - "step": 12798 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019567104106067782, - "loss": 1.728, - "step": 12799 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019567070283433674, - "loss": 1.5151, - "step": 12800 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019567036460799568, - "loss": 1.5874, - "step": 12801 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019567002638165462, - "loss": 1.7014, - "step": 12802 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019566968815531354, - "loss": 1.7728, - "step": 12803 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019566934992897248, - "loss": 1.5176, - "step": 12804 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019566901170263142, - "loss": 1.6602, - "step": 12805 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019566867347629033, - "loss": 1.6046, - "step": 12806 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019566833524994927, - "loss": 1.6313, - "step": 12807 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019566799702360821, - "loss": 1.57, - "step": 12808 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019566765879726713, - "loss": 1.5453, - "step": 12809 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019566732057092607, - "loss": 1.6897, - "step": 12810 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195666982344585, - "loss": 1.5419, - "step": 12811 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019566664411824392, - "loss": 1.6792, - "step": 12812 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019566630589190287, - "loss": 1.7671, - "step": 12813 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956659676655618, - "loss": 1.4109, - "step": 12814 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019566562943922072, - "loss": 1.6797, - "step": 12815 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019566529121287966, - "loss": 1.5103, - "step": 12816 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956649529865386, - "loss": 1.6884, - "step": 12817 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019566461476019752, - "loss": 1.5181, - "step": 12818 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019566427653385646, - "loss": 1.7385, - "step": 12819 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956639383075154, - "loss": 1.599, - "step": 12820 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956636000811743, - "loss": 1.623, - "step": 12821 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019566326185483325, - "loss": 1.7168, - "step": 12822 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956629236284922, - "loss": 1.6805, - "step": 12823 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956625854021511, - "loss": 1.6345, - "step": 12824 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019566224717581005, - "loss": 1.4991, - "step": 12825 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195661908949469, - "loss": 1.615, - "step": 12826 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956615707231279, - "loss": 1.6785, - "step": 12827 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019566123249678685, - "loss": 1.6036, - "step": 12828 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956608942704458, - "loss": 1.5389, - "step": 12829 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956605560441047, - "loss": 1.6982, - "step": 12830 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019566021781776364, - "loss": 1.5357, - "step": 12831 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019565987959142258, - "loss": 1.6678, - "step": 12832 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956595413650815, - "loss": 1.6622, - "step": 12833 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019565920313874044, - "loss": 1.5877, - "step": 12834 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956588649123994, - "loss": 1.6681, - "step": 12835 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019565852668605832, - "loss": 1.6908, - "step": 12836 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019565818845971726, - "loss": 1.7642, - "step": 12837 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956578502333762, - "loss": 1.5501, - "step": 12838 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019565751200703512, - "loss": 1.7813, - "step": 12839 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019565717378069406, - "loss": 1.7262, - "step": 12840 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195656835554353, - "loss": 1.7494, - "step": 12841 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019565649732801191, - "loss": 1.6903, - "step": 12842 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019565615910167086, - "loss": 1.5551, - "step": 12843 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956558208753298, - "loss": 1.6287, - "step": 12844 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956554826489887, - "loss": 1.5549, - "step": 12845 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019565514442264765, - "loss": 1.6053, - "step": 12846 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956548061963066, - "loss": 1.6189, - "step": 12847 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956544679699655, - "loss": 1.625, - "step": 12848 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019565412974362445, - "loss": 1.6025, - "step": 12849 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956537915172834, - "loss": 1.6625, - "step": 12850 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956534532909423, - "loss": 1.5378, - "step": 12851 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019565311506460124, - "loss": 1.5827, - "step": 12852 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019565277683826019, - "loss": 1.6008, - "step": 12853 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956524386119191, - "loss": 1.4124, - "step": 12854 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019565210038557804, - "loss": 1.5051, - "step": 12855 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019565176215923698, - "loss": 1.6215, - "step": 12856 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956514239328959, - "loss": 1.7853, - "step": 12857 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019565108570655484, - "loss": 1.6189, - "step": 12858 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019565074748021378, - "loss": 1.6414, - "step": 12859 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956504092538727, - "loss": 1.5904, - "step": 12860 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019565007102753163, - "loss": 1.6188, - "step": 12861 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019564973280119057, - "loss": 1.6859, - "step": 12862 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956493945748495, - "loss": 1.5235, - "step": 12863 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019564905634850843, - "loss": 1.667, - "step": 12864 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019564871812216737, - "loss": 1.6945, - "step": 12865 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019564837989582628, - "loss": 1.5261, - "step": 12866 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019564804166948523, - "loss": 1.7486, - "step": 12867 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019564770344314417, - "loss": 1.6259, - "step": 12868 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019564736521680308, - "loss": 1.6927, - "step": 12869 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019564702699046202, - "loss": 1.6274, - "step": 12870 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019564668876412096, - "loss": 1.5665, - "step": 12871 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019564635053777988, - "loss": 1.6446, - "step": 12872 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019564601231143882, - "loss": 1.6106, - "step": 12873 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019564567408509776, - "loss": 1.6844, - "step": 12874 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019564533585875667, - "loss": 1.7214, - "step": 12875 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019564499763241561, - "loss": 1.6337, - "step": 12876 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019564465940607456, - "loss": 1.7316, - "step": 12877 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019564432117973347, - "loss": 1.4633, - "step": 12878 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956439829533924, - "loss": 1.4827, - "step": 12879 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019564364472705135, - "loss": 1.6852, - "step": 12880 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019564330650071027, - "loss": 1.5842, - "step": 12881 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956429682743692, - "loss": 1.6341, - "step": 12882 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019564263004802815, - "loss": 1.5955, - "step": 12883 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019564229182168706, - "loss": 1.575, - "step": 12884 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195641953595346, - "loss": 1.7259, - "step": 12885 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019564161536900494, - "loss": 1.602, - "step": 12886 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019564127714266386, - "loss": 1.7207, - "step": 12887 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956409389163228, - "loss": 1.5402, - "step": 12888 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019564060068998174, - "loss": 1.6472, - "step": 12889 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019564026246364066, - "loss": 1.7659, - "step": 12890 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019563992423729962, - "loss": 1.6809, - "step": 12891 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019563958601095856, - "loss": 1.492, - "step": 12892 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019563924778461748, - "loss": 1.3711, - "step": 12893 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019563890955827642, - "loss": 1.4908, - "step": 12894 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019563857133193536, - "loss": 1.5518, - "step": 12895 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019563823310559428, - "loss": 1.4901, - "step": 12896 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019563789487925322, - "loss": 1.5629, - "step": 12897 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019563755665291216, - "loss": 1.5287, - "step": 12898 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019563721842657107, - "loss": 1.5702, - "step": 12899 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019563688020023, - "loss": 1.7525, - "step": 12900 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019563654197388895, - "loss": 1.5376, - "step": 12901 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019563620374754787, - "loss": 1.7452, - "step": 12902 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956358655212068, - "loss": 1.5988, - "step": 12903 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019563552729486575, - "loss": 1.533, - "step": 12904 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019563518906852466, - "loss": 1.4728, - "step": 12905 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956348508421836, - "loss": 1.7266, - "step": 12906 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019563451261584255, - "loss": 1.5339, - "step": 12907 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019563417438950146, - "loss": 1.5319, - "step": 12908 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956338361631604, - "loss": 1.5875, - "step": 12909 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019563349793681934, - "loss": 1.6586, - "step": 12910 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019563315971047826, - "loss": 1.6322, - "step": 12911 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956328214841372, - "loss": 1.6099, - "step": 12912 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019563248325779614, - "loss": 1.7431, - "step": 12913 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019563214503145505, - "loss": 1.4601, - "step": 12914 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195631806805114, - "loss": 1.6542, - "step": 12915 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019563146857877294, - "loss": 1.618, - "step": 12916 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019563113035243185, - "loss": 1.65, - "step": 12917 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956307921260908, - "loss": 1.7206, - "step": 12918 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019563045389974973, - "loss": 1.752, - "step": 12919 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019563011567340865, - "loss": 1.5077, - "step": 12920 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956297774470676, - "loss": 1.6099, - "step": 12921 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019562943922072653, - "loss": 1.5808, - "step": 12922 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019562910099438544, - "loss": 1.5231, - "step": 12923 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019562876276804438, - "loss": 1.5447, - "step": 12924 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019562842454170332, - "loss": 1.6321, - "step": 12925 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019562808631536224, - "loss": 1.617, - "step": 12926 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019562774808902118, - "loss": 1.6102, - "step": 12927 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019562740986268012, - "loss": 1.6179, - "step": 12928 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019562707163633903, - "loss": 1.5933, - "step": 12929 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019562673340999798, - "loss": 1.558, - "step": 12930 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019562639518365692, - "loss": 1.6539, - "step": 12931 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019562605695731583, - "loss": 1.6655, - "step": 12932 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019562571873097477, - "loss": 1.6465, - "step": 12933 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956253805046337, - "loss": 1.588, - "step": 12934 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019562504227829263, - "loss": 1.5262, - "step": 12935 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019562470405195157, - "loss": 1.7051, - "step": 12936 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956243658256105, - "loss": 1.6251, - "step": 12937 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019562402759926942, - "loss": 1.5882, - "step": 12938 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019562368937292836, - "loss": 1.4271, - "step": 12939 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956233511465873, - "loss": 1.6755, - "step": 12940 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019562301292024622, - "loss": 1.4963, - "step": 12941 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019562267469390516, - "loss": 1.7532, - "step": 12942 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956223364675641, - "loss": 1.638, - "step": 12943 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019562199824122302, - "loss": 1.5433, - "step": 12944 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019562166001488196, - "loss": 1.6097, - "step": 12945 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956213217885409, - "loss": 1.5061, - "step": 12946 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019562098356219984, - "loss": 1.7236, - "step": 12947 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019562064533585878, - "loss": 1.562, - "step": 12948 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019562030710951772, - "loss": 1.5281, - "step": 12949 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019561996888317664, - "loss": 1.6803, - "step": 12950 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019561963065683558, - "loss": 1.6617, - "step": 12951 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019561929243049452, - "loss": 1.5941, - "step": 12952 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019561895420415343, - "loss": 1.614, - "step": 12953 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019561861597781237, - "loss": 1.7662, - "step": 12954 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019561827775147131, - "loss": 1.6144, - "step": 12955 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019561793952513023, - "loss": 1.5387, - "step": 12956 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019561760129878917, - "loss": 1.4649, - "step": 12957 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956172630724481, - "loss": 1.7316, - "step": 12958 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019561692484610702, - "loss": 1.6842, - "step": 12959 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019561658661976597, - "loss": 1.5874, - "step": 12960 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956162483934249, - "loss": 1.6706, - "step": 12961 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019561591016708382, - "loss": 1.6068, - "step": 12962 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019561557194074276, - "loss": 1.6294, - "step": 12963 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956152337144017, - "loss": 1.677, - "step": 12964 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019561489548806062, - "loss": 1.5976, - "step": 12965 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019561455726171956, - "loss": 1.4837, - "step": 12966 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956142190353785, - "loss": 1.5599, - "step": 12967 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019561388080903741, - "loss": 1.6602, - "step": 12968 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019561354258269635, - "loss": 1.7157, - "step": 12969 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956132043563553, - "loss": 1.5113, - "step": 12970 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956128661300142, - "loss": 1.704, - "step": 12971 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019561252790367315, - "loss": 1.6152, - "step": 12972 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956121896773321, - "loss": 1.5966, - "step": 12973 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195611851450991, - "loss": 1.5983, - "step": 12974 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019561151322464995, - "loss": 1.626, - "step": 12975 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956111749983089, - "loss": 1.644, - "step": 12976 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956108367719678, - "loss": 1.6578, - "step": 12977 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019561049854562674, - "loss": 1.7303, - "step": 12978 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019561016031928568, - "loss": 1.5073, - "step": 12979 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956098220929446, - "loss": 1.7272, - "step": 12980 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019560948386660354, - "loss": 1.6344, - "step": 12981 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019560914564026248, - "loss": 1.6689, - "step": 12982 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956088074139214, - "loss": 1.7638, - "step": 12983 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019560846918758034, - "loss": 1.5686, - "step": 12984 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019560813096123928, - "loss": 1.6285, - "step": 12985 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956077927348982, - "loss": 1.7233, - "step": 12986 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019560745450855713, - "loss": 1.7484, - "step": 12987 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019560711628221607, - "loss": 1.5514, - "step": 12988 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195606778055875, - "loss": 1.5562, - "step": 12989 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019560643982953393, - "loss": 1.8074, - "step": 12990 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019560610160319287, - "loss": 1.6456, - "step": 12991 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019560576337685178, - "loss": 1.5888, - "step": 12992 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019560542515051073, - "loss": 1.485, - "step": 12993 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019560508692416967, - "loss": 1.6104, - "step": 12994 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019560474869782858, - "loss": 1.5662, - "step": 12995 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019560441047148752, - "loss": 1.6202, - "step": 12996 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019560407224514646, - "loss": 1.6308, - "step": 12997 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019560373401880538, - "loss": 1.6767, - "step": 12998 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019560339579246432, - "loss": 1.6121, - "step": 12999 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019560305756612326, - "loss": 1.5731, - "step": 13000 - }, - { - "epoch": 0.04, - "eval_loss": 1.48095703125, - "mrc_em": 18.0, - "mrc_f1": 26.515873015873023, - "nli_acc": 0.13, - "step": 13000, - "sts_f1": 0.6212121212121212, - "ynat_f1": 0.7033504213420324 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019560271933978217, - "loss": 1.572, - "step": 13001 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019560238111344111, - "loss": 1.6562, - "step": 13002 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019560204288710006, - "loss": 1.5724, - "step": 13003 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195601704660759, - "loss": 1.6105, - "step": 13004 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019560136643441794, - "loss": 1.6752, - "step": 13005 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019560102820807685, - "loss": 1.8778, - "step": 13006 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001956006899817358, - "loss": 1.543, - "step": 13007 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019560035175539473, - "loss": 1.5433, - "step": 13008 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019560001352905365, - "loss": 1.6228, - "step": 13009 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955996753027126, - "loss": 1.4753, - "step": 13010 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019559933707637153, - "loss": 1.522, - "step": 13011 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019559899885003044, - "loss": 1.6226, - "step": 13012 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019559866062368939, - "loss": 1.5852, - "step": 13013 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019559832239734833, - "loss": 1.5366, - "step": 13014 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019559798417100724, - "loss": 1.7195, - "step": 13015 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019559764594466618, - "loss": 1.6247, - "step": 13016 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019559730771832512, - "loss": 1.5821, - "step": 13017 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019559696949198404, - "loss": 1.6866, - "step": 13018 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019559663126564298, - "loss": 1.4261, - "step": 13019 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019559629303930192, - "loss": 1.6375, - "step": 13020 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019559595481296083, - "loss": 1.6077, - "step": 13021 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019559561658661977, - "loss": 1.5852, - "step": 13022 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019559527836027872, - "loss": 1.5106, - "step": 13023 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019559494013393763, - "loss": 1.6098, - "step": 13024 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019559460190759657, - "loss": 1.6196, - "step": 13025 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955942636812555, - "loss": 1.6417, - "step": 13026 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019559392545491445, - "loss": 1.5689, - "step": 13027 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019559358722857337, - "loss": 1.646, - "step": 13028 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955932490022323, - "loss": 1.7361, - "step": 13029 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019559291077589125, - "loss": 1.5806, - "step": 13030 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019559257254955016, - "loss": 1.4845, - "step": 13031 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955922343232091, - "loss": 1.688, - "step": 13032 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019559189609686805, - "loss": 1.5654, - "step": 13033 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019559155787052696, - "loss": 1.6283, - "step": 13034 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955912196441859, - "loss": 1.6236, - "step": 13035 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019559088141784484, - "loss": 1.6965, - "step": 13036 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019559054319150376, - "loss": 1.7158, - "step": 13037 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955902049651627, - "loss": 1.6277, - "step": 13038 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019558986673882164, - "loss": 1.6034, - "step": 13039 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019558952851248055, - "loss": 1.6569, - "step": 13040 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955891902861395, - "loss": 1.5536, - "step": 13041 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019558885205979843, - "loss": 1.6632, - "step": 13042 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019558851383345735, - "loss": 1.6083, - "step": 13043 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955881756071163, - "loss": 1.5128, - "step": 13044 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019558783738077523, - "loss": 1.6163, - "step": 13045 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019558749915443414, - "loss": 1.6423, - "step": 13046 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019558716092809309, - "loss": 1.6458, - "step": 13047 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019558682270175203, - "loss": 1.7961, - "step": 13048 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019558648447541094, - "loss": 1.5939, - "step": 13049 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019558614624906988, - "loss": 1.4824, - "step": 13050 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019558580802272882, - "loss": 1.6232, - "step": 13051 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019558546979638774, - "loss": 1.5273, - "step": 13052 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019558513157004668, - "loss": 1.7443, - "step": 13053 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019558479334370562, - "loss": 1.7738, - "step": 13054 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019558445511736453, - "loss": 1.5242, - "step": 13055 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019558411689102347, - "loss": 1.8682, - "step": 13056 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019558377866468242, - "loss": 1.7138, - "step": 13057 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019558344043834133, - "loss": 1.6341, - "step": 13058 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955831022120003, - "loss": 1.603, - "step": 13059 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955827639856592, - "loss": 1.5613, - "step": 13060 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019558242575931815, - "loss": 1.6161, - "step": 13061 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955820875329771, - "loss": 1.5775, - "step": 13062 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195581749306636, - "loss": 1.6143, - "step": 13063 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019558141108029495, - "loss": 1.5692, - "step": 13064 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955810728539539, - "loss": 1.4983, - "step": 13065 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955807346276128, - "loss": 1.6233, - "step": 13066 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019558039640127175, - "loss": 1.712, - "step": 13067 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955800581749307, - "loss": 1.4227, - "step": 13068 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955797199485896, - "loss": 1.4915, - "step": 13069 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019557938172224854, - "loss": 1.5066, - "step": 13070 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019557904349590748, - "loss": 1.6438, - "step": 13071 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955787052695664, - "loss": 1.5804, - "step": 13072 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019557836704322534, - "loss": 1.6615, - "step": 13073 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019557802881688428, - "loss": 1.6248, - "step": 13074 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955776905905432, - "loss": 1.5017, - "step": 13075 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019557735236420214, - "loss": 1.6394, - "step": 13076 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019557701413786108, - "loss": 1.6099, - "step": 13077 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019557667591152, - "loss": 1.5382, - "step": 13078 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019557633768517893, - "loss": 1.6712, - "step": 13079 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019557599945883787, - "loss": 1.6195, - "step": 13080 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955756612324968, - "loss": 1.5598, - "step": 13081 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019557532300615573, - "loss": 1.488, - "step": 13082 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019557498477981467, - "loss": 1.7443, - "step": 13083 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019557464655347358, - "loss": 1.6961, - "step": 13084 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019557430832713252, - "loss": 1.795, - "step": 13085 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019557397010079147, - "loss": 1.64, - "step": 13086 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019557363187445038, - "loss": 1.5827, - "step": 13087 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019557329364810932, - "loss": 1.6108, - "step": 13088 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019557295542176826, - "loss": 1.6511, - "step": 13089 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019557261719542718, - "loss": 1.6426, - "step": 13090 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019557227896908612, - "loss": 1.6275, - "step": 13091 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019557194074274506, - "loss": 1.5271, - "step": 13092 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019557160251640397, - "loss": 1.6674, - "step": 13093 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955712642900629, - "loss": 1.5134, - "step": 13094 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019557092606372185, - "loss": 1.6006, - "step": 13095 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019557058783738077, - "loss": 1.5059, - "step": 13096 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955702496110397, - "loss": 1.6367, - "step": 13097 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019556991138469865, - "loss": 1.5666, - "step": 13098 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019556957315835756, - "loss": 1.6627, - "step": 13099 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955692349320165, - "loss": 1.6625, - "step": 13100 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019556889670567545, - "loss": 1.4979, - "step": 13101 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019556855847933436, - "loss": 1.7821, - "step": 13102 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955682202529933, - "loss": 1.5515, - "step": 13103 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019556788202665224, - "loss": 1.5675, - "step": 13104 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019556754380031116, - "loss": 1.6408, - "step": 13105 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955672055739701, - "loss": 1.6134, - "step": 13106 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019556686734762904, - "loss": 1.7193, - "step": 13107 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019556652912128795, - "loss": 1.4909, - "step": 13108 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955661908949469, - "loss": 1.5369, - "step": 13109 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019556585266860584, - "loss": 1.4841, - "step": 13110 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019556551444226478, - "loss": 1.7022, - "step": 13111 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955651762159237, - "loss": 1.6429, - "step": 13112 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019556483798958263, - "loss": 1.6433, - "step": 13113 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019556449976324157, - "loss": 1.4941, - "step": 13114 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019556416153690051, - "loss": 1.7394, - "step": 13115 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019556382331055946, - "loss": 1.73, - "step": 13116 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019556348508421837, - "loss": 1.486, - "step": 13117 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955631468578773, - "loss": 1.6054, - "step": 13118 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019556280863153625, - "loss": 1.523, - "step": 13119 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019556247040519517, - "loss": 1.5708, - "step": 13120 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955621321788541, - "loss": 1.6758, - "step": 13121 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019556179395251305, - "loss": 1.7504, - "step": 13122 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019556145572617196, - "loss": 1.6835, - "step": 13123 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955611174998309, - "loss": 1.6006, - "step": 13124 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019556077927348984, - "loss": 1.6593, - "step": 13125 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019556044104714876, - "loss": 1.6342, - "step": 13126 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955601028208077, - "loss": 1.3899, - "step": 13127 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019555976459446664, - "loss": 1.538, - "step": 13128 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019555942636812555, - "loss": 1.4984, - "step": 13129 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955590881417845, - "loss": 1.5418, - "step": 13130 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019555874991544344, - "loss": 1.7026, - "step": 13131 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019555841168910235, - "loss": 1.6036, - "step": 13132 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955580734627613, - "loss": 1.6185, - "step": 13133 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019555773523642023, - "loss": 1.6157, - "step": 13134 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019555739701007915, - "loss": 1.6544, - "step": 13135 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955570587837381, - "loss": 1.7472, - "step": 13136 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019555672055739703, - "loss": 1.6939, - "step": 13137 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019555638233105594, - "loss": 1.6072, - "step": 13138 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019555604410471488, - "loss": 1.559, - "step": 13139 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019555570587837383, - "loss": 1.5021, - "step": 13140 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019555536765203274, - "loss": 1.5552, - "step": 13141 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019555502942569168, - "loss": 1.5579, - "step": 13142 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019555469119935062, - "loss": 1.666, - "step": 13143 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019555435297300954, - "loss": 1.4916, - "step": 13144 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019555401474666848, - "loss": 1.6849, - "step": 13145 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019555367652032742, - "loss": 1.5483, - "step": 13146 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019555333829398633, - "loss": 1.5731, - "step": 13147 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019555300006764527, - "loss": 1.6615, - "step": 13148 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019555266184130421, - "loss": 1.76, - "step": 13149 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019555232361496313, - "loss": 1.4716, - "step": 13150 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019555198538862207, - "loss": 1.6767, - "step": 13151 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195551647162281, - "loss": 1.5917, - "step": 13152 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019555130893593993, - "loss": 1.5691, - "step": 13153 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019555097070959887, - "loss": 1.4368, - "step": 13154 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955506324832578, - "loss": 1.6748, - "step": 13155 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019555029425691672, - "loss": 1.6028, - "step": 13156 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019554995603057566, - "loss": 1.5538, - "step": 13157 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955496178042346, - "loss": 1.6282, - "step": 13158 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019554927957789352, - "loss": 1.6279, - "step": 13159 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019554894135155246, - "loss": 1.3023, - "step": 13160 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955486031252114, - "loss": 1.5694, - "step": 13161 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019554826489887031, - "loss": 1.704, - "step": 13162 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019554792667252926, - "loss": 1.7488, - "step": 13163 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955475884461882, - "loss": 1.6574, - "step": 13164 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955472502198471, - "loss": 1.6465, - "step": 13165 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019554691199350605, - "loss": 1.5631, - "step": 13166 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195546573767165, - "loss": 1.6657, - "step": 13167 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955462355408239, - "loss": 1.6597, - "step": 13168 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019554589731448285, - "loss": 1.4994, - "step": 13169 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955455590881418, - "loss": 1.6454, - "step": 13170 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955452208618007, - "loss": 1.5543, - "step": 13171 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019554488263545967, - "loss": 1.6737, - "step": 13172 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955445444091186, - "loss": 1.7133, - "step": 13173 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019554420618277753, - "loss": 1.6196, - "step": 13174 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019554386795643647, - "loss": 1.5873, - "step": 13175 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955435297300954, - "loss": 1.714, - "step": 13176 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019554319150375432, - "loss": 1.5457, - "step": 13177 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019554285327741326, - "loss": 1.5504, - "step": 13178 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955425150510722, - "loss": 1.7611, - "step": 13179 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019554217682473112, - "loss": 1.5886, - "step": 13180 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019554183859839006, - "loss": 1.5632, - "step": 13181 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195541500372049, - "loss": 1.778, - "step": 13182 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019554116214570792, - "loss": 1.6748, - "step": 13183 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019554082391936686, - "loss": 1.5334, - "step": 13184 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955404856930258, - "loss": 1.6923, - "step": 13185 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955401474666847, - "loss": 1.5194, - "step": 13186 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019553980924034365, - "loss": 1.635, - "step": 13187 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955394710140026, - "loss": 1.4818, - "step": 13188 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955391327876615, - "loss": 1.6962, - "step": 13189 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019553879456132045, - "loss": 1.6936, - "step": 13190 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955384563349794, - "loss": 1.8847, - "step": 13191 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955381181086383, - "loss": 1.581, - "step": 13192 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019553777988229725, - "loss": 1.4889, - "step": 13193 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955374416559562, - "loss": 1.6036, - "step": 13194 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955371034296151, - "loss": 1.669, - "step": 13195 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019553676520327404, - "loss": 1.6305, - "step": 13196 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019553642697693298, - "loss": 1.4651, - "step": 13197 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955360887505919, - "loss": 1.6804, - "step": 13198 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019553575052425084, - "loss": 1.6547, - "step": 13199 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019553541229790978, - "loss": 1.6614, - "step": 13200 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955350740715687, - "loss": 1.567, - "step": 13201 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019553473584522763, - "loss": 1.6233, - "step": 13202 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019553439761888658, - "loss": 1.6447, - "step": 13203 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955340593925455, - "loss": 1.6743, - "step": 13204 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019553372116620443, - "loss": 1.6727, - "step": 13205 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019553338293986337, - "loss": 1.5501, - "step": 13206 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019553304471352229, - "loss": 1.6669, - "step": 13207 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019553270648718123, - "loss": 1.5513, - "step": 13208 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019553236826084017, - "loss": 1.6892, - "step": 13209 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019553203003449908, - "loss": 1.6515, - "step": 13210 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019553169180815802, - "loss": 1.6454, - "step": 13211 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019553135358181696, - "loss": 1.5679, - "step": 13212 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019553101535547588, - "loss": 1.7175, - "step": 13213 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019553067712913482, - "loss": 1.5416, - "step": 13214 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019553033890279376, - "loss": 1.6551, - "step": 13215 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019553000067645267, - "loss": 1.5952, - "step": 13216 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019552966245011162, - "loss": 1.6768, - "step": 13217 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019552932422377056, - "loss": 1.5202, - "step": 13218 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019552898599742947, - "loss": 1.6162, - "step": 13219 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955286477710884, - "loss": 1.8205, - "step": 13220 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019552830954474735, - "loss": 1.6066, - "step": 13221 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019552797131840627, - "loss": 1.5762, - "step": 13222 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955276330920652, - "loss": 1.8843, - "step": 13223 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019552729486572415, - "loss": 1.4652, - "step": 13224 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019552695663938306, - "loss": 1.3879, - "step": 13225 - }, - { - "epoch": 0.04, - "learning_rate": 0.000195526618413042, - "loss": 1.631, - "step": 13226 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019552628018670095, - "loss": 1.5924, - "step": 13227 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955259419603599, - "loss": 1.6624, - "step": 13228 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019552560373401883, - "loss": 1.733, - "step": 13229 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019552526550767777, - "loss": 1.4964, - "step": 13230 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019552492728133668, - "loss": 1.7276, - "step": 13231 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019552458905499562, - "loss": 1.713, - "step": 13232 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019552425082865457, - "loss": 1.5278, - "step": 13233 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019552391260231348, - "loss": 1.4339, - "step": 13234 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019552357437597242, - "loss": 1.6101, - "step": 13235 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019552323614963136, - "loss": 1.612, - "step": 13236 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019552289792329028, - "loss": 1.5878, - "step": 13237 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019552255969694922, - "loss": 1.5799, - "step": 13238 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019552222147060816, - "loss": 1.6478, - "step": 13239 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019552188324426707, - "loss": 1.5471, - "step": 13240 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019552154501792601, - "loss": 1.5798, - "step": 13241 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019552120679158495, - "loss": 1.6686, - "step": 13242 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019552086856524387, - "loss": 1.7793, - "step": 13243 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955205303389028, - "loss": 1.8114, - "step": 13244 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019552019211256175, - "loss": 1.6478, - "step": 13245 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019551985388622067, - "loss": 1.557, - "step": 13246 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955195156598796, - "loss": 1.625, - "step": 13247 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019551917743353855, - "loss": 1.5506, - "step": 13248 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019551883920719746, - "loss": 1.5009, - "step": 13249 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955185009808564, - "loss": 1.5576, - "step": 13250 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019551816275451534, - "loss": 1.5741, - "step": 13251 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019551782452817426, - "loss": 1.7734, - "step": 13252 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955174863018332, - "loss": 1.6088, - "step": 13253 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019551714807549214, - "loss": 1.5311, - "step": 13254 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019551680984915105, - "loss": 1.5661, - "step": 13255 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019551647162281, - "loss": 1.5105, - "step": 13256 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019551613339646894, - "loss": 1.4878, - "step": 13257 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019551579517012785, - "loss": 1.6891, - "step": 13258 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955154569437868, - "loss": 1.7214, - "step": 13259 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019551511871744573, - "loss": 1.6409, - "step": 13260 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019551478049110465, - "loss": 1.8046, - "step": 13261 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955144422647636, - "loss": 1.5434, - "step": 13262 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019551410403842253, - "loss": 1.553, - "step": 13263 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019551376581208144, - "loss": 1.7179, - "step": 13264 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019551342758574038, - "loss": 1.6902, - "step": 13265 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019551308935939933, - "loss": 1.4395, - "step": 13266 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019551275113305824, - "loss": 1.754, - "step": 13267 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019551241290671718, - "loss": 1.5761, - "step": 13268 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019551207468037612, - "loss": 1.4653, - "step": 13269 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019551173645403504, - "loss": 1.6233, - "step": 13270 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019551139822769398, - "loss": 1.4721, - "step": 13271 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019551106000135292, - "loss": 1.6345, - "step": 13272 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019551072177501183, - "loss": 1.7146, - "step": 13273 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019551038354867077, - "loss": 1.5436, - "step": 13274 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019551004532232971, - "loss": 1.7267, - "step": 13275 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019550970709598863, - "loss": 1.5008, - "step": 13276 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019550936886964757, - "loss": 1.6944, - "step": 13277 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955090306433065, - "loss": 1.6116, - "step": 13278 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019550869241696542, - "loss": 1.6157, - "step": 13279 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019550835419062437, - "loss": 1.6979, - "step": 13280 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955080159642833, - "loss": 1.4796, - "step": 13281 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019550767773794222, - "loss": 1.579, - "step": 13282 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019550733951160116, - "loss": 1.4815, - "step": 13283 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955070012852601, - "loss": 1.6188, - "step": 13284 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019550666305891904, - "loss": 1.6226, - "step": 13285 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019550632483257799, - "loss": 1.7029, - "step": 13286 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955059866062369, - "loss": 1.6503, - "step": 13287 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019550564837989584, - "loss": 1.7109, - "step": 13288 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019550531015355478, - "loss": 1.6587, - "step": 13289 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955049719272137, - "loss": 1.5075, - "step": 13290 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019550463370087264, - "loss": 1.8337, - "step": 13291 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019550429547453158, - "loss": 1.553, - "step": 13292 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019550395724819052, - "loss": 1.7904, - "step": 13293 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019550361902184943, - "loss": 1.5819, - "step": 13294 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019550328079550837, - "loss": 1.5566, - "step": 13295 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019550294256916732, - "loss": 1.6305, - "step": 13296 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019550260434282623, - "loss": 1.5313, - "step": 13297 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019550226611648517, - "loss": 1.5049, - "step": 13298 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955019278901441, - "loss": 1.5011, - "step": 13299 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019550158966380303, - "loss": 1.4266, - "step": 13300 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019550125143746197, - "loss": 1.5233, - "step": 13301 - }, - { - "epoch": 0.04, - "learning_rate": 0.0001955009132111209, - "loss": 1.5432, - "step": 13302 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019550057498477982, - "loss": 1.7061, - "step": 13303 - }, - { - "epoch": 0.04, - "learning_rate": 0.00019550023675843876, - "loss": 1.7918, - "step": 13304 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954998985320977, - "loss": 1.7977, - "step": 13305 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019549956030575662, - "loss": 1.5033, - "step": 13306 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019549922207941556, - "loss": 1.6197, - "step": 13307 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954988838530745, - "loss": 1.5304, - "step": 13308 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019549854562673341, - "loss": 1.7773, - "step": 13309 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019549820740039236, - "loss": 1.7083, - "step": 13310 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954978691740513, - "loss": 1.5606, - "step": 13311 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954975309477102, - "loss": 1.62, - "step": 13312 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019549719272136915, - "loss": 1.5762, - "step": 13313 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954968544950281, - "loss": 1.494, - "step": 13314 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195496516268687, - "loss": 1.6945, - "step": 13315 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019549617804234595, - "loss": 1.6842, - "step": 13316 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954958398160049, - "loss": 1.6345, - "step": 13317 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954955015896638, - "loss": 1.6405, - "step": 13318 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019549516336332274, - "loss": 1.6079, - "step": 13319 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019549482513698169, - "loss": 1.6393, - "step": 13320 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954944869106406, - "loss": 1.6065, - "step": 13321 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019549414868429954, - "loss": 1.6337, - "step": 13322 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019549381045795848, - "loss": 1.5859, - "step": 13323 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954934722316174, - "loss": 1.6809, - "step": 13324 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019549313400527634, - "loss": 1.6135, - "step": 13325 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019549279577893528, - "loss": 1.5724, - "step": 13326 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954924575525942, - "loss": 1.6359, - "step": 13327 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019549211932625313, - "loss": 1.7374, - "step": 13328 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019549178109991208, - "loss": 1.51, - "step": 13329 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195491442873571, - "loss": 1.6349, - "step": 13330 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019549110464722993, - "loss": 1.5701, - "step": 13331 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019549076642088887, - "loss": 1.7516, - "step": 13332 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019549042819454779, - "loss": 1.6918, - "step": 13333 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019549008996820673, - "loss": 1.5029, - "step": 13334 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019548975174186567, - "loss": 1.5815, - "step": 13335 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019548941351552458, - "loss": 1.5446, - "step": 13336 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019548907528918352, - "loss": 1.492, - "step": 13337 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019548873706284246, - "loss": 1.64, - "step": 13338 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019548839883650138, - "loss": 1.4958, - "step": 13339 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019548806061016035, - "loss": 1.655, - "step": 13340 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019548772238381926, - "loss": 1.6468, - "step": 13341 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954873841574782, - "loss": 1.4934, - "step": 13342 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019548704593113714, - "loss": 1.6716, - "step": 13343 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019548670770479606, - "loss": 1.4543, - "step": 13344 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195486369478455, - "loss": 1.7613, - "step": 13345 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019548603125211394, - "loss": 1.6071, - "step": 13346 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019548569302577285, - "loss": 1.6327, - "step": 13347 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954853547994318, - "loss": 1.5499, - "step": 13348 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019548501657309074, - "loss": 1.6079, - "step": 13349 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019548467834674965, - "loss": 1.6186, - "step": 13350 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954843401204086, - "loss": 1.6462, - "step": 13351 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019548400189406753, - "loss": 1.6673, - "step": 13352 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019548366366772645, - "loss": 1.5677, - "step": 13353 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954833254413854, - "loss": 1.4286, - "step": 13354 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019548298721504433, - "loss": 1.502, - "step": 13355 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019548264898870324, - "loss": 1.76, - "step": 13356 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019548231076236218, - "loss": 1.5499, - "step": 13357 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019548197253602112, - "loss": 1.7736, - "step": 13358 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019548163430968004, - "loss": 1.5348, - "step": 13359 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019548129608333898, - "loss": 1.6417, - "step": 13360 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019548095785699792, - "loss": 1.6569, - "step": 13361 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019548061963065683, - "loss": 1.5623, - "step": 13362 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019548028140431578, - "loss": 1.7737, - "step": 13363 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019547994317797472, - "loss": 1.5907, - "step": 13364 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019547960495163363, - "loss": 1.4751, - "step": 13365 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019547926672529257, - "loss": 1.4844, - "step": 13366 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954789284989515, - "loss": 1.6758, - "step": 13367 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019547859027261043, - "loss": 1.4681, - "step": 13368 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019547825204626937, - "loss": 1.6036, - "step": 13369 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954779138199283, - "loss": 1.7595, - "step": 13370 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019547757559358722, - "loss": 1.6797, - "step": 13371 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019547723736724616, - "loss": 1.6574, - "step": 13372 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954768991409051, - "loss": 1.5203, - "step": 13373 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019547656091456402, - "loss": 1.6527, - "step": 13374 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019547622268822296, - "loss": 1.5947, - "step": 13375 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954758844618819, - "loss": 1.5451, - "step": 13376 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019547554623554082, - "loss": 1.6547, - "step": 13377 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019547520800919976, - "loss": 1.7415, - "step": 13378 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954748697828587, - "loss": 1.3256, - "step": 13379 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019547453155651764, - "loss": 1.6335, - "step": 13380 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019547419333017655, - "loss": 1.6341, - "step": 13381 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954738551038355, - "loss": 1.6006, - "step": 13382 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019547351687749444, - "loss": 1.5899, - "step": 13383 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019547317865115335, - "loss": 1.7725, - "step": 13384 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954728404248123, - "loss": 1.5092, - "step": 13385 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019547250219847123, - "loss": 1.6551, - "step": 13386 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019547216397213015, - "loss": 1.5449, - "step": 13387 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954718257457891, - "loss": 1.6891, - "step": 13388 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019547148751944803, - "loss": 1.5319, - "step": 13389 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019547114929310694, - "loss": 1.6646, - "step": 13390 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019547081106676588, - "loss": 1.5341, - "step": 13391 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019547047284042482, - "loss": 1.6022, - "step": 13392 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019547013461408374, - "loss": 1.5717, - "step": 13393 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019546979638774268, - "loss": 1.5929, - "step": 13394 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019546945816140162, - "loss": 1.4941, - "step": 13395 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019546911993506056, - "loss": 1.6384, - "step": 13396 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954687817087195, - "loss": 1.7916, - "step": 13397 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019546844348237842, - "loss": 1.6327, - "step": 13398 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019546810525603736, - "loss": 1.6793, - "step": 13399 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954677670296963, - "loss": 1.6293, - "step": 13400 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019546742880335521, - "loss": 1.7093, - "step": 13401 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019546709057701415, - "loss": 1.7263, - "step": 13402 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954667523506731, - "loss": 1.6165, - "step": 13403 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195466414124332, - "loss": 1.456, - "step": 13404 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019546607589799095, - "loss": 1.6715, - "step": 13405 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954657376716499, - "loss": 1.5638, - "step": 13406 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954653994453088, - "loss": 1.5445, - "step": 13407 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019546506121896775, - "loss": 1.6752, - "step": 13408 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954647229926267, - "loss": 1.5736, - "step": 13409 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954643847662856, - "loss": 1.5578, - "step": 13410 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019546404653994454, - "loss": 1.7918, - "step": 13411 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019546370831360348, - "loss": 1.6437, - "step": 13412 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954633700872624, - "loss": 1.6863, - "step": 13413 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019546303186092134, - "loss": 1.5584, - "step": 13414 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019546269363458028, - "loss": 1.7387, - "step": 13415 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954623554082392, - "loss": 1.7735, - "step": 13416 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019546201718189814, - "loss": 1.6664, - "step": 13417 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019546167895555708, - "loss": 1.6351, - "step": 13418 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195461340729216, - "loss": 1.7114, - "step": 13419 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019546100250287493, - "loss": 1.5977, - "step": 13420 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019546066427653387, - "loss": 1.7778, - "step": 13421 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954603260501928, - "loss": 1.706, - "step": 13422 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019545998782385173, - "loss": 1.526, - "step": 13423 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019545964959751067, - "loss": 1.6348, - "step": 13424 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019545931137116958, - "loss": 1.6514, - "step": 13425 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019545897314482853, - "loss": 1.5787, - "step": 13426 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019545863491848747, - "loss": 1.6564, - "step": 13427 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019545829669214638, - "loss": 1.5704, - "step": 13428 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019545795846580532, - "loss": 1.6125, - "step": 13429 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019545762023946426, - "loss": 1.554, - "step": 13430 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019545728201312318, - "loss": 1.6878, - "step": 13431 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019545694378678212, - "loss": 1.626, - "step": 13432 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019545660556044106, - "loss": 1.6609, - "step": 13433 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019545626733409997, - "loss": 1.5592, - "step": 13434 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019545592910775891, - "loss": 1.5387, - "step": 13435 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019545559088141786, - "loss": 1.6935, - "step": 13436 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019545525265507677, - "loss": 1.4937, - "step": 13437 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954549144287357, - "loss": 1.5595, - "step": 13438 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019545457620239465, - "loss": 1.6414, - "step": 13439 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019545423797605357, - "loss": 1.5896, - "step": 13440 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954538997497125, - "loss": 1.481, - "step": 13441 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019545356152337145, - "loss": 1.7247, - "step": 13442 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019545322329703036, - "loss": 1.6169, - "step": 13443 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954528850706893, - "loss": 1.569, - "step": 13444 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019545254684434824, - "loss": 1.6123, - "step": 13445 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019545220861800716, - "loss": 1.6769, - "step": 13446 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954518703916661, - "loss": 1.702, - "step": 13447 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019545153216532504, - "loss": 1.52, - "step": 13448 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019545119393898395, - "loss": 1.6464, - "step": 13449 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954508557126429, - "loss": 1.7681, - "step": 13450 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019545051748630184, - "loss": 1.6771, - "step": 13451 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019545017925996078, - "loss": 1.5861, - "step": 13452 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019544984103361972, - "loss": 1.6879, - "step": 13453 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019544950280727866, - "loss": 1.6102, - "step": 13454 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019544916458093757, - "loss": 1.4998, - "step": 13455 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019544882635459652, - "loss": 1.6617, - "step": 13456 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019544848812825546, - "loss": 1.568, - "step": 13457 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019544814990191437, - "loss": 1.4836, - "step": 13458 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954478116755733, - "loss": 1.6386, - "step": 13459 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019544747344923225, - "loss": 1.5078, - "step": 13460 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019544713522289117, - "loss": 1.6289, - "step": 13461 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954467969965501, - "loss": 1.6494, - "step": 13462 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019544645877020905, - "loss": 1.5356, - "step": 13463 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019544612054386796, - "loss": 1.5253, - "step": 13464 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954457823175269, - "loss": 1.6264, - "step": 13465 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019544544409118585, - "loss": 1.7757, - "step": 13466 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019544510586484476, - "loss": 1.523, - "step": 13467 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954447676385037, - "loss": 1.6554, - "step": 13468 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019544442941216264, - "loss": 1.486, - "step": 13469 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019544409118582156, - "loss": 1.5611, - "step": 13470 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954437529594805, - "loss": 1.6373, - "step": 13471 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019544341473313944, - "loss": 1.6493, - "step": 13472 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019544307650679835, - "loss": 1.489, - "step": 13473 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954427382804573, - "loss": 1.758, - "step": 13474 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019544240005411623, - "loss": 1.7987, - "step": 13475 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019544206182777515, - "loss": 1.6395, - "step": 13476 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954417236014341, - "loss": 1.7424, - "step": 13477 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019544138537509303, - "loss": 1.6585, - "step": 13478 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019544104714875194, - "loss": 1.5361, - "step": 13479 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019544070892241089, - "loss": 1.5195, - "step": 13480 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019544037069606983, - "loss": 1.5239, - "step": 13481 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019544003246972874, - "loss": 1.6501, - "step": 13482 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019543969424338768, - "loss": 1.5815, - "step": 13483 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019543935601704662, - "loss": 1.4778, - "step": 13484 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019543901779070554, - "loss": 1.6527, - "step": 13485 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019543867956436448, - "loss": 1.5508, - "step": 13486 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019543834133802342, - "loss": 1.7128, - "step": 13487 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019543800311168233, - "loss": 1.555, - "step": 13488 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019543766488534127, - "loss": 1.581, - "step": 13489 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019543732665900022, - "loss": 1.6482, - "step": 13490 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019543698843265913, - "loss": 1.4862, - "step": 13491 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019543665020631807, - "loss": 1.7428, - "step": 13492 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195436311979977, - "loss": 1.471, - "step": 13493 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019543597375363593, - "loss": 1.7817, - "step": 13494 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019543563552729487, - "loss": 1.5885, - "step": 13495 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954352973009538, - "loss": 1.5617, - "step": 13496 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019543495907461272, - "loss": 1.5773, - "step": 13497 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019543462084827166, - "loss": 1.6031, - "step": 13498 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954342826219306, - "loss": 1.5214, - "step": 13499 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019543394439558952, - "loss": 1.7332, - "step": 13500 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019543360616924846, - "loss": 1.6568, - "step": 13501 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954332679429074, - "loss": 1.6819, - "step": 13502 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019543292971656632, - "loss": 1.4878, - "step": 13503 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019543259149022526, - "loss": 1.4572, - "step": 13504 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954322532638842, - "loss": 1.5119, - "step": 13505 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954319150375431, - "loss": 1.7369, - "step": 13506 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019543157681120205, - "loss": 1.6544, - "step": 13507 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195431238584861, - "loss": 1.6646, - "step": 13508 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019543090035851994, - "loss": 1.5212, - "step": 13509 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019543056213217888, - "loss": 1.7044, - "step": 13510 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019543022390583782, - "loss": 1.7431, - "step": 13511 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019542988567949673, - "loss": 1.7379, - "step": 13512 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019542954745315567, - "loss": 1.6583, - "step": 13513 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019542920922681461, - "loss": 1.7582, - "step": 13514 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019542887100047353, - "loss": 1.7237, - "step": 13515 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019542853277413247, - "loss": 1.742, - "step": 13516 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954281945477914, - "loss": 1.6007, - "step": 13517 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019542785632145032, - "loss": 1.5905, - "step": 13518 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019542751809510927, - "loss": 1.7091, - "step": 13519 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954271798687682, - "loss": 1.5779, - "step": 13520 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019542684164242712, - "loss": 1.6, - "step": 13521 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019542650341608606, - "loss": 1.5852, - "step": 13522 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195426165189745, - "loss": 1.566, - "step": 13523 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019542582696340392, - "loss": 1.5576, - "step": 13524 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019542548873706286, - "loss": 1.6957, - "step": 13525 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954251505107218, - "loss": 1.4599, - "step": 13526 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954248122843807, - "loss": 1.6697, - "step": 13527 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019542447405803965, - "loss": 1.7043, - "step": 13528 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954241358316986, - "loss": 1.8165, - "step": 13529 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954237976053575, - "loss": 1.6688, - "step": 13530 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019542345937901645, - "loss": 1.7227, - "step": 13531 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954231211526754, - "loss": 1.3969, - "step": 13532 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954227829263343, - "loss": 1.6551, - "step": 13533 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019542244469999325, - "loss": 1.6944, - "step": 13534 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954221064736522, - "loss": 1.6949, - "step": 13535 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954217682473111, - "loss": 1.7185, - "step": 13536 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019542143002097004, - "loss": 1.6867, - "step": 13537 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019542109179462898, - "loss": 1.53, - "step": 13538 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954207535682879, - "loss": 1.592, - "step": 13539 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019542041534194684, - "loss": 1.7138, - "step": 13540 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019542007711560578, - "loss": 1.5786, - "step": 13541 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954197388892647, - "loss": 1.6431, - "step": 13542 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019541940066292364, - "loss": 1.6334, - "step": 13543 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019541906243658258, - "loss": 1.5662, - "step": 13544 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954187242102415, - "loss": 1.614, - "step": 13545 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019541838598390043, - "loss": 1.4998, - "step": 13546 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019541804775755937, - "loss": 1.7355, - "step": 13547 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954177095312183, - "loss": 1.4942, - "step": 13548 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019541737130487723, - "loss": 1.477, - "step": 13549 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019541703307853617, - "loss": 1.4917, - "step": 13550 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019541669485219508, - "loss": 1.4335, - "step": 13551 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019541635662585402, - "loss": 1.6616, - "step": 13552 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019541601839951297, - "loss": 1.7638, - "step": 13553 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019541568017317188, - "loss": 1.6569, - "step": 13554 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019541534194683082, - "loss": 1.6267, - "step": 13555 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019541500372048976, - "loss": 1.6711, - "step": 13556 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019541466549414868, - "loss": 1.7527, - "step": 13557 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019541432726780762, - "loss": 1.5858, - "step": 13558 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019541398904146656, - "loss": 1.4922, - "step": 13559 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019541365081512547, - "loss": 1.5257, - "step": 13560 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019541331258878441, - "loss": 1.5216, - "step": 13561 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019541297436244335, - "loss": 1.4924, - "step": 13562 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019541263613610227, - "loss": 1.6183, - "step": 13563 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954122979097612, - "loss": 1.7099, - "step": 13564 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019541195968342018, - "loss": 1.6042, - "step": 13565 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954116214570791, - "loss": 1.686, - "step": 13566 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019541128323073803, - "loss": 1.5355, - "step": 13567 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019541094500439697, - "loss": 1.7152, - "step": 13568 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954106067780559, - "loss": 1.5748, - "step": 13569 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019541026855171483, - "loss": 1.4866, - "step": 13570 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019540993032537377, - "loss": 1.6326, - "step": 13571 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019540959209903268, - "loss": 1.6491, - "step": 13572 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019540925387269163, - "loss": 1.6829, - "step": 13573 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019540891564635057, - "loss": 1.6392, - "step": 13574 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019540857742000948, - "loss": 1.6032, - "step": 13575 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019540823919366842, - "loss": 1.6471, - "step": 13576 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019540790096732736, - "loss": 1.638, - "step": 13577 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019540756274098628, - "loss": 1.6298, - "step": 13578 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019540722451464522, - "loss": 1.5228, - "step": 13579 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019540688628830416, - "loss": 1.5171, - "step": 13580 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019540654806196307, - "loss": 1.5341, - "step": 13581 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019540620983562201, - "loss": 1.441, - "step": 13582 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019540587160928096, - "loss": 1.5925, - "step": 13583 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019540553338293987, - "loss": 1.6124, - "step": 13584 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954051951565988, - "loss": 1.6701, - "step": 13585 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019540485693025775, - "loss": 1.5696, - "step": 13586 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019540451870391667, - "loss": 1.497, - "step": 13587 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954041804775756, - "loss": 1.6883, - "step": 13588 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019540384225123455, - "loss": 1.599, - "step": 13589 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019540350402489346, - "loss": 1.4635, - "step": 13590 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954031657985524, - "loss": 1.6195, - "step": 13591 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019540282757221135, - "loss": 1.6784, - "step": 13592 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019540248934587026, - "loss": 1.6355, - "step": 13593 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954021511195292, - "loss": 1.4416, - "step": 13594 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019540181289318814, - "loss": 1.5524, - "step": 13595 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019540147466684706, - "loss": 1.5896, - "step": 13596 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195401136440506, - "loss": 1.5531, - "step": 13597 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019540079821416494, - "loss": 1.7841, - "step": 13598 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019540045998782385, - "loss": 1.6742, - "step": 13599 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001954001217614828, - "loss": 1.6457, - "step": 13600 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019539978353514173, - "loss": 1.6236, - "step": 13601 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019539944530880065, - "loss": 1.5175, - "step": 13602 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953991070824596, - "loss": 1.5959, - "step": 13603 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019539876885611853, - "loss": 1.6735, - "step": 13604 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019539843062977744, - "loss": 1.7182, - "step": 13605 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019539809240343639, - "loss": 1.5147, - "step": 13606 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019539775417709533, - "loss": 1.5399, - "step": 13607 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019539741595075424, - "loss": 1.6213, - "step": 13608 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019539707772441318, - "loss": 1.7917, - "step": 13609 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019539673949807212, - "loss": 1.5176, - "step": 13610 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019539640127173104, - "loss": 1.6397, - "step": 13611 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019539606304538998, - "loss": 1.602, - "step": 13612 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019539572481904892, - "loss": 1.7404, - "step": 13613 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019539538659270783, - "loss": 1.4586, - "step": 13614 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019539504836636677, - "loss": 1.6748, - "step": 13615 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019539471014002572, - "loss": 1.6035, - "step": 13616 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019539437191368463, - "loss": 1.5383, - "step": 13617 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019539403368734357, - "loss": 1.6072, - "step": 13618 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953936954610025, - "loss": 1.5399, - "step": 13619 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019539335723466143, - "loss": 1.5762, - "step": 13620 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953930190083204, - "loss": 1.5526, - "step": 13621 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953926807819793, - "loss": 1.6369, - "step": 13622 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019539234255563825, - "loss": 1.5665, - "step": 13623 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953920043292972, - "loss": 1.7362, - "step": 13624 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953916661029561, - "loss": 1.4727, - "step": 13625 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019539132787661505, - "loss": 1.6078, - "step": 13626 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195390989650274, - "loss": 1.5645, - "step": 13627 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953906514239329, - "loss": 1.5778, - "step": 13628 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019539031319759184, - "loss": 1.5201, - "step": 13629 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019538997497125078, - "loss": 1.629, - "step": 13630 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953896367449097, - "loss": 1.5479, - "step": 13631 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019538929851856864, - "loss": 1.6003, - "step": 13632 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019538896029222758, - "loss": 1.4947, - "step": 13633 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953886220658865, - "loss": 1.6069, - "step": 13634 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019538828383954543, - "loss": 1.722, - "step": 13635 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019538794561320438, - "loss": 1.6012, - "step": 13636 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953876073868633, - "loss": 1.6883, - "step": 13637 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019538726916052223, - "loss": 1.5858, - "step": 13638 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019538693093418117, - "loss": 1.6037, - "step": 13639 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019538659270784009, - "loss": 1.6096, - "step": 13640 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019538625448149903, - "loss": 1.5879, - "step": 13641 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019538591625515797, - "loss": 1.4753, - "step": 13642 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019538557802881688, - "loss": 1.597, - "step": 13643 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019538523980247582, - "loss": 1.7485, - "step": 13644 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019538490157613476, - "loss": 1.6297, - "step": 13645 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953845633497937, - "loss": 1.6151, - "step": 13646 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019538422512345262, - "loss": 1.5833, - "step": 13647 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019538388689711156, - "loss": 1.5831, - "step": 13648 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953835486707705, - "loss": 1.5319, - "step": 13649 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019538321044442942, - "loss": 1.6025, - "step": 13650 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019538287221808836, - "loss": 1.6117, - "step": 13651 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953825339917473, - "loss": 1.5757, - "step": 13652 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953821957654062, - "loss": 1.7215, - "step": 13653 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019538185753906515, - "loss": 1.7087, - "step": 13654 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953815193127241, - "loss": 1.4806, - "step": 13655 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195381181086383, - "loss": 1.6567, - "step": 13656 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019538084286004195, - "loss": 1.6347, - "step": 13657 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953805046337009, - "loss": 1.7459, - "step": 13658 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953801664073598, - "loss": 1.7047, - "step": 13659 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019537982818101875, - "loss": 1.7005, - "step": 13660 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953794899546777, - "loss": 1.5477, - "step": 13661 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953791517283366, - "loss": 1.5913, - "step": 13662 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019537881350199554, - "loss": 1.6774, - "step": 13663 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019537847527565448, - "loss": 1.6091, - "step": 13664 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953781370493134, - "loss": 1.6234, - "step": 13665 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019537779882297234, - "loss": 1.6499, - "step": 13666 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019537746059663128, - "loss": 1.7139, - "step": 13667 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953771223702902, - "loss": 1.7443, - "step": 13668 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019537678414394914, - "loss": 1.6831, - "step": 13669 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019537644591760808, - "loss": 1.5544, - "step": 13670 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195376107691267, - "loss": 1.7236, - "step": 13671 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019537576946492593, - "loss": 1.6867, - "step": 13672 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019537543123858487, - "loss": 1.3672, - "step": 13673 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953750930122438, - "loss": 1.6875, - "step": 13674 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019537475478590273, - "loss": 1.7375, - "step": 13675 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019537441655956167, - "loss": 1.3863, - "step": 13676 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953740783332206, - "loss": 1.4451, - "step": 13677 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019537374010687955, - "loss": 1.5888, - "step": 13678 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019537340188053847, - "loss": 1.6255, - "step": 13679 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953730636541974, - "loss": 1.6976, - "step": 13680 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019537272542785635, - "loss": 1.5047, - "step": 13681 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019537238720151526, - "loss": 1.5783, - "step": 13682 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953720489751742, - "loss": 1.7665, - "step": 13683 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019537171074883314, - "loss": 1.6194, - "step": 13684 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019537137252249206, - "loss": 1.6244, - "step": 13685 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195371034296151, - "loss": 1.7383, - "step": 13686 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019537069606980994, - "loss": 1.6756, - "step": 13687 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019537035784346885, - "loss": 1.7174, - "step": 13688 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953700196171278, - "loss": 1.6431, - "step": 13689 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019536968139078674, - "loss": 1.5835, - "step": 13690 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019536934316444565, - "loss": 1.5357, - "step": 13691 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953690049381046, - "loss": 1.6925, - "step": 13692 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019536866671176353, - "loss": 1.5775, - "step": 13693 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019536832848542245, - "loss": 1.5434, - "step": 13694 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953679902590814, - "loss": 1.6037, - "step": 13695 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019536765203274033, - "loss": 1.4893, - "step": 13696 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019536731380639924, - "loss": 1.6017, - "step": 13697 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019536697558005818, - "loss": 1.5193, - "step": 13698 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019536663735371713, - "loss": 1.6113, - "step": 13699 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019536629912737604, - "loss": 1.7245, - "step": 13700 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019536596090103498, - "loss": 1.542, - "step": 13701 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019536562267469392, - "loss": 1.6341, - "step": 13702 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019536528444835284, - "loss": 1.5565, - "step": 13703 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019536494622201178, - "loss": 1.7234, - "step": 13704 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019536460799567072, - "loss": 1.7541, - "step": 13705 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019536426976932963, - "loss": 1.7182, - "step": 13706 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019536393154298857, - "loss": 1.6024, - "step": 13707 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019536359331664751, - "loss": 1.7434, - "step": 13708 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019536325509030643, - "loss": 1.6382, - "step": 13709 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019536291686396537, - "loss": 1.5459, - "step": 13710 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953625786376243, - "loss": 1.7502, - "step": 13711 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019536224041128322, - "loss": 1.4844, - "step": 13712 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019536190218494217, - "loss": 1.7345, - "step": 13713 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953615639586011, - "loss": 1.6238, - "step": 13714 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019536122573226002, - "loss": 1.6286, - "step": 13715 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019536088750591896, - "loss": 1.4457, - "step": 13716 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953605492795779, - "loss": 1.5583, - "step": 13717 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019536021105323682, - "loss": 1.725, - "step": 13718 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019535987282689576, - "loss": 1.6289, - "step": 13719 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953595346005547, - "loss": 1.5303, - "step": 13720 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019535919637421361, - "loss": 1.4038, - "step": 13721 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019535885814787255, - "loss": 1.4961, - "step": 13722 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953585199215315, - "loss": 1.8612, - "step": 13723 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953581816951904, - "loss": 1.5367, - "step": 13724 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019535784346884935, - "loss": 1.6567, - "step": 13725 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953575052425083, - "loss": 1.4748, - "step": 13726 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953571670161672, - "loss": 1.6485, - "step": 13727 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019535682878982615, - "loss": 1.6601, - "step": 13728 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953564905634851, - "loss": 1.504, - "step": 13729 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019535615233714403, - "loss": 1.6253, - "step": 13730 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019535581411080294, - "loss": 1.7354, - "step": 13731 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019535547588446188, - "loss": 1.6711, - "step": 13732 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019535513765812083, - "loss": 1.7837, - "step": 13733 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019535479943177977, - "loss": 1.6248, - "step": 13734 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953544612054387, - "loss": 1.6488, - "step": 13735 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019535412297909762, - "loss": 1.5205, - "step": 13736 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019535378475275656, - "loss": 1.6242, - "step": 13737 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953534465264155, - "loss": 1.571, - "step": 13738 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019535310830007442, - "loss": 1.6224, - "step": 13739 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019535277007373336, - "loss": 1.5711, - "step": 13740 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953524318473923, - "loss": 1.6148, - "step": 13741 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019535209362105121, - "loss": 1.5055, - "step": 13742 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019535175539471016, - "loss": 1.6752, - "step": 13743 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953514171683691, - "loss": 1.4675, - "step": 13744 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195351078942028, - "loss": 1.645, - "step": 13745 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019535074071568695, - "loss": 1.5668, - "step": 13746 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953504024893459, - "loss": 1.6168, - "step": 13747 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953500642630048, - "loss": 1.5877, - "step": 13748 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019534972603666375, - "loss": 1.8269, - "step": 13749 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953493878103227, - "loss": 1.5461, - "step": 13750 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953490495839816, - "loss": 1.7611, - "step": 13751 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019534871135764054, - "loss": 1.5649, - "step": 13752 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019534837313129949, - "loss": 1.62, - "step": 13753 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953480349049584, - "loss": 1.6412, - "step": 13754 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019534769667861734, - "loss": 1.6369, - "step": 13755 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019534735845227628, - "loss": 1.654, - "step": 13756 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953470202259352, - "loss": 1.5946, - "step": 13757 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019534668199959414, - "loss": 1.6384, - "step": 13758 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019534634377325308, - "loss": 1.4364, - "step": 13759 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195346005546912, - "loss": 1.6239, - "step": 13760 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019534566732057093, - "loss": 1.3663, - "step": 13761 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019534532909422988, - "loss": 1.5948, - "step": 13762 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953449908678888, - "loss": 1.5916, - "step": 13763 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019534465264154773, - "loss": 1.6406, - "step": 13764 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019534431441520667, - "loss": 1.5488, - "step": 13765 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019534397618886559, - "loss": 1.6696, - "step": 13766 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019534363796252453, - "loss": 1.4341, - "step": 13767 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019534329973618347, - "loss": 1.5169, - "step": 13768 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019534296150984238, - "loss": 1.5663, - "step": 13769 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019534262328350132, - "loss": 1.5811, - "step": 13770 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019534228505716026, - "loss": 1.6056, - "step": 13771 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019534194683081918, - "loss": 1.7549, - "step": 13772 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019534160860447812, - "loss": 1.6257, - "step": 13773 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019534127037813706, - "loss": 1.7392, - "step": 13774 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019534093215179597, - "loss": 1.7647, - "step": 13775 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019534059392545492, - "loss": 1.7253, - "step": 13776 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019534025569911386, - "loss": 1.4875, - "step": 13777 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019533991747277277, - "loss": 1.6626, - "step": 13778 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953395792464317, - "loss": 1.6789, - "step": 13779 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019533924102009065, - "loss": 1.4631, - "step": 13780 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019533890279374957, - "loss": 1.6728, - "step": 13781 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953385645674085, - "loss": 1.7556, - "step": 13782 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019533822634106745, - "loss": 1.7011, - "step": 13783 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019533788811472636, - "loss": 1.7101, - "step": 13784 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953375498883853, - "loss": 1.6735, - "step": 13785 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019533721166204425, - "loss": 1.5179, - "step": 13786 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019533687343570316, - "loss": 1.651, - "step": 13787 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953365352093621, - "loss": 1.5234, - "step": 13788 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019533619698302107, - "loss": 1.6097, - "step": 13789 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019533585875667998, - "loss": 1.635, - "step": 13790 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019533552053033892, - "loss": 1.541, - "step": 13791 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019533518230399787, - "loss": 1.5348, - "step": 13792 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019533484407765678, - "loss": 1.5846, - "step": 13793 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019533450585131572, - "loss": 1.6894, - "step": 13794 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019533416762497466, - "loss": 1.6954, - "step": 13795 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019533382939863358, - "loss": 1.6871, - "step": 13796 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019533349117229252, - "loss": 1.549, - "step": 13797 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019533315294595146, - "loss": 1.4772, - "step": 13798 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019533281471961037, - "loss": 1.7256, - "step": 13799 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953324764932693, - "loss": 1.4989, - "step": 13800 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019533213826692825, - "loss": 1.4964, - "step": 13801 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019533180004058717, - "loss": 1.5728, - "step": 13802 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953314618142461, - "loss": 1.4513, - "step": 13803 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019533112358790505, - "loss": 1.6614, - "step": 13804 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019533078536156396, - "loss": 1.7106, - "step": 13805 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953304471352229, - "loss": 1.5682, - "step": 13806 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019533010890888185, - "loss": 1.5968, - "step": 13807 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019532977068254076, - "loss": 1.5117, - "step": 13808 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953294324561997, - "loss": 1.6713, - "step": 13809 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019532909422985864, - "loss": 1.5788, - "step": 13810 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019532875600351756, - "loss": 1.5325, - "step": 13811 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953284177771765, - "loss": 1.6496, - "step": 13812 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019532807955083544, - "loss": 1.6727, - "step": 13813 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019532774132449435, - "loss": 1.5867, - "step": 13814 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953274030981533, - "loss": 1.5925, - "step": 13815 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019532706487181224, - "loss": 1.8732, - "step": 13816 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019532672664547115, - "loss": 1.5115, - "step": 13817 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953263884191301, - "loss": 1.846, - "step": 13818 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019532605019278903, - "loss": 1.6291, - "step": 13819 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019532571196644795, - "loss": 1.7071, - "step": 13820 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953253737401069, - "loss": 1.5208, - "step": 13821 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019532503551376583, - "loss": 1.6135, - "step": 13822 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019532469728742474, - "loss": 1.7662, - "step": 13823 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019532435906108368, - "loss": 1.5025, - "step": 13824 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019532402083474262, - "loss": 1.5057, - "step": 13825 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019532368260840154, - "loss": 1.696, - "step": 13826 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019532334438206048, - "loss": 1.6574, - "step": 13827 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019532300615571942, - "loss": 1.7518, - "step": 13828 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019532266792937834, - "loss": 1.6808, - "step": 13829 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019532232970303728, - "loss": 1.6949, - "step": 13830 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019532199147669622, - "loss": 1.6228, - "step": 13831 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019532165325035513, - "loss": 1.5156, - "step": 13832 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019532131502401407, - "loss": 1.6578, - "step": 13833 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019532097679767301, - "loss": 1.5094, - "step": 13834 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019532063857133193, - "loss": 1.6554, - "step": 13835 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019532030034499087, - "loss": 1.5553, - "step": 13836 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953199621186498, - "loss": 1.6982, - "step": 13837 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019531962389230872, - "loss": 1.7532, - "step": 13838 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019531928566596767, - "loss": 1.6297, - "step": 13839 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953189474396266, - "loss": 1.6509, - "step": 13840 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019531860921328552, - "loss": 1.6342, - "step": 13841 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019531827098694446, - "loss": 1.7254, - "step": 13842 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953179327606034, - "loss": 1.6986, - "step": 13843 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019531759453426232, - "loss": 1.6284, - "step": 13844 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019531725630792126, - "loss": 1.7968, - "step": 13845 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019531691808158023, - "loss": 1.5894, - "step": 13846 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019531657985523914, - "loss": 1.6194, - "step": 13847 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019531624162889808, - "loss": 1.7593, - "step": 13848 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019531590340255702, - "loss": 1.7912, - "step": 13849 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019531556517621594, - "loss": 1.8787, - "step": 13850 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019531522694987488, - "loss": 1.5336, - "step": 13851 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019531488872353382, - "loss": 1.544, - "step": 13852 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019531455049719273, - "loss": 1.576, - "step": 13853 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019531421227085167, - "loss": 1.6529, - "step": 13854 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019531387404451061, - "loss": 1.7501, - "step": 13855 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019531353581816953, - "loss": 1.6989, - "step": 13856 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019531319759182847, - "loss": 1.5389, - "step": 13857 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953128593654874, - "loss": 1.6456, - "step": 13858 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019531252113914633, - "loss": 1.6487, - "step": 13859 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019531218291280527, - "loss": 1.534, - "step": 13860 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953118446864642, - "loss": 1.6025, - "step": 13861 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019531150646012312, - "loss": 1.6187, - "step": 13862 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019531116823378206, - "loss": 1.601, - "step": 13863 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195310830007441, - "loss": 1.6133, - "step": 13864 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019531049178109992, - "loss": 1.5251, - "step": 13865 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019531015355475886, - "loss": 1.466, - "step": 13866 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953098153284178, - "loss": 1.5428, - "step": 13867 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019530947710207671, - "loss": 1.5476, - "step": 13868 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019530913887573566, - "loss": 1.5825, - "step": 13869 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953088006493946, - "loss": 1.5369, - "step": 13870 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953084624230535, - "loss": 1.4742, - "step": 13871 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019530812419671245, - "loss": 1.7105, - "step": 13872 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953077859703714, - "loss": 1.7499, - "step": 13873 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953074477440303, - "loss": 1.5741, - "step": 13874 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019530710951768925, - "loss": 1.5604, - "step": 13875 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953067712913482, - "loss": 1.3564, - "step": 13876 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953064330650071, - "loss": 1.6046, - "step": 13877 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019530609483866604, - "loss": 1.5169, - "step": 13878 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019530575661232499, - "loss": 1.5429, - "step": 13879 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953054183859839, - "loss": 1.6453, - "step": 13880 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019530508015964284, - "loss": 1.6856, - "step": 13881 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019530474193330178, - "loss": 1.6367, - "step": 13882 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953044037069607, - "loss": 1.5493, - "step": 13883 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019530406548061964, - "loss": 1.3916, - "step": 13884 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019530372725427858, - "loss": 1.5858, - "step": 13885 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953033890279375, - "loss": 1.6171, - "step": 13886 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019530305080159643, - "loss": 1.4923, - "step": 13887 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019530271257525537, - "loss": 1.5439, - "step": 13888 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001953023743489143, - "loss": 1.6365, - "step": 13889 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019530203612257323, - "loss": 1.5961, - "step": 13890 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019530169789623217, - "loss": 1.5228, - "step": 13891 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019530135966989108, - "loss": 1.5653, - "step": 13892 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019530102144355003, - "loss": 1.8206, - "step": 13893 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019530068321720897, - "loss": 1.6435, - "step": 13894 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019530034499086788, - "loss": 1.6058, - "step": 13895 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019530000676452682, - "loss": 1.6608, - "step": 13896 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019529966853818576, - "loss": 1.6766, - "step": 13897 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019529933031184468, - "loss": 1.6188, - "step": 13898 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019529899208550362, - "loss": 1.5422, - "step": 13899 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019529865385916256, - "loss": 1.6549, - "step": 13900 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019529831563282147, - "loss": 1.7659, - "step": 13901 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019529797740648044, - "loss": 1.5726, - "step": 13902 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019529763918013936, - "loss": 1.575, - "step": 13903 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952973009537983, - "loss": 1.6671, - "step": 13904 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019529696272745724, - "loss": 1.6039, - "step": 13905 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019529662450111615, - "loss": 1.6086, - "step": 13906 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952962862747751, - "loss": 1.6251, - "step": 13907 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019529594804843403, - "loss": 1.5743, - "step": 13908 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019529560982209295, - "loss": 1.7497, - "step": 13909 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952952715957519, - "loss": 1.5948, - "step": 13910 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019529493336941083, - "loss": 1.781, - "step": 13911 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019529459514306974, - "loss": 1.7306, - "step": 13912 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019529425691672869, - "loss": 1.5419, - "step": 13913 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019529391869038763, - "loss": 1.7363, - "step": 13914 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019529358046404657, - "loss": 1.7036, - "step": 13915 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019529324223770548, - "loss": 1.734, - "step": 13916 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019529290401136442, - "loss": 1.5463, - "step": 13917 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019529256578502336, - "loss": 1.5026, - "step": 13918 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019529222755868228, - "loss": 1.6636, - "step": 13919 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019529188933234122, - "loss": 1.7619, - "step": 13920 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019529155110600016, - "loss": 1.596, - "step": 13921 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019529121287965908, - "loss": 1.6707, - "step": 13922 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019529087465331802, - "loss": 1.4784, - "step": 13923 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019529053642697696, - "loss": 1.696, - "step": 13924 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019529019820063587, - "loss": 1.7008, - "step": 13925 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952898599742948, - "loss": 1.5711, - "step": 13926 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019528952174795375, - "loss": 1.6875, - "step": 13927 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019528918352161267, - "loss": 1.476, - "step": 13928 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952888452952716, - "loss": 1.5165, - "step": 13929 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019528850706893055, - "loss": 1.7401, - "step": 13930 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019528816884258946, - "loss": 1.5538, - "step": 13931 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952878306162484, - "loss": 1.6168, - "step": 13932 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019528749238990735, - "loss": 1.6324, - "step": 13933 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019528715416356626, - "loss": 1.6407, - "step": 13934 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952868159372252, - "loss": 1.5053, - "step": 13935 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019528647771088414, - "loss": 1.5829, - "step": 13936 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019528613948454306, - "loss": 1.5544, - "step": 13937 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195285801258202, - "loss": 1.6983, - "step": 13938 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019528546303186094, - "loss": 1.4133, - "step": 13939 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019528512480551985, - "loss": 1.5731, - "step": 13940 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952847865791788, - "loss": 1.6231, - "step": 13941 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019528444835283774, - "loss": 1.6195, - "step": 13942 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019528411012649665, - "loss": 1.6745, - "step": 13943 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952837719001556, - "loss": 1.6443, - "step": 13944 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019528343367381453, - "loss": 1.5211, - "step": 13945 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019528309544747345, - "loss": 1.683, - "step": 13946 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952827572211324, - "loss": 1.64, - "step": 13947 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019528241899479133, - "loss": 1.6209, - "step": 13948 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019528208076845024, - "loss": 1.5616, - "step": 13949 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019528174254210918, - "loss": 1.5679, - "step": 13950 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019528140431576812, - "loss": 1.4584, - "step": 13951 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019528106608942704, - "loss": 1.5726, - "step": 13952 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019528072786308598, - "loss": 1.6076, - "step": 13953 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019528038963674492, - "loss": 1.454, - "step": 13954 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019528005141040383, - "loss": 1.7521, - "step": 13955 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019527971318406278, - "loss": 1.6606, - "step": 13956 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019527937495772172, - "loss": 1.7388, - "step": 13957 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019527903673138066, - "loss": 1.665, - "step": 13958 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952786985050396, - "loss": 1.4403, - "step": 13959 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952783602786985, - "loss": 1.5315, - "step": 13960 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019527802205235745, - "loss": 1.6937, - "step": 13961 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952776838260164, - "loss": 1.6431, - "step": 13962 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952773455996753, - "loss": 1.4781, - "step": 13963 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019527700737333425, - "loss": 1.6042, - "step": 13964 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952766691469932, - "loss": 1.6018, - "step": 13965 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952763309206521, - "loss": 1.7032, - "step": 13966 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019527599269431105, - "loss": 1.7138, - "step": 13967 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019527565446797, - "loss": 1.5313, - "step": 13968 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952753162416289, - "loss": 1.5154, - "step": 13969 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019527497801528784, - "loss": 1.429, - "step": 13970 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019527463978894678, - "loss": 1.5361, - "step": 13971 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952743015626057, - "loss": 1.5101, - "step": 13972 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019527396333626464, - "loss": 1.4585, - "step": 13973 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019527362510992358, - "loss": 1.7181, - "step": 13974 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952732868835825, - "loss": 1.4576, - "step": 13975 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019527294865724144, - "loss": 1.539, - "step": 13976 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019527261043090038, - "loss": 1.4964, - "step": 13977 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952722722045593, - "loss": 1.6695, - "step": 13978 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019527193397821823, - "loss": 1.692, - "step": 13979 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019527159575187717, - "loss": 1.5925, - "step": 13980 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952712575255361, - "loss": 1.6642, - "step": 13981 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019527091929919503, - "loss": 1.4928, - "step": 13982 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019527058107285397, - "loss": 1.5202, - "step": 13983 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019527024284651288, - "loss": 1.4592, - "step": 13984 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019526990462017182, - "loss": 1.6138, - "step": 13985 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019526956639383077, - "loss": 1.7318, - "step": 13986 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019526922816748968, - "loss": 1.6739, - "step": 13987 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019526888994114862, - "loss": 1.631, - "step": 13988 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019526855171480756, - "loss": 1.6763, - "step": 13989 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019526821348846648, - "loss": 1.662, - "step": 13990 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019526787526212542, - "loss": 1.6089, - "step": 13991 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019526753703578436, - "loss": 1.5686, - "step": 13992 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019526719880944327, - "loss": 1.5956, - "step": 13993 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019526686058310221, - "loss": 1.5202, - "step": 13994 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019526652235676115, - "loss": 1.6001, - "step": 13995 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019526618413042007, - "loss": 1.5828, - "step": 13996 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195265845904079, - "loss": 1.6455, - "step": 13997 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019526550767773795, - "loss": 1.644, - "step": 13998 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952651694513969, - "loss": 1.575, - "step": 13999 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952648312250558, - "loss": 1.5542, - "step": 14000 - }, - { - "epoch": 0.05, - "eval_loss": 1.3529334435096154, - "mrc_em": 11.0, - "mrc_f1": 22.928571428571434, - "nli_acc": 0.28, - "step": 14000, - "sts_f1": 0.6277372262773723, - "ynat_f1": 0.7309515878796776 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019526449299871475, - "loss": 1.5647, - "step": 14001 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952641547723737, - "loss": 1.6939, - "step": 14002 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952638165460326, - "loss": 1.6111, - "step": 14003 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019526347831969154, - "loss": 1.6216, - "step": 14004 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019526314009335048, - "loss": 1.728, - "step": 14005 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952628018670094, - "loss": 1.5576, - "step": 14006 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019526246364066834, - "loss": 1.5674, - "step": 14007 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019526212541432728, - "loss": 1.6417, - "step": 14008 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952617871879862, - "loss": 1.6668, - "step": 14009 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019526144896164514, - "loss": 1.6267, - "step": 14010 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019526111073530408, - "loss": 1.6376, - "step": 14011 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195260772508963, - "loss": 1.6272, - "step": 14012 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019526043428262193, - "loss": 1.6583, - "step": 14013 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019526009605628087, - "loss": 1.6888, - "step": 14014 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019525975782993981, - "loss": 1.6705, - "step": 14015 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019525941960359876, - "loss": 1.6309, - "step": 14016 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019525908137725767, - "loss": 1.5251, - "step": 14017 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952587431509166, - "loss": 1.6955, - "step": 14018 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019525840492457555, - "loss": 1.5047, - "step": 14019 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019525806669823447, - "loss": 1.6388, - "step": 14020 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952577284718934, - "loss": 1.5985, - "step": 14021 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019525739024555235, - "loss": 1.4071, - "step": 14022 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019525705201921126, - "loss": 1.6852, - "step": 14023 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952567137928702, - "loss": 1.5893, - "step": 14024 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019525637556652915, - "loss": 1.5202, - "step": 14025 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019525603734018806, - "loss": 1.7653, - "step": 14026 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195255699113847, - "loss": 1.72, - "step": 14027 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019525536088750594, - "loss": 1.5595, - "step": 14028 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019525502266116486, - "loss": 1.5721, - "step": 14029 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952546844348238, - "loss": 1.5358, - "step": 14030 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019525434620848274, - "loss": 1.7244, - "step": 14031 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019525400798214165, - "loss": 1.6216, - "step": 14032 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952536697558006, - "loss": 1.7246, - "step": 14033 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019525333152945953, - "loss": 1.6383, - "step": 14034 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019525299330311845, - "loss": 1.7035, - "step": 14035 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952526550767774, - "loss": 1.5993, - "step": 14036 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019525231685043633, - "loss": 1.4571, - "step": 14037 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019525197862409524, - "loss": 1.6376, - "step": 14038 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019525164039775419, - "loss": 1.5571, - "step": 14039 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019525130217141313, - "loss": 1.5738, - "step": 14040 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019525096394507204, - "loss": 1.5485, - "step": 14041 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019525062571873098, - "loss": 1.4406, - "step": 14042 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019525028749238992, - "loss": 1.5362, - "step": 14043 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019524994926604884, - "loss": 1.6281, - "step": 14044 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019524961103970778, - "loss": 1.6239, - "step": 14045 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019524927281336672, - "loss": 1.6609, - "step": 14046 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019524893458702563, - "loss": 1.6679, - "step": 14047 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019524859636068457, - "loss": 1.5918, - "step": 14048 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019524825813434352, - "loss": 1.5824, - "step": 14049 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019524791990800243, - "loss": 1.6908, - "step": 14050 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019524758168166137, - "loss": 1.5125, - "step": 14051 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952472434553203, - "loss": 1.614, - "step": 14052 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019524690522897923, - "loss": 1.5738, - "step": 14053 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019524656700263817, - "loss": 1.4855, - "step": 14054 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952462287762971, - "loss": 1.5369, - "step": 14055 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019524589054995602, - "loss": 1.5663, - "step": 14056 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019524555232361496, - "loss": 1.5841, - "step": 14057 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952452140972739, - "loss": 1.6229, - "step": 14058 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019524487587093282, - "loss": 1.6276, - "step": 14059 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019524453764459176, - "loss": 1.5213, - "step": 14060 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952441994182507, - "loss": 1.7336, - "step": 14061 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019524386119190961, - "loss": 1.795, - "step": 14062 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019524352296556856, - "loss": 1.6948, - "step": 14063 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952431847392275, - "loss": 1.4466, - "step": 14064 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952428465128864, - "loss": 1.7884, - "step": 14065 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019524250828654535, - "loss": 1.5513, - "step": 14066 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952421700602043, - "loss": 1.593, - "step": 14067 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952418318338632, - "loss": 1.6939, - "step": 14068 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019524149360752215, - "loss": 1.5601, - "step": 14069 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019524115538118112, - "loss": 1.6174, - "step": 14070 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019524081715484003, - "loss": 1.5775, - "step": 14071 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019524047892849897, - "loss": 1.5487, - "step": 14072 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952401407021579, - "loss": 1.7351, - "step": 14073 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019523980247581683, - "loss": 1.5112, - "step": 14074 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019523946424947577, - "loss": 1.7321, - "step": 14075 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952391260231347, - "loss": 1.6005, - "step": 14076 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019523878779679362, - "loss": 1.8083, - "step": 14077 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019523844957045256, - "loss": 1.5818, - "step": 14078 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952381113441115, - "loss": 1.75, - "step": 14079 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019523777311777042, - "loss": 1.7224, - "step": 14080 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019523743489142936, - "loss": 1.6923, - "step": 14081 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952370966650883, - "loss": 1.6127, - "step": 14082 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019523675843874722, - "loss": 1.6543, - "step": 14083 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019523642021240616, - "loss": 1.7767, - "step": 14084 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952360819860651, - "loss": 1.6162, - "step": 14085 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195235743759724, - "loss": 1.6973, - "step": 14086 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019523540553338295, - "loss": 1.4647, - "step": 14087 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952350673070419, - "loss": 1.8212, - "step": 14088 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952347290807008, - "loss": 1.6151, - "step": 14089 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019523439085435975, - "loss": 1.6832, - "step": 14090 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952340526280187, - "loss": 1.6002, - "step": 14091 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952337144016776, - "loss": 1.5355, - "step": 14092 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019523337617533655, - "loss": 1.5581, - "step": 14093 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952330379489955, - "loss": 1.6634, - "step": 14094 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952326997226544, - "loss": 1.6604, - "step": 14095 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019523236149631334, - "loss": 1.59, - "step": 14096 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019523202326997228, - "loss": 1.606, - "step": 14097 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952316850436312, - "loss": 1.5204, - "step": 14098 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019523134681729014, - "loss": 1.633, - "step": 14099 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019523100859094908, - "loss": 1.5388, - "step": 14100 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195230670364608, - "loss": 1.6658, - "step": 14101 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019523033213826694, - "loss": 1.569, - "step": 14102 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019522999391192588, - "loss": 1.5168, - "step": 14103 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952296556855848, - "loss": 1.5204, - "step": 14104 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019522931745924373, - "loss": 1.5116, - "step": 14105 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019522897923290267, - "loss": 1.707, - "step": 14106 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952286410065616, - "loss": 1.6316, - "step": 14107 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019522830278022053, - "loss": 1.6729, - "step": 14108 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019522796455387947, - "loss": 1.5548, - "step": 14109 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019522762632753838, - "loss": 1.621, - "step": 14110 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019522728810119732, - "loss": 1.6752, - "step": 14111 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019522694987485627, - "loss": 1.5982, - "step": 14112 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019522661164851518, - "loss": 1.6745, - "step": 14113 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019522627342217412, - "loss": 1.6014, - "step": 14114 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019522593519583306, - "loss": 1.5327, - "step": 14115 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019522559696949198, - "loss": 1.618, - "step": 14116 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019522525874315092, - "loss": 1.496, - "step": 14117 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019522492051680986, - "loss": 1.7073, - "step": 14118 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019522458229046877, - "loss": 1.6353, - "step": 14119 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952242440641277, - "loss": 1.521, - "step": 14120 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019522390583778665, - "loss": 1.5979, - "step": 14121 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019522356761144557, - "loss": 1.6087, - "step": 14122 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952232293851045, - "loss": 1.6866, - "step": 14123 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019522289115876345, - "loss": 1.5906, - "step": 14124 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019522255293242236, - "loss": 1.7789, - "step": 14125 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019522221470608133, - "loss": 1.5285, - "step": 14126 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019522187647974027, - "loss": 1.7281, - "step": 14127 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952215382533992, - "loss": 1.5784, - "step": 14128 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019522120002705813, - "loss": 1.4855, - "step": 14129 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019522086180071707, - "loss": 1.7629, - "step": 14130 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019522052357437598, - "loss": 1.7392, - "step": 14131 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019522018534803493, - "loss": 1.6138, - "step": 14132 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019521984712169387, - "loss": 1.5289, - "step": 14133 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019521950889535278, - "loss": 1.5381, - "step": 14134 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019521917066901172, - "loss": 1.7553, - "step": 14135 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019521883244267066, - "loss": 1.6854, - "step": 14136 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019521849421632958, - "loss": 1.723, - "step": 14137 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019521815598998852, - "loss": 1.594, - "step": 14138 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019521781776364746, - "loss": 1.5475, - "step": 14139 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019521747953730637, - "loss": 1.6043, - "step": 14140 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019521714131096531, - "loss": 1.6823, - "step": 14141 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019521680308462426, - "loss": 1.5985, - "step": 14142 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019521646485828317, - "loss": 1.6498, - "step": 14143 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952161266319421, - "loss": 1.6246, - "step": 14144 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019521578840560105, - "loss": 1.6375, - "step": 14145 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019521545017925997, - "loss": 1.6547, - "step": 14146 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952151119529189, - "loss": 1.7652, - "step": 14147 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019521477372657785, - "loss": 1.6431, - "step": 14148 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019521443550023676, - "loss": 1.6948, - "step": 14149 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952140972738957, - "loss": 1.5702, - "step": 14150 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019521375904755464, - "loss": 1.7207, - "step": 14151 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019521342082121356, - "loss": 1.5662, - "step": 14152 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952130825948725, - "loss": 1.5496, - "step": 14153 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019521274436853144, - "loss": 1.4949, - "step": 14154 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019521240614219035, - "loss": 1.5282, - "step": 14155 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952120679158493, - "loss": 1.5335, - "step": 14156 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019521172968950824, - "loss": 1.6336, - "step": 14157 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019521139146316715, - "loss": 1.5302, - "step": 14158 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952110532368261, - "loss": 1.692, - "step": 14159 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019521071501048503, - "loss": 1.6667, - "step": 14160 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019521037678414395, - "loss": 1.6609, - "step": 14161 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952100385578029, - "loss": 1.657, - "step": 14162 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019520970033146183, - "loss": 1.4651, - "step": 14163 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019520936210512074, - "loss": 1.6814, - "step": 14164 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019520902387877968, - "loss": 1.664, - "step": 14165 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019520868565243863, - "loss": 1.6556, - "step": 14166 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019520834742609754, - "loss": 1.5487, - "step": 14167 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019520800919975648, - "loss": 1.5388, - "step": 14168 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019520767097341542, - "loss": 1.6995, - "step": 14169 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019520733274707434, - "loss": 1.5495, - "step": 14170 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019520699452073328, - "loss": 1.5967, - "step": 14171 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019520665629439222, - "loss": 1.5558, - "step": 14172 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019520631806805113, - "loss": 1.7502, - "step": 14173 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019520597984171007, - "loss": 1.5404, - "step": 14174 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019520564161536901, - "loss": 1.6216, - "step": 14175 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019520530338902793, - "loss": 1.5045, - "step": 14176 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019520496516268687, - "loss": 1.7194, - "step": 14177 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952046269363458, - "loss": 1.6285, - "step": 14178 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019520428871000473, - "loss": 1.7542, - "step": 14179 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019520395048366367, - "loss": 1.7121, - "step": 14180 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952036122573226, - "loss": 1.6874, - "step": 14181 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019520327403098152, - "loss": 1.6817, - "step": 14182 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001952029358046405, - "loss": 1.7179, - "step": 14183 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019520259757829943, - "loss": 1.7344, - "step": 14184 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019520225935195834, - "loss": 1.7158, - "step": 14185 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019520192112561729, - "loss": 1.5653, - "step": 14186 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019520158289927623, - "loss": 1.6685, - "step": 14187 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019520124467293514, - "loss": 1.5004, - "step": 14188 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019520090644659408, - "loss": 1.6755, - "step": 14189 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019520056822025302, - "loss": 1.7017, - "step": 14190 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019520022999391194, - "loss": 1.6649, - "step": 14191 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019519989176757088, - "loss": 1.5267, - "step": 14192 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019519955354122982, - "loss": 1.5854, - "step": 14193 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019519921531488873, - "loss": 1.5682, - "step": 14194 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019519887708854768, - "loss": 1.6699, - "step": 14195 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019519853886220662, - "loss": 1.5756, - "step": 14196 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019519820063586553, - "loss": 1.6181, - "step": 14197 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019519786240952447, - "loss": 1.8198, - "step": 14198 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951975241831834, - "loss": 1.5935, - "step": 14199 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019519718595684233, - "loss": 1.6087, - "step": 14200 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019519684773050127, - "loss": 1.5917, - "step": 14201 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951965095041602, - "loss": 1.6443, - "step": 14202 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019519617127781912, - "loss": 1.5763, - "step": 14203 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019519583305147806, - "loss": 1.6444, - "step": 14204 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195195494825137, - "loss": 1.5372, - "step": 14205 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019519515659879592, - "loss": 1.6726, - "step": 14206 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019519481837245486, - "loss": 1.6898, - "step": 14207 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951944801461138, - "loss": 1.6039, - "step": 14208 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019519414191977272, - "loss": 1.7016, - "step": 14209 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019519380369343166, - "loss": 1.6557, - "step": 14210 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951934654670906, - "loss": 1.5974, - "step": 14211 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951931272407495, - "loss": 1.6012, - "step": 14212 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019519278901440845, - "loss": 1.6306, - "step": 14213 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951924507880674, - "loss": 1.7192, - "step": 14214 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951921125617263, - "loss": 1.6076, - "step": 14215 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019519177433538525, - "loss": 1.7326, - "step": 14216 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951914361090442, - "loss": 1.5548, - "step": 14217 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951910978827031, - "loss": 1.6729, - "step": 14218 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019519075965636205, - "loss": 1.6668, - "step": 14219 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195190421430021, - "loss": 1.6028, - "step": 14220 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951900832036799, - "loss": 1.6664, - "step": 14221 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019518974497733884, - "loss": 1.5936, - "step": 14222 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019518940675099778, - "loss": 1.657, - "step": 14223 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951890685246567, - "loss": 1.6654, - "step": 14224 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019518873029831564, - "loss": 1.6643, - "step": 14225 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019518839207197458, - "loss": 1.8172, - "step": 14226 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951880538456335, - "loss": 1.6846, - "step": 14227 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019518771561929243, - "loss": 1.5421, - "step": 14228 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019518737739295138, - "loss": 1.7354, - "step": 14229 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951870391666103, - "loss": 1.6001, - "step": 14230 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019518670094026923, - "loss": 1.6734, - "step": 14231 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019518636271392817, - "loss": 1.5071, - "step": 14232 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019518602448758709, - "loss": 1.6765, - "step": 14233 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019518568626124603, - "loss": 1.7229, - "step": 14234 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019518534803490497, - "loss": 1.5476, - "step": 14235 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019518500980856388, - "loss": 1.6046, - "step": 14236 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019518467158222282, - "loss": 1.6856, - "step": 14237 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019518433335588176, - "loss": 1.5674, - "step": 14238 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951839951295407, - "loss": 1.7084, - "step": 14239 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019518365690319965, - "loss": 1.5716, - "step": 14240 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019518331867685856, - "loss": 1.6412, - "step": 14241 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951829804505175, - "loss": 1.6743, - "step": 14242 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019518264222417644, - "loss": 1.4956, - "step": 14243 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019518230399783536, - "loss": 1.5107, - "step": 14244 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951819657714943, - "loss": 1.5295, - "step": 14245 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019518162754515324, - "loss": 1.4468, - "step": 14246 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019518128931881215, - "loss": 1.5881, - "step": 14247 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951809510924711, - "loss": 1.3928, - "step": 14248 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019518061286613004, - "loss": 1.6316, - "step": 14249 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019518027463978895, - "loss": 1.7391, - "step": 14250 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951799364134479, - "loss": 1.4473, - "step": 14251 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019517959818710683, - "loss": 1.4964, - "step": 14252 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019517925996076575, - "loss": 1.5473, - "step": 14253 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951789217344247, - "loss": 1.4639, - "step": 14254 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019517858350808363, - "loss": 1.6339, - "step": 14255 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019517824528174254, - "loss": 1.5248, - "step": 14256 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019517790705540148, - "loss": 1.4807, - "step": 14257 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019517756882906042, - "loss": 1.7402, - "step": 14258 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019517723060271934, - "loss": 1.5498, - "step": 14259 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019517689237637828, - "loss": 1.4739, - "step": 14260 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019517655415003722, - "loss": 1.5435, - "step": 14261 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019517621592369614, - "loss": 1.5784, - "step": 14262 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019517587769735508, - "loss": 1.491, - "step": 14263 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019517553947101402, - "loss": 1.6225, - "step": 14264 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019517520124467296, - "loss": 1.5234, - "step": 14265 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019517486301833187, - "loss": 1.5186, - "step": 14266 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019517452479199081, - "loss": 1.5642, - "step": 14267 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019517418656564975, - "loss": 1.5535, - "step": 14268 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019517384833930867, - "loss": 1.5844, - "step": 14269 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951735101129676, - "loss": 1.7069, - "step": 14270 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019517317188662655, - "loss": 1.5533, - "step": 14271 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019517283366028547, - "loss": 1.4608, - "step": 14272 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951724954339444, - "loss": 1.6128, - "step": 14273 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019517215720760335, - "loss": 1.7965, - "step": 14274 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019517181898126226, - "loss": 1.5598, - "step": 14275 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951714807549212, - "loss": 1.3974, - "step": 14276 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019517114252858014, - "loss": 1.4298, - "step": 14277 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019517080430223906, - "loss": 1.5272, - "step": 14278 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195170466075898, - "loss": 1.6995, - "step": 14279 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019517012784955694, - "loss": 1.6341, - "step": 14280 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019516978962321585, - "loss": 1.5699, - "step": 14281 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951694513968748, - "loss": 1.5487, - "step": 14282 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019516911317053374, - "loss": 1.5931, - "step": 14283 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019516877494419265, - "loss": 1.7594, - "step": 14284 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951684367178516, - "loss": 1.5051, - "step": 14285 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019516809849151053, - "loss": 1.6756, - "step": 14286 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019516776026516945, - "loss": 1.7163, - "step": 14287 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951674220388284, - "loss": 1.5189, - "step": 14288 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019516708381248733, - "loss": 1.5265, - "step": 14289 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019516674558614624, - "loss": 1.5581, - "step": 14290 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019516640735980518, - "loss": 1.5391, - "step": 14291 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019516606913346413, - "loss": 1.6001, - "step": 14292 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019516573090712304, - "loss": 1.4697, - "step": 14293 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019516539268078198, - "loss": 1.5527, - "step": 14294 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019516505445444092, - "loss": 1.6449, - "step": 14295 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019516471622809986, - "loss": 1.7192, - "step": 14296 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951643780017588, - "loss": 1.5485, - "step": 14297 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019516403977541772, - "loss": 1.6614, - "step": 14298 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019516370154907666, - "loss": 1.6932, - "step": 14299 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951633633227356, - "loss": 1.5515, - "step": 14300 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019516302509639451, - "loss": 1.6554, - "step": 14301 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019516268687005346, - "loss": 1.513, - "step": 14302 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951623486437124, - "loss": 1.5998, - "step": 14303 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951620104173713, - "loss": 1.5921, - "step": 14304 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019516167219103025, - "loss": 1.6854, - "step": 14305 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951613339646892, - "loss": 1.4794, - "step": 14306 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951609957383481, - "loss": 1.6141, - "step": 14307 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019516065751200705, - "loss": 1.6448, - "step": 14308 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195160319285666, - "loss": 1.7188, - "step": 14309 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951599810593249, - "loss": 1.6475, - "step": 14310 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019515964283298384, - "loss": 1.6812, - "step": 14311 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019515930460664279, - "loss": 1.5113, - "step": 14312 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951589663803017, - "loss": 1.5461, - "step": 14313 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019515862815396064, - "loss": 1.6725, - "step": 14314 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019515828992761958, - "loss": 1.7292, - "step": 14315 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951579517012785, - "loss": 1.4735, - "step": 14316 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019515761347493744, - "loss": 1.6306, - "step": 14317 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019515727524859638, - "loss": 1.6209, - "step": 14318 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951569370222553, - "loss": 1.6554, - "step": 14319 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019515659879591423, - "loss": 1.4155, - "step": 14320 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019515626056957317, - "loss": 1.6225, - "step": 14321 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951559223432321, - "loss": 1.5354, - "step": 14322 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019515558411689103, - "loss": 1.6142, - "step": 14323 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019515524589054997, - "loss": 1.5697, - "step": 14324 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019515490766420888, - "loss": 1.6432, - "step": 14325 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019515456943786783, - "loss": 1.6081, - "step": 14326 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019515423121152677, - "loss": 1.5058, - "step": 14327 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019515389298518568, - "loss": 1.6321, - "step": 14328 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019515355475884462, - "loss": 1.6938, - "step": 14329 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019515321653250356, - "loss": 1.7176, - "step": 14330 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019515287830616248, - "loss": 1.6648, - "step": 14331 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019515254007982142, - "loss": 1.6438, - "step": 14332 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019515220185348036, - "loss": 1.6232, - "step": 14333 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019515186362713927, - "loss": 1.6032, - "step": 14334 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019515152540079821, - "loss": 1.3656, - "step": 14335 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019515118717445716, - "loss": 1.6611, - "step": 14336 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019515084894811607, - "loss": 1.7381, - "step": 14337 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195150510721775, - "loss": 1.5727, - "step": 14338 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019515017249543395, - "loss": 1.701, - "step": 14339 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019514983426909287, - "loss": 1.509, - "step": 14340 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951494960427518, - "loss": 1.56, - "step": 14341 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019514915781641075, - "loss": 1.634, - "step": 14342 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019514881959006966, - "loss": 1.6176, - "step": 14343 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951484813637286, - "loss": 1.5659, - "step": 14344 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019514814313738754, - "loss": 1.5902, - "step": 14345 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019514780491104646, - "loss": 1.5789, - "step": 14346 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951474666847054, - "loss": 1.6094, - "step": 14347 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019514712845836434, - "loss": 1.6024, - "step": 14348 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019514679023202326, - "loss": 1.642, - "step": 14349 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951464520056822, - "loss": 1.5787, - "step": 14350 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019514611377934116, - "loss": 1.6178, - "step": 14351 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019514577555300008, - "loss": 1.7061, - "step": 14352 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019514543732665902, - "loss": 1.7003, - "step": 14353 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019514509910031796, - "loss": 1.6581, - "step": 14354 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019514476087397688, - "loss": 1.6738, - "step": 14355 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019514442264763582, - "loss": 1.5862, - "step": 14356 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019514408442129476, - "loss": 1.6296, - "step": 14357 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019514374619495367, - "loss": 1.3599, - "step": 14358 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951434079686126, - "loss": 1.6109, - "step": 14359 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019514306974227155, - "loss": 1.6973, - "step": 14360 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019514273151593047, - "loss": 1.667, - "step": 14361 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951423932895894, - "loss": 1.5531, - "step": 14362 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019514205506324835, - "loss": 1.8242, - "step": 14363 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019514171683690726, - "loss": 1.6718, - "step": 14364 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951413786105662, - "loss": 1.4611, - "step": 14365 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019514104038422515, - "loss": 1.5478, - "step": 14366 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019514070215788406, - "loss": 1.6453, - "step": 14367 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195140363931543, - "loss": 1.5725, - "step": 14368 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019514002570520194, - "loss": 1.6622, - "step": 14369 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019513968747886086, - "loss": 1.6679, - "step": 14370 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951393492525198, - "loss": 1.6621, - "step": 14371 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019513901102617874, - "loss": 1.5874, - "step": 14372 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019513867279983765, - "loss": 1.5667, - "step": 14373 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951383345734966, - "loss": 1.6171, - "step": 14374 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019513799634715554, - "loss": 1.4746, - "step": 14375 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019513765812081445, - "loss": 1.49, - "step": 14376 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951373198944734, - "loss": 1.7929, - "step": 14377 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019513698166813233, - "loss": 1.5305, - "step": 14378 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019513664344179125, - "loss": 1.5715, - "step": 14379 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951363052154502, - "loss": 1.4867, - "step": 14380 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019513596698910913, - "loss": 1.6509, - "step": 14381 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019513562876276804, - "loss": 1.6025, - "step": 14382 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019513529053642698, - "loss": 1.5012, - "step": 14383 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019513495231008592, - "loss": 1.5487, - "step": 14384 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019513461408374484, - "loss": 1.5416, - "step": 14385 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019513427585740378, - "loss": 1.6827, - "step": 14386 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019513393763106272, - "loss": 1.6025, - "step": 14387 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019513359940472163, - "loss": 1.7533, - "step": 14388 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019513326117838058, - "loss": 1.7077, - "step": 14389 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019513292295203952, - "loss": 1.5414, - "step": 14390 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019513258472569843, - "loss": 1.5397, - "step": 14391 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019513224649935737, - "loss": 1.6777, - "step": 14392 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951319082730163, - "loss": 1.4193, - "step": 14393 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019513157004667523, - "loss": 1.4995, - "step": 14394 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019513123182033417, - "loss": 1.5998, - "step": 14395 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951308935939931, - "loss": 1.5394, - "step": 14396 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019513055536765202, - "loss": 1.5697, - "step": 14397 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019513021714131096, - "loss": 1.5324, - "step": 14398 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951298789149699, - "loss": 1.5776, - "step": 14399 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019512954068862882, - "loss": 1.5754, - "step": 14400 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019512920246228776, - "loss": 1.8549, - "step": 14401 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951288642359467, - "loss": 1.7955, - "step": 14402 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019512852600960562, - "loss": 1.7563, - "step": 14403 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019512818778326456, - "loss": 1.4863, - "step": 14404 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951278495569235, - "loss": 1.5771, - "step": 14405 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951275113305824, - "loss": 1.6267, - "step": 14406 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019512717310424138, - "loss": 1.8475, - "step": 14407 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019512683487790032, - "loss": 1.6119, - "step": 14408 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019512649665155924, - "loss": 1.6664, - "step": 14409 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019512615842521818, - "loss": 1.5711, - "step": 14410 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019512582019887712, - "loss": 1.6721, - "step": 14411 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019512548197253603, - "loss": 1.7191, - "step": 14412 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019512514374619497, - "loss": 1.7436, - "step": 14413 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019512480551985391, - "loss": 1.7165, - "step": 14414 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019512446729351283, - "loss": 1.4352, - "step": 14415 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019512412906717177, - "loss": 1.6302, - "step": 14416 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951237908408307, - "loss": 1.4963, - "step": 14417 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019512345261448962, - "loss": 1.5245, - "step": 14418 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019512311438814857, - "loss": 1.5615, - "step": 14419 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951227761618075, - "loss": 1.5672, - "step": 14420 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019512243793546642, - "loss": 1.6242, - "step": 14421 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019512209970912536, - "loss": 1.6742, - "step": 14422 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951217614827843, - "loss": 1.6094, - "step": 14423 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019512142325644322, - "loss": 1.6497, - "step": 14424 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019512108503010216, - "loss": 1.5796, - "step": 14425 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951207468037611, - "loss": 1.5403, - "step": 14426 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019512040857742001, - "loss": 1.6228, - "step": 14427 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019512007035107895, - "loss": 1.7105, - "step": 14428 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951197321247379, - "loss": 1.4232, - "step": 14429 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951193938983968, - "loss": 1.7925, - "step": 14430 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019511905567205575, - "loss": 1.5594, - "step": 14431 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951187174457147, - "loss": 1.5948, - "step": 14432 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951183792193736, - "loss": 1.7192, - "step": 14433 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019511804099303255, - "loss": 1.6582, - "step": 14434 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951177027666915, - "loss": 1.7623, - "step": 14435 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951173645403504, - "loss": 1.5157, - "step": 14436 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019511702631400934, - "loss": 1.6544, - "step": 14437 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019511668808766828, - "loss": 1.5617, - "step": 14438 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951163498613272, - "loss": 1.6223, - "step": 14439 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019511601163498614, - "loss": 1.6011, - "step": 14440 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019511567340864508, - "loss": 1.6103, - "step": 14441 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195115335182304, - "loss": 1.7138, - "step": 14442 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019511499695596294, - "loss": 1.4992, - "step": 14443 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019511465872962188, - "loss": 1.6, - "step": 14444 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951143205032808, - "loss": 1.6208, - "step": 14445 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019511398227693973, - "loss": 1.5146, - "step": 14446 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019511364405059867, - "loss": 1.665, - "step": 14447 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951133058242576, - "loss": 1.6077, - "step": 14448 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019511296759791653, - "loss": 1.8171, - "step": 14449 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019511262937157547, - "loss": 1.3819, - "step": 14450 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019511229114523438, - "loss": 1.6471, - "step": 14451 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019511195291889333, - "loss": 1.4293, - "step": 14452 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019511161469255227, - "loss": 1.5313, - "step": 14453 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019511127646621118, - "loss": 1.755, - "step": 14454 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019511093823987012, - "loss": 1.632, - "step": 14455 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019511060001352906, - "loss": 1.5616, - "step": 14456 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019511026178718798, - "loss": 1.6522, - "step": 14457 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019510992356084692, - "loss": 1.6376, - "step": 14458 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019510958533450586, - "loss": 1.5466, - "step": 14459 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019510924710816477, - "loss": 1.4972, - "step": 14460 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019510890888182371, - "loss": 1.4405, - "step": 14461 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019510857065548266, - "loss": 1.541, - "step": 14462 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951082324291416, - "loss": 1.6188, - "step": 14463 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019510789420280054, - "loss": 1.5962, - "step": 14464 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019510755597645948, - "loss": 1.5412, - "step": 14465 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951072177501184, - "loss": 1.6244, - "step": 14466 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019510687952377733, - "loss": 1.7769, - "step": 14467 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019510654129743628, - "loss": 1.5283, - "step": 14468 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951062030710952, - "loss": 1.7323, - "step": 14469 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019510586484475413, - "loss": 1.4809, - "step": 14470 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019510552661841307, - "loss": 1.6151, - "step": 14471 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019510518839207199, - "loss": 1.5475, - "step": 14472 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019510485016573093, - "loss": 1.6196, - "step": 14473 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019510451193938987, - "loss": 1.6385, - "step": 14474 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019510417371304878, - "loss": 1.8293, - "step": 14475 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019510383548670772, - "loss": 1.582, - "step": 14476 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019510349726036666, - "loss": 1.6336, - "step": 14477 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019510315903402558, - "loss": 1.7781, - "step": 14478 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019510282080768452, - "loss": 1.4528, - "step": 14479 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019510248258134346, - "loss": 1.6698, - "step": 14480 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019510214435500237, - "loss": 1.6774, - "step": 14481 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019510180612866132, - "loss": 1.5211, - "step": 14482 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019510146790232026, - "loss": 1.5562, - "step": 14483 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019510112967597917, - "loss": 1.5119, - "step": 14484 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001951007914496381, - "loss": 1.776, - "step": 14485 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019510045322329705, - "loss": 1.682, - "step": 14486 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019510011499695597, - "loss": 1.4632, - "step": 14487 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950997767706149, - "loss": 1.6405, - "step": 14488 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019509943854427385, - "loss": 1.6229, - "step": 14489 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019509910031793276, - "loss": 1.5739, - "step": 14490 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950987620915917, - "loss": 1.6911, - "step": 14491 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019509842386525065, - "loss": 1.64, - "step": 14492 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019509808563890956, - "loss": 1.6132, - "step": 14493 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950977474125685, - "loss": 1.6434, - "step": 14494 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019509740918622744, - "loss": 1.6819, - "step": 14495 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019509707095988636, - "loss": 1.6701, - "step": 14496 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950967327335453, - "loss": 1.6696, - "step": 14497 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019509639450720424, - "loss": 1.7868, - "step": 14498 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019509605628086315, - "loss": 1.6032, - "step": 14499 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950957180545221, - "loss": 1.6254, - "step": 14500 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019509537982818103, - "loss": 1.7796, - "step": 14501 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019509504160183995, - "loss": 1.6312, - "step": 14502 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950947033754989, - "loss": 1.7343, - "step": 14503 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019509436514915783, - "loss": 1.653, - "step": 14504 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019509402692281674, - "loss": 1.6771, - "step": 14505 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019509368869647569, - "loss": 1.5506, - "step": 14506 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019509335047013463, - "loss": 1.4961, - "step": 14507 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019509301224379354, - "loss": 1.7328, - "step": 14508 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019509267401745248, - "loss": 1.5168, - "step": 14509 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019509233579111142, - "loss": 1.6746, - "step": 14510 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019509199756477034, - "loss": 1.5041, - "step": 14511 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019509165933842928, - "loss": 1.7178, - "step": 14512 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019509132111208822, - "loss": 1.6053, - "step": 14513 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019509098288574713, - "loss": 1.7149, - "step": 14514 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019509064465940608, - "loss": 1.4956, - "step": 14515 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019509030643306502, - "loss": 1.7439, - "step": 14516 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019508996820672393, - "loss": 1.5262, - "step": 14517 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019508962998038287, - "loss": 1.5694, - "step": 14518 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950892917540418, - "loss": 1.6491, - "step": 14519 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019508895352770075, - "loss": 1.5599, - "step": 14520 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950886153013597, - "loss": 1.531, - "step": 14521 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950882770750186, - "loss": 1.6318, - "step": 14522 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019508793884867755, - "loss": 1.5584, - "step": 14523 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950876006223365, - "loss": 1.7654, - "step": 14524 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950872623959954, - "loss": 1.5171, - "step": 14525 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019508692416965435, - "loss": 1.5328, - "step": 14526 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950865859433133, - "loss": 1.6697, - "step": 14527 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950862477169722, - "loss": 1.7513, - "step": 14528 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019508590949063114, - "loss": 1.7204, - "step": 14529 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019508557126429008, - "loss": 1.7145, - "step": 14530 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195085233037949, - "loss": 1.6491, - "step": 14531 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019508489481160794, - "loss": 1.6232, - "step": 14532 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019508455658526688, - "loss": 1.652, - "step": 14533 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019508421835892582, - "loss": 1.5568, - "step": 14534 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019508388013258474, - "loss": 1.5419, - "step": 14535 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019508354190624368, - "loss": 1.4654, - "step": 14536 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019508320367990262, - "loss": 1.6363, - "step": 14537 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019508286545356153, - "loss": 1.542, - "step": 14538 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019508252722722047, - "loss": 1.5944, - "step": 14539 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019508218900087941, - "loss": 1.5766, - "step": 14540 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019508185077453833, - "loss": 1.6576, - "step": 14541 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019508151254819727, - "loss": 1.6193, - "step": 14542 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950811743218562, - "loss": 1.5387, - "step": 14543 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019508083609551512, - "loss": 1.448, - "step": 14544 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019508049786917407, - "loss": 1.6758, - "step": 14545 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195080159642833, - "loss": 1.7638, - "step": 14546 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019507982141649192, - "loss": 1.484, - "step": 14547 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019507948319015086, - "loss": 1.6663, - "step": 14548 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950791449638098, - "loss": 1.6428, - "step": 14549 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019507880673746872, - "loss": 1.5022, - "step": 14550 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019507846851112766, - "loss": 1.6055, - "step": 14551 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950781302847866, - "loss": 1.5786, - "step": 14552 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950777920584455, - "loss": 1.5348, - "step": 14553 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019507745383210445, - "loss": 1.6945, - "step": 14554 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950771156057634, - "loss": 1.695, - "step": 14555 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950767773794223, - "loss": 1.5261, - "step": 14556 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019507643915308125, - "loss": 1.5917, - "step": 14557 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950761009267402, - "loss": 1.5891, - "step": 14558 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950757627003991, - "loss": 1.5146, - "step": 14559 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019507542447405805, - "loss": 1.7147, - "step": 14560 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195075086247717, - "loss": 1.5089, - "step": 14561 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950747480213759, - "loss": 1.7042, - "step": 14562 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019507440979503484, - "loss": 1.6364, - "step": 14563 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019507407156869378, - "loss": 1.6234, - "step": 14564 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950737333423527, - "loss": 1.6613, - "step": 14565 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019507339511601164, - "loss": 1.5906, - "step": 14566 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019507305688967058, - "loss": 1.7423, - "step": 14567 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950727186633295, - "loss": 1.7067, - "step": 14568 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019507238043698844, - "loss": 1.5246, - "step": 14569 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019507204221064738, - "loss": 1.8324, - "step": 14570 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950717039843063, - "loss": 1.5698, - "step": 14571 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019507136575796523, - "loss": 1.6927, - "step": 14572 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019507102753162417, - "loss": 1.4767, - "step": 14573 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950706893052831, - "loss": 1.4915, - "step": 14574 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019507035107894203, - "loss": 1.7276, - "step": 14575 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019507001285260097, - "loss": 1.7315, - "step": 14576 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950696746262599, - "loss": 1.6048, - "step": 14577 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019506933639991885, - "loss": 1.7909, - "step": 14578 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019506899817357777, - "loss": 1.5645, - "step": 14579 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950686599472367, - "loss": 1.6383, - "step": 14580 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019506832172089565, - "loss": 1.398, - "step": 14581 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019506798349455456, - "loss": 1.6103, - "step": 14582 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950676452682135, - "loss": 1.5074, - "step": 14583 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019506730704187244, - "loss": 1.6461, - "step": 14584 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019506696881553136, - "loss": 1.3658, - "step": 14585 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950666305891903, - "loss": 1.4619, - "step": 14586 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019506629236284924, - "loss": 1.6851, - "step": 14587 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019506595413650815, - "loss": 1.6539, - "step": 14588 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950656159101671, - "loss": 1.7399, - "step": 14589 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019506527768382604, - "loss": 1.5737, - "step": 14590 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019506493945748495, - "loss": 1.4068, - "step": 14591 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950646012311439, - "loss": 1.6501, - "step": 14592 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019506426300480283, - "loss": 1.5082, - "step": 14593 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019506392477846175, - "loss": 1.5826, - "step": 14594 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950635865521207, - "loss": 1.5758, - "step": 14595 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019506324832577963, - "loss": 1.6975, - "step": 14596 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019506291009943854, - "loss": 1.7104, - "step": 14597 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019506257187309748, - "loss": 1.4804, - "step": 14598 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019506223364675643, - "loss": 1.6399, - "step": 14599 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019506189542041534, - "loss": 1.6178, - "step": 14600 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019506155719407428, - "loss": 1.5918, - "step": 14601 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019506121896773322, - "loss": 1.3696, - "step": 14602 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019506088074139214, - "loss": 1.6264, - "step": 14603 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019506054251505108, - "loss": 1.5146, - "step": 14604 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019506020428871002, - "loss": 1.6453, - "step": 14605 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019505986606236893, - "loss": 1.5179, - "step": 14606 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019505952783602787, - "loss": 1.5373, - "step": 14607 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019505918960968681, - "loss": 1.4054, - "step": 14608 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019505885138334573, - "loss": 1.7297, - "step": 14609 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019505851315700467, - "loss": 1.6852, - "step": 14610 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950581749306636, - "loss": 1.6109, - "step": 14611 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019505783670432253, - "loss": 1.6498, - "step": 14612 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019505749847798147, - "loss": 1.6643, - "step": 14613 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950571602516404, - "loss": 1.4928, - "step": 14614 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019505682202529932, - "loss": 1.5648, - "step": 14615 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019505648379895826, - "loss": 1.5267, - "step": 14616 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950561455726172, - "loss": 1.6129, - "step": 14617 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019505580734627615, - "loss": 1.6136, - "step": 14618 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019505546911993506, - "loss": 1.6421, - "step": 14619 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195055130893594, - "loss": 1.4131, - "step": 14620 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019505479266725294, - "loss": 1.5068, - "step": 14621 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019505445444091186, - "loss": 1.3972, - "step": 14622 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950541162145708, - "loss": 1.5754, - "step": 14623 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019505377798822974, - "loss": 1.3967, - "step": 14624 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019505343976188865, - "loss": 1.5249, - "step": 14625 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950531015355476, - "loss": 1.5202, - "step": 14626 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019505276330920653, - "loss": 1.5369, - "step": 14627 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019505242508286545, - "loss": 1.484, - "step": 14628 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950520868565244, - "loss": 1.6776, - "step": 14629 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019505174863018333, - "loss": 1.7946, - "step": 14630 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019505141040384224, - "loss": 1.6965, - "step": 14631 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950510721775012, - "loss": 1.5042, - "step": 14632 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019505073395116013, - "loss": 1.561, - "step": 14633 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019505039572481907, - "loss": 1.6361, - "step": 14634 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195050057498478, - "loss": 1.6519, - "step": 14635 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019504971927213692, - "loss": 1.6343, - "step": 14636 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019504938104579586, - "loss": 1.4876, - "step": 14637 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950490428194548, - "loss": 1.5613, - "step": 14638 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019504870459311372, - "loss": 1.6986, - "step": 14639 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019504836636677266, - "loss": 1.6158, - "step": 14640 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950480281404316, - "loss": 1.4474, - "step": 14641 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019504768991409052, - "loss": 1.6453, - "step": 14642 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019504735168774946, - "loss": 1.5223, - "step": 14643 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950470134614084, - "loss": 1.6223, - "step": 14644 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950466752350673, - "loss": 1.6935, - "step": 14645 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019504633700872625, - "loss": 1.5799, - "step": 14646 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950459987823852, - "loss": 1.5935, - "step": 14647 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950456605560441, - "loss": 1.4312, - "step": 14648 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019504532232970305, - "loss": 1.4831, - "step": 14649 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195044984103362, - "loss": 1.5265, - "step": 14650 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950446458770209, - "loss": 1.6841, - "step": 14651 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019504430765067985, - "loss": 1.6202, - "step": 14652 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950439694243388, - "loss": 1.6145, - "step": 14653 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950436311979977, - "loss": 1.6715, - "step": 14654 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019504329297165664, - "loss": 1.7636, - "step": 14655 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019504295474531558, - "loss": 1.7261, - "step": 14656 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950426165189745, - "loss": 1.6842, - "step": 14657 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019504227829263344, - "loss": 1.4964, - "step": 14658 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019504194006629238, - "loss": 1.718, - "step": 14659 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950416018399513, - "loss": 1.7276, - "step": 14660 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019504126361361023, - "loss": 1.6666, - "step": 14661 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019504092538726918, - "loss": 1.5911, - "step": 14662 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950405871609281, - "loss": 1.7373, - "step": 14663 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019504024893458703, - "loss": 1.7495, - "step": 14664 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019503991070824597, - "loss": 1.5792, - "step": 14665 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019503957248190489, - "loss": 1.6079, - "step": 14666 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019503923425556383, - "loss": 1.6134, - "step": 14667 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019503889602922277, - "loss": 1.6066, - "step": 14668 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019503855780288168, - "loss": 1.6117, - "step": 14669 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019503821957654062, - "loss": 1.7222, - "step": 14670 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019503788135019956, - "loss": 1.7578, - "step": 14671 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019503754312385848, - "loss": 1.6782, - "step": 14672 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019503720489751742, - "loss": 1.5655, - "step": 14673 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019503686667117636, - "loss": 1.689, - "step": 14674 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019503652844483527, - "loss": 1.6103, - "step": 14675 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019503619021849422, - "loss": 1.6572, - "step": 14676 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019503585199215316, - "loss": 1.5183, - "step": 14677 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019503551376581207, - "loss": 1.7468, - "step": 14678 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195035175539471, - "loss": 1.5146, - "step": 14679 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019503483731312995, - "loss": 1.6148, - "step": 14680 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019503449908678887, - "loss": 1.6476, - "step": 14681 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950341608604478, - "loss": 1.585, - "step": 14682 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019503382263410675, - "loss": 1.6073, - "step": 14683 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019503348440776566, - "loss": 1.7266, - "step": 14684 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950331461814246, - "loss": 1.6547, - "step": 14685 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019503280795508355, - "loss": 1.6498, - "step": 14686 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019503246972874246, - "loss": 1.5941, - "step": 14687 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019503213150240143, - "loss": 1.7696, - "step": 14688 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019503179327606037, - "loss": 1.5494, - "step": 14689 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019503145504971928, - "loss": 1.5681, - "step": 14690 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019503111682337822, - "loss": 1.6982, - "step": 14691 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019503077859703717, - "loss": 1.5461, - "step": 14692 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019503044037069608, - "loss": 1.662, - "step": 14693 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019503010214435502, - "loss": 1.5084, - "step": 14694 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019502976391801396, - "loss": 1.5076, - "step": 14695 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019502942569167288, - "loss": 1.6991, - "step": 14696 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019502908746533182, - "loss": 1.6173, - "step": 14697 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019502874923899076, - "loss": 1.5617, - "step": 14698 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019502841101264967, - "loss": 1.4678, - "step": 14699 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019502807278630861, - "loss": 1.5314, - "step": 14700 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019502773455996755, - "loss": 1.5263, - "step": 14701 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019502739633362647, - "loss": 1.6747, - "step": 14702 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950270581072854, - "loss": 1.6043, - "step": 14703 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019502671988094435, - "loss": 1.5832, - "step": 14704 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019502638165460327, - "loss": 1.7228, - "step": 14705 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950260434282622, - "loss": 1.6744, - "step": 14706 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019502570520192115, - "loss": 1.6727, - "step": 14707 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019502536697558006, - "loss": 1.6767, - "step": 14708 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195025028749239, - "loss": 1.6764, - "step": 14709 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019502469052289794, - "loss": 1.537, - "step": 14710 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019502435229655686, - "loss": 1.6255, - "step": 14711 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950240140702158, - "loss": 1.5909, - "step": 14712 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019502367584387474, - "loss": 1.3756, - "step": 14713 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019502333761753365, - "loss": 1.5065, - "step": 14714 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950229993911926, - "loss": 1.4731, - "step": 14715 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019502266116485154, - "loss": 1.5546, - "step": 14716 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019502232293851045, - "loss": 1.5516, - "step": 14717 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950219847121694, - "loss": 1.5509, - "step": 14718 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019502164648582833, - "loss": 1.7252, - "step": 14719 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019502130825948725, - "loss": 1.7003, - "step": 14720 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950209700331462, - "loss": 1.6185, - "step": 14721 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019502063180680513, - "loss": 1.7086, - "step": 14722 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019502029358046404, - "loss": 1.5881, - "step": 14723 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019501995535412298, - "loss": 1.467, - "step": 14724 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019501961712778193, - "loss": 1.647, - "step": 14725 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019501927890144084, - "loss": 1.5798, - "step": 14726 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019501894067509978, - "loss": 1.5401, - "step": 14727 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019501860244875872, - "loss": 1.6075, - "step": 14728 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019501826422241764, - "loss": 1.5825, - "step": 14729 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019501792599607658, - "loss": 1.6495, - "step": 14730 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019501758776973552, - "loss": 1.7311, - "step": 14731 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019501724954339443, - "loss": 1.5639, - "step": 14732 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019501691131705337, - "loss": 1.6207, - "step": 14733 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019501657309071231, - "loss": 1.4873, - "step": 14734 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019501623486437123, - "loss": 1.6205, - "step": 14735 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019501589663803017, - "loss": 1.4469, - "step": 14736 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950155584116891, - "loss": 1.7349, - "step": 14737 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019501522018534802, - "loss": 1.7063, - "step": 14738 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019501488195900697, - "loss": 1.4924, - "step": 14739 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950145437326659, - "loss": 1.5991, - "step": 14740 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019501420550632482, - "loss": 1.6653, - "step": 14741 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019501386727998376, - "loss": 1.4989, - "step": 14742 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950135290536427, - "loss": 1.6638, - "step": 14743 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019501319082730164, - "loss": 1.5431, - "step": 14744 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019501285260096059, - "loss": 1.5992, - "step": 14745 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019501251437461953, - "loss": 1.6321, - "step": 14746 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019501217614827844, - "loss": 1.4302, - "step": 14747 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019501183792193738, - "loss": 1.6771, - "step": 14748 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019501149969559632, - "loss": 1.5862, - "step": 14749 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019501116146925524, - "loss": 1.6445, - "step": 14750 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019501082324291418, - "loss": 1.4996, - "step": 14751 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019501048501657312, - "loss": 1.6227, - "step": 14752 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019501014679023203, - "loss": 1.8161, - "step": 14753 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019500980856389097, - "loss": 1.6184, - "step": 14754 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019500947033754992, - "loss": 1.743, - "step": 14755 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019500913211120883, - "loss": 1.5344, - "step": 14756 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019500879388486777, - "loss": 1.5991, - "step": 14757 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950084556585267, - "loss": 1.5522, - "step": 14758 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019500811743218563, - "loss": 1.5845, - "step": 14759 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019500777920584457, - "loss": 1.6212, - "step": 14760 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950074409795035, - "loss": 1.6697, - "step": 14761 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019500710275316242, - "loss": 1.717, - "step": 14762 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019500676452682136, - "loss": 1.619, - "step": 14763 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950064263004803, - "loss": 1.4523, - "step": 14764 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019500608807413922, - "loss": 1.52, - "step": 14765 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019500574984779816, - "loss": 1.6153, - "step": 14766 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950054116214571, - "loss": 1.6996, - "step": 14767 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019500507339511601, - "loss": 1.5251, - "step": 14768 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019500473516877496, - "loss": 1.4292, - "step": 14769 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950043969424339, - "loss": 1.5105, - "step": 14770 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950040587160928, - "loss": 1.4001, - "step": 14771 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019500372048975175, - "loss": 1.5946, - "step": 14772 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950033822634107, - "loss": 1.5457, - "step": 14773 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950030440370696, - "loss": 1.6743, - "step": 14774 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019500270581072855, - "loss": 1.5619, - "step": 14775 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950023675843875, - "loss": 1.2782, - "step": 14776 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950020293580464, - "loss": 1.6336, - "step": 14777 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019500169113170534, - "loss": 1.5183, - "step": 14778 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019500135290536429, - "loss": 1.5956, - "step": 14779 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001950010146790232, - "loss": 1.6559, - "step": 14780 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019500067645268214, - "loss": 1.5946, - "step": 14781 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019500033822634108, - "loss": 1.6278, - "step": 14782 - }, - { - "epoch": 0.05, - "learning_rate": 0.000195, - "loss": 1.5684, - "step": 14783 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019499966177365894, - "loss": 1.5245, - "step": 14784 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019499932354731788, - "loss": 1.6184, - "step": 14785 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949989853209768, - "loss": 1.6277, - "step": 14786 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019499864709463573, - "loss": 1.4774, - "step": 14787 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019499830886829468, - "loss": 1.6558, - "step": 14788 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949979706419536, - "loss": 1.6582, - "step": 14789 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019499763241561253, - "loss": 1.7687, - "step": 14790 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019499729418927147, - "loss": 1.6768, - "step": 14791 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019499695596293039, - "loss": 1.4765, - "step": 14792 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019499661773658933, - "loss": 1.4567, - "step": 14793 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019499627951024827, - "loss": 1.3928, - "step": 14794 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019499594128390718, - "loss": 1.7947, - "step": 14795 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019499560305756612, - "loss": 1.6609, - "step": 14796 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019499526483122506, - "loss": 1.5895, - "step": 14797 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019499492660488398, - "loss": 1.6373, - "step": 14798 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019499458837854292, - "loss": 1.7195, - "step": 14799 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019499425015220186, - "loss": 1.7507, - "step": 14800 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949939119258608, - "loss": 1.63, - "step": 14801 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019499357369951974, - "loss": 1.6, - "step": 14802 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019499323547317868, - "loss": 1.7567, - "step": 14803 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949928972468376, - "loss": 1.6091, - "step": 14804 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019499255902049654, - "loss": 1.6361, - "step": 14805 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019499222079415548, - "loss": 1.6399, - "step": 14806 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949918825678144, - "loss": 1.6837, - "step": 14807 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019499154434147334, - "loss": 1.6187, - "step": 14808 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019499120611513228, - "loss": 1.7074, - "step": 14809 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949908678887912, - "loss": 1.5646, - "step": 14810 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019499052966245013, - "loss": 1.7056, - "step": 14811 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019499019143610907, - "loss": 1.6533, - "step": 14812 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194989853209768, - "loss": 1.6123, - "step": 14813 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019498951498342693, - "loss": 1.5046, - "step": 14814 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019498917675708587, - "loss": 1.45, - "step": 14815 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019498883853074478, - "loss": 1.6378, - "step": 14816 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019498850030440372, - "loss": 1.631, - "step": 14817 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019498816207806267, - "loss": 1.4051, - "step": 14818 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019498782385172158, - "loss": 1.6073, - "step": 14819 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019498748562538052, - "loss": 1.6153, - "step": 14820 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019498714739903946, - "loss": 1.6302, - "step": 14821 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019498680917269838, - "loss": 1.6808, - "step": 14822 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019498647094635732, - "loss": 1.5416, - "step": 14823 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019498613272001626, - "loss": 1.6791, - "step": 14824 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019498579449367517, - "loss": 1.652, - "step": 14825 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949854562673341, - "loss": 1.4899, - "step": 14826 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019498511804099305, - "loss": 1.5707, - "step": 14827 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019498477981465197, - "loss": 1.6738, - "step": 14828 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949844415883109, - "loss": 1.5813, - "step": 14829 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019498410336196985, - "loss": 1.6179, - "step": 14830 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019498376513562876, - "loss": 1.6786, - "step": 14831 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949834269092877, - "loss": 1.5317, - "step": 14832 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019498308868294665, - "loss": 1.6067, - "step": 14833 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019498275045660556, - "loss": 1.6272, - "step": 14834 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949824122302645, - "loss": 1.4045, - "step": 14835 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019498207400392344, - "loss": 1.6413, - "step": 14836 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019498173577758236, - "loss": 1.5939, - "step": 14837 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949813975512413, - "loss": 1.5226, - "step": 14838 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019498105932490024, - "loss": 1.6884, - "step": 14839 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019498072109855915, - "loss": 1.4982, - "step": 14840 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949803828722181, - "loss": 1.5332, - "step": 14841 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019498004464587704, - "loss": 1.5609, - "step": 14842 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019497970641953595, - "loss": 1.4246, - "step": 14843 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949793681931949, - "loss": 1.5518, - "step": 14844 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019497902996685383, - "loss": 1.6408, - "step": 14845 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019497869174051275, - "loss": 1.5071, - "step": 14846 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949783535141717, - "loss": 1.5191, - "step": 14847 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019497801528783063, - "loss": 1.7268, - "step": 14848 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019497767706148954, - "loss": 1.6096, - "step": 14849 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019497733883514848, - "loss": 1.5182, - "step": 14850 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019497700060880742, - "loss": 1.4315, - "step": 14851 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019497666238246634, - "loss": 1.4718, - "step": 14852 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019497632415612528, - "loss": 1.5123, - "step": 14853 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019497598592978422, - "loss": 1.6071, - "step": 14854 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019497564770344314, - "loss": 1.6862, - "step": 14855 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019497530947710208, - "loss": 1.5283, - "step": 14856 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019497497125076102, - "loss": 1.5816, - "step": 14857 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019497463302441996, - "loss": 1.6273, - "step": 14858 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949742947980789, - "loss": 1.544, - "step": 14859 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019497395657173781, - "loss": 1.55, - "step": 14860 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019497361834539675, - "loss": 1.6953, - "step": 14861 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949732801190557, - "loss": 1.5391, - "step": 14862 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949729418927146, - "loss": 1.6192, - "step": 14863 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019497260366637355, - "loss": 1.4979, - "step": 14864 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949722654400325, - "loss": 1.5011, - "step": 14865 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949719272136914, - "loss": 1.5847, - "step": 14866 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019497158898735035, - "loss": 1.5746, - "step": 14867 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949712507610093, - "loss": 1.5912, - "step": 14868 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949709125346682, - "loss": 1.7078, - "step": 14869 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019497057430832714, - "loss": 1.5633, - "step": 14870 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019497023608198608, - "loss": 1.4549, - "step": 14871 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194969897855645, - "loss": 1.4814, - "step": 14872 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019496955962930394, - "loss": 1.5604, - "step": 14873 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019496922140296288, - "loss": 1.8312, - "step": 14874 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949688831766218, - "loss": 1.5533, - "step": 14875 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019496854495028074, - "loss": 1.6467, - "step": 14876 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019496820672393968, - "loss": 1.6372, - "step": 14877 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949678684975986, - "loss": 1.7499, - "step": 14878 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019496753027125753, - "loss": 1.7478, - "step": 14879 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019496719204491647, - "loss": 1.5837, - "step": 14880 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949668538185754, - "loss": 1.3892, - "step": 14881 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019496651559223433, - "loss": 1.6065, - "step": 14882 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019496617736589327, - "loss": 1.6123, - "step": 14883 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019496583913955218, - "loss": 1.7181, - "step": 14884 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019496550091321113, - "loss": 1.6679, - "step": 14885 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019496516268687007, - "loss": 1.6977, - "step": 14886 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194964824460529, - "loss": 1.6955, - "step": 14887 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019496448623418792, - "loss": 1.6175, - "step": 14888 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019496414800784686, - "loss": 1.5872, - "step": 14889 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949638097815058, - "loss": 1.568, - "step": 14890 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019496347155516472, - "loss": 1.6926, - "step": 14891 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019496313332882366, - "loss": 1.4902, - "step": 14892 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949627951024826, - "loss": 1.5828, - "step": 14893 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019496245687614151, - "loss": 1.5481, - "step": 14894 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019496211864980046, - "loss": 1.5075, - "step": 14895 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949617804234594, - "loss": 1.637, - "step": 14896 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949614421971183, - "loss": 1.6099, - "step": 14897 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019496110397077725, - "loss": 1.6287, - "step": 14898 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949607657444362, - "loss": 1.594, - "step": 14899 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949604275180951, - "loss": 1.642, - "step": 14900 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019496008929175405, - "loss": 1.5747, - "step": 14901 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194959751065413, - "loss": 1.6007, - "step": 14902 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949594128390719, - "loss": 1.6746, - "step": 14903 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019495907461273084, - "loss": 1.6057, - "step": 14904 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019495873638638979, - "loss": 1.5145, - "step": 14905 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949583981600487, - "loss": 1.535, - "step": 14906 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019495805993370764, - "loss": 1.6101, - "step": 14907 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019495772170736658, - "loss": 1.492, - "step": 14908 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949573834810255, - "loss": 1.6729, - "step": 14909 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019495704525468444, - "loss": 1.5675, - "step": 14910 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019495670702834338, - "loss": 1.4139, - "step": 14911 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949563688020023, - "loss": 1.5234, - "step": 14912 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019495603057566126, - "loss": 1.5517, - "step": 14913 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019495569234932017, - "loss": 1.6249, - "step": 14914 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019495535412297912, - "loss": 1.7863, - "step": 14915 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019495501589663806, - "loss": 1.6506, - "step": 14916 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019495467767029697, - "loss": 1.594, - "step": 14917 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949543394439559, - "loss": 1.5216, - "step": 14918 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019495400121761485, - "loss": 1.5258, - "step": 14919 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019495366299127377, - "loss": 1.5523, - "step": 14920 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949533247649327, - "loss": 1.5886, - "step": 14921 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019495298653859165, - "loss": 1.5459, - "step": 14922 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019495264831225056, - "loss": 1.7337, - "step": 14923 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949523100859095, - "loss": 1.7251, - "step": 14924 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019495197185956845, - "loss": 1.7414, - "step": 14925 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019495163363322736, - "loss": 1.7455, - "step": 14926 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949512954068863, - "loss": 1.7486, - "step": 14927 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019495095718054524, - "loss": 1.6797, - "step": 14928 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019495061895420416, - "loss": 1.526, - "step": 14929 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949502807278631, - "loss": 1.6393, - "step": 14930 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019494994250152204, - "loss": 1.5584, - "step": 14931 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019494960427518095, - "loss": 1.5977, - "step": 14932 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949492660488399, - "loss": 1.7973, - "step": 14933 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019494892782249883, - "loss": 1.5302, - "step": 14934 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019494858959615775, - "loss": 1.6044, - "step": 14935 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949482513698167, - "loss": 1.8357, - "step": 14936 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019494791314347563, - "loss": 1.4135, - "step": 14937 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019494757491713454, - "loss": 1.4902, - "step": 14938 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019494723669079349, - "loss": 1.5732, - "step": 14939 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019494689846445243, - "loss": 1.4851, - "step": 14940 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019494656023811134, - "loss": 1.6457, - "step": 14941 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019494622201177028, - "loss": 1.6195, - "step": 14942 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019494588378542922, - "loss": 1.6637, - "step": 14943 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019494554555908814, - "loss": 1.6029, - "step": 14944 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019494520733274708, - "loss": 1.5599, - "step": 14945 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019494486910640602, - "loss": 1.6814, - "step": 14946 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019494453088006493, - "loss": 1.5655, - "step": 14947 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019494419265372388, - "loss": 1.5467, - "step": 14948 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019494385442738282, - "loss": 1.734, - "step": 14949 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019494351620104173, - "loss": 1.5649, - "step": 14950 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019494317797470067, - "loss": 1.5319, - "step": 14951 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949428397483596, - "loss": 1.7373, - "step": 14952 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019494250152201853, - "loss": 1.4899, - "step": 14953 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019494216329567747, - "loss": 1.671, - "step": 14954 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949418250693364, - "loss": 1.804, - "step": 14955 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019494148684299532, - "loss": 1.5024, - "step": 14956 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019494114861665426, - "loss": 1.6198, - "step": 14957 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949408103903132, - "loss": 1.7483, - "step": 14958 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019494047216397212, - "loss": 1.6707, - "step": 14959 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019494013393763106, - "loss": 1.6424, - "step": 14960 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019493979571129, - "loss": 1.6534, - "step": 14961 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019493945748494892, - "loss": 1.7253, - "step": 14962 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019493911925860786, - "loss": 1.5439, - "step": 14963 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949387810322668, - "loss": 1.5036, - "step": 14964 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949384428059257, - "loss": 1.6266, - "step": 14965 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019493810457958465, - "loss": 1.7087, - "step": 14966 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949377663532436, - "loss": 1.5595, - "step": 14967 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949374281269025, - "loss": 1.6572, - "step": 14968 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019493708990056148, - "loss": 1.6116, - "step": 14969 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019493675167422042, - "loss": 1.6056, - "step": 14970 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019493641344787933, - "loss": 1.5923, - "step": 14971 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019493607522153827, - "loss": 1.5738, - "step": 14972 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019493573699519721, - "loss": 1.6555, - "step": 14973 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019493539876885613, - "loss": 1.4962, - "step": 14974 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019493506054251507, - "loss": 1.5071, - "step": 14975 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194934722316174, - "loss": 1.531, - "step": 14976 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019493438408983292, - "loss": 1.5666, - "step": 14977 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019493404586349187, - "loss": 1.6045, - "step": 14978 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949337076371508, - "loss": 1.6296, - "step": 14979 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019493336941080972, - "loss": 1.6681, - "step": 14980 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019493303118446866, - "loss": 1.5345, - "step": 14981 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949326929581276, - "loss": 1.5583, - "step": 14982 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019493235473178652, - "loss": 1.571, - "step": 14983 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019493201650544546, - "loss": 1.7496, - "step": 14984 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949316782791044, - "loss": 1.4969, - "step": 14985 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949313400527633, - "loss": 1.5433, - "step": 14986 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019493100182642225, - "loss": 1.6161, - "step": 14987 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949306636000812, - "loss": 1.5797, - "step": 14988 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949303253737401, - "loss": 1.4996, - "step": 14989 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019492998714739905, - "loss": 1.551, - "step": 14990 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194929648921058, - "loss": 1.6389, - "step": 14991 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949293106947169, - "loss": 1.681, - "step": 14992 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019492897246837585, - "loss": 1.5472, - "step": 14993 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949286342420348, - "loss": 1.7178, - "step": 14994 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949282960156937, - "loss": 1.6585, - "step": 14995 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019492795778935264, - "loss": 1.5307, - "step": 14996 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019492761956301158, - "loss": 1.7559, - "step": 14997 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949272813366705, - "loss": 1.6511, - "step": 14998 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019492694311032944, - "loss": 1.6273, - "step": 14999 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019492660488398838, - "loss": 1.6857, - "step": 15000 - }, - { - "epoch": 0.05, - "eval_loss": 1.3643329326923077, - "mrc_em": 20.0, - "mrc_f1": 27.590476190476192, - "nli_acc": 0.19, - "step": 15000, - "sts_f1": 0.5964912280701754, - "ynat_f1": 0.6068766376167691 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949262666576473, - "loss": 1.7955, - "step": 15001 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019492592843130624, - "loss": 1.6418, - "step": 15002 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019492559020496518, - "loss": 1.6539, - "step": 15003 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949252519786241, - "loss": 1.6716, - "step": 15004 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019492491375228303, - "loss": 1.6407, - "step": 15005 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019492457552594197, - "loss": 1.5713, - "step": 15006 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949242372996009, - "loss": 1.7427, - "step": 15007 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019492389907325983, - "loss": 1.5878, - "step": 15008 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019492356084691877, - "loss": 1.6027, - "step": 15009 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019492322262057768, - "loss": 1.6507, - "step": 15010 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019492288439423662, - "loss": 1.5454, - "step": 15011 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019492254616789557, - "loss": 1.7423, - "step": 15012 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019492220794155448, - "loss": 1.7099, - "step": 15013 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019492186971521342, - "loss": 1.4892, - "step": 15014 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019492153148887236, - "loss": 1.584, - "step": 15015 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019492119326253128, - "loss": 1.5307, - "step": 15016 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019492085503619022, - "loss": 1.5817, - "step": 15017 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019492051680984916, - "loss": 1.6452, - "step": 15018 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019492017858350807, - "loss": 1.423, - "step": 15019 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019491984035716701, - "loss": 1.6084, - "step": 15020 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019491950213082595, - "loss": 1.5351, - "step": 15021 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019491916390448487, - "loss": 1.7589, - "step": 15022 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949188256781438, - "loss": 1.6561, - "step": 15023 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019491848745180275, - "loss": 1.5419, - "step": 15024 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949181492254617, - "loss": 1.6676, - "step": 15025 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019491781099912063, - "loss": 1.6467, - "step": 15026 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019491747277277957, - "loss": 1.6062, - "step": 15027 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949171345464385, - "loss": 1.7063, - "step": 15028 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019491679632009743, - "loss": 1.5751, - "step": 15029 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019491645809375637, - "loss": 1.7663, - "step": 15030 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019491611986741528, - "loss": 1.6793, - "step": 15031 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019491578164107423, - "loss": 1.6186, - "step": 15032 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019491544341473317, - "loss": 1.6236, - "step": 15033 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019491510518839208, - "loss": 1.5363, - "step": 15034 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019491476696205102, - "loss": 1.6594, - "step": 15035 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019491442873570996, - "loss": 1.6716, - "step": 15036 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019491409050936888, - "loss": 1.4385, - "step": 15037 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019491375228302782, - "loss": 1.6626, - "step": 15038 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019491341405668676, - "loss": 1.6535, - "step": 15039 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019491307583034567, - "loss": 1.6301, - "step": 15040 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019491273760400461, - "loss": 1.5978, - "step": 15041 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019491239937766356, - "loss": 1.6188, - "step": 15042 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019491206115132247, - "loss": 1.5928, - "step": 15043 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949117229249814, - "loss": 1.4575, - "step": 15044 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019491138469864035, - "loss": 1.7243, - "step": 15045 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019491104647229927, - "loss": 1.5553, - "step": 15046 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949107082459582, - "loss": 1.4809, - "step": 15047 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019491037001961715, - "loss": 1.5562, - "step": 15048 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019491003179327606, - "loss": 1.6422, - "step": 15049 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194909693566935, - "loss": 1.5898, - "step": 15050 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019490935534059395, - "loss": 1.6578, - "step": 15051 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019490901711425286, - "loss": 1.6857, - "step": 15052 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949086788879118, - "loss": 1.6649, - "step": 15053 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019490834066157074, - "loss": 1.6667, - "step": 15054 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019490800243522966, - "loss": 1.5763, - "step": 15055 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949076642088886, - "loss": 1.5437, - "step": 15056 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019490732598254754, - "loss": 1.5183, - "step": 15057 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019490698775620645, - "loss": 1.619, - "step": 15058 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949066495298654, - "loss": 1.5473, - "step": 15059 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019490631130352433, - "loss": 1.6504, - "step": 15060 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019490597307718325, - "loss": 1.6182, - "step": 15061 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949056348508422, - "loss": 1.8448, - "step": 15062 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019490529662450113, - "loss": 1.6142, - "step": 15063 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019490495839816004, - "loss": 1.5373, - "step": 15064 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019490462017181899, - "loss": 1.5902, - "step": 15065 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019490428194547793, - "loss": 1.6835, - "step": 15066 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019490394371913684, - "loss": 1.6096, - "step": 15067 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019490360549279578, - "loss": 1.5568, - "step": 15068 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019490326726645472, - "loss": 1.4938, - "step": 15069 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019490292904011364, - "loss": 1.5486, - "step": 15070 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019490259081377258, - "loss": 1.8496, - "step": 15071 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019490225258743152, - "loss": 1.6606, - "step": 15072 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019490191436109043, - "loss": 1.5806, - "step": 15073 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019490157613474937, - "loss": 1.4649, - "step": 15074 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019490123790840832, - "loss": 1.4966, - "step": 15075 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019490089968206723, - "loss": 1.5047, - "step": 15076 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019490056145572617, - "loss": 1.5112, - "step": 15077 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001949002232293851, - "loss": 1.4465, - "step": 15078 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019489988500304403, - "loss": 1.3893, - "step": 15079 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019489954677670297, - "loss": 1.4894, - "step": 15080 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019489920855036194, - "loss": 1.5048, - "step": 15081 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019489887032402085, - "loss": 1.7735, - "step": 15082 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948985320976798, - "loss": 1.543, - "step": 15083 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019489819387133873, - "loss": 1.4728, - "step": 15084 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019489785564499765, - "loss": 1.5911, - "step": 15085 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948975174186566, - "loss": 1.8016, - "step": 15086 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019489717919231553, - "loss": 1.6714, - "step": 15087 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019489684096597444, - "loss": 1.5516, - "step": 15088 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019489650273963338, - "loss": 1.3871, - "step": 15089 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019489616451329232, - "loss": 1.7177, - "step": 15090 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019489582628695124, - "loss": 1.7749, - "step": 15091 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019489548806061018, - "loss": 1.735, - "step": 15092 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019489514983426912, - "loss": 1.7547, - "step": 15093 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019489481160792803, - "loss": 1.6583, - "step": 15094 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019489447338158698, - "loss": 1.6876, - "step": 15095 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019489413515524592, - "loss": 1.5798, - "step": 15096 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019489379692890483, - "loss": 1.5627, - "step": 15097 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019489345870256377, - "loss": 1.6738, - "step": 15098 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948931204762227, - "loss": 1.6174, - "step": 15099 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019489278224988163, - "loss": 1.6575, - "step": 15100 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019489244402354057, - "loss": 1.6248, - "step": 15101 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948921057971995, - "loss": 1.6625, - "step": 15102 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019489176757085842, - "loss": 1.5683, - "step": 15103 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019489142934451736, - "loss": 1.6406, - "step": 15104 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948910911181763, - "loss": 1.6204, - "step": 15105 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019489075289183522, - "loss": 1.6453, - "step": 15106 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019489041466549416, - "loss": 1.5248, - "step": 15107 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948900764391531, - "loss": 1.6069, - "step": 15108 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019488973821281202, - "loss": 1.6056, - "step": 15109 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019488939998647096, - "loss": 1.5332, - "step": 15110 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948890617601299, - "loss": 1.8132, - "step": 15111 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948887235337888, - "loss": 1.4637, - "step": 15112 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019488838530744775, - "loss": 1.5609, - "step": 15113 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948880470811067, - "loss": 1.459, - "step": 15114 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948877088547656, - "loss": 1.6605, - "step": 15115 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019488737062842455, - "loss": 1.6242, - "step": 15116 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948870324020835, - "loss": 1.5049, - "step": 15117 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948866941757424, - "loss": 1.7379, - "step": 15118 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019488635594940135, - "loss": 1.6591, - "step": 15119 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948860177230603, - "loss": 1.7206, - "step": 15120 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948856794967192, - "loss": 1.6127, - "step": 15121 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019488534127037814, - "loss": 1.7408, - "step": 15122 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019488500304403708, - "loss": 1.5081, - "step": 15123 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194884664817696, - "loss": 1.6837, - "step": 15124 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019488432659135494, - "loss": 1.7025, - "step": 15125 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019488398836501388, - "loss": 1.5743, - "step": 15126 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948836501386728, - "loss": 1.4546, - "step": 15127 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019488331191233174, - "loss": 1.6421, - "step": 15128 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019488297368599068, - "loss": 1.7009, - "step": 15129 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948826354596496, - "loss": 1.7283, - "step": 15130 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019488229723330853, - "loss": 1.502, - "step": 15131 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019488195900696747, - "loss": 1.5863, - "step": 15132 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948816207806264, - "loss": 1.5542, - "step": 15133 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019488128255428533, - "loss": 1.6611, - "step": 15134 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019488094432794427, - "loss": 1.4752, - "step": 15135 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019488060610160318, - "loss": 1.585, - "step": 15136 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019488026787526215, - "loss": 1.7749, - "step": 15137 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019487992964892107, - "loss": 1.5293, - "step": 15138 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019487959142258, - "loss": 1.6655, - "step": 15139 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019487925319623895, - "loss": 1.6658, - "step": 15140 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019487891496989786, - "loss": 1.6799, - "step": 15141 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948785767435568, - "loss": 1.7624, - "step": 15142 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019487823851721574, - "loss": 1.5285, - "step": 15143 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019487790029087466, - "loss": 1.6224, - "step": 15144 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948775620645336, - "loss": 1.6208, - "step": 15145 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019487722383819254, - "loss": 1.5711, - "step": 15146 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019487688561185145, - "loss": 1.6419, - "step": 15147 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948765473855104, - "loss": 1.566, - "step": 15148 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019487620915916934, - "loss": 1.494, - "step": 15149 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019487587093282825, - "loss": 1.6398, - "step": 15150 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948755327064872, - "loss": 1.6285, - "step": 15151 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019487519448014613, - "loss": 1.8805, - "step": 15152 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019487485625380507, - "loss": 1.6806, - "step": 15153 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194874518027464, - "loss": 1.8132, - "step": 15154 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019487417980112293, - "loss": 1.6763, - "step": 15155 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019487384157478187, - "loss": 1.4588, - "step": 15156 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019487350334844078, - "loss": 1.6206, - "step": 15157 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019487316512209973, - "loss": 1.4929, - "step": 15158 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019487282689575867, - "loss": 1.4439, - "step": 15159 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019487248866941758, - "loss": 1.5777, - "step": 15160 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019487215044307652, - "loss": 1.5577, - "step": 15161 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019487181221673546, - "loss": 1.6406, - "step": 15162 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019487147399039438, - "loss": 1.6701, - "step": 15163 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019487113576405332, - "loss": 1.7109, - "step": 15164 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019487079753771226, - "loss": 1.5064, - "step": 15165 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019487045931137117, - "loss": 1.6126, - "step": 15166 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019487012108503011, - "loss": 1.5099, - "step": 15167 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019486978285868906, - "loss": 1.5117, - "step": 15168 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019486944463234797, - "loss": 1.5284, - "step": 15169 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948691064060069, - "loss": 1.5597, - "step": 15170 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019486876817966585, - "loss": 1.6059, - "step": 15171 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019486842995332477, - "loss": 1.5408, - "step": 15172 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948680917269837, - "loss": 1.6239, - "step": 15173 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019486775350064265, - "loss": 1.6697, - "step": 15174 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019486741527430156, - "loss": 1.4682, - "step": 15175 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948670770479605, - "loss": 1.7372, - "step": 15176 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019486673882161944, - "loss": 1.5404, - "step": 15177 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019486640059527836, - "loss": 1.6284, - "step": 15178 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948660623689373, - "loss": 1.5293, - "step": 15179 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019486572414259624, - "loss": 1.6667, - "step": 15180 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019486538591625515, - "loss": 1.6847, - "step": 15181 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948650476899141, - "loss": 1.5347, - "step": 15182 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019486470946357304, - "loss": 1.6073, - "step": 15183 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019486437123723195, - "loss": 1.7517, - "step": 15184 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948640330108909, - "loss": 1.684, - "step": 15185 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019486369478454983, - "loss": 1.7087, - "step": 15186 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019486335655820875, - "loss": 1.5505, - "step": 15187 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948630183318677, - "loss": 1.6931, - "step": 15188 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019486268010552663, - "loss": 1.5788, - "step": 15189 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019486234187918554, - "loss": 1.7727, - "step": 15190 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019486200365284448, - "loss": 1.6145, - "step": 15191 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019486166542650343, - "loss": 1.7148, - "step": 15192 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019486132720016234, - "loss": 1.6129, - "step": 15193 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948609889738213, - "loss": 1.5686, - "step": 15194 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019486065074748022, - "loss": 1.6239, - "step": 15195 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019486031252113916, - "loss": 1.3539, - "step": 15196 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948599742947981, - "loss": 1.4094, - "step": 15197 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019485963606845702, - "loss": 1.6594, - "step": 15198 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019485929784211596, - "loss": 1.5519, - "step": 15199 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948589596157749, - "loss": 1.8191, - "step": 15200 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019485862138943381, - "loss": 1.6901, - "step": 15201 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019485828316309276, - "loss": 1.459, - "step": 15202 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948579449367517, - "loss": 1.6423, - "step": 15203 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948576067104106, - "loss": 1.7118, - "step": 15204 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019485726848406955, - "loss": 1.6186, - "step": 15205 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948569302577285, - "loss": 1.7867, - "step": 15206 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948565920313874, - "loss": 1.6132, - "step": 15207 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019485625380504635, - "loss": 1.6019, - "step": 15208 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948559155787053, - "loss": 1.7749, - "step": 15209 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948555773523642, - "loss": 1.6095, - "step": 15210 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019485523912602315, - "loss": 1.573, - "step": 15211 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019485490089968209, - "loss": 1.5934, - "step": 15212 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194854562673341, - "loss": 1.6893, - "step": 15213 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019485422444699994, - "loss": 1.4019, - "step": 15214 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019485388622065888, - "loss": 1.5968, - "step": 15215 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948535479943178, - "loss": 1.678, - "step": 15216 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019485320976797674, - "loss": 1.4994, - "step": 15217 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019485287154163568, - "loss": 1.5446, - "step": 15218 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948525333152946, - "loss": 1.5502, - "step": 15219 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019485219508895353, - "loss": 1.6388, - "step": 15220 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019485185686261248, - "loss": 1.5725, - "step": 15221 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948515186362714, - "loss": 1.8331, - "step": 15222 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019485118040993033, - "loss": 1.5376, - "step": 15223 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019485084218358927, - "loss": 1.5002, - "step": 15224 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019485050395724819, - "loss": 1.4581, - "step": 15225 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019485016573090713, - "loss": 1.5406, - "step": 15226 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019484982750456607, - "loss": 1.6464, - "step": 15227 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019484948927822498, - "loss": 1.5049, - "step": 15228 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019484915105188392, - "loss": 1.5065, - "step": 15229 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019484881282554286, - "loss": 1.6302, - "step": 15230 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019484847459920178, - "loss": 1.6879, - "step": 15231 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019484813637286072, - "loss": 1.7156, - "step": 15232 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019484779814651966, - "loss": 1.4387, - "step": 15233 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019484745992017857, - "loss": 1.5953, - "step": 15234 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019484712169383752, - "loss": 1.6518, - "step": 15235 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019484678346749646, - "loss": 1.7, - "step": 15236 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019484644524115537, - "loss": 1.5086, - "step": 15237 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948461070148143, - "loss": 1.5873, - "step": 15238 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019484576878847325, - "loss": 1.6098, - "step": 15239 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948454305621322, - "loss": 1.5467, - "step": 15240 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948450923357911, - "loss": 1.5883, - "step": 15241 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019484475410945005, - "loss": 1.5411, - "step": 15242 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194844415883109, - "loss": 1.4708, - "step": 15243 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948440776567679, - "loss": 1.6329, - "step": 15244 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019484373943042685, - "loss": 1.7098, - "step": 15245 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948434012040858, - "loss": 1.4853, - "step": 15246 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948430629777447, - "loss": 1.6539, - "step": 15247 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019484272475140364, - "loss": 1.6312, - "step": 15248 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019484238652506258, - "loss": 1.6762, - "step": 15249 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019484204829872152, - "loss": 1.5273, - "step": 15250 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019484171007238047, - "loss": 1.5193, - "step": 15251 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019484137184603938, - "loss": 1.4732, - "step": 15252 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019484103361969832, - "loss": 1.5162, - "step": 15253 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019484069539335726, - "loss": 1.5365, - "step": 15254 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019484035716701618, - "loss": 1.5263, - "step": 15255 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019484001894067512, - "loss": 1.3958, - "step": 15256 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019483968071433406, - "loss": 1.4244, - "step": 15257 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019483934248799297, - "loss": 1.6794, - "step": 15258 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948390042616519, - "loss": 1.5495, - "step": 15259 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019483866603531085, - "loss": 1.608, - "step": 15260 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019483832780896977, - "loss": 1.6446, - "step": 15261 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948379895826287, - "loss": 1.6426, - "step": 15262 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019483765135628765, - "loss": 1.7144, - "step": 15263 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019483731312994656, - "loss": 1.5012, - "step": 15264 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948369749036055, - "loss": 1.5403, - "step": 15265 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019483663667726445, - "loss": 1.5925, - "step": 15266 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019483629845092336, - "loss": 1.5531, - "step": 15267 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948359602245823, - "loss": 1.5363, - "step": 15268 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019483562199824124, - "loss": 1.4227, - "step": 15269 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019483528377190016, - "loss": 1.6788, - "step": 15270 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948349455455591, - "loss": 1.6225, - "step": 15271 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019483460731921804, - "loss": 1.5047, - "step": 15272 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019483426909287695, - "loss": 1.6501, - "step": 15273 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948339308665359, - "loss": 1.4866, - "step": 15274 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019483359264019484, - "loss": 1.6618, - "step": 15275 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019483325441385375, - "loss": 1.5665, - "step": 15276 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948329161875127, - "loss": 1.7392, - "step": 15277 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019483257796117163, - "loss": 1.6195, - "step": 15278 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019483223973483055, - "loss": 1.7532, - "step": 15279 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948319015084895, - "loss": 1.5073, - "step": 15280 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019483156328214843, - "loss": 1.5247, - "step": 15281 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019483122505580734, - "loss": 1.5628, - "step": 15282 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019483088682946628, - "loss": 1.5298, - "step": 15283 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019483054860312522, - "loss": 1.5833, - "step": 15284 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019483021037678414, - "loss": 1.6319, - "step": 15285 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019482987215044308, - "loss": 1.6688, - "step": 15286 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019482953392410202, - "loss": 1.6043, - "step": 15287 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019482919569776094, - "loss": 1.6042, - "step": 15288 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019482885747141988, - "loss": 1.5824, - "step": 15289 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019482851924507882, - "loss": 1.5291, - "step": 15290 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019482818101873773, - "loss": 1.6631, - "step": 15291 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019482784279239667, - "loss": 1.4346, - "step": 15292 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019482750456605561, - "loss": 1.4986, - "step": 15293 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019482716633971453, - "loss": 1.61, - "step": 15294 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019482682811337347, - "loss": 1.7302, - "step": 15295 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948264898870324, - "loss": 1.5115, - "step": 15296 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019482615166069132, - "loss": 1.8446, - "step": 15297 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019482581343435027, - "loss": 1.6213, - "step": 15298 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948254752080092, - "loss": 1.6368, - "step": 15299 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019482513698166812, - "loss": 1.5003, - "step": 15300 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019482479875532706, - "loss": 1.6956, - "step": 15301 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194824460528986, - "loss": 1.5418, - "step": 15302 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019482412230264492, - "loss": 1.6011, - "step": 15303 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019482378407630386, - "loss": 1.6287, - "step": 15304 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948234458499628, - "loss": 1.5825, - "step": 15305 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019482310762362174, - "loss": 1.6836, - "step": 15306 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019482276939728068, - "loss": 1.6301, - "step": 15307 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019482243117093962, - "loss": 1.5125, - "step": 15308 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019482209294459854, - "loss": 1.7212, - "step": 15309 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019482175471825748, - "loss": 1.4736, - "step": 15310 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019482141649191642, - "loss": 1.528, - "step": 15311 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019482107826557533, - "loss": 1.5055, - "step": 15312 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019482074003923427, - "loss": 1.7102, - "step": 15313 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019482040181289322, - "loss": 1.5535, - "step": 15314 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019482006358655213, - "loss": 1.6705, - "step": 15315 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019481972536021107, - "loss": 1.575, - "step": 15316 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019481938713387, - "loss": 1.5118, - "step": 15317 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019481904890752893, - "loss": 1.5522, - "step": 15318 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019481871068118787, - "loss": 1.4505, - "step": 15319 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948183724548468, - "loss": 1.5142, - "step": 15320 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019481803422850572, - "loss": 1.635, - "step": 15321 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019481769600216466, - "loss": 1.6642, - "step": 15322 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948173577758236, - "loss": 1.4046, - "step": 15323 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019481701954948252, - "loss": 1.6709, - "step": 15324 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019481668132314146, - "loss": 1.6429, - "step": 15325 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948163430968004, - "loss": 1.6815, - "step": 15326 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019481600487045931, - "loss": 1.7203, - "step": 15327 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019481566664411826, - "loss": 1.5802, - "step": 15328 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948153284177772, - "loss": 1.7152, - "step": 15329 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948149901914361, - "loss": 1.5101, - "step": 15330 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019481465196509505, - "loss": 1.6675, - "step": 15331 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194814313738754, - "loss": 1.5805, - "step": 15332 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948139755124129, - "loss": 1.7872, - "step": 15333 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019481363728607185, - "loss": 1.5527, - "step": 15334 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948132990597308, - "loss": 1.6134, - "step": 15335 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948129608333897, - "loss": 1.6885, - "step": 15336 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019481262260704864, - "loss": 1.6477, - "step": 15337 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019481228438070759, - "loss": 1.5947, - "step": 15338 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948119461543665, - "loss": 1.6349, - "step": 15339 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019481160792802544, - "loss": 1.6623, - "step": 15340 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019481126970168438, - "loss": 1.6116, - "step": 15341 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948109314753433, - "loss": 1.618, - "step": 15342 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019481059324900224, - "loss": 1.7139, - "step": 15343 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019481025502266118, - "loss": 1.563, - "step": 15344 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948099167963201, - "loss": 1.5197, - "step": 15345 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019480957856997903, - "loss": 1.6976, - "step": 15346 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019480924034363797, - "loss": 1.69, - "step": 15347 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948089021172969, - "loss": 1.663, - "step": 15348 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019480856389095583, - "loss": 1.594, - "step": 15349 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019480822566461477, - "loss": 1.6259, - "step": 15350 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019480788743827368, - "loss": 1.5799, - "step": 15351 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019480754921193263, - "loss": 1.7862, - "step": 15352 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019480721098559157, - "loss": 1.5493, - "step": 15353 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019480687275925048, - "loss": 1.4875, - "step": 15354 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019480653453290942, - "loss": 1.6828, - "step": 15355 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019480619630656836, - "loss": 1.6845, - "step": 15356 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019480585808022728, - "loss": 1.8163, - "step": 15357 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019480551985388622, - "loss": 1.7881, - "step": 15358 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019480518162754516, - "loss": 1.7497, - "step": 15359 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019480484340120407, - "loss": 1.4273, - "step": 15360 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019480450517486301, - "loss": 1.4667, - "step": 15361 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019480416694852198, - "loss": 1.6959, - "step": 15362 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948038287221809, - "loss": 1.5415, - "step": 15363 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019480349049583984, - "loss": 1.6323, - "step": 15364 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019480315226949878, - "loss": 1.5822, - "step": 15365 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948028140431577, - "loss": 1.5568, - "step": 15366 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019480247581681663, - "loss": 1.4708, - "step": 15367 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019480213759047558, - "loss": 1.5687, - "step": 15368 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001948017993641345, - "loss": 1.672, - "step": 15369 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019480146113779343, - "loss": 1.8951, - "step": 15370 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019480112291145237, - "loss": 1.5983, - "step": 15371 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019480078468511129, - "loss": 1.6034, - "step": 15372 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019480044645877023, - "loss": 1.3158, - "step": 15373 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019480010823242917, - "loss": 1.5877, - "step": 15374 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019479977000608808, - "loss": 1.7153, - "step": 15375 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019479943177974702, - "loss": 1.4759, - "step": 15376 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019479909355340596, - "loss": 1.5868, - "step": 15377 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019479875532706488, - "loss": 1.5786, - "step": 15378 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019479841710072382, - "loss": 1.7, - "step": 15379 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019479807887438276, - "loss": 1.6483, - "step": 15380 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019479774064804168, - "loss": 1.7374, - "step": 15381 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019479740242170062, - "loss": 1.4994, - "step": 15382 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019479706419535956, - "loss": 1.7455, - "step": 15383 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019479672596901847, - "loss": 1.5743, - "step": 15384 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947963877426774, - "loss": 1.6849, - "step": 15385 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019479604951633635, - "loss": 1.6713, - "step": 15386 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019479571128999527, - "loss": 1.3395, - "step": 15387 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947953730636542, - "loss": 1.5249, - "step": 15388 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019479503483731315, - "loss": 1.6431, - "step": 15389 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019479469661097206, - "loss": 1.5511, - "step": 15390 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194794358384631, - "loss": 1.7075, - "step": 15391 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019479402015828995, - "loss": 1.703, - "step": 15392 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019479368193194886, - "loss": 1.7223, - "step": 15393 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947933437056078, - "loss": 1.5663, - "step": 15394 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019479300547926674, - "loss": 1.7395, - "step": 15395 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019479266725292566, - "loss": 1.6858, - "step": 15396 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947923290265846, - "loss": 1.6921, - "step": 15397 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019479199080024354, - "loss": 1.5434, - "step": 15398 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019479165257390245, - "loss": 1.6503, - "step": 15399 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947913143475614, - "loss": 1.7689, - "step": 15400 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019479097612122034, - "loss": 1.582, - "step": 15401 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019479063789487925, - "loss": 1.4698, - "step": 15402 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947902996685382, - "loss": 1.6788, - "step": 15403 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019478996144219713, - "loss": 1.7019, - "step": 15404 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019478962321585605, - "loss": 1.4945, - "step": 15405 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194789284989515, - "loss": 1.706, - "step": 15406 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019478894676317393, - "loss": 1.4784, - "step": 15407 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019478860853683284, - "loss": 1.5761, - "step": 15408 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019478827031049178, - "loss": 1.6378, - "step": 15409 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019478793208415072, - "loss": 1.728, - "step": 15410 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019478759385780964, - "loss": 1.5352, - "step": 15411 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019478725563146858, - "loss": 1.6864, - "step": 15412 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019478691740512752, - "loss": 1.6427, - "step": 15413 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019478657917878643, - "loss": 1.6128, - "step": 15414 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019478624095244538, - "loss": 1.6035, - "step": 15415 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019478590272610432, - "loss": 1.4588, - "step": 15416 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019478556449976323, - "loss": 1.3768, - "step": 15417 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947852262734222, - "loss": 1.5021, - "step": 15418 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947848880470811, - "loss": 1.637, - "step": 15419 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019478454982074005, - "loss": 1.5395, - "step": 15420 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194784211594399, - "loss": 1.5632, - "step": 15421 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019478387336805794, - "loss": 1.6045, - "step": 15422 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019478353514171685, - "loss": 1.7032, - "step": 15423 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947831969153758, - "loss": 1.6535, - "step": 15424 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019478285868903473, - "loss": 1.6164, - "step": 15425 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019478252046269365, - "loss": 1.588, - "step": 15426 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947821822363526, - "loss": 1.4699, - "step": 15427 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019478184401001153, - "loss": 1.6955, - "step": 15428 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019478150578367044, - "loss": 1.6616, - "step": 15429 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019478116755732938, - "loss": 1.5994, - "step": 15430 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019478082933098833, - "loss": 1.5645, - "step": 15431 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019478049110464724, - "loss": 1.581, - "step": 15432 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019478015287830618, - "loss": 1.5744, - "step": 15433 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019477981465196512, - "loss": 1.8013, - "step": 15434 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019477947642562404, - "loss": 1.6815, - "step": 15435 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019477913819928298, - "loss": 1.6105, - "step": 15436 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019477879997294192, - "loss": 1.5617, - "step": 15437 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019477846174660083, - "loss": 1.6468, - "step": 15438 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019477812352025977, - "loss": 1.6209, - "step": 15439 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019477778529391871, - "loss": 1.762, - "step": 15440 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019477744706757763, - "loss": 1.6835, - "step": 15441 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019477710884123657, - "loss": 1.6788, - "step": 15442 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947767706148955, - "loss": 1.5729, - "step": 15443 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019477643238855442, - "loss": 1.404, - "step": 15444 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019477609416221337, - "loss": 1.6624, - "step": 15445 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947757559358723, - "loss": 1.6356, - "step": 15446 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019477541770953122, - "loss": 1.5583, - "step": 15447 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019477507948319016, - "loss": 1.6742, - "step": 15448 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947747412568491, - "loss": 1.6752, - "step": 15449 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019477440303050802, - "loss": 1.5903, - "step": 15450 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019477406480416696, - "loss": 1.4249, - "step": 15451 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947737265778259, - "loss": 1.5601, - "step": 15452 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019477338835148481, - "loss": 1.4414, - "step": 15453 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019477305012514375, - "loss": 1.4115, - "step": 15454 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947727118988027, - "loss": 1.5366, - "step": 15455 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947723736724616, - "loss": 1.5584, - "step": 15456 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019477203544612055, - "loss": 1.4082, - "step": 15457 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947716972197795, - "loss": 1.5693, - "step": 15458 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947713589934384, - "loss": 1.4799, - "step": 15459 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019477102076709735, - "loss": 1.7337, - "step": 15460 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947706825407563, - "loss": 1.8041, - "step": 15461 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947703443144152, - "loss": 1.506, - "step": 15462 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019477000608807414, - "loss": 1.4922, - "step": 15463 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019476966786173308, - "loss": 1.6752, - "step": 15464 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194769329635392, - "loss": 1.4601, - "step": 15465 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019476899140905094, - "loss": 1.509, - "step": 15466 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019476865318270988, - "loss": 1.571, - "step": 15467 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947683149563688, - "loss": 1.4948, - "step": 15468 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019476797673002774, - "loss": 1.4461, - "step": 15469 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019476763850368668, - "loss": 1.6968, - "step": 15470 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947673002773456, - "loss": 1.7574, - "step": 15471 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019476696205100453, - "loss": 1.5866, - "step": 15472 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019476662382466347, - "loss": 1.5256, - "step": 15473 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019476628559832241, - "loss": 1.6746, - "step": 15474 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019476594737198136, - "loss": 1.4542, - "step": 15475 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019476560914564027, - "loss": 1.6798, - "step": 15476 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947652709192992, - "loss": 1.5761, - "step": 15477 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019476493269295815, - "loss": 1.576, - "step": 15478 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019476459446661707, - "loss": 1.5227, - "step": 15479 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194764256240276, - "loss": 1.6316, - "step": 15480 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019476391801393495, - "loss": 1.5626, - "step": 15481 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019476357978759386, - "loss": 1.593, - "step": 15482 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947632415612528, - "loss": 1.589, - "step": 15483 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019476290333491175, - "loss": 1.5407, - "step": 15484 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019476256510857066, - "loss": 1.5181, - "step": 15485 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947622268822296, - "loss": 1.6268, - "step": 15486 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019476188865588854, - "loss": 1.5138, - "step": 15487 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019476155042954746, - "loss": 1.5185, - "step": 15488 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947612122032064, - "loss": 1.5735, - "step": 15489 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019476087397686534, - "loss": 1.6878, - "step": 15490 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019476053575052425, - "loss": 1.4726, - "step": 15491 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947601975241832, - "loss": 1.5626, - "step": 15492 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019475985929784213, - "loss": 1.6191, - "step": 15493 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019475952107150105, - "loss": 1.6306, - "step": 15494 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019475918284516, - "loss": 1.5032, - "step": 15495 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019475884461881893, - "loss": 1.6575, - "step": 15496 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019475850639247784, - "loss": 1.7056, - "step": 15497 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019475816816613679, - "loss": 1.6967, - "step": 15498 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019475782993979573, - "loss": 1.534, - "step": 15499 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019475749171345464, - "loss": 1.5012, - "step": 15500 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019475715348711358, - "loss": 1.5888, - "step": 15501 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019475681526077252, - "loss": 1.6175, - "step": 15502 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019475647703443144, - "loss": 1.4291, - "step": 15503 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019475613880809038, - "loss": 1.6108, - "step": 15504 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019475580058174932, - "loss": 1.5935, - "step": 15505 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019475546235540826, - "loss": 1.5811, - "step": 15506 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019475512412906717, - "loss": 1.5646, - "step": 15507 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019475478590272612, - "loss": 1.8066, - "step": 15508 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019475444767638506, - "loss": 1.6227, - "step": 15509 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019475410945004397, - "loss": 1.4313, - "step": 15510 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947537712237029, - "loss": 1.6664, - "step": 15511 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019475343299736185, - "loss": 1.7038, - "step": 15512 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019475309477102077, - "loss": 1.5149, - "step": 15513 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947527565446797, - "loss": 1.5715, - "step": 15514 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019475241831833865, - "loss": 1.5764, - "step": 15515 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019475208009199756, - "loss": 1.6569, - "step": 15516 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947517418656565, - "loss": 1.6746, - "step": 15517 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019475140363931545, - "loss": 1.5425, - "step": 15518 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019475106541297436, - "loss": 1.5867, - "step": 15519 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947507271866333, - "loss": 1.7, - "step": 15520 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019475038896029224, - "loss": 1.4444, - "step": 15521 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019475005073395116, - "loss": 1.5614, - "step": 15522 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947497125076101, - "loss": 1.5753, - "step": 15523 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019474937428126904, - "loss": 1.6982, - "step": 15524 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019474903605492795, - "loss": 1.645, - "step": 15525 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947486978285869, - "loss": 1.7688, - "step": 15526 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019474835960224583, - "loss": 1.4678, - "step": 15527 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019474802137590475, - "loss": 1.5418, - "step": 15528 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947476831495637, - "loss": 1.5822, - "step": 15529 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019474734492322263, - "loss": 1.5413, - "step": 15530 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019474700669688157, - "loss": 1.5492, - "step": 15531 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947466684705405, - "loss": 1.761, - "step": 15532 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019474633024419943, - "loss": 1.7001, - "step": 15533 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019474599201785837, - "loss": 1.7195, - "step": 15534 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947456537915173, - "loss": 1.6548, - "step": 15535 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019474531556517622, - "loss": 1.5852, - "step": 15536 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019474497733883516, - "loss": 1.4584, - "step": 15537 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947446391124941, - "loss": 1.5904, - "step": 15538 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019474430088615302, - "loss": 1.5717, - "step": 15539 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019474396265981196, - "loss": 1.6922, - "step": 15540 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947436244334709, - "loss": 1.5154, - "step": 15541 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019474328620712982, - "loss": 1.607, - "step": 15542 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019474294798078876, - "loss": 1.5853, - "step": 15543 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947426097544477, - "loss": 1.5958, - "step": 15544 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947422715281066, - "loss": 1.7502, - "step": 15545 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019474193330176555, - "loss": 1.6421, - "step": 15546 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947415950754245, - "loss": 1.6247, - "step": 15547 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947412568490834, - "loss": 1.5706, - "step": 15548 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019474091862274235, - "loss": 1.6276, - "step": 15549 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947405803964013, - "loss": 1.6339, - "step": 15550 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947402421700602, - "loss": 1.6281, - "step": 15551 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019473990394371915, - "loss": 1.5784, - "step": 15552 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947395657173781, - "loss": 1.8138, - "step": 15553 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194739227491037, - "loss": 1.5546, - "step": 15554 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019473888926469594, - "loss": 1.6124, - "step": 15555 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019473855103835488, - "loss": 1.4314, - "step": 15556 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947382128120138, - "loss": 1.6418, - "step": 15557 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019473787458567274, - "loss": 1.5531, - "step": 15558 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019473753635933168, - "loss": 1.6538, - "step": 15559 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947371981329906, - "loss": 1.578, - "step": 15560 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019473685990664954, - "loss": 1.5584, - "step": 15561 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019473652168030848, - "loss": 1.6814, - "step": 15562 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947361834539674, - "loss": 1.7041, - "step": 15563 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019473584522762633, - "loss": 1.6333, - "step": 15564 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019473550700128527, - "loss": 1.466, - "step": 15565 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947351687749442, - "loss": 1.6864, - "step": 15566 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019473483054860313, - "loss": 1.5611, - "step": 15567 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019473449232226207, - "loss": 1.5985, - "step": 15568 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019473415409592098, - "loss": 1.5545, - "step": 15569 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019473381586957992, - "loss": 1.5994, - "step": 15570 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019473347764323887, - "loss": 1.513, - "step": 15571 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019473313941689778, - "loss": 1.5469, - "step": 15572 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019473280119055672, - "loss": 1.8604, - "step": 15573 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019473246296421566, - "loss": 1.5683, - "step": 15574 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019473212473787458, - "loss": 1.6962, - "step": 15575 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019473178651153352, - "loss": 1.597, - "step": 15576 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019473144828519246, - "loss": 1.5278, - "step": 15577 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019473111005885137, - "loss": 1.604, - "step": 15578 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947307718325103, - "loss": 1.6078, - "step": 15579 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019473043360616925, - "loss": 1.7202, - "step": 15580 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019473009537982817, - "loss": 1.602, - "step": 15581 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947297571534871, - "loss": 1.5307, - "step": 15582 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019472941892714605, - "loss": 1.6172, - "step": 15583 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019472908070080496, - "loss": 1.5995, - "step": 15584 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947287424744639, - "loss": 1.5925, - "step": 15585 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019472840424812285, - "loss": 1.492, - "step": 15586 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947280660217818, - "loss": 1.6558, - "step": 15587 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019472772779544073, - "loss": 1.7231, - "step": 15588 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019472738956909967, - "loss": 1.5795, - "step": 15589 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019472705134275858, - "loss": 1.7542, - "step": 15590 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019472671311641753, - "loss": 1.4424, - "step": 15591 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019472637489007647, - "loss": 1.5219, - "step": 15592 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019472603666373538, - "loss": 1.5822, - "step": 15593 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019472569843739432, - "loss": 1.5682, - "step": 15594 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019472536021105326, - "loss": 1.5407, - "step": 15595 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019472502198471218, - "loss": 1.7273, - "step": 15596 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019472468375837112, - "loss": 1.7487, - "step": 15597 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019472434553203006, - "loss": 1.7229, - "step": 15598 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019472400730568897, - "loss": 1.4104, - "step": 15599 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019472366907934791, - "loss": 1.6123, - "step": 15600 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019472333085300686, - "loss": 1.5992, - "step": 15601 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019472299262666577, - "loss": 1.6506, - "step": 15602 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947226544003247, - "loss": 1.577, - "step": 15603 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019472231617398365, - "loss": 1.5117, - "step": 15604 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019472197794764257, - "loss": 1.4559, - "step": 15605 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947216397213015, - "loss": 1.7123, - "step": 15606 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019472130149496045, - "loss": 1.628, - "step": 15607 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019472096326861936, - "loss": 1.5771, - "step": 15608 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947206250422783, - "loss": 1.4802, - "step": 15609 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019472028681593724, - "loss": 1.6209, - "step": 15610 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019471994858959616, - "loss": 1.679, - "step": 15611 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947196103632551, - "loss": 1.5641, - "step": 15612 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019471927213691404, - "loss": 1.5689, - "step": 15613 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019471893391057295, - "loss": 1.6931, - "step": 15614 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947185956842319, - "loss": 1.8385, - "step": 15615 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019471825745789084, - "loss": 1.6006, - "step": 15616 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019471791923154975, - "loss": 1.5661, - "step": 15617 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947175810052087, - "loss": 1.6962, - "step": 15618 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019471724277886763, - "loss": 1.5692, - "step": 15619 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019471690455252655, - "loss": 1.6504, - "step": 15620 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947165663261855, - "loss": 1.5953, - "step": 15621 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019471622809984443, - "loss": 1.5258, - "step": 15622 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019471588987350334, - "loss": 1.5091, - "step": 15623 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019471555164716228, - "loss": 1.5409, - "step": 15624 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019471521342082123, - "loss": 1.7591, - "step": 15625 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019471487519448014, - "loss": 1.6314, - "step": 15626 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019471453696813908, - "loss": 1.5935, - "step": 15627 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019471419874179802, - "loss": 1.3931, - "step": 15628 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019471386051545694, - "loss": 1.4379, - "step": 15629 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019471352228911588, - "loss": 1.7803, - "step": 15630 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019471318406277482, - "loss": 1.5771, - "step": 15631 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019471284583643373, - "loss": 1.6793, - "step": 15632 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019471250761009267, - "loss": 1.4494, - "step": 15633 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019471216938375161, - "loss": 1.5837, - "step": 15634 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019471183115741053, - "loss": 1.3876, - "step": 15635 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019471149293106947, - "loss": 1.5341, - "step": 15636 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947111547047284, - "loss": 1.5211, - "step": 15637 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019471081647838733, - "loss": 1.6297, - "step": 15638 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019471047825204627, - "loss": 1.5399, - "step": 15639 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947101400257052, - "loss": 1.5741, - "step": 15640 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019470980179936412, - "loss": 1.4768, - "step": 15641 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019470946357302306, - "loss": 1.6204, - "step": 15642 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019470912534668203, - "loss": 1.5127, - "step": 15643 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019470878712034095, - "loss": 1.6228, - "step": 15644 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019470844889399989, - "loss": 1.5335, - "step": 15645 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019470811066765883, - "loss": 1.6835, - "step": 15646 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019470777244131774, - "loss": 1.5425, - "step": 15647 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019470743421497668, - "loss": 1.5494, - "step": 15648 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019470709598863562, - "loss": 1.5208, - "step": 15649 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019470675776229454, - "loss": 1.6834, - "step": 15650 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019470641953595348, - "loss": 1.654, - "step": 15651 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019470608130961242, - "loss": 1.5007, - "step": 15652 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019470574308327133, - "loss": 1.353, - "step": 15653 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019470540485693028, - "loss": 1.5026, - "step": 15654 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019470506663058922, - "loss": 1.5924, - "step": 15655 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019470472840424813, - "loss": 1.4357, - "step": 15656 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019470439017790707, - "loss": 1.6183, - "step": 15657 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194704051951566, - "loss": 1.6696, - "step": 15658 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019470371372522493, - "loss": 1.5466, - "step": 15659 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019470337549888387, - "loss": 1.6564, - "step": 15660 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947030372725428, - "loss": 1.7216, - "step": 15661 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019470269904620172, - "loss": 1.5298, - "step": 15662 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019470236081986066, - "loss": 1.4756, - "step": 15663 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947020225935196, - "loss": 1.6143, - "step": 15664 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019470168436717852, - "loss": 1.7133, - "step": 15665 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019470134614083746, - "loss": 1.5492, - "step": 15666 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001947010079144964, - "loss": 1.708, - "step": 15667 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019470066968815532, - "loss": 1.665, - "step": 15668 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019470033146181426, - "loss": 1.6178, - "step": 15669 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946999932354732, - "loss": 1.7094, - "step": 15670 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946996550091321, - "loss": 1.7079, - "step": 15671 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019469931678279105, - "loss": 1.5086, - "step": 15672 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019469897855645, - "loss": 1.641, - "step": 15673 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946986403301089, - "loss": 1.5768, - "step": 15674 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019469830210376785, - "loss": 1.5799, - "step": 15675 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946979638774268, - "loss": 1.5675, - "step": 15676 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946976256510857, - "loss": 1.638, - "step": 15677 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019469728742474465, - "loss": 1.3986, - "step": 15678 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946969491984036, - "loss": 1.6724, - "step": 15679 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946966109720625, - "loss": 1.757, - "step": 15680 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019469627274572144, - "loss": 1.5173, - "step": 15681 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019469593451938038, - "loss": 1.592, - "step": 15682 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946955962930393, - "loss": 1.5823, - "step": 15683 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019469525806669824, - "loss": 1.5698, - "step": 15684 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019469491984035718, - "loss": 1.6598, - "step": 15685 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946945816140161, - "loss": 1.5699, - "step": 15686 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019469424338767503, - "loss": 1.4823, - "step": 15687 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019469390516133398, - "loss": 1.5727, - "step": 15688 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946935669349929, - "loss": 1.5337, - "step": 15689 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019469322870865183, - "loss": 1.5347, - "step": 15690 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019469289048231077, - "loss": 1.587, - "step": 15691 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019469255225596969, - "loss": 1.6722, - "step": 15692 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019469221402962863, - "loss": 1.5457, - "step": 15693 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019469187580328757, - "loss": 1.7303, - "step": 15694 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019469153757694648, - "loss": 1.6132, - "step": 15695 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019469119935060542, - "loss": 1.5068, - "step": 15696 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019469086112426436, - "loss": 1.6502, - "step": 15697 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019469052289792328, - "loss": 1.3778, - "step": 15698 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019469018467158225, - "loss": 1.5211, - "step": 15699 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946898464452412, - "loss": 1.6572, - "step": 15700 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946895082189001, - "loss": 1.4813, - "step": 15701 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019468916999255904, - "loss": 1.4561, - "step": 15702 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019468883176621798, - "loss": 1.5795, - "step": 15703 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946884935398769, - "loss": 1.5913, - "step": 15704 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019468815531353584, - "loss": 1.6056, - "step": 15705 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019468781708719478, - "loss": 1.6867, - "step": 15706 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946874788608537, - "loss": 1.5183, - "step": 15707 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019468714063451264, - "loss": 1.4376, - "step": 15708 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019468680240817158, - "loss": 1.6486, - "step": 15709 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946864641818305, - "loss": 1.6314, - "step": 15710 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019468612595548943, - "loss": 1.4838, - "step": 15711 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019468578772914837, - "loss": 1.8203, - "step": 15712 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946854495028073, - "loss": 1.6728, - "step": 15713 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019468511127646623, - "loss": 1.6356, - "step": 15714 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019468477305012517, - "loss": 1.5319, - "step": 15715 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019468443482378408, - "loss": 1.6664, - "step": 15716 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019468409659744302, - "loss": 1.5297, - "step": 15717 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019468375837110197, - "loss": 1.6717, - "step": 15718 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019468342014476088, - "loss": 1.428, - "step": 15719 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019468308191841982, - "loss": 1.707, - "step": 15720 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019468274369207876, - "loss": 1.5196, - "step": 15721 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019468240546573768, - "loss": 1.4974, - "step": 15722 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019468206723939662, - "loss": 1.4286, - "step": 15723 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019468172901305556, - "loss": 1.6404, - "step": 15724 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019468139078671447, - "loss": 1.7156, - "step": 15725 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019468105256037341, - "loss": 1.714, - "step": 15726 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019468071433403235, - "loss": 1.6051, - "step": 15727 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019468037610769127, - "loss": 1.4715, - "step": 15728 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946800378813502, - "loss": 1.7292, - "step": 15729 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019467969965500915, - "loss": 1.5373, - "step": 15730 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019467936142866807, - "loss": 1.5715, - "step": 15731 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194679023202327, - "loss": 1.6081, - "step": 15732 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019467868497598595, - "loss": 1.5962, - "step": 15733 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019467834674964486, - "loss": 1.582, - "step": 15734 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946780085233038, - "loss": 1.6926, - "step": 15735 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019467767029696274, - "loss": 1.619, - "step": 15736 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019467733207062166, - "loss": 1.6449, - "step": 15737 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946769938442806, - "loss": 1.6101, - "step": 15738 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019467665561793954, - "loss": 1.7911, - "step": 15739 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019467631739159845, - "loss": 1.615, - "step": 15740 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946759791652574, - "loss": 1.7848, - "step": 15741 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019467564093891634, - "loss": 1.532, - "step": 15742 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019467530271257525, - "loss": 1.7326, - "step": 15743 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946749644862342, - "loss": 1.4229, - "step": 15744 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019467462625989313, - "loss": 1.7231, - "step": 15745 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019467428803355205, - "loss": 1.6248, - "step": 15746 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194673949807211, - "loss": 1.7816, - "step": 15747 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019467361158086993, - "loss": 1.5789, - "step": 15748 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019467327335452884, - "loss": 1.6369, - "step": 15749 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019467293512818778, - "loss": 1.5107, - "step": 15750 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019467259690184673, - "loss": 1.6134, - "step": 15751 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019467225867550564, - "loss": 1.6395, - "step": 15752 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019467192044916458, - "loss": 1.5612, - "step": 15753 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019467158222282352, - "loss": 1.6121, - "step": 15754 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019467124399648246, - "loss": 1.5819, - "step": 15755 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946709057701414, - "loss": 1.6109, - "step": 15756 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019467056754380032, - "loss": 1.7657, - "step": 15757 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019467022931745926, - "loss": 1.6026, - "step": 15758 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946698910911182, - "loss": 1.4271, - "step": 15759 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019466955286477711, - "loss": 1.7499, - "step": 15760 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019466921463843606, - "loss": 1.5221, - "step": 15761 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194668876412095, - "loss": 1.5626, - "step": 15762 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946685381857539, - "loss": 1.688, - "step": 15763 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019466819995941285, - "loss": 1.5715, - "step": 15764 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946678617330718, - "loss": 1.5539, - "step": 15765 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946675235067307, - "loss": 1.507, - "step": 15766 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019466718528038965, - "loss": 1.6148, - "step": 15767 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946668470540486, - "loss": 1.6736, - "step": 15768 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946665088277075, - "loss": 1.6931, - "step": 15769 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019466617060136644, - "loss": 1.6735, - "step": 15770 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019466583237502539, - "loss": 1.5828, - "step": 15771 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946654941486843, - "loss": 1.5766, - "step": 15772 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019466515592234324, - "loss": 1.6833, - "step": 15773 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019466481769600218, - "loss": 1.7803, - "step": 15774 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019466447946966112, - "loss": 1.5854, - "step": 15775 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019466414124332004, - "loss": 1.5943, - "step": 15776 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019466380301697898, - "loss": 1.6676, - "step": 15777 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019466346479063792, - "loss": 1.6979, - "step": 15778 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019466312656429683, - "loss": 1.6668, - "step": 15779 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019466278833795577, - "loss": 1.5448, - "step": 15780 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019466245011161472, - "loss": 1.5527, - "step": 15781 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019466211188527363, - "loss": 1.474, - "step": 15782 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019466177365893257, - "loss": 1.625, - "step": 15783 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946614354325915, - "loss": 1.5102, - "step": 15784 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019466109720625043, - "loss": 1.5995, - "step": 15785 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019466075897990937, - "loss": 1.5721, - "step": 15786 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946604207535683, - "loss": 1.5511, - "step": 15787 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019466008252722722, - "loss": 1.6069, - "step": 15788 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019465974430088616, - "loss": 1.5941, - "step": 15789 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946594060745451, - "loss": 1.7487, - "step": 15790 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019465906784820402, - "loss": 1.6706, - "step": 15791 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019465872962186296, - "loss": 1.6271, - "step": 15792 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946583913955219, - "loss": 1.5463, - "step": 15793 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019465805316918081, - "loss": 1.6568, - "step": 15794 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019465771494283976, - "loss": 1.5449, - "step": 15795 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946573767164987, - "loss": 1.5351, - "step": 15796 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946570384901576, - "loss": 1.4077, - "step": 15797 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019465670026381655, - "loss": 1.6798, - "step": 15798 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946563620374755, - "loss": 1.6298, - "step": 15799 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946560238111344, - "loss": 1.7828, - "step": 15800 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019465568558479335, - "loss": 1.7323, - "step": 15801 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946553473584523, - "loss": 1.575, - "step": 15802 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946550091321112, - "loss": 1.6143, - "step": 15803 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019465467090577014, - "loss": 1.7239, - "step": 15804 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019465433267942909, - "loss": 1.6244, - "step": 15805 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194653994453088, - "loss": 1.5757, - "step": 15806 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019465365622674694, - "loss": 1.4789, - "step": 15807 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019465331800040588, - "loss": 1.5848, - "step": 15808 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946529797740648, - "loss": 1.6921, - "step": 15809 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019465264154772374, - "loss": 1.4741, - "step": 15810 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019465230332138268, - "loss": 1.6014, - "step": 15811 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019465196509504162, - "loss": 1.5197, - "step": 15812 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019465162686870056, - "loss": 1.4266, - "step": 15813 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019465128864235948, - "loss": 1.6435, - "step": 15814 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019465095041601842, - "loss": 1.5237, - "step": 15815 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019465061218967736, - "loss": 1.6029, - "step": 15816 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019465027396333627, - "loss": 1.5421, - "step": 15817 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946499357369952, - "loss": 1.5373, - "step": 15818 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019464959751065415, - "loss": 1.5939, - "step": 15819 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019464925928431307, - "loss": 1.4427, - "step": 15820 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194648921057972, - "loss": 1.5189, - "step": 15821 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019464858283163095, - "loss": 1.5574, - "step": 15822 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019464824460528986, - "loss": 1.4302, - "step": 15823 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946479063789488, - "loss": 1.5375, - "step": 15824 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019464756815260775, - "loss": 1.6572, - "step": 15825 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019464722992626666, - "loss": 1.6376, - "step": 15826 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946468916999256, - "loss": 1.5608, - "step": 15827 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019464655347358454, - "loss": 1.623, - "step": 15828 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019464621524724346, - "loss": 1.543, - "step": 15829 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946458770209024, - "loss": 1.5319, - "step": 15830 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019464553879456134, - "loss": 1.661, - "step": 15831 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019464520056822025, - "loss": 1.4965, - "step": 15832 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946448623418792, - "loss": 1.5014, - "step": 15833 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019464452411553814, - "loss": 1.755, - "step": 15834 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019464418588919705, - "loss": 1.546, - "step": 15835 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194643847662856, - "loss": 1.649, - "step": 15836 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019464350943651493, - "loss": 1.6427, - "step": 15837 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019464317121017385, - "loss": 1.6108, - "step": 15838 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946428329838328, - "loss": 1.6993, - "step": 15839 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019464249475749173, - "loss": 1.449, - "step": 15840 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019464215653115064, - "loss": 1.4979, - "step": 15841 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019464181830480958, - "loss": 1.538, - "step": 15842 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019464148007846852, - "loss": 1.4598, - "step": 15843 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019464114185212744, - "loss": 1.7532, - "step": 15844 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019464080362578638, - "loss": 1.643, - "step": 15845 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019464046539944532, - "loss": 1.6664, - "step": 15846 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019464012717310423, - "loss": 1.6258, - "step": 15847 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019463978894676318, - "loss": 1.6835, - "step": 15848 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019463945072042212, - "loss": 1.5212, - "step": 15849 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019463911249408103, - "loss": 1.6398, - "step": 15850 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019463877426773997, - "loss": 1.5787, - "step": 15851 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946384360413989, - "loss": 1.5827, - "step": 15852 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019463809781505783, - "loss": 1.5602, - "step": 15853 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019463775958871677, - "loss": 1.5961, - "step": 15854 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946374213623757, - "loss": 1.5046, - "step": 15855 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019463708313603462, - "loss": 1.6511, - "step": 15856 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019463674490969356, - "loss": 1.6017, - "step": 15857 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946364066833525, - "loss": 1.6894, - "step": 15858 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019463606845701145, - "loss": 1.5985, - "step": 15859 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019463573023067036, - "loss": 1.6026, - "step": 15860 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946353920043293, - "loss": 1.6542, - "step": 15861 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019463505377798824, - "loss": 1.6609, - "step": 15862 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019463471555164716, - "loss": 1.5605, - "step": 15863 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946343773253061, - "loss": 1.6092, - "step": 15864 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019463403909896504, - "loss": 1.5135, - "step": 15865 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019463370087262395, - "loss": 1.618, - "step": 15866 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946333626462829, - "loss": 1.5683, - "step": 15867 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019463302441994184, - "loss": 1.5194, - "step": 15868 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019463268619360078, - "loss": 1.6269, - "step": 15869 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019463234796725972, - "loss": 1.6659, - "step": 15870 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019463200974091863, - "loss": 1.5626, - "step": 15871 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019463167151457757, - "loss": 1.4672, - "step": 15872 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019463133328823651, - "loss": 1.5268, - "step": 15873 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019463099506189543, - "loss": 1.6612, - "step": 15874 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019463065683555437, - "loss": 1.5652, - "step": 15875 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946303186092133, - "loss": 1.4564, - "step": 15876 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019462998038287222, - "loss": 1.6339, - "step": 15877 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019462964215653117, - "loss": 1.5369, - "step": 15878 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946293039301901, - "loss": 1.5353, - "step": 15879 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019462896570384902, - "loss": 1.5591, - "step": 15880 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019462862747750796, - "loss": 1.5934, - "step": 15881 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946282892511669, - "loss": 1.6526, - "step": 15882 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019462795102482582, - "loss": 1.7402, - "step": 15883 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019462761279848476, - "loss": 1.4879, - "step": 15884 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946272745721437, - "loss": 1.6831, - "step": 15885 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019462693634580261, - "loss": 1.499, - "step": 15886 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019462659811946155, - "loss": 1.6283, - "step": 15887 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946262598931205, - "loss": 1.5454, - "step": 15888 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946259216667794, - "loss": 1.4911, - "step": 15889 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019462558344043835, - "loss": 1.5749, - "step": 15890 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946252452140973, - "loss": 1.7177, - "step": 15891 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946249069877562, - "loss": 1.6622, - "step": 15892 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019462456876141515, - "loss": 1.555, - "step": 15893 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946242305350741, - "loss": 1.6478, - "step": 15894 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194623892308733, - "loss": 1.5043, - "step": 15895 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019462355408239194, - "loss": 1.5547, - "step": 15896 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019462321585605088, - "loss": 1.6511, - "step": 15897 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946228776297098, - "loss": 1.6372, - "step": 15898 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019462253940336874, - "loss": 1.4164, - "step": 15899 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019462220117702768, - "loss": 1.4692, - "step": 15900 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946218629506866, - "loss": 1.4762, - "step": 15901 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019462152472434554, - "loss": 1.6728, - "step": 15902 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019462118649800448, - "loss": 1.7157, - "step": 15903 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946208482716634, - "loss": 1.7389, - "step": 15904 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019462051004532233, - "loss": 1.4806, - "step": 15905 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019462017181898127, - "loss": 1.834, - "step": 15906 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946198335926402, - "loss": 1.3495, - "step": 15907 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019461949536629913, - "loss": 1.5785, - "step": 15908 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019461915713995807, - "loss": 1.5623, - "step": 15909 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019461881891361698, - "loss": 1.527, - "step": 15910 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019461848068727593, - "loss": 1.6352, - "step": 15911 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019461814246093487, - "loss": 1.6479, - "step": 15912 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019461780423459378, - "loss": 1.725, - "step": 15913 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019461746600825272, - "loss": 1.5882, - "step": 15914 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019461712778191166, - "loss": 1.6916, - "step": 15915 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019461678955557058, - "loss": 1.6779, - "step": 15916 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019461645132922952, - "loss": 1.5606, - "step": 15917 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019461611310288846, - "loss": 1.6499, - "step": 15918 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019461577487654737, - "loss": 1.5846, - "step": 15919 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019461543665020631, - "loss": 1.7195, - "step": 15920 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019461509842386526, - "loss": 1.674, - "step": 15921 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019461476019752417, - "loss": 1.5452, - "step": 15922 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946144219711831, - "loss": 1.5366, - "step": 15923 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019461408374484208, - "loss": 1.4744, - "step": 15924 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194613745518501, - "loss": 1.6074, - "step": 15925 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019461340729215993, - "loss": 1.7325, - "step": 15926 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019461306906581888, - "loss": 1.5734, - "step": 15927 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946127308394778, - "loss": 1.704, - "step": 15928 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019461239261313673, - "loss": 1.6263, - "step": 15929 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019461205438679567, - "loss": 1.6436, - "step": 15930 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019461171616045459, - "loss": 1.7206, - "step": 15931 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019461137793411353, - "loss": 1.5627, - "step": 15932 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019461103970777247, - "loss": 1.5966, - "step": 15933 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019461070148143138, - "loss": 1.4894, - "step": 15934 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019461036325509032, - "loss": 1.5387, - "step": 15935 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019461002502874926, - "loss": 1.6802, - "step": 15936 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019460968680240818, - "loss": 1.6515, - "step": 15937 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019460934857606712, - "loss": 1.6504, - "step": 15938 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019460901034972606, - "loss": 1.4728, - "step": 15939 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019460867212338497, - "loss": 1.6185, - "step": 15940 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019460833389704392, - "loss": 1.5754, - "step": 15941 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019460799567070286, - "loss": 1.6777, - "step": 15942 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019460765744436177, - "loss": 1.6492, - "step": 15943 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946073192180207, - "loss": 1.5088, - "step": 15944 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019460698099167965, - "loss": 1.436, - "step": 15945 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019460664276533857, - "loss": 1.5089, - "step": 15946 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946063045389975, - "loss": 1.6559, - "step": 15947 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019460596631265645, - "loss": 1.5132, - "step": 15948 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019460562808631536, - "loss": 1.712, - "step": 15949 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946052898599743, - "loss": 1.4721, - "step": 15950 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019460495163363325, - "loss": 1.4593, - "step": 15951 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019460461340729216, - "loss": 1.57, - "step": 15952 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946042751809511, - "loss": 1.4584, - "step": 15953 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019460393695461004, - "loss": 1.672, - "step": 15954 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019460359872826896, - "loss": 1.536, - "step": 15955 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946032605019279, - "loss": 1.5555, - "step": 15956 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019460292227558684, - "loss": 1.6434, - "step": 15957 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019460258404924575, - "loss": 1.5295, - "step": 15958 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946022458229047, - "loss": 1.6328, - "step": 15959 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019460190759656363, - "loss": 1.742, - "step": 15960 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019460156937022255, - "loss": 1.6676, - "step": 15961 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001946012311438815, - "loss": 1.6071, - "step": 15962 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019460089291754043, - "loss": 1.6217, - "step": 15963 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019460055469119934, - "loss": 1.5218, - "step": 15964 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019460021646485829, - "loss": 1.4223, - "step": 15965 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019459987823851723, - "loss": 1.8531, - "step": 15966 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019459954001217614, - "loss": 1.6722, - "step": 15967 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019459920178583508, - "loss": 1.671, - "step": 15968 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019459886355949402, - "loss": 1.6587, - "step": 15969 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019459852533315294, - "loss": 1.6943, - "step": 15970 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019459818710681188, - "loss": 1.5966, - "step": 15971 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019459784888047082, - "loss": 1.4761, - "step": 15972 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019459751065412973, - "loss": 1.7822, - "step": 15973 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019459717242778868, - "loss": 1.6678, - "step": 15974 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019459683420144762, - "loss": 1.6161, - "step": 15975 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019459649597510653, - "loss": 1.4341, - "step": 15976 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019459615774876547, - "loss": 1.5576, - "step": 15977 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945958195224244, - "loss": 1.5481, - "step": 15978 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019459548129608333, - "loss": 1.6811, - "step": 15979 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945951430697423, - "loss": 1.6137, - "step": 15980 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019459480484340124, - "loss": 1.5214, - "step": 15981 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019459446661706015, - "loss": 1.6443, - "step": 15982 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945941283907191, - "loss": 1.5872, - "step": 15983 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019459379016437803, - "loss": 1.5977, - "step": 15984 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019459345193803695, - "loss": 1.6691, - "step": 15985 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945931137116959, - "loss": 1.7555, - "step": 15986 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019459277548535483, - "loss": 1.6497, - "step": 15987 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019459243725901374, - "loss": 1.5292, - "step": 15988 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019459209903267268, - "loss": 1.6792, - "step": 15989 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019459176080633162, - "loss": 1.775, - "step": 15990 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019459142257999054, - "loss": 1.5213, - "step": 15991 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019459108435364948, - "loss": 1.6355, - "step": 15992 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019459074612730842, - "loss": 1.7036, - "step": 15993 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019459040790096734, - "loss": 1.6026, - "step": 15994 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019459006967462628, - "loss": 1.5581, - "step": 15995 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019458973144828522, - "loss": 1.5171, - "step": 15996 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019458939322194413, - "loss": 1.7268, - "step": 15997 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019458905499560307, - "loss": 1.6063, - "step": 15998 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019458871676926201, - "loss": 1.62, - "step": 15999 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019458837854292093, - "loss": 1.6667, - "step": 16000 - }, - { - "epoch": 0.05, - "eval_loss": 1.4827223557692308, - "mrc_em": 22.0, - "mrc_f1": 28.85238095238095, - "nli_acc": 0.08, - "step": 16000, - "sts_f1": 0.6206896551724138, - "ynat_f1": 0.6216462753016904 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019458804031657987, - "loss": 1.6584, - "step": 16001 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945877020902388, - "loss": 1.4671, - "step": 16002 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019458736386389772, - "loss": 1.4486, - "step": 16003 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019458702563755667, - "loss": 1.5862, - "step": 16004 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945866874112156, - "loss": 1.4962, - "step": 16005 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019458634918487452, - "loss": 1.4342, - "step": 16006 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019458601095853346, - "loss": 1.5533, - "step": 16007 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945856727321924, - "loss": 1.519, - "step": 16008 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019458533450585132, - "loss": 1.6731, - "step": 16009 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019458499627951026, - "loss": 1.5765, - "step": 16010 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945846580531692, - "loss": 1.595, - "step": 16011 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945843198268281, - "loss": 1.4564, - "step": 16012 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019458398160048705, - "loss": 1.5567, - "step": 16013 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194583643374146, - "loss": 1.6277, - "step": 16014 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945833051478049, - "loss": 1.4784, - "step": 16015 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019458296692146385, - "loss": 1.5806, - "step": 16016 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945826286951228, - "loss": 1.6526, - "step": 16017 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945822904687817, - "loss": 1.5973, - "step": 16018 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019458195224244065, - "loss": 1.6001, - "step": 16019 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945816140160996, - "loss": 1.5257, - "step": 16020 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945812757897585, - "loss": 1.6389, - "step": 16021 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019458093756341744, - "loss": 1.7547, - "step": 16022 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019458059933707638, - "loss": 1.5299, - "step": 16023 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945802611107353, - "loss": 1.5844, - "step": 16024 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019457992288439424, - "loss": 1.6353, - "step": 16025 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019457958465805318, - "loss": 1.4859, - "step": 16026 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945792464317121, - "loss": 1.5794, - "step": 16027 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019457890820537104, - "loss": 1.7297, - "step": 16028 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019457856997902998, - "loss": 1.6186, - "step": 16029 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945782317526889, - "loss": 1.7025, - "step": 16030 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019457789352634783, - "loss": 1.6495, - "step": 16031 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019457755530000677, - "loss": 1.5825, - "step": 16032 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945772170736657, - "loss": 1.3483, - "step": 16033 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019457687884732463, - "loss": 1.5367, - "step": 16034 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019457654062098357, - "loss": 1.5191, - "step": 16035 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945762023946425, - "loss": 1.7689, - "step": 16036 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019457586416830145, - "loss": 1.6818, - "step": 16037 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019457552594196037, - "loss": 1.5442, - "step": 16038 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945751877156193, - "loss": 1.5499, - "step": 16039 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019457484948927825, - "loss": 1.6177, - "step": 16040 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945745112629372, - "loss": 1.6095, - "step": 16041 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945741730365961, - "loss": 1.6205, - "step": 16042 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019457383481025504, - "loss": 1.6421, - "step": 16043 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019457349658391399, - "loss": 1.5414, - "step": 16044 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945731583575729, - "loss": 1.668, - "step": 16045 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019457282013123184, - "loss": 1.6188, - "step": 16046 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019457248190489078, - "loss": 1.5892, - "step": 16047 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945721436785497, - "loss": 1.744, - "step": 16048 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019457180545220864, - "loss": 1.6126, - "step": 16049 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019457146722586758, - "loss": 2.0112, - "step": 16050 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945711289995265, - "loss": 1.6344, - "step": 16051 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019457079077318543, - "loss": 1.632, - "step": 16052 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019457045254684437, - "loss": 1.5222, - "step": 16053 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945701143205033, - "loss": 1.5636, - "step": 16054 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019456977609416223, - "loss": 1.6167, - "step": 16055 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019456943786782117, - "loss": 1.5491, - "step": 16056 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019456909964148008, - "loss": 1.5461, - "step": 16057 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019456876141513903, - "loss": 1.7225, - "step": 16058 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019456842318879797, - "loss": 1.5817, - "step": 16059 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019456808496245688, - "loss": 1.5058, - "step": 16060 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019456774673611582, - "loss": 1.5639, - "step": 16061 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019456740850977476, - "loss": 1.6896, - "step": 16062 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019456707028343368, - "loss": 1.6425, - "step": 16063 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019456673205709262, - "loss": 1.3804, - "step": 16064 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019456639383075156, - "loss": 1.5076, - "step": 16065 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019456605560441047, - "loss": 1.5753, - "step": 16066 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019456571737806941, - "loss": 1.5218, - "step": 16067 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019456537915172836, - "loss": 1.5562, - "step": 16068 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019456504092538727, - "loss": 1.5319, - "step": 16069 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945647026990462, - "loss": 1.6767, - "step": 16070 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019456436447270515, - "loss": 1.5516, - "step": 16071 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019456402624636407, - "loss": 1.516, - "step": 16072 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194563688020023, - "loss": 1.713, - "step": 16073 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019456334979368195, - "loss": 1.6262, - "step": 16074 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019456301156734086, - "loss": 1.7604, - "step": 16075 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945626733409998, - "loss": 1.5225, - "step": 16076 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019456233511465875, - "loss": 1.5657, - "step": 16077 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019456199688831766, - "loss": 1.5282, - "step": 16078 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945616586619766, - "loss": 1.56, - "step": 16079 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019456132043563554, - "loss": 1.5644, - "step": 16080 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019456098220929446, - "loss": 1.6374, - "step": 16081 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945606439829534, - "loss": 1.6519, - "step": 16082 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019456030575661234, - "loss": 1.6053, - "step": 16083 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019455996753027125, - "loss": 1.6022, - "step": 16084 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945596293039302, - "loss": 1.5116, - "step": 16085 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019455929107758913, - "loss": 1.4756, - "step": 16086 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019455895285124805, - "loss": 1.6704, - "step": 16087 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194558614624907, - "loss": 1.5312, - "step": 16088 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019455827639856593, - "loss": 1.6823, - "step": 16089 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019455793817222484, - "loss": 1.4998, - "step": 16090 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019455759994588379, - "loss": 1.6952, - "step": 16091 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019455726171954273, - "loss": 1.4771, - "step": 16092 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019455692349320167, - "loss": 1.5169, - "step": 16093 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945565852668606, - "loss": 1.566, - "step": 16094 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019455624704051952, - "loss": 1.7657, - "step": 16095 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019455590881417846, - "loss": 1.7858, - "step": 16096 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945555705878374, - "loss": 1.6217, - "step": 16097 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019455523236149632, - "loss": 1.5552, - "step": 16098 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019455489413515526, - "loss": 1.5925, - "step": 16099 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945545559088142, - "loss": 1.5865, - "step": 16100 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019455421768247312, - "loss": 1.645, - "step": 16101 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019455387945613206, - "loss": 1.6207, - "step": 16102 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194553541229791, - "loss": 1.5257, - "step": 16103 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945532030034499, - "loss": 1.7799, - "step": 16104 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019455286477710885, - "loss": 1.5073, - "step": 16105 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945525265507678, - "loss": 1.6218, - "step": 16106 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945521883244267, - "loss": 1.6693, - "step": 16107 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019455185009808565, - "loss": 1.5625, - "step": 16108 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945515118717446, - "loss": 1.5293, - "step": 16109 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945511736454035, - "loss": 1.4076, - "step": 16110 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019455083541906245, - "loss": 1.627, - "step": 16111 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945504971927214, - "loss": 1.5573, - "step": 16112 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945501589663803, - "loss": 1.6184, - "step": 16113 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019454982074003924, - "loss": 1.5952, - "step": 16114 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019454948251369818, - "loss": 1.5319, - "step": 16115 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945491442873571, - "loss": 1.7207, - "step": 16116 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019454880606101604, - "loss": 1.7316, - "step": 16117 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019454846783467498, - "loss": 1.6292, - "step": 16118 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945481296083339, - "loss": 1.6174, - "step": 16119 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019454779138199283, - "loss": 1.7303, - "step": 16120 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019454745315565178, - "loss": 1.554, - "step": 16121 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945471149293107, - "loss": 1.3701, - "step": 16122 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019454677670296963, - "loss": 1.7285, - "step": 16123 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019454643847662857, - "loss": 1.5638, - "step": 16124 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945461002502875, - "loss": 1.5977, - "step": 16125 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019454576202394643, - "loss": 1.4987, - "step": 16126 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019454542379760537, - "loss": 1.5181, - "step": 16127 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945450855712643, - "loss": 1.4014, - "step": 16128 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019454474734492322, - "loss": 1.7027, - "step": 16129 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019454440911858216, - "loss": 1.4857, - "step": 16130 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945440708922411, - "loss": 1.5626, - "step": 16131 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019454373266590002, - "loss": 1.6213, - "step": 16132 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019454339443955896, - "loss": 1.6849, - "step": 16133 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945430562132179, - "loss": 1.5863, - "step": 16134 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019454271798687682, - "loss": 1.5059, - "step": 16135 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019454237976053576, - "loss": 1.7675, - "step": 16136 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945420415341947, - "loss": 1.5908, - "step": 16137 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945417033078536, - "loss": 1.7447, - "step": 16138 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019454136508151255, - "loss": 1.5962, - "step": 16139 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945410268551715, - "loss": 1.6217, - "step": 16140 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945406886288304, - "loss": 1.4949, - "step": 16141 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019454035040248935, - "loss": 1.4832, - "step": 16142 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945400121761483, - "loss": 1.7527, - "step": 16143 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945396739498072, - "loss": 1.7233, - "step": 16144 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019453933572346615, - "loss": 1.7013, - "step": 16145 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945389974971251, - "loss": 1.4699, - "step": 16146 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194538659270784, - "loss": 1.6004, - "step": 16147 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019453832104444297, - "loss": 1.5514, - "step": 16148 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019453798281810188, - "loss": 1.5127, - "step": 16149 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019453764459176082, - "loss": 1.6331, - "step": 16150 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019453730636541977, - "loss": 1.5861, - "step": 16151 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019453696813907868, - "loss": 1.6113, - "step": 16152 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019453662991273762, - "loss": 1.6532, - "step": 16153 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019453629168639656, - "loss": 1.5994, - "step": 16154 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019453595346005548, - "loss": 1.5388, - "step": 16155 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019453561523371442, - "loss": 1.6227, - "step": 16156 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019453527700737336, - "loss": 1.7755, - "step": 16157 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019453493878103227, - "loss": 1.4772, - "step": 16158 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019453460055469121, - "loss": 1.4952, - "step": 16159 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019453426232835015, - "loss": 1.6687, - "step": 16160 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019453392410200907, - "loss": 1.7703, - "step": 16161 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194533585875668, - "loss": 1.5378, - "step": 16162 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019453324764932695, - "loss": 1.6519, - "step": 16163 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019453290942298587, - "loss": 1.4941, - "step": 16164 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945325711966448, - "loss": 1.6024, - "step": 16165 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019453223297030375, - "loss": 1.551, - "step": 16166 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019453189474396266, - "loss": 1.7179, - "step": 16167 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945315565176216, - "loss": 1.5239, - "step": 16168 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019453121829128054, - "loss": 1.6229, - "step": 16169 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019453088006493946, - "loss": 1.6132, - "step": 16170 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945305418385984, - "loss": 1.481, - "step": 16171 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019453020361225734, - "loss": 1.5469, - "step": 16172 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019452986538591625, - "loss": 1.669, - "step": 16173 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945295271595752, - "loss": 1.4888, - "step": 16174 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019452918893323414, - "loss": 1.465, - "step": 16175 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019452885070689305, - "loss": 1.6241, - "step": 16176 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194528512480552, - "loss": 1.4988, - "step": 16177 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019452817425421093, - "loss": 1.5371, - "step": 16178 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019452783602786985, - "loss": 1.5683, - "step": 16179 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945274978015288, - "loss": 1.6278, - "step": 16180 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019452715957518773, - "loss": 1.5992, - "step": 16181 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019452682134884664, - "loss": 1.6322, - "step": 16182 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019452648312250558, - "loss": 1.7451, - "step": 16183 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019452614489616453, - "loss": 1.5687, - "step": 16184 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019452580666982344, - "loss": 1.5441, - "step": 16185 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019452546844348238, - "loss": 1.6383, - "step": 16186 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019452513021714132, - "loss": 1.5837, - "step": 16187 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019452479199080024, - "loss": 1.5091, - "step": 16188 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019452445376445918, - "loss": 1.7122, - "step": 16189 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019452411553811812, - "loss": 1.5067, - "step": 16190 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019452377731177703, - "loss": 1.5549, - "step": 16191 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019452343908543597, - "loss": 1.519, - "step": 16192 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019452310085909491, - "loss": 1.5723, - "step": 16193 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019452276263275383, - "loss": 1.6751, - "step": 16194 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019452242440641277, - "loss": 1.7042, - "step": 16195 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945220861800717, - "loss": 1.4567, - "step": 16196 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019452174795373062, - "loss": 1.4892, - "step": 16197 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019452140972738957, - "loss": 1.6093, - "step": 16198 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945210715010485, - "loss": 1.7695, - "step": 16199 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019452073327470742, - "loss": 1.6794, - "step": 16200 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019452039504836636, - "loss": 1.6711, - "step": 16201 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945200568220253, - "loss": 1.6267, - "step": 16202 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019451971859568422, - "loss": 1.4723, - "step": 16203 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019451938036934316, - "loss": 1.6408, - "step": 16204 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019451904214300213, - "loss": 1.7133, - "step": 16205 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019451870391666104, - "loss": 1.7133, - "step": 16206 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019451836569031998, - "loss": 1.5028, - "step": 16207 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019451802746397892, - "loss": 1.5812, - "step": 16208 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019451768923763784, - "loss": 1.5986, - "step": 16209 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019451735101129678, - "loss": 1.5318, - "step": 16210 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019451701278495572, - "loss": 1.5539, - "step": 16211 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019451667455861463, - "loss": 1.5722, - "step": 16212 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019451633633227357, - "loss": 1.7377, - "step": 16213 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019451599810593252, - "loss": 1.6105, - "step": 16214 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019451565987959143, - "loss": 1.5955, - "step": 16215 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019451532165325037, - "loss": 1.5931, - "step": 16216 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945149834269093, - "loss": 1.4843, - "step": 16217 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019451464520056823, - "loss": 1.594, - "step": 16218 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019451430697422717, - "loss": 1.5513, - "step": 16219 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945139687478861, - "loss": 1.5662, - "step": 16220 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019451363052154502, - "loss": 1.6915, - "step": 16221 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019451329229520396, - "loss": 1.6423, - "step": 16222 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945129540688629, - "loss": 1.4961, - "step": 16223 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019451261584252182, - "loss": 1.6348, - "step": 16224 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019451227761618076, - "loss": 1.5813, - "step": 16225 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945119393898397, - "loss": 1.5215, - "step": 16226 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019451160116349861, - "loss": 1.714, - "step": 16227 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019451126293715756, - "loss": 1.51, - "step": 16228 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945109247108165, - "loss": 1.636, - "step": 16229 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945105864844754, - "loss": 1.3649, - "step": 16230 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019451024825813435, - "loss": 1.3609, - "step": 16231 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945099100317933, - "loss": 1.5467, - "step": 16232 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945095718054522, - "loss": 1.5115, - "step": 16233 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019450923357911115, - "loss": 1.5168, - "step": 16234 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945088953527701, - "loss": 1.7039, - "step": 16235 - }, - { - "epoch": 0.05, - "learning_rate": 0.000194508557126429, - "loss": 1.7158, - "step": 16236 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019450821890008795, - "loss": 1.6249, - "step": 16237 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019450788067374689, - "loss": 1.4874, - "step": 16238 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945075424474058, - "loss": 1.7987, - "step": 16239 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019450720422106474, - "loss": 1.7357, - "step": 16240 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019450686599472368, - "loss": 1.6308, - "step": 16241 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945065277683826, - "loss": 1.4227, - "step": 16242 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019450618954204154, - "loss": 1.5195, - "step": 16243 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019450585131570048, - "loss": 1.4547, - "step": 16244 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945055130893594, - "loss": 1.6103, - "step": 16245 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019450517486301833, - "loss": 1.7627, - "step": 16246 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019450483663667728, - "loss": 1.8183, - "step": 16247 - }, - { - "epoch": 0.05, - "learning_rate": 0.0001945044984103362, - "loss": 1.6666, - "step": 16248 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019450416018399513, - "loss": 1.6091, - "step": 16249 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019450382195765407, - "loss": 1.6736, - "step": 16250 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019450348373131299, - "loss": 1.5889, - "step": 16251 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019450314550497193, - "loss": 1.59, - "step": 16252 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019450280727863087, - "loss": 1.7104, - "step": 16253 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019450246905228978, - "loss": 1.6001, - "step": 16254 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019450213082594872, - "loss": 1.3788, - "step": 16255 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019450179259960766, - "loss": 1.7116, - "step": 16256 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019450145437326658, - "loss": 1.6933, - "step": 16257 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019450111614692552, - "loss": 1.5606, - "step": 16258 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019450077792058446, - "loss": 1.6481, - "step": 16259 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019450043969424337, - "loss": 1.5547, - "step": 16260 - }, - { - "epoch": 0.05, - "learning_rate": 0.00019450010146790234, - "loss": 1.6346, - "step": 16261 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019449976324156128, - "loss": 1.6243, - "step": 16262 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944994250152202, - "loss": 1.8542, - "step": 16263 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019449908678887914, - "loss": 1.592, - "step": 16264 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019449874856253808, - "loss": 1.625, - "step": 16265 - }, - { - "epoch": 0.06, - "learning_rate": 0.000194498410336197, - "loss": 1.5857, - "step": 16266 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019449807210985594, - "loss": 1.5449, - "step": 16267 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019449773388351488, - "loss": 1.5712, - "step": 16268 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944973956571738, - "loss": 1.5936, - "step": 16269 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019449705743083273, - "loss": 1.7013, - "step": 16270 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019449671920449167, - "loss": 1.5871, - "step": 16271 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944963809781506, - "loss": 1.5313, - "step": 16272 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019449604275180953, - "loss": 1.7253, - "step": 16273 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019449570452546847, - "loss": 1.5547, - "step": 16274 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019449536629912738, - "loss": 1.6457, - "step": 16275 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019449502807278632, - "loss": 1.6695, - "step": 16276 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019449468984644527, - "loss": 1.4958, - "step": 16277 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019449435162010418, - "loss": 1.6809, - "step": 16278 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019449401339376312, - "loss": 1.6709, - "step": 16279 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019449367516742206, - "loss": 1.6498, - "step": 16280 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019449333694108098, - "loss": 1.6924, - "step": 16281 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019449299871473992, - "loss": 1.6831, - "step": 16282 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019449266048839886, - "loss": 1.7133, - "step": 16283 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019449232226205777, - "loss": 1.573, - "step": 16284 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944919840357167, - "loss": 1.4551, - "step": 16285 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019449164580937565, - "loss": 1.6737, - "step": 16286 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019449130758303457, - "loss": 1.7781, - "step": 16287 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944909693566935, - "loss": 1.5897, - "step": 16288 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019449063113035245, - "loss": 1.697, - "step": 16289 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019449029290401136, - "loss": 1.5953, - "step": 16290 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944899546776703, - "loss": 1.6861, - "step": 16291 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019448961645132925, - "loss": 1.6928, - "step": 16292 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019448927822498816, - "loss": 1.5228, - "step": 16293 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944889399986471, - "loss": 1.7786, - "step": 16294 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019448860177230604, - "loss": 1.6411, - "step": 16295 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019448826354596496, - "loss": 1.6328, - "step": 16296 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944879253196239, - "loss": 1.5072, - "step": 16297 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019448758709328284, - "loss": 1.6275, - "step": 16298 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019448724886694175, - "loss": 1.637, - "step": 16299 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944869106406007, - "loss": 1.8086, - "step": 16300 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019448657241425964, - "loss": 1.5549, - "step": 16301 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019448623418791855, - "loss": 1.6482, - "step": 16302 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944858959615775, - "loss": 1.6145, - "step": 16303 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019448555773523643, - "loss": 1.4215, - "step": 16304 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019448521950889535, - "loss": 1.4241, - "step": 16305 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944848812825543, - "loss": 1.4608, - "step": 16306 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019448454305621323, - "loss": 1.6561, - "step": 16307 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019448420482987214, - "loss": 1.5358, - "step": 16308 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019448386660353108, - "loss": 1.4836, - "step": 16309 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019448352837719002, - "loss": 1.4193, - "step": 16310 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019448319015084894, - "loss": 1.6012, - "step": 16311 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019448285192450788, - "loss": 1.6032, - "step": 16312 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019448251369816682, - "loss": 1.5702, - "step": 16313 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019448217547182574, - "loss": 1.6602, - "step": 16314 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019448183724548468, - "loss": 1.6616, - "step": 16315 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019448149901914362, - "loss": 1.6371, - "step": 16316 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019448116079280256, - "loss": 1.6474, - "step": 16317 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944808225664615, - "loss": 1.6882, - "step": 16318 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019448048434012044, - "loss": 1.7084, - "step": 16319 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019448014611377935, - "loss": 1.7831, - "step": 16320 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944798078874383, - "loss": 1.6709, - "step": 16321 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019447946966109724, - "loss": 1.5693, - "step": 16322 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019447913143475615, - "loss": 1.7659, - "step": 16323 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944787932084151, - "loss": 1.754, - "step": 16324 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019447845498207403, - "loss": 1.6454, - "step": 16325 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019447811675573295, - "loss": 1.6111, - "step": 16326 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944777785293919, - "loss": 1.693, - "step": 16327 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019447744030305083, - "loss": 1.6462, - "step": 16328 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019447710207670974, - "loss": 1.5023, - "step": 16329 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019447676385036868, - "loss": 1.6553, - "step": 16330 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019447642562402763, - "loss": 1.6674, - "step": 16331 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019447608739768654, - "loss": 1.5965, - "step": 16332 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019447574917134548, - "loss": 1.5351, - "step": 16333 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019447541094500442, - "loss": 1.7511, - "step": 16334 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019447507271866334, - "loss": 1.9015, - "step": 16335 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019447473449232228, - "loss": 1.5838, - "step": 16336 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019447439626598122, - "loss": 1.5047, - "step": 16337 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019447405803964013, - "loss": 1.6359, - "step": 16338 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019447371981329907, - "loss": 1.6261, - "step": 16339 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019447338158695802, - "loss": 1.6834, - "step": 16340 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019447304336061693, - "loss": 1.6049, - "step": 16341 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019447270513427587, - "loss": 1.6349, - "step": 16342 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944723669079348, - "loss": 1.6308, - "step": 16343 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019447202868159373, - "loss": 1.6724, - "step": 16344 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019447169045525267, - "loss": 1.7383, - "step": 16345 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944713522289116, - "loss": 1.5272, - "step": 16346 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019447101400257052, - "loss": 1.5972, - "step": 16347 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019447067577622946, - "loss": 1.556, - "step": 16348 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944703375498884, - "loss": 1.7804, - "step": 16349 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019446999932354732, - "loss": 1.6508, - "step": 16350 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019446966109720626, - "loss": 1.5743, - "step": 16351 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944693228708652, - "loss": 1.5816, - "step": 16352 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019446898464452411, - "loss": 1.6934, - "step": 16353 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019446864641818306, - "loss": 1.6659, - "step": 16354 - }, - { - "epoch": 0.06, - "learning_rate": 0.000194468308191842, - "loss": 1.6815, - "step": 16355 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944679699655009, - "loss": 1.4886, - "step": 16356 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019446763173915985, - "loss": 1.5495, - "step": 16357 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944672935128188, - "loss": 1.7427, - "step": 16358 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944669552864777, - "loss": 1.5202, - "step": 16359 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019446661706013665, - "loss": 1.6398, - "step": 16360 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944662788337956, - "loss": 1.607, - "step": 16361 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944659406074545, - "loss": 1.4831, - "step": 16362 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019446560238111344, - "loss": 1.7162, - "step": 16363 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019446526415477239, - "loss": 1.7708, - "step": 16364 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944649259284313, - "loss": 1.808, - "step": 16365 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019446458770209024, - "loss": 1.4722, - "step": 16366 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019446424947574918, - "loss": 1.6201, - "step": 16367 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944639112494081, - "loss": 1.5933, - "step": 16368 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019446357302306704, - "loss": 1.5792, - "step": 16369 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019446323479672598, - "loss": 1.6438, - "step": 16370 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944628965703849, - "loss": 1.7171, - "step": 16371 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019446255834404383, - "loss": 1.5077, - "step": 16372 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019446222011770277, - "loss": 1.5591, - "step": 16373 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019446188189136172, - "loss": 1.5762, - "step": 16374 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019446154366502066, - "loss": 1.4454, - "step": 16375 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019446120543867957, - "loss": 1.5116, - "step": 16376 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944608672123385, - "loss": 1.5476, - "step": 16377 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019446052898599745, - "loss": 1.8895, - "step": 16378 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019446019075965637, - "loss": 1.5897, - "step": 16379 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944598525333153, - "loss": 1.572, - "step": 16380 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019445951430697425, - "loss": 1.6616, - "step": 16381 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019445917608063316, - "loss": 1.4561, - "step": 16382 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944588378542921, - "loss": 1.6614, - "step": 16383 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019445849962795105, - "loss": 1.6835, - "step": 16384 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019445816140160996, - "loss": 1.6287, - "step": 16385 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944578231752689, - "loss": 1.4476, - "step": 16386 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019445748494892784, - "loss": 1.5308, - "step": 16387 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019445714672258676, - "loss": 1.5291, - "step": 16388 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944568084962457, - "loss": 1.3201, - "step": 16389 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019445647026990464, - "loss": 1.5291, - "step": 16390 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019445613204356355, - "loss": 1.7396, - "step": 16391 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944557938172225, - "loss": 1.5489, - "step": 16392 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019445545559088143, - "loss": 1.5972, - "step": 16393 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019445511736454038, - "loss": 1.7291, - "step": 16394 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944547791381993, - "loss": 1.8048, - "step": 16395 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019445444091185823, - "loss": 1.549, - "step": 16396 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019445410268551717, - "loss": 1.5711, - "step": 16397 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019445376445917609, - "loss": 1.7555, - "step": 16398 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019445342623283503, - "loss": 1.7061, - "step": 16399 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019445308800649397, - "loss": 1.632, - "step": 16400 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019445274978015288, - "loss": 1.6035, - "step": 16401 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019445241155381182, - "loss": 1.6656, - "step": 16402 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019445207332747076, - "loss": 1.6219, - "step": 16403 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019445173510112968, - "loss": 1.6297, - "step": 16404 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019445139687478862, - "loss": 1.6428, - "step": 16405 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019445105864844756, - "loss": 1.4336, - "step": 16406 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019445072042210648, - "loss": 1.6883, - "step": 16407 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019445038219576542, - "loss": 1.651, - "step": 16408 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019445004396942436, - "loss": 1.6297, - "step": 16409 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019444970574308327, - "loss": 1.6143, - "step": 16410 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944493675167422, - "loss": 1.5268, - "step": 16411 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019444902929040115, - "loss": 1.6549, - "step": 16412 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019444869106406007, - "loss": 1.5146, - "step": 16413 - }, - { - "epoch": 0.06, - "learning_rate": 0.000194448352837719, - "loss": 1.7672, - "step": 16414 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019444801461137795, - "loss": 1.585, - "step": 16415 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019444767638503686, - "loss": 1.6365, - "step": 16416 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944473381586958, - "loss": 1.6224, - "step": 16417 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019444699993235475, - "loss": 1.7519, - "step": 16418 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019444666170601366, - "loss": 1.4876, - "step": 16419 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944463234796726, - "loss": 1.7857, - "step": 16420 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019444598525333154, - "loss": 1.5025, - "step": 16421 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019444564702699046, - "loss": 1.5739, - "step": 16422 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944453088006494, - "loss": 1.4741, - "step": 16423 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019444497057430834, - "loss": 1.5446, - "step": 16424 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019444463234796725, - "loss": 1.4716, - "step": 16425 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944442941216262, - "loss": 1.731, - "step": 16426 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019444395589528514, - "loss": 1.52, - "step": 16427 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019444361766894405, - "loss": 1.5627, - "step": 16428 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019444327944260302, - "loss": 1.3896, - "step": 16429 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019444294121626193, - "loss": 1.7323, - "step": 16430 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019444260298992087, - "loss": 1.6176, - "step": 16431 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019444226476357981, - "loss": 1.613, - "step": 16432 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019444192653723873, - "loss": 1.6216, - "step": 16433 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019444158831089767, - "loss": 1.659, - "step": 16434 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944412500845566, - "loss": 1.6079, - "step": 16435 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019444091185821552, - "loss": 1.779, - "step": 16436 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019444057363187447, - "loss": 1.6268, - "step": 16437 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944402354055334, - "loss": 1.6188, - "step": 16438 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019443989717919232, - "loss": 1.5776, - "step": 16439 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019443955895285126, - "loss": 1.6697, - "step": 16440 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944392207265102, - "loss": 1.4924, - "step": 16441 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019443888250016912, - "loss": 1.4695, - "step": 16442 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019443854427382806, - "loss": 1.7093, - "step": 16443 - }, - { - "epoch": 0.06, - "learning_rate": 0.000194438206047487, - "loss": 1.6945, - "step": 16444 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944378678211459, - "loss": 1.6502, - "step": 16445 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019443752959480485, - "loss": 1.6647, - "step": 16446 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944371913684638, - "loss": 1.4097, - "step": 16447 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944368531421227, - "loss": 1.5503, - "step": 16448 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019443651491578165, - "loss": 1.5748, - "step": 16449 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944361766894406, - "loss": 1.6747, - "step": 16450 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944358384630995, - "loss": 1.5501, - "step": 16451 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019443550023675845, - "loss": 1.6208, - "step": 16452 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944351620104174, - "loss": 1.589, - "step": 16453 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944348237840763, - "loss": 1.6018, - "step": 16454 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019443448555773524, - "loss": 1.7451, - "step": 16455 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019443414733139418, - "loss": 1.4885, - "step": 16456 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944338091050531, - "loss": 1.5473, - "step": 16457 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019443347087871204, - "loss": 1.5037, - "step": 16458 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019443313265237098, - "loss": 1.5449, - "step": 16459 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944327944260299, - "loss": 1.7003, - "step": 16460 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019443245619968884, - "loss": 1.5587, - "step": 16461 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019443211797334778, - "loss": 1.5622, - "step": 16462 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944317797470067, - "loss": 1.5947, - "step": 16463 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019443144152066563, - "loss": 1.5924, - "step": 16464 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019443110329432457, - "loss": 1.6028, - "step": 16465 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944307650679835, - "loss": 1.6586, - "step": 16466 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019443042684164243, - "loss": 1.6034, - "step": 16467 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019443008861530137, - "loss": 1.5956, - "step": 16468 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019442975038896028, - "loss": 1.5916, - "step": 16469 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019442941216261922, - "loss": 1.5695, - "step": 16470 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019442907393627817, - "loss": 1.4747, - "step": 16471 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019442873570993708, - "loss": 1.6, - "step": 16472 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019442839748359602, - "loss": 1.745, - "step": 16473 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019442805925725496, - "loss": 1.7257, - "step": 16474 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019442772103091388, - "loss": 1.604, - "step": 16475 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019442738280457282, - "loss": 1.5391, - "step": 16476 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019442704457823176, - "loss": 1.4379, - "step": 16477 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944267063518907, - "loss": 1.641, - "step": 16478 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019442636812554961, - "loss": 1.6155, - "step": 16479 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019442602989920855, - "loss": 1.5979, - "step": 16480 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944256916728675, - "loss": 1.6528, - "step": 16481 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944253534465264, - "loss": 1.5602, - "step": 16482 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019442501522018535, - "loss": 1.5863, - "step": 16483 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944246769938443, - "loss": 1.559, - "step": 16484 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019442433876750323, - "loss": 1.516, - "step": 16485 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019442400054116217, - "loss": 1.4922, - "step": 16486 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944236623148211, - "loss": 1.6882, - "step": 16487 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019442332408848003, - "loss": 1.6439, - "step": 16488 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019442298586213897, - "loss": 1.8284, - "step": 16489 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019442264763579788, - "loss": 1.479, - "step": 16490 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019442230940945683, - "loss": 1.5779, - "step": 16491 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019442197118311577, - "loss": 1.6774, - "step": 16492 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019442163295677468, - "loss": 1.6354, - "step": 16493 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019442129473043362, - "loss": 1.6698, - "step": 16494 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019442095650409256, - "loss": 1.6176, - "step": 16495 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019442061827775148, - "loss": 1.513, - "step": 16496 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019442028005141042, - "loss": 1.5361, - "step": 16497 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019441994182506936, - "loss": 1.5717, - "step": 16498 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019441960359872827, - "loss": 1.6006, - "step": 16499 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019441926537238721, - "loss": 1.6527, - "step": 16500 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019441892714604616, - "loss": 1.5485, - "step": 16501 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019441858891970507, - "loss": 1.5781, - "step": 16502 - }, - { - "epoch": 0.06, - "learning_rate": 0.000194418250693364, - "loss": 1.448, - "step": 16503 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019441791246702295, - "loss": 1.6918, - "step": 16504 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019441757424068187, - "loss": 1.4797, - "step": 16505 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944172360143408, - "loss": 1.5606, - "step": 16506 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019441689778799975, - "loss": 1.4297, - "step": 16507 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019441655956165866, - "loss": 1.6312, - "step": 16508 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944162213353176, - "loss": 1.6086, - "step": 16509 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019441588310897655, - "loss": 1.5521, - "step": 16510 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019441554488263546, - "loss": 1.5782, - "step": 16511 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944152066562944, - "loss": 1.7165, - "step": 16512 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019441486842995334, - "loss": 1.4654, - "step": 16513 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019441453020361226, - "loss": 1.6525, - "step": 16514 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944141919772712, - "loss": 1.6545, - "step": 16515 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019441385375093014, - "loss": 1.6418, - "step": 16516 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019441351552458905, - "loss": 1.5421, - "step": 16517 - }, - { - "epoch": 0.06, - "learning_rate": 0.000194413177298248, - "loss": 1.6489, - "step": 16518 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019441283907190693, - "loss": 1.4914, - "step": 16519 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019441250084556585, - "loss": 1.7151, - "step": 16520 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944121626192248, - "loss": 1.5196, - "step": 16521 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019441182439288373, - "loss": 1.5851, - "step": 16522 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019441148616654264, - "loss": 1.5769, - "step": 16523 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019441114794020159, - "loss": 1.5322, - "step": 16524 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019441080971386053, - "loss": 1.472, - "step": 16525 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019441047148751944, - "loss": 1.6178, - "step": 16526 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019441013326117838, - "loss": 1.7964, - "step": 16527 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019440979503483732, - "loss": 1.4413, - "step": 16528 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019440945680849624, - "loss": 1.7107, - "step": 16529 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019440911858215518, - "loss": 1.5262, - "step": 16530 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019440878035581412, - "loss": 1.5143, - "step": 16531 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019440844212947303, - "loss": 1.5598, - "step": 16532 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019440810390313197, - "loss": 1.6521, - "step": 16533 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019440776567679092, - "loss": 1.6146, - "step": 16534 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019440742745044983, - "loss": 1.6519, - "step": 16535 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019440708922410877, - "loss": 1.643, - "step": 16536 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944067509977677, - "loss": 1.5887, - "step": 16537 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019440641277142663, - "loss": 1.6401, - "step": 16538 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019440607454508557, - "loss": 1.5686, - "step": 16539 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944057363187445, - "loss": 1.5731, - "step": 16540 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019440539809240342, - "loss": 1.6864, - "step": 16541 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944050598660624, - "loss": 1.385, - "step": 16542 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019440472163972133, - "loss": 1.411, - "step": 16543 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019440438341338025, - "loss": 1.439, - "step": 16544 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001944040451870392, - "loss": 1.6544, - "step": 16545 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019440370696069813, - "loss": 1.6796, - "step": 16546 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019440336873435704, - "loss": 1.5874, - "step": 16547 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019440303050801598, - "loss": 1.6253, - "step": 16548 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019440269228167492, - "loss": 1.4465, - "step": 16549 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019440235405533384, - "loss": 1.7408, - "step": 16550 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019440201582899278, - "loss": 1.6925, - "step": 16551 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019440167760265172, - "loss": 1.5125, - "step": 16552 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019440133937631063, - "loss": 1.5413, - "step": 16553 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019440100114996958, - "loss": 1.5785, - "step": 16554 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019440066292362852, - "loss": 1.5888, - "step": 16555 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019440032469728743, - "loss": 1.6813, - "step": 16556 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019439998647094637, - "loss": 1.5652, - "step": 16557 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943996482446053, - "loss": 1.5589, - "step": 16558 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019439931001826423, - "loss": 1.682, - "step": 16559 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019439897179192317, - "loss": 1.5957, - "step": 16560 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943986335655821, - "loss": 1.7402, - "step": 16561 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019439829533924102, - "loss": 1.6726, - "step": 16562 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019439795711289996, - "loss": 1.587, - "step": 16563 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943976188865589, - "loss": 1.6849, - "step": 16564 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019439728066021782, - "loss": 1.6431, - "step": 16565 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019439694243387676, - "loss": 1.6749, - "step": 16566 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943966042075357, - "loss": 1.5604, - "step": 16567 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019439626598119462, - "loss": 1.5434, - "step": 16568 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019439592775485356, - "loss": 1.6653, - "step": 16569 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943955895285125, - "loss": 1.6047, - "step": 16570 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943952513021714, - "loss": 1.3822, - "step": 16571 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019439491307583035, - "loss": 1.6352, - "step": 16572 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943945748494893, - "loss": 1.6397, - "step": 16573 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943942366231482, - "loss": 1.411, - "step": 16574 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019439389839680715, - "loss": 1.7624, - "step": 16575 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943935601704661, - "loss": 1.5136, - "step": 16576 - }, - { - "epoch": 0.06, - "learning_rate": 0.000194393221944125, - "loss": 1.5041, - "step": 16577 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019439288371778395, - "loss": 1.6869, - "step": 16578 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943925454914429, - "loss": 1.655, - "step": 16579 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943922072651018, - "loss": 1.6295, - "step": 16580 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019439186903876074, - "loss": 1.567, - "step": 16581 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019439153081241968, - "loss": 1.5351, - "step": 16582 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943911925860786, - "loss": 1.6622, - "step": 16583 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019439085435973754, - "loss": 1.3953, - "step": 16584 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019439051613339648, - "loss": 1.6276, - "step": 16585 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943901779070554, - "loss": 1.5795, - "step": 16586 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019438983968071434, - "loss": 1.555, - "step": 16587 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019438950145437328, - "loss": 1.5385, - "step": 16588 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943891632280322, - "loss": 1.5587, - "step": 16589 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019438882500169113, - "loss": 1.5557, - "step": 16590 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019438848677535007, - "loss": 1.6178, - "step": 16591 - }, - { - "epoch": 0.06, - "learning_rate": 0.000194388148549009, - "loss": 1.5465, - "step": 16592 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019438781032266793, - "loss": 1.5598, - "step": 16593 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019438747209632687, - "loss": 1.5129, - "step": 16594 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019438713386998578, - "loss": 1.6057, - "step": 16595 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019438679564364472, - "loss": 1.623, - "step": 16596 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019438645741730367, - "loss": 1.6742, - "step": 16597 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943861191909626, - "loss": 1.6527, - "step": 16598 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019438578096462155, - "loss": 1.5569, - "step": 16599 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943854427382805, - "loss": 1.8057, - "step": 16600 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943851045119394, - "loss": 1.5581, - "step": 16601 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019438476628559834, - "loss": 1.4879, - "step": 16602 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019438442805925729, - "loss": 1.7023, - "step": 16603 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943840898329162, - "loss": 1.581, - "step": 16604 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019438375160657514, - "loss": 1.6323, - "step": 16605 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019438341338023408, - "loss": 1.597, - "step": 16606 - }, - { - "epoch": 0.06, - "learning_rate": 0.000194383075153893, - "loss": 1.6244, - "step": 16607 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019438273692755194, - "loss": 1.5212, - "step": 16608 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019438239870121088, - "loss": 1.5787, - "step": 16609 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943820604748698, - "loss": 1.5819, - "step": 16610 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019438172224852873, - "loss": 1.6588, - "step": 16611 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019438138402218767, - "loss": 1.7212, - "step": 16612 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943810457958466, - "loss": 1.6081, - "step": 16613 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019438070756950553, - "loss": 1.5741, - "step": 16614 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019438036934316447, - "loss": 1.6216, - "step": 16615 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019438003111682338, - "loss": 1.6522, - "step": 16616 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019437969289048233, - "loss": 1.7536, - "step": 16617 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019437935466414127, - "loss": 1.5613, - "step": 16618 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019437901643780018, - "loss": 1.6926, - "step": 16619 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019437867821145912, - "loss": 1.6509, - "step": 16620 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019437833998511806, - "loss": 1.6482, - "step": 16621 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019437800175877698, - "loss": 1.6727, - "step": 16622 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019437766353243592, - "loss": 1.5523, - "step": 16623 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019437732530609486, - "loss": 1.6094, - "step": 16624 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019437698707975377, - "loss": 1.6095, - "step": 16625 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019437664885341271, - "loss": 1.5814, - "step": 16626 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019437631062707166, - "loss": 1.6799, - "step": 16627 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019437597240073057, - "loss": 1.4094, - "step": 16628 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943756341743895, - "loss": 1.7507, - "step": 16629 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019437529594804845, - "loss": 1.5707, - "step": 16630 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019437495772170737, - "loss": 1.5954, - "step": 16631 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943746194953663, - "loss": 1.5169, - "step": 16632 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019437428126902525, - "loss": 1.7325, - "step": 16633 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019437394304268416, - "loss": 1.633, - "step": 16634 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943736048163431, - "loss": 1.4729, - "step": 16635 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019437326659000204, - "loss": 1.4775, - "step": 16636 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019437292836366096, - "loss": 1.5931, - "step": 16637 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943725901373199, - "loss": 1.7165, - "step": 16638 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019437225191097884, - "loss": 1.611, - "step": 16639 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019437191368463775, - "loss": 1.688, - "step": 16640 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943715754582967, - "loss": 1.4525, - "step": 16641 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019437123723195564, - "loss": 1.6287, - "step": 16642 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019437089900561455, - "loss": 1.5814, - "step": 16643 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943705607792735, - "loss": 1.5579, - "step": 16644 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019437022255293243, - "loss": 1.5629, - "step": 16645 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019436988432659135, - "loss": 1.7257, - "step": 16646 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943695461002503, - "loss": 1.4868, - "step": 16647 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019436920787390923, - "loss": 1.5384, - "step": 16648 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019436886964756814, - "loss": 1.7942, - "step": 16649 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019436853142122708, - "loss": 1.6548, - "step": 16650 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019436819319488603, - "loss": 1.5759, - "step": 16651 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019436785496854494, - "loss": 1.6317, - "step": 16652 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019436751674220388, - "loss": 1.5219, - "step": 16653 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019436717851586282, - "loss": 1.5968, - "step": 16654 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019436684028952176, - "loss": 1.5487, - "step": 16655 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943665020631807, - "loss": 1.5646, - "step": 16656 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019436616383683962, - "loss": 1.6076, - "step": 16657 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019436582561049856, - "loss": 1.5858, - "step": 16658 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943654873841575, - "loss": 1.5623, - "step": 16659 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019436514915781644, - "loss": 1.5605, - "step": 16660 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019436481093147536, - "loss": 1.4606, - "step": 16661 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943644727051343, - "loss": 1.6709, - "step": 16662 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019436413447879324, - "loss": 1.7001, - "step": 16663 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019436379625245215, - "loss": 1.575, - "step": 16664 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943634580261111, - "loss": 1.7709, - "step": 16665 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019436311979977003, - "loss": 1.7152, - "step": 16666 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019436278157342895, - "loss": 1.5608, - "step": 16667 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943624433470879, - "loss": 1.6762, - "step": 16668 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019436210512074683, - "loss": 1.4812, - "step": 16669 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019436176689440575, - "loss": 1.5594, - "step": 16670 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019436142866806469, - "loss": 1.4197, - "step": 16671 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019436109044172363, - "loss": 1.6017, - "step": 16672 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019436075221538254, - "loss": 1.6461, - "step": 16673 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019436041398904148, - "loss": 1.7157, - "step": 16674 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019436007576270042, - "loss": 1.5028, - "step": 16675 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019435973753635934, - "loss": 1.5071, - "step": 16676 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019435939931001828, - "loss": 1.5633, - "step": 16677 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019435906108367722, - "loss": 1.5796, - "step": 16678 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019435872285733613, - "loss": 1.5946, - "step": 16679 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019435838463099508, - "loss": 1.5786, - "step": 16680 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019435804640465402, - "loss": 1.3952, - "step": 16681 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019435770817831293, - "loss": 1.6469, - "step": 16682 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019435736995197187, - "loss": 1.5824, - "step": 16683 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943570317256308, - "loss": 1.6319, - "step": 16684 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019435669349928973, - "loss": 1.4352, - "step": 16685 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019435635527294867, - "loss": 1.5878, - "step": 16686 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943560170466076, - "loss": 1.5906, - "step": 16687 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019435567882026652, - "loss": 1.6674, - "step": 16688 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019435534059392546, - "loss": 1.6114, - "step": 16689 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943550023675844, - "loss": 1.626, - "step": 16690 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019435466414124332, - "loss": 1.8223, - "step": 16691 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019435432591490226, - "loss": 1.5645, - "step": 16692 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943539876885612, - "loss": 1.6194, - "step": 16693 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019435364946222012, - "loss": 1.8168, - "step": 16694 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019435331123587906, - "loss": 1.6499, - "step": 16695 - }, - { - "epoch": 0.06, - "learning_rate": 0.000194352973009538, - "loss": 1.6517, - "step": 16696 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943526347831969, - "loss": 1.5231, - "step": 16697 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019435229655685585, - "loss": 1.4809, - "step": 16698 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943519583305148, - "loss": 1.5124, - "step": 16699 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943516201041737, - "loss": 1.6903, - "step": 16700 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019435128187783265, - "loss": 1.7112, - "step": 16701 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943509436514916, - "loss": 1.7201, - "step": 16702 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943506054251505, - "loss": 1.6019, - "step": 16703 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019435026719880945, - "loss": 1.5548, - "step": 16704 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943499289724684, - "loss": 1.6311, - "step": 16705 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943495907461273, - "loss": 1.6837, - "step": 16706 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019434925251978624, - "loss": 1.6506, - "step": 16707 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019434891429344518, - "loss": 1.5411, - "step": 16708 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943485760671041, - "loss": 1.6592, - "step": 16709 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019434823784076307, - "loss": 1.6016, - "step": 16710 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019434789961442198, - "loss": 1.5376, - "step": 16711 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019434756138808092, - "loss": 1.5921, - "step": 16712 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019434722316173986, - "loss": 1.5756, - "step": 16713 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019434688493539878, - "loss": 1.6872, - "step": 16714 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019434654670905772, - "loss": 1.5873, - "step": 16715 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019434620848271666, - "loss": 1.481, - "step": 16716 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019434587025637557, - "loss": 1.7325, - "step": 16717 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943455320300345, - "loss": 1.5612, - "step": 16718 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019434519380369345, - "loss": 1.6665, - "step": 16719 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019434485557735237, - "loss": 1.5154, - "step": 16720 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943445173510113, - "loss": 1.6063, - "step": 16721 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019434417912467025, - "loss": 1.6224, - "step": 16722 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019434384089832916, - "loss": 1.5572, - "step": 16723 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943435026719881, - "loss": 1.4756, - "step": 16724 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019434316444564705, - "loss": 1.5807, - "step": 16725 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019434282621930596, - "loss": 1.6557, - "step": 16726 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943424879929649, - "loss": 1.6126, - "step": 16727 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019434214976662384, - "loss": 1.6298, - "step": 16728 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019434181154028276, - "loss": 1.5717, - "step": 16729 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943414733139417, - "loss": 1.6381, - "step": 16730 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019434113508760064, - "loss": 1.653, - "step": 16731 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019434079686125955, - "loss": 1.5161, - "step": 16732 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943404586349185, - "loss": 1.6278, - "step": 16733 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019434012040857744, - "loss": 1.6756, - "step": 16734 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019433978218223635, - "loss": 1.6627, - "step": 16735 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943394439558953, - "loss": 1.5824, - "step": 16736 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019433910572955423, - "loss": 1.6148, - "step": 16737 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019433876750321315, - "loss": 1.6634, - "step": 16738 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943384292768721, - "loss": 1.5675, - "step": 16739 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019433809105053103, - "loss": 1.6204, - "step": 16740 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019433775282418994, - "loss": 1.637, - "step": 16741 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019433741459784888, - "loss": 1.5426, - "step": 16742 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019433707637150782, - "loss": 1.5156, - "step": 16743 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019433673814516674, - "loss": 1.5073, - "step": 16744 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019433639991882568, - "loss": 1.58, - "step": 16745 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019433606169248462, - "loss": 1.6171, - "step": 16746 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019433572346614356, - "loss": 1.5677, - "step": 16747 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019433538523980248, - "loss": 1.6394, - "step": 16748 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019433504701346142, - "loss": 1.5788, - "step": 16749 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019433470878712036, - "loss": 1.729, - "step": 16750 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019433437056077927, - "loss": 1.6516, - "step": 16751 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019433403233443821, - "loss": 1.6209, - "step": 16752 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019433369410809715, - "loss": 1.6542, - "step": 16753 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019433335588175607, - "loss": 1.813, - "step": 16754 - }, - { - "epoch": 0.06, - "learning_rate": 0.000194333017655415, - "loss": 1.7269, - "step": 16755 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019433267942907395, - "loss": 1.4597, - "step": 16756 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019433234120273287, - "loss": 1.6204, - "step": 16757 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943320029763918, - "loss": 1.4763, - "step": 16758 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019433166475005075, - "loss": 1.5932, - "step": 16759 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019433132652370966, - "loss": 1.5223, - "step": 16760 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943309882973686, - "loss": 1.7137, - "step": 16761 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019433065007102754, - "loss": 1.5749, - "step": 16762 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019433031184468646, - "loss": 1.5107, - "step": 16763 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943299736183454, - "loss": 1.5951, - "step": 16764 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019432963539200434, - "loss": 1.5342, - "step": 16765 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019432929716566328, - "loss": 1.5298, - "step": 16766 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019432895893932222, - "loss": 1.5062, - "step": 16767 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019432862071298114, - "loss": 1.6184, - "step": 16768 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019432828248664008, - "loss": 1.4647, - "step": 16769 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019432794426029902, - "loss": 1.7437, - "step": 16770 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019432760603395793, - "loss": 1.6176, - "step": 16771 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019432726780761687, - "loss": 1.5657, - "step": 16772 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019432692958127582, - "loss": 1.5887, - "step": 16773 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019432659135493473, - "loss": 1.6999, - "step": 16774 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019432625312859367, - "loss": 1.4911, - "step": 16775 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943259149022526, - "loss": 1.591, - "step": 16776 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019432557667591153, - "loss": 1.5417, - "step": 16777 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019432523844957047, - "loss": 1.6862, - "step": 16778 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943249002232294, - "loss": 1.6074, - "step": 16779 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019432456199688832, - "loss": 1.6235, - "step": 16780 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019432422377054726, - "loss": 1.3839, - "step": 16781 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943238855442062, - "loss": 1.6338, - "step": 16782 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019432354731786512, - "loss": 1.5857, - "step": 16783 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019432320909152406, - "loss": 1.6591, - "step": 16784 - }, - { - "epoch": 0.06, - "learning_rate": 0.000194322870865183, - "loss": 1.4689, - "step": 16785 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019432253263884191, - "loss": 1.6221, - "step": 16786 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019432219441250086, - "loss": 1.6346, - "step": 16787 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943218561861598, - "loss": 1.6656, - "step": 16788 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943215179598187, - "loss": 1.4461, - "step": 16789 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019432117973347765, - "loss": 1.5803, - "step": 16790 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943208415071366, - "loss": 1.4992, - "step": 16791 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943205032807955, - "loss": 1.5746, - "step": 16792 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019432016505445445, - "loss": 1.5784, - "step": 16793 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943198268281134, - "loss": 1.6036, - "step": 16794 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943194886017723, - "loss": 1.6132, - "step": 16795 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019431915037543124, - "loss": 1.5473, - "step": 16796 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019431881214909019, - "loss": 1.6399, - "step": 16797 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943184739227491, - "loss": 1.5502, - "step": 16798 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019431813569640804, - "loss": 1.7231, - "step": 16799 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019431779747006698, - "loss": 1.4952, - "step": 16800 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943174592437259, - "loss": 1.6899, - "step": 16801 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019431712101738484, - "loss": 1.569, - "step": 16802 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019431678279104378, - "loss": 1.5182, - "step": 16803 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943164445647027, - "loss": 1.5868, - "step": 16804 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019431610633836163, - "loss": 1.5869, - "step": 16805 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019431576811202057, - "loss": 1.5607, - "step": 16806 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943154298856795, - "loss": 1.4621, - "step": 16807 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019431509165933843, - "loss": 1.6754, - "step": 16808 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019431475343299737, - "loss": 1.4651, - "step": 16809 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019431441520665628, - "loss": 1.5627, - "step": 16810 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019431407698031523, - "loss": 1.6755, - "step": 16811 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019431373875397417, - "loss": 1.4917, - "step": 16812 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019431340052763308, - "loss": 1.6602, - "step": 16813 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019431306230129202, - "loss": 1.6436, - "step": 16814 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019431272407495096, - "loss": 1.6304, - "step": 16815 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019431238584860988, - "loss": 1.5404, - "step": 16816 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019431204762226882, - "loss": 1.4872, - "step": 16817 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019431170939592776, - "loss": 1.6133, - "step": 16818 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019431137116958667, - "loss": 1.377, - "step": 16819 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019431103294324561, - "loss": 1.7394, - "step": 16820 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019431069471690456, - "loss": 1.6944, - "step": 16821 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943103564905635, - "loss": 1.5763, - "step": 16822 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019431001826422244, - "loss": 1.6169, - "step": 16823 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019430968003788138, - "loss": 1.6252, - "step": 16824 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943093418115403, - "loss": 1.7058, - "step": 16825 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019430900358519923, - "loss": 1.8039, - "step": 16826 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019430866535885818, - "loss": 1.7593, - "step": 16827 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943083271325171, - "loss": 1.495, - "step": 16828 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019430798890617603, - "loss": 1.5908, - "step": 16829 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019430765067983497, - "loss": 1.5049, - "step": 16830 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019430731245349389, - "loss": 1.6677, - "step": 16831 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019430697422715283, - "loss": 1.4179, - "step": 16832 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019430663600081177, - "loss": 1.3451, - "step": 16833 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019430629777447068, - "loss": 1.5337, - "step": 16834 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019430595954812962, - "loss": 1.4464, - "step": 16835 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019430562132178856, - "loss": 1.3762, - "step": 16836 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019430528309544748, - "loss": 1.5517, - "step": 16837 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019430494486910642, - "loss": 1.3979, - "step": 16838 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019430460664276536, - "loss": 1.5593, - "step": 16839 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019430426841642428, - "loss": 1.5773, - "step": 16840 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019430393019008322, - "loss": 1.5717, - "step": 16841 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019430359196374216, - "loss": 1.6162, - "step": 16842 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019430325373740107, - "loss": 1.5498, - "step": 16843 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019430291551106, - "loss": 1.552, - "step": 16844 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019430257728471895, - "loss": 1.5725, - "step": 16845 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019430223905837787, - "loss": 1.6158, - "step": 16846 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943019008320368, - "loss": 1.6164, - "step": 16847 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019430156260569575, - "loss": 1.5593, - "step": 16848 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019430122437935466, - "loss": 1.7356, - "step": 16849 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001943008861530136, - "loss": 1.6118, - "step": 16850 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019430054792667255, - "loss": 1.7968, - "step": 16851 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019430020970033146, - "loss": 1.671, - "step": 16852 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942998714739904, - "loss": 1.6539, - "step": 16853 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019429953324764934, - "loss": 1.6266, - "step": 16854 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019429919502130826, - "loss": 1.5193, - "step": 16855 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942988567949672, - "loss": 1.5389, - "step": 16856 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019429851856862614, - "loss": 1.4529, - "step": 16857 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019429818034228505, - "loss": 1.4911, - "step": 16858 - }, - { - "epoch": 0.06, - "learning_rate": 0.000194297842115944, - "loss": 1.601, - "step": 16859 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019429750388960294, - "loss": 1.5524, - "step": 16860 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019429716566326185, - "loss": 1.7728, - "step": 16861 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942968274369208, - "loss": 1.6222, - "step": 16862 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019429648921057973, - "loss": 1.7417, - "step": 16863 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019429615098423865, - "loss": 1.4345, - "step": 16864 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942958127578976, - "loss": 1.5494, - "step": 16865 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019429547453155653, - "loss": 1.7654, - "step": 16866 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019429513630521544, - "loss": 1.3956, - "step": 16867 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019429479807887438, - "loss": 1.5613, - "step": 16868 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019429445985253332, - "loss": 1.5807, - "step": 16869 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019429412162619224, - "loss": 1.6529, - "step": 16870 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019429378339985118, - "loss": 1.6127, - "step": 16871 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019429344517351012, - "loss": 1.5127, - "step": 16872 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019429310694716903, - "loss": 1.5867, - "step": 16873 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019429276872082798, - "loss": 1.4446, - "step": 16874 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019429243049448692, - "loss": 1.6074, - "step": 16875 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019429209226814583, - "loss": 1.6134, - "step": 16876 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019429175404180477, - "loss": 1.6303, - "step": 16877 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942914158154637, - "loss": 1.5422, - "step": 16878 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019429107758912265, - "loss": 1.5, - "step": 16879 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942907393627816, - "loss": 1.8392, - "step": 16880 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019429040113644054, - "loss": 1.6418, - "step": 16881 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019429006291009945, - "loss": 1.7448, - "step": 16882 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942897246837584, - "loss": 1.7013, - "step": 16883 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019428938645741733, - "loss": 1.5396, - "step": 16884 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019428904823107625, - "loss": 1.5095, - "step": 16885 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942887100047352, - "loss": 1.7005, - "step": 16886 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019428837177839413, - "loss": 1.5066, - "step": 16887 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019428803355205304, - "loss": 1.6483, - "step": 16888 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019428769532571198, - "loss": 1.5922, - "step": 16889 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019428735709937093, - "loss": 1.5688, - "step": 16890 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019428701887302984, - "loss": 1.5224, - "step": 16891 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019428668064668878, - "loss": 1.7637, - "step": 16892 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019428634242034772, - "loss": 1.3429, - "step": 16893 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019428600419400664, - "loss": 1.5173, - "step": 16894 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019428566596766558, - "loss": 1.7569, - "step": 16895 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019428532774132452, - "loss": 1.6154, - "step": 16896 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019428498951498343, - "loss": 1.4934, - "step": 16897 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019428465128864237, - "loss": 1.6647, - "step": 16898 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019428431306230131, - "loss": 1.6882, - "step": 16899 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019428397483596023, - "loss": 1.731, - "step": 16900 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019428363660961917, - "loss": 1.5713, - "step": 16901 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942832983832781, - "loss": 1.5903, - "step": 16902 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019428296015693702, - "loss": 1.6011, - "step": 16903 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019428262193059597, - "loss": 1.5815, - "step": 16904 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942822837042549, - "loss": 1.555, - "step": 16905 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019428194547791382, - "loss": 1.6365, - "step": 16906 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019428160725157276, - "loss": 1.395, - "step": 16907 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942812690252317, - "loss": 1.6837, - "step": 16908 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019428093079889062, - "loss": 1.4365, - "step": 16909 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019428059257254956, - "loss": 1.5994, - "step": 16910 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942802543462085, - "loss": 1.6424, - "step": 16911 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019427991611986741, - "loss": 1.5747, - "step": 16912 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019427957789352635, - "loss": 1.5453, - "step": 16913 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942792396671853, - "loss": 1.6209, - "step": 16914 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942789014408442, - "loss": 1.4806, - "step": 16915 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019427856321450315, - "loss": 1.7914, - "step": 16916 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942782249881621, - "loss": 1.6384, - "step": 16917 - }, - { - "epoch": 0.06, - "learning_rate": 0.000194277886761821, - "loss": 1.5889, - "step": 16918 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019427754853547995, - "loss": 1.5717, - "step": 16919 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942772103091389, - "loss": 1.4428, - "step": 16920 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942768720827978, - "loss": 1.5458, - "step": 16921 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019427653385645674, - "loss": 1.5153, - "step": 16922 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019427619563011568, - "loss": 1.4852, - "step": 16923 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942758574037746, - "loss": 1.6573, - "step": 16924 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019427551917743354, - "loss": 1.747, - "step": 16925 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019427518095109248, - "loss": 1.7562, - "step": 16926 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942748427247514, - "loss": 1.6575, - "step": 16927 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019427450449841034, - "loss": 1.5702, - "step": 16928 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019427416627206928, - "loss": 1.623, - "step": 16929 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942738280457282, - "loss": 1.7646, - "step": 16930 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019427348981938713, - "loss": 1.6145, - "step": 16931 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019427315159304607, - "loss": 1.7228, - "step": 16932 - }, - { - "epoch": 0.06, - "learning_rate": 0.000194272813366705, - "loss": 1.5476, - "step": 16933 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019427247514036393, - "loss": 1.4902, - "step": 16934 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942721369140229, - "loss": 1.5024, - "step": 16935 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942717986876818, - "loss": 1.461, - "step": 16936 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019427146046134075, - "loss": 1.6732, - "step": 16937 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942711222349997, - "loss": 1.5579, - "step": 16938 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942707840086586, - "loss": 1.5293, - "step": 16939 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019427044578231755, - "loss": 1.6207, - "step": 16940 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942701075559765, - "loss": 1.3927, - "step": 16941 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942697693296354, - "loss": 1.4913, - "step": 16942 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019426943110329435, - "loss": 1.4646, - "step": 16943 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019426909287695329, - "loss": 1.5773, - "step": 16944 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942687546506122, - "loss": 1.5162, - "step": 16945 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019426841642427114, - "loss": 1.5222, - "step": 16946 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019426807819793008, - "loss": 1.4941, - "step": 16947 - }, - { - "epoch": 0.06, - "learning_rate": 0.000194267739971589, - "loss": 1.6049, - "step": 16948 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019426740174524794, - "loss": 1.5732, - "step": 16949 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019426706351890688, - "loss": 1.5728, - "step": 16950 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942667252925658, - "loss": 1.6484, - "step": 16951 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019426638706622473, - "loss": 1.6358, - "step": 16952 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019426604883988368, - "loss": 1.5777, - "step": 16953 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942657106135426, - "loss": 1.465, - "step": 16954 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019426537238720153, - "loss": 1.6581, - "step": 16955 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019426503416086047, - "loss": 1.5218, - "step": 16956 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019426469593451939, - "loss": 1.3798, - "step": 16957 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019426435770817833, - "loss": 1.6257, - "step": 16958 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019426401948183727, - "loss": 1.4885, - "step": 16959 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019426368125549618, - "loss": 1.6686, - "step": 16960 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019426334302915512, - "loss": 1.5471, - "step": 16961 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019426300480281406, - "loss": 1.5656, - "step": 16962 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019426266657647298, - "loss": 1.4598, - "step": 16963 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019426232835013192, - "loss": 1.7052, - "step": 16964 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019426199012379086, - "loss": 1.5028, - "step": 16965 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019426165189744977, - "loss": 1.5083, - "step": 16966 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019426131367110872, - "loss": 1.4634, - "step": 16967 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019426097544476766, - "loss": 1.6532, - "step": 16968 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019426063721842657, - "loss": 1.7073, - "step": 16969 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942602989920855, - "loss": 1.5, - "step": 16970 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019425996076574445, - "loss": 1.4975, - "step": 16971 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019425962253940337, - "loss": 1.5988, - "step": 16972 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942592843130623, - "loss": 1.6275, - "step": 16973 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019425894608672125, - "loss": 1.6565, - "step": 16974 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019425860786038016, - "loss": 1.6098, - "step": 16975 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942582696340391, - "loss": 1.759, - "step": 16976 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019425793140769805, - "loss": 1.5152, - "step": 16977 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019425759318135696, - "loss": 1.5545, - "step": 16978 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942572549550159, - "loss": 1.6786, - "step": 16979 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019425691672867484, - "loss": 1.5473, - "step": 16980 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019425657850233376, - "loss": 1.5613, - "step": 16981 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942562402759927, - "loss": 1.5174, - "step": 16982 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019425590204965164, - "loss": 1.5963, - "step": 16983 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019425556382331055, - "loss": 1.718, - "step": 16984 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942552255969695, - "loss": 1.5597, - "step": 16985 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019425488737062843, - "loss": 1.6392, - "step": 16986 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019425454914428735, - "loss": 1.6135, - "step": 16987 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942542109179463, - "loss": 1.5073, - "step": 16988 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019425387269160523, - "loss": 1.6491, - "step": 16989 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019425353446526414, - "loss": 1.6478, - "step": 16990 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942531962389231, - "loss": 1.7593, - "step": 16991 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019425285801258203, - "loss": 1.8035, - "step": 16992 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019425251978624097, - "loss": 1.4956, - "step": 16993 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942521815598999, - "loss": 1.5658, - "step": 16994 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019425184333355882, - "loss": 1.545, - "step": 16995 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019425150510721776, - "loss": 1.6245, - "step": 16996 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942511668808767, - "loss": 1.532, - "step": 16997 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019425082865453562, - "loss": 1.4833, - "step": 16998 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019425049042819456, - "loss": 1.6457, - "step": 16999 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942501522018535, - "loss": 1.5311, - "step": 17000 - }, - { - "epoch": 0.06, - "eval_loss": 1.4079402043269231, - "mrc_em": 18.0, - "mrc_f1": 24.519047619047623, - "nli_acc": 0.11, - "step": 17000, - "sts_f1": 0.6170212765957447, - "ynat_f1": 0.6837244266836712 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019424981397551242, - "loss": 1.6517, - "step": 17001 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019424947574917136, - "loss": 1.5249, - "step": 17002 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942491375228303, - "loss": 1.7853, - "step": 17003 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942487992964892, - "loss": 1.7454, - "step": 17004 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019424846107014815, - "loss": 1.7339, - "step": 17005 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942481228438071, - "loss": 1.6497, - "step": 17006 - }, - { - "epoch": 0.06, - "learning_rate": 0.000194247784617466, - "loss": 1.5817, - "step": 17007 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019424744639112495, - "loss": 1.6335, - "step": 17008 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942471081647839, - "loss": 1.7041, - "step": 17009 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942467699384428, - "loss": 1.4697, - "step": 17010 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019424643171210175, - "loss": 1.4166, - "step": 17011 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942460934857607, - "loss": 1.5105, - "step": 17012 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019424575525941963, - "loss": 1.7172, - "step": 17013 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019424541703307854, - "loss": 1.6034, - "step": 17014 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019424507880673748, - "loss": 1.6168, - "step": 17015 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019424474058039642, - "loss": 1.7647, - "step": 17016 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019424440235405534, - "loss": 1.5897, - "step": 17017 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019424406412771428, - "loss": 1.4791, - "step": 17018 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019424372590137322, - "loss": 1.4549, - "step": 17019 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019424338767503214, - "loss": 1.5555, - "step": 17020 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019424304944869108, - "loss": 1.6657, - "step": 17021 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019424271122235002, - "loss": 1.6502, - "step": 17022 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019424237299600893, - "loss": 1.6008, - "step": 17023 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019424203476966787, - "loss": 1.6049, - "step": 17024 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019424169654332681, - "loss": 1.5025, - "step": 17025 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019424135831698573, - "loss": 1.6567, - "step": 17026 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019424102009064467, - "loss": 1.6113, - "step": 17027 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942406818643036, - "loss": 1.6833, - "step": 17028 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019424034363796252, - "loss": 1.5229, - "step": 17029 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019424000541162147, - "loss": 1.713, - "step": 17030 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942396671852804, - "loss": 1.4014, - "step": 17031 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019423932895893932, - "loss": 1.641, - "step": 17032 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019423899073259826, - "loss": 1.4888, - "step": 17033 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942386525062572, - "loss": 1.5963, - "step": 17034 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019423831427991612, - "loss": 1.5962, - "step": 17035 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019423797605357506, - "loss": 1.5883, - "step": 17036 - }, - { - "epoch": 0.06, - "learning_rate": 0.000194237637827234, - "loss": 1.7251, - "step": 17037 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942372996008929, - "loss": 1.6483, - "step": 17038 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019423696137455185, - "loss": 1.5932, - "step": 17039 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942366231482108, - "loss": 1.7904, - "step": 17040 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942362849218697, - "loss": 1.6194, - "step": 17041 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019423594669552865, - "loss": 1.5812, - "step": 17042 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942356084691876, - "loss": 1.6883, - "step": 17043 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942352702428465, - "loss": 1.563, - "step": 17044 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019423493201650545, - "loss": 1.5852, - "step": 17045 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942345937901644, - "loss": 1.6549, - "step": 17046 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019423425556382333, - "loss": 1.6077, - "step": 17047 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019423391733748227, - "loss": 1.5955, - "step": 17048 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019423357911114118, - "loss": 1.6831, - "step": 17049 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019423324088480013, - "loss": 1.6708, - "step": 17050 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019423290265845907, - "loss": 1.6267, - "step": 17051 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019423256443211798, - "loss": 1.639, - "step": 17052 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019423222620577692, - "loss": 1.6122, - "step": 17053 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019423188797943586, - "loss": 1.6676, - "step": 17054 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019423154975309478, - "loss": 1.5847, - "step": 17055 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019423121152675372, - "loss": 1.6782, - "step": 17056 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019423087330041266, - "loss": 1.4161, - "step": 17057 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019423053507407157, - "loss": 1.5758, - "step": 17058 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019423019684773051, - "loss": 1.567, - "step": 17059 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019422985862138946, - "loss": 1.8028, - "step": 17060 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019422952039504837, - "loss": 1.6191, - "step": 17061 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942291821687073, - "loss": 1.7825, - "step": 17062 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019422884394236625, - "loss": 1.5306, - "step": 17063 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019422850571602517, - "loss": 1.6778, - "step": 17064 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942281674896841, - "loss": 1.4535, - "step": 17065 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019422782926334305, - "loss": 1.6409, - "step": 17066 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019422749103700196, - "loss": 1.7006, - "step": 17067 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942271528106609, - "loss": 1.4456, - "step": 17068 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019422681458431984, - "loss": 1.6611, - "step": 17069 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019422647635797876, - "loss": 1.5844, - "step": 17070 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942261381316377, - "loss": 1.6161, - "step": 17071 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019422579990529664, - "loss": 1.5631, - "step": 17072 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019422546167895555, - "loss": 1.604, - "step": 17073 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942251234526145, - "loss": 1.6638, - "step": 17074 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019422478522627344, - "loss": 1.5719, - "step": 17075 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019422444699993235, - "loss": 1.4736, - "step": 17076 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942241087735913, - "loss": 1.6727, - "step": 17077 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019422377054725023, - "loss": 1.4903, - "step": 17078 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019422343232090915, - "loss": 1.6874, - "step": 17079 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942230940945681, - "loss": 1.8539, - "step": 17080 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019422275586822703, - "loss": 1.5528, - "step": 17081 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019422241764188594, - "loss": 1.6426, - "step": 17082 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019422207941554488, - "loss": 1.493, - "step": 17083 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019422174118920383, - "loss": 1.5038, - "step": 17084 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019422140296286274, - "loss": 1.5112, - "step": 17085 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019422106473652168, - "loss": 1.6158, - "step": 17086 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019422072651018062, - "loss": 1.4579, - "step": 17087 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019422038828383954, - "loss": 1.5814, - "step": 17088 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019422005005749848, - "loss": 1.5893, - "step": 17089 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019421971183115742, - "loss": 1.5802, - "step": 17090 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019421937360481633, - "loss": 1.5515, - "step": 17091 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019421903537847527, - "loss": 1.5956, - "step": 17092 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019421869715213421, - "loss": 1.6081, - "step": 17093 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019421835892579313, - "loss": 1.7386, - "step": 17094 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019421802069945207, - "loss": 1.624, - "step": 17095 - }, - { - "epoch": 0.06, - "learning_rate": 0.000194217682473111, - "loss": 1.58, - "step": 17096 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019421734424676995, - "loss": 1.5802, - "step": 17097 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019421700602042887, - "loss": 1.6698, - "step": 17098 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942166677940878, - "loss": 1.5773, - "step": 17099 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019421632956774675, - "loss": 1.5874, - "step": 17100 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019421599134140566, - "loss": 1.7084, - "step": 17101 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942156531150646, - "loss": 1.6613, - "step": 17102 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019421531488872355, - "loss": 1.562, - "step": 17103 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019421497666238249, - "loss": 1.6432, - "step": 17104 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019421463843604143, - "loss": 1.5574, - "step": 17105 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019421430020970034, - "loss": 1.5628, - "step": 17106 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019421396198335928, - "loss": 1.5498, - "step": 17107 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019421362375701822, - "loss": 1.735, - "step": 17108 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019421328553067714, - "loss": 1.4529, - "step": 17109 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019421294730433608, - "loss": 1.6212, - "step": 17110 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019421260907799502, - "loss": 1.6992, - "step": 17111 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019421227085165393, - "loss": 1.6185, - "step": 17112 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019421193262531288, - "loss": 1.5382, - "step": 17113 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019421159439897182, - "loss": 1.573, - "step": 17114 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019421125617263073, - "loss": 1.5306, - "step": 17115 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019421091794628967, - "loss": 1.6725, - "step": 17116 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942105797199486, - "loss": 1.5255, - "step": 17117 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019421024149360753, - "loss": 1.628, - "step": 17118 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019420990326726647, - "loss": 1.5817, - "step": 17119 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942095650409254, - "loss": 1.4197, - "step": 17120 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019420922681458432, - "loss": 1.4736, - "step": 17121 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019420888858824326, - "loss": 1.7062, - "step": 17122 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942085503619022, - "loss": 1.4726, - "step": 17123 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019420821213556112, - "loss": 1.5909, - "step": 17124 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019420787390922006, - "loss": 1.6044, - "step": 17125 - }, - { - "epoch": 0.06, - "learning_rate": 0.000194207535682879, - "loss": 1.5199, - "step": 17126 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019420719745653792, - "loss": 1.7125, - "step": 17127 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019420685923019686, - "loss": 1.6771, - "step": 17128 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942065210038558, - "loss": 1.507, - "step": 17129 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942061827775147, - "loss": 1.5514, - "step": 17130 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019420584455117365, - "loss": 1.7199, - "step": 17131 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942055063248326, - "loss": 1.6195, - "step": 17132 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942051680984915, - "loss": 1.4146, - "step": 17133 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019420482987215045, - "loss": 1.6179, - "step": 17134 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942044916458094, - "loss": 1.6503, - "step": 17135 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942041534194683, - "loss": 1.3202, - "step": 17136 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019420381519312725, - "loss": 1.5901, - "step": 17137 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942034769667862, - "loss": 1.5637, - "step": 17138 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942031387404451, - "loss": 1.5975, - "step": 17139 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019420280051410404, - "loss": 1.5882, - "step": 17140 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019420246228776298, - "loss": 1.6142, - "step": 17141 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942021240614219, - "loss": 1.4817, - "step": 17142 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019420178583508084, - "loss": 1.5329, - "step": 17143 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019420144760873978, - "loss": 1.5573, - "step": 17144 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942011093823987, - "loss": 1.6315, - "step": 17145 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019420077115605763, - "loss": 1.449, - "step": 17146 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019420043292971658, - "loss": 1.6561, - "step": 17147 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001942000947033755, - "loss": 1.6037, - "step": 17148 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019419975647703443, - "loss": 1.4747, - "step": 17149 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019419941825069337, - "loss": 1.5327, - "step": 17150 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019419908002435229, - "loss": 1.5059, - "step": 17151 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019419874179801123, - "loss": 1.5297, - "step": 17152 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019419840357167017, - "loss": 1.4551, - "step": 17153 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019419806534532908, - "loss": 1.7468, - "step": 17154 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019419772711898802, - "loss": 1.5949, - "step": 17155 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019419738889264696, - "loss": 1.5636, - "step": 17156 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019419705066630588, - "loss": 1.5414, - "step": 17157 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019419671243996482, - "loss": 1.4778, - "step": 17158 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941963742136238, - "loss": 1.489, - "step": 17159 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941960359872827, - "loss": 1.7138, - "step": 17160 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019419569776094164, - "loss": 1.6264, - "step": 17161 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019419535953460058, - "loss": 1.7134, - "step": 17162 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941950213082595, - "loss": 1.68, - "step": 17163 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019419468308191844, - "loss": 1.5746, - "step": 17164 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019419434485557738, - "loss": 1.6425, - "step": 17165 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941940066292363, - "loss": 1.5743, - "step": 17166 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019419366840289524, - "loss": 1.6145, - "step": 17167 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019419333017655418, - "loss": 1.5204, - "step": 17168 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941929919502131, - "loss": 1.673, - "step": 17169 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019419265372387203, - "loss": 1.6364, - "step": 17170 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019419231549753097, - "loss": 1.6278, - "step": 17171 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941919772711899, - "loss": 1.6837, - "step": 17172 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019419163904484883, - "loss": 1.4849, - "step": 17173 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019419130081850777, - "loss": 1.4589, - "step": 17174 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019419096259216668, - "loss": 1.6271, - "step": 17175 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019419062436582562, - "loss": 1.531, - "step": 17176 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019419028613948457, - "loss": 1.6854, - "step": 17177 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019418994791314348, - "loss": 1.5511, - "step": 17178 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019418960968680242, - "loss": 1.65, - "step": 17179 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019418927146046136, - "loss": 1.4514, - "step": 17180 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019418893323412028, - "loss": 1.5554, - "step": 17181 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019418859500777922, - "loss": 1.6891, - "step": 17182 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019418825678143816, - "loss": 1.498, - "step": 17183 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019418791855509707, - "loss": 1.6807, - "step": 17184 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019418758032875601, - "loss": 1.8064, - "step": 17185 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019418724210241495, - "loss": 1.635, - "step": 17186 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019418690387607387, - "loss": 1.6577, - "step": 17187 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941865656497328, - "loss": 1.6178, - "step": 17188 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019418622742339175, - "loss": 1.5541, - "step": 17189 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019418588919705067, - "loss": 1.578, - "step": 17190 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941855509707096, - "loss": 1.5734, - "step": 17191 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019418521274436855, - "loss": 1.4941, - "step": 17192 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019418487451802746, - "loss": 1.5719, - "step": 17193 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941845362916864, - "loss": 1.6855, - "step": 17194 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019418419806534534, - "loss": 1.6112, - "step": 17195 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019418385983900426, - "loss": 1.6574, - "step": 17196 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941835216126632, - "loss": 1.795, - "step": 17197 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019418318338632214, - "loss": 1.4871, - "step": 17198 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019418284515998105, - "loss": 1.5044, - "step": 17199 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019418250693364, - "loss": 1.5923, - "step": 17200 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019418216870729894, - "loss": 1.4906, - "step": 17201 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019418183048095785, - "loss": 1.6525, - "step": 17202 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941814922546168, - "loss": 1.684, - "step": 17203 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019418115402827573, - "loss": 1.5263, - "step": 17204 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019418081580193465, - "loss": 1.6638, - "step": 17205 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941804775755936, - "loss": 1.5115, - "step": 17206 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019418013934925253, - "loss": 1.5453, - "step": 17207 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019417980112291144, - "loss": 1.4834, - "step": 17208 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019417946289657038, - "loss": 1.4566, - "step": 17209 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019417912467022933, - "loss": 1.519, - "step": 17210 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019417878644388824, - "loss": 1.4263, - "step": 17211 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019417844821754718, - "loss": 1.4697, - "step": 17212 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019417810999120612, - "loss": 1.4773, - "step": 17213 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019417777176486504, - "loss": 1.6507, - "step": 17214 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019417743353852398, - "loss": 1.5694, - "step": 17215 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019417709531218295, - "loss": 1.5301, - "step": 17216 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019417675708584186, - "loss": 1.6484, - "step": 17217 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941764188595008, - "loss": 1.6122, - "step": 17218 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019417608063315974, - "loss": 1.6561, - "step": 17219 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019417574240681866, - "loss": 1.6369, - "step": 17220 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941754041804776, - "loss": 1.5516, - "step": 17221 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019417506595413654, - "loss": 1.6284, - "step": 17222 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019417472772779545, - "loss": 1.5469, - "step": 17223 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941743895014544, - "loss": 1.4255, - "step": 17224 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019417405127511333, - "loss": 1.6166, - "step": 17225 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019417371304877225, - "loss": 1.5301, - "step": 17226 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941733748224312, - "loss": 1.5465, - "step": 17227 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019417303659609013, - "loss": 1.5971, - "step": 17228 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019417269836974904, - "loss": 1.5318, - "step": 17229 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019417236014340799, - "loss": 1.6088, - "step": 17230 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019417202191706693, - "loss": 1.7919, - "step": 17231 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019417168369072584, - "loss": 1.6317, - "step": 17232 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019417134546438478, - "loss": 1.4969, - "step": 17233 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019417100723804372, - "loss": 1.6251, - "step": 17234 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019417066901170264, - "loss": 1.5283, - "step": 17235 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019417033078536158, - "loss": 1.4751, - "step": 17236 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019416999255902052, - "loss": 1.6317, - "step": 17237 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019416965433267943, - "loss": 1.6386, - "step": 17238 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019416931610633837, - "loss": 1.6836, - "step": 17239 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019416897787999732, - "loss": 1.5384, - "step": 17240 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019416863965365623, - "loss": 1.5576, - "step": 17241 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019416830142731517, - "loss": 1.4886, - "step": 17242 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941679632009741, - "loss": 1.5834, - "step": 17243 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019416762497463303, - "loss": 1.396, - "step": 17244 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019416728674829197, - "loss": 1.5508, - "step": 17245 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941669485219509, - "loss": 1.396, - "step": 17246 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019416661029560982, - "loss": 1.6555, - "step": 17247 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019416627206926876, - "loss": 1.4311, - "step": 17248 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941659338429277, - "loss": 1.7162, - "step": 17249 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019416559561658662, - "loss": 1.6854, - "step": 17250 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019416525739024556, - "loss": 1.5074, - "step": 17251 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941649191639045, - "loss": 1.5276, - "step": 17252 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019416458093756341, - "loss": 1.7112, - "step": 17253 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019416424271122236, - "loss": 1.5676, - "step": 17254 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941639044848813, - "loss": 1.5489, - "step": 17255 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941635662585402, - "loss": 1.6, - "step": 17256 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019416322803219915, - "loss": 1.5347, - "step": 17257 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941628898058581, - "loss": 1.485, - "step": 17258 - }, - { - "epoch": 0.06, - "learning_rate": 0.000194162551579517, - "loss": 1.6086, - "step": 17259 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019416221335317595, - "loss": 1.5451, - "step": 17260 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941618751268349, - "loss": 1.5128, - "step": 17261 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941615369004938, - "loss": 1.4018, - "step": 17262 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019416119867415275, - "loss": 1.6113, - "step": 17263 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019416086044781169, - "loss": 1.5203, - "step": 17264 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941605222214706, - "loss": 1.5179, - "step": 17265 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019416018399512954, - "loss": 1.6337, - "step": 17266 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019415984576878848, - "loss": 1.6879, - "step": 17267 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941595075424474, - "loss": 1.6192, - "step": 17268 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019415916931610634, - "loss": 1.6155, - "step": 17269 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019415883108976528, - "loss": 1.5699, - "step": 17270 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941584928634242, - "loss": 1.6071, - "step": 17271 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019415815463708316, - "loss": 1.5327, - "step": 17272 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019415781641074208, - "loss": 1.4573, - "step": 17273 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019415747818440102, - "loss": 1.457, - "step": 17274 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019415713995805996, - "loss": 1.6592, - "step": 17275 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019415680173171887, - "loss": 1.6873, - "step": 17276 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941564635053778, - "loss": 1.602, - "step": 17277 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019415612527903675, - "loss": 1.6691, - "step": 17278 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019415578705269567, - "loss": 1.5506, - "step": 17279 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941554488263546, - "loss": 1.3267, - "step": 17280 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019415511060001355, - "loss": 1.73, - "step": 17281 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941547723736725, - "loss": 1.6353, - "step": 17282 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941544341473314, - "loss": 1.6116, - "step": 17283 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019415409592099035, - "loss": 1.3537, - "step": 17284 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941537576946493, - "loss": 1.6054, - "step": 17285 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941534194683082, - "loss": 1.6196, - "step": 17286 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019415308124196714, - "loss": 1.5839, - "step": 17287 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019415274301562608, - "loss": 1.7397, - "step": 17288 - }, - { - "epoch": 0.06, - "learning_rate": 0.000194152404789285, - "loss": 1.4651, - "step": 17289 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019415206656294394, - "loss": 1.6331, - "step": 17290 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019415172833660288, - "loss": 1.7031, - "step": 17291 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941513901102618, - "loss": 1.5087, - "step": 17292 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019415105188392074, - "loss": 1.6298, - "step": 17293 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019415071365757968, - "loss": 1.6725, - "step": 17294 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941503754312386, - "loss": 1.5697, - "step": 17295 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019415003720489753, - "loss": 1.6629, - "step": 17296 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019414969897855647, - "loss": 1.6082, - "step": 17297 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941493607522154, - "loss": 1.5423, - "step": 17298 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019414902252587433, - "loss": 1.6221, - "step": 17299 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019414868429953327, - "loss": 1.5016, - "step": 17300 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019414834607319218, - "loss": 1.6426, - "step": 17301 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019414800784685112, - "loss": 1.5032, - "step": 17302 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019414766962051007, - "loss": 1.6297, - "step": 17303 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019414733139416898, - "loss": 1.7639, - "step": 17304 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019414699316782792, - "loss": 1.6128, - "step": 17305 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019414665494148686, - "loss": 1.696, - "step": 17306 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019414631671514578, - "loss": 1.5256, - "step": 17307 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019414597848880472, - "loss": 1.4716, - "step": 17308 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019414564026246366, - "loss": 1.539, - "step": 17309 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019414530203612257, - "loss": 1.6471, - "step": 17310 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941449638097815, - "loss": 1.5787, - "step": 17311 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019414462558344045, - "loss": 1.4525, - "step": 17312 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019414428735709937, - "loss": 1.5542, - "step": 17313 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941439491307583, - "loss": 1.4878, - "step": 17314 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019414361090441725, - "loss": 1.7531, - "step": 17315 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019414327267807616, - "loss": 1.7373, - "step": 17316 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941429344517351, - "loss": 1.6167, - "step": 17317 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019414259622539405, - "loss": 1.5191, - "step": 17318 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019414225799905296, - "loss": 1.4634, - "step": 17319 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941419197727119, - "loss": 1.5665, - "step": 17320 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019414158154637084, - "loss": 1.4906, - "step": 17321 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019414124332002976, - "loss": 1.597, - "step": 17322 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941409050936887, - "loss": 1.5856, - "step": 17323 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019414056686734764, - "loss": 1.7244, - "step": 17324 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019414022864100655, - "loss": 1.5725, - "step": 17325 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941398904146655, - "loss": 1.6575, - "step": 17326 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019413955218832444, - "loss": 1.6929, - "step": 17327 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019413921396198338, - "loss": 1.6507, - "step": 17328 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019413887573564232, - "loss": 1.7336, - "step": 17329 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019413853750930123, - "loss": 1.5951, - "step": 17330 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019413819928296017, - "loss": 1.6061, - "step": 17331 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019413786105661911, - "loss": 1.61, - "step": 17332 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019413752283027803, - "loss": 1.466, - "step": 17333 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019413718460393697, - "loss": 1.497, - "step": 17334 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941368463775959, - "loss": 1.5774, - "step": 17335 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019413650815125482, - "loss": 1.6128, - "step": 17336 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019413616992491377, - "loss": 1.7088, - "step": 17337 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941358316985727, - "loss": 1.4843, - "step": 17338 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019413549347223162, - "loss": 1.6399, - "step": 17339 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019413515524589056, - "loss": 1.6916, - "step": 17340 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941348170195495, - "loss": 1.6318, - "step": 17341 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019413447879320842, - "loss": 1.8208, - "step": 17342 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019413414056686736, - "loss": 1.537, - "step": 17343 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941338023405263, - "loss": 1.6109, - "step": 17344 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019413346411418521, - "loss": 1.6594, - "step": 17345 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019413312588784415, - "loss": 1.5908, - "step": 17346 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941327876615031, - "loss": 1.6763, - "step": 17347 - }, - { - "epoch": 0.06, - "learning_rate": 0.000194132449435162, - "loss": 1.5933, - "step": 17348 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019413211120882095, - "loss": 1.735, - "step": 17349 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941317729824799, - "loss": 1.568, - "step": 17350 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941314347561388, - "loss": 1.6429, - "step": 17351 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019413109652979775, - "loss": 1.5783, - "step": 17352 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941307583034567, - "loss": 1.6329, - "step": 17353 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941304200771156, - "loss": 1.639, - "step": 17354 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019413008185077454, - "loss": 1.5674, - "step": 17355 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019412974362443348, - "loss": 1.6152, - "step": 17356 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941294053980924, - "loss": 1.7284, - "step": 17357 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019412906717175134, - "loss": 1.7017, - "step": 17358 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019412872894541028, - "loss": 1.6842, - "step": 17359 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941283907190692, - "loss": 1.6083, - "step": 17360 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019412805249272814, - "loss": 1.5088, - "step": 17361 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019412771426638708, - "loss": 1.6923, - "step": 17362 - }, - { - "epoch": 0.06, - "learning_rate": 0.000194127376040046, - "loss": 1.5492, - "step": 17363 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019412703781370493, - "loss": 1.7747, - "step": 17364 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019412669958736387, - "loss": 1.6185, - "step": 17365 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019412636136102282, - "loss": 1.6335, - "step": 17366 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019412602313468173, - "loss": 1.5077, - "step": 17367 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019412568490834067, - "loss": 1.5545, - "step": 17368 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941253466819996, - "loss": 1.5823, - "step": 17369 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019412500845565853, - "loss": 1.6252, - "step": 17370 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019412467022931747, - "loss": 1.6874, - "step": 17371 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941243320029764, - "loss": 1.7422, - "step": 17372 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019412399377663532, - "loss": 1.5546, - "step": 17373 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019412365555029426, - "loss": 1.4368, - "step": 17374 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941233173239532, - "loss": 1.5484, - "step": 17375 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019412297909761212, - "loss": 1.6418, - "step": 17376 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019412264087127106, - "loss": 1.738, - "step": 17377 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019412230264493, - "loss": 1.6173, - "step": 17378 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019412196441858891, - "loss": 1.4621, - "step": 17379 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019412162619224786, - "loss": 1.552, - "step": 17380 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941212879659068, - "loss": 1.6762, - "step": 17381 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941209497395657, - "loss": 1.4921, - "step": 17382 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019412061151322465, - "loss": 1.6371, - "step": 17383 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941202732868836, - "loss": 1.5887, - "step": 17384 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019411993506054253, - "loss": 1.6032, - "step": 17385 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019411959683420148, - "loss": 1.7052, - "step": 17386 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941192586078604, - "loss": 1.6512, - "step": 17387 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019411892038151933, - "loss": 1.5711, - "step": 17388 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019411858215517827, - "loss": 1.6982, - "step": 17389 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019411824392883719, - "loss": 1.5611, - "step": 17390 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019411790570249613, - "loss": 1.6515, - "step": 17391 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019411756747615507, - "loss": 1.653, - "step": 17392 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019411722924981398, - "loss": 1.6675, - "step": 17393 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019411689102347292, - "loss": 1.5177, - "step": 17394 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019411655279713186, - "loss": 1.5433, - "step": 17395 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019411621457079078, - "loss": 1.481, - "step": 17396 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019411587634444972, - "loss": 1.552, - "step": 17397 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019411553811810866, - "loss": 1.6045, - "step": 17398 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019411519989176757, - "loss": 1.4951, - "step": 17399 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019411486166542652, - "loss": 1.4649, - "step": 17400 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019411452343908546, - "loss": 1.5702, - "step": 17401 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019411418521274437, - "loss": 1.701, - "step": 17402 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941138469864033, - "loss": 1.5821, - "step": 17403 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019411350876006225, - "loss": 1.6076, - "step": 17404 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019411317053372117, - "loss": 1.776, - "step": 17405 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941128323073801, - "loss": 1.5483, - "step": 17406 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019411249408103905, - "loss": 1.5176, - "step": 17407 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019411215585469796, - "loss": 1.5246, - "step": 17408 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941118176283569, - "loss": 1.5158, - "step": 17409 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019411147940201585, - "loss": 1.5974, - "step": 17410 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019411114117567476, - "loss": 1.4851, - "step": 17411 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941108029493337, - "loss": 1.4181, - "step": 17412 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019411046472299264, - "loss": 1.4293, - "step": 17413 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019411012649665156, - "loss": 1.7536, - "step": 17414 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941097882703105, - "loss": 1.5532, - "step": 17415 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019410945004396944, - "loss": 1.82, - "step": 17416 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019410911181762835, - "loss": 1.5552, - "step": 17417 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941087735912873, - "loss": 1.5965, - "step": 17418 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019410843536494623, - "loss": 1.6482, - "step": 17419 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019410809713860515, - "loss": 1.7515, - "step": 17420 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941077589122641, - "loss": 1.3974, - "step": 17421 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019410742068592303, - "loss": 1.5085, - "step": 17422 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019410708245958194, - "loss": 1.6611, - "step": 17423 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019410674423324089, - "loss": 1.5256, - "step": 17424 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019410640600689983, - "loss": 1.7385, - "step": 17425 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019410606778055874, - "loss": 1.435, - "step": 17426 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019410572955421768, - "loss": 1.604, - "step": 17427 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019410539132787662, - "loss": 1.5744, - "step": 17428 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019410505310153554, - "loss": 1.6008, - "step": 17429 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019410471487519448, - "loss": 1.8345, - "step": 17430 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019410437664885342, - "loss": 1.6639, - "step": 17431 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019410403842251233, - "loss": 1.6691, - "step": 17432 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019410370019617128, - "loss": 1.7109, - "step": 17433 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019410336196983022, - "loss": 1.4438, - "step": 17434 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019410302374348913, - "loss": 1.5651, - "step": 17435 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019410268551714807, - "loss": 1.5336, - "step": 17436 - }, - { - "epoch": 0.06, - "learning_rate": 0.000194102347290807, - "loss": 1.5383, - "step": 17437 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019410200906446593, - "loss": 1.3539, - "step": 17438 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019410167083812487, - "loss": 1.6241, - "step": 17439 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019410133261178384, - "loss": 1.7396, - "step": 17440 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019410099438544275, - "loss": 1.5297, - "step": 17441 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001941006561591017, - "loss": 1.6102, - "step": 17442 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019410031793276063, - "loss": 1.6208, - "step": 17443 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019409997970641955, - "loss": 1.5112, - "step": 17444 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940996414800785, - "loss": 1.4862, - "step": 17445 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019409930325373743, - "loss": 1.5262, - "step": 17446 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019409896502739634, - "loss": 1.5127, - "step": 17447 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019409862680105528, - "loss": 1.616, - "step": 17448 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019409828857471422, - "loss": 1.5545, - "step": 17449 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019409795034837314, - "loss": 1.7367, - "step": 17450 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019409761212203208, - "loss": 1.5836, - "step": 17451 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019409727389569102, - "loss": 1.5824, - "step": 17452 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019409693566934994, - "loss": 1.6521, - "step": 17453 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019409659744300888, - "loss": 1.5049, - "step": 17454 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019409625921666782, - "loss": 1.4445, - "step": 17455 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019409592099032673, - "loss": 1.5899, - "step": 17456 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019409558276398567, - "loss": 1.4521, - "step": 17457 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019409524453764461, - "loss": 1.6518, - "step": 17458 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019409490631130353, - "loss": 1.5289, - "step": 17459 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019409456808496247, - "loss": 1.581, - "step": 17460 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940942298586214, - "loss": 1.5708, - "step": 17461 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019409389163228032, - "loss": 1.5568, - "step": 17462 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019409355340593927, - "loss": 1.7081, - "step": 17463 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940932151795982, - "loss": 1.4758, - "step": 17464 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019409287695325712, - "loss": 1.5458, - "step": 17465 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019409253872691606, - "loss": 1.7803, - "step": 17466 - }, - { - "epoch": 0.06, - "learning_rate": 0.000194092200500575, - "loss": 1.5193, - "step": 17467 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019409186227423392, - "loss": 1.6796, - "step": 17468 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019409152404789286, - "loss": 1.5441, - "step": 17469 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940911858215518, - "loss": 1.6696, - "step": 17470 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940908475952107, - "loss": 1.4767, - "step": 17471 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019409050936886965, - "loss": 1.568, - "step": 17472 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940901711425286, - "loss": 1.5038, - "step": 17473 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940898329161875, - "loss": 1.7494, - "step": 17474 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019408949468984645, - "loss": 1.6245, - "step": 17475 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940891564635054, - "loss": 1.605, - "step": 17476 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940888182371643, - "loss": 1.6926, - "step": 17477 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019408848001082325, - "loss": 1.5232, - "step": 17478 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940881417844822, - "loss": 1.4897, - "step": 17479 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940878035581411, - "loss": 1.4899, - "step": 17480 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019408746533180004, - "loss": 1.6526, - "step": 17481 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019408712710545898, - "loss": 1.5678, - "step": 17482 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940867888791179, - "loss": 1.5232, - "step": 17483 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019408645065277684, - "loss": 1.5124, - "step": 17484 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019408611242643578, - "loss": 1.5024, - "step": 17485 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940857742000947, - "loss": 1.4762, - "step": 17486 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019408543597375364, - "loss": 1.5396, - "step": 17487 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019408509774741258, - "loss": 1.6259, - "step": 17488 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940847595210715, - "loss": 1.6639, - "step": 17489 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019408442129473043, - "loss": 1.6269, - "step": 17490 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019408408306838937, - "loss": 1.6518, - "step": 17491 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940837448420483, - "loss": 1.5101, - "step": 17492 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019408340661570723, - "loss": 1.4032, - "step": 17493 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019408306838936617, - "loss": 1.5348, - "step": 17494 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019408273016302508, - "loss": 1.841, - "step": 17495 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019408239193668405, - "loss": 1.6082, - "step": 17496 - }, - { - "epoch": 0.06, - "learning_rate": 0.000194082053710343, - "loss": 1.5135, - "step": 17497 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940817154840019, - "loss": 1.8011, - "step": 17498 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019408137725766085, - "loss": 1.5647, - "step": 17499 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940810390313198, - "loss": 1.6814, - "step": 17500 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940807008049787, - "loss": 1.6254, - "step": 17501 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019408036257863764, - "loss": 1.5634, - "step": 17502 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019408002435229659, - "loss": 1.5777, - "step": 17503 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940796861259555, - "loss": 1.7187, - "step": 17504 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019407934789961444, - "loss": 1.6258, - "step": 17505 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019407900967327338, - "loss": 1.61, - "step": 17506 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940786714469323, - "loss": 1.5766, - "step": 17507 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019407833322059124, - "loss": 1.5173, - "step": 17508 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019407799499425018, - "loss": 1.5907, - "step": 17509 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940776567679091, - "loss": 1.7409, - "step": 17510 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019407731854156803, - "loss": 1.6067, - "step": 17511 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019407698031522697, - "loss": 1.618, - "step": 17512 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940766420888859, - "loss": 1.5063, - "step": 17513 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019407630386254483, - "loss": 1.6092, - "step": 17514 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019407596563620377, - "loss": 1.4098, - "step": 17515 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019407562740986268, - "loss": 1.7345, - "step": 17516 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019407528918352163, - "loss": 1.5133, - "step": 17517 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019407495095718057, - "loss": 1.6348, - "step": 17518 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019407461273083948, - "loss": 1.4697, - "step": 17519 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019407427450449842, - "loss": 1.5473, - "step": 17520 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019407393627815736, - "loss": 1.5547, - "step": 17521 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019407359805181628, - "loss": 1.571, - "step": 17522 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019407325982547522, - "loss": 1.4826, - "step": 17523 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019407292159913416, - "loss": 1.4897, - "step": 17524 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019407258337279307, - "loss": 1.4928, - "step": 17525 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019407224514645202, - "loss": 1.4724, - "step": 17526 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019407190692011096, - "loss": 1.5993, - "step": 17527 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019407156869376987, - "loss": 1.4925, - "step": 17528 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940712304674288, - "loss": 1.6041, - "step": 17529 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019407089224108775, - "loss": 1.5027, - "step": 17530 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019407055401474667, - "loss": 1.6949, - "step": 17531 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940702157884056, - "loss": 1.624, - "step": 17532 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019406987756206455, - "loss": 1.5046, - "step": 17533 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019406953933572346, - "loss": 1.5687, - "step": 17534 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940692011093824, - "loss": 1.6516, - "step": 17535 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019406886288304135, - "loss": 1.5304, - "step": 17536 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019406852465670026, - "loss": 1.5346, - "step": 17537 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940681864303592, - "loss": 1.62, - "step": 17538 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019406784820401814, - "loss": 1.6229, - "step": 17539 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019406750997767706, - "loss": 1.5623, - "step": 17540 - }, - { - "epoch": 0.06, - "learning_rate": 0.000194067171751336, - "loss": 1.5092, - "step": 17541 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019406683352499494, - "loss": 1.6952, - "step": 17542 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019406649529865385, - "loss": 1.627, - "step": 17543 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940661570723128, - "loss": 1.6431, - "step": 17544 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019406581884597173, - "loss": 1.6207, - "step": 17545 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019406548061963065, - "loss": 1.5405, - "step": 17546 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940651423932896, - "loss": 1.562, - "step": 17547 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019406480416694853, - "loss": 1.6017, - "step": 17548 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019406446594060744, - "loss": 1.5298, - "step": 17549 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019406412771426639, - "loss": 1.6819, - "step": 17550 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019406378948792533, - "loss": 1.6916, - "step": 17551 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019406345126158424, - "loss": 1.4923, - "step": 17552 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940631130352432, - "loss": 1.5525, - "step": 17553 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019406277480890215, - "loss": 1.6995, - "step": 17554 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019406243658256106, - "loss": 1.5467, - "step": 17555 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019406209835622, - "loss": 1.5571, - "step": 17556 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019406176012987895, - "loss": 1.5539, - "step": 17557 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019406142190353786, - "loss": 1.6594, - "step": 17558 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940610836771968, - "loss": 1.5957, - "step": 17559 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019406074545085574, - "loss": 1.5399, - "step": 17560 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019406040722451466, - "loss": 1.5535, - "step": 17561 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940600689981736, - "loss": 1.6493, - "step": 17562 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019405973077183254, - "loss": 1.721, - "step": 17563 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019405939254549145, - "loss": 1.6917, - "step": 17564 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940590543191504, - "loss": 1.7166, - "step": 17565 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019405871609280934, - "loss": 1.6644, - "step": 17566 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019405837786646825, - "loss": 1.5538, - "step": 17567 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940580396401272, - "loss": 1.4034, - "step": 17568 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019405770141378613, - "loss": 1.7624, - "step": 17569 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019405736318744505, - "loss": 1.6308, - "step": 17570 - }, - { - "epoch": 0.06, - "learning_rate": 0.000194057024961104, - "loss": 1.5206, - "step": 17571 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019405668673476293, - "loss": 1.6963, - "step": 17572 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019405634850842184, - "loss": 1.7281, - "step": 17573 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019405601028208078, - "loss": 1.6949, - "step": 17574 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019405567205573972, - "loss": 1.5833, - "step": 17575 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019405533382939864, - "loss": 1.5971, - "step": 17576 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019405499560305758, - "loss": 1.7048, - "step": 17577 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019405465737671652, - "loss": 1.598, - "step": 17578 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019405431915037543, - "loss": 1.5537, - "step": 17579 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019405398092403438, - "loss": 1.4851, - "step": 17580 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019405364269769332, - "loss": 1.6878, - "step": 17581 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019405330447135223, - "loss": 1.5258, - "step": 17582 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019405296624501117, - "loss": 1.6152, - "step": 17583 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940526280186701, - "loss": 1.6347, - "step": 17584 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019405228979232903, - "loss": 1.5324, - "step": 17585 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019405195156598797, - "loss": 1.6051, - "step": 17586 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940516133396469, - "loss": 1.494, - "step": 17587 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019405127511330582, - "loss": 1.5614, - "step": 17588 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019405093688696476, - "loss": 1.6094, - "step": 17589 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940505986606237, - "loss": 1.4783, - "step": 17590 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019405026043428262, - "loss": 1.4737, - "step": 17591 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019404992220794156, - "loss": 1.7323, - "step": 17592 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940495839816005, - "loss": 1.5558, - "step": 17593 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019404924575525942, - "loss": 1.5751, - "step": 17594 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019404890752891836, - "loss": 1.4922, - "step": 17595 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940485693025773, - "loss": 1.6563, - "step": 17596 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940482310762362, - "loss": 1.6403, - "step": 17597 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019404789284989515, - "loss": 1.5091, - "step": 17598 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940475546235541, - "loss": 1.7648, - "step": 17599 - }, - { - "epoch": 0.06, - "learning_rate": 0.000194047216397213, - "loss": 1.5553, - "step": 17600 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019404687817087195, - "loss": 1.5462, - "step": 17601 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940465399445309, - "loss": 1.5398, - "step": 17602 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940462017181898, - "loss": 1.5414, - "step": 17603 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019404586349184875, - "loss": 1.4234, - "step": 17604 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940455252655077, - "loss": 1.5185, - "step": 17605 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940451870391666, - "loss": 1.6682, - "step": 17606 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019404484881282554, - "loss": 1.5741, - "step": 17607 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019404451058648448, - "loss": 1.5723, - "step": 17608 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019404417236014342, - "loss": 1.6622, - "step": 17609 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019404383413380237, - "loss": 1.5871, - "step": 17610 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019404349590746128, - "loss": 1.43, - "step": 17611 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019404315768112022, - "loss": 1.4579, - "step": 17612 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019404281945477916, - "loss": 1.6525, - "step": 17613 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019404248122843808, - "loss": 1.5324, - "step": 17614 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019404214300209702, - "loss": 1.4349, - "step": 17615 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019404180477575596, - "loss": 1.6863, - "step": 17616 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019404146654941487, - "loss": 1.646, - "step": 17617 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019404112832307381, - "loss": 1.5967, - "step": 17618 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019404079009673275, - "loss": 1.4972, - "step": 17619 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019404045187039167, - "loss": 1.5756, - "step": 17620 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940401136440506, - "loss": 1.6869, - "step": 17621 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019403977541770955, - "loss": 1.6951, - "step": 17622 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019403943719136847, - "loss": 1.6612, - "step": 17623 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940390989650274, - "loss": 1.6614, - "step": 17624 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019403876073868635, - "loss": 1.4576, - "step": 17625 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019403842251234526, - "loss": 1.7653, - "step": 17626 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940380842860042, - "loss": 1.6639, - "step": 17627 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019403774605966314, - "loss": 1.7167, - "step": 17628 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019403740783332206, - "loss": 1.6623, - "step": 17629 - }, - { - "epoch": 0.06, - "learning_rate": 0.000194037069606981, - "loss": 1.633, - "step": 17630 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019403673138063994, - "loss": 1.5888, - "step": 17631 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019403639315429885, - "loss": 1.6272, - "step": 17632 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940360549279578, - "loss": 1.4365, - "step": 17633 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019403571670161674, - "loss": 1.7231, - "step": 17634 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019403537847527568, - "loss": 1.6387, - "step": 17635 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940350402489346, - "loss": 1.6366, - "step": 17636 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019403470202259353, - "loss": 1.6697, - "step": 17637 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019403436379625247, - "loss": 1.5947, - "step": 17638 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940340255699114, - "loss": 1.5719, - "step": 17639 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019403368734357033, - "loss": 1.5322, - "step": 17640 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019403334911722927, - "loss": 1.621, - "step": 17641 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019403301089088818, - "loss": 1.4437, - "step": 17642 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019403267266454713, - "loss": 1.6097, - "step": 17643 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019403233443820607, - "loss": 1.4318, - "step": 17644 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019403199621186498, - "loss": 1.763, - "step": 17645 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019403165798552392, - "loss": 1.4506, - "step": 17646 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019403131975918286, - "loss": 1.6791, - "step": 17647 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019403098153284178, - "loss": 1.6204, - "step": 17648 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019403064330650072, - "loss": 1.5969, - "step": 17649 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019403030508015966, - "loss": 1.6483, - "step": 17650 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019402996685381857, - "loss": 1.4151, - "step": 17651 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019402962862747751, - "loss": 1.5903, - "step": 17652 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019402929040113646, - "loss": 1.6106, - "step": 17653 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019402895217479537, - "loss": 1.421, - "step": 17654 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940286139484543, - "loss": 1.4662, - "step": 17655 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019402827572211325, - "loss": 1.6744, - "step": 17656 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019402793749577217, - "loss": 1.6886, - "step": 17657 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940275992694311, - "loss": 1.4783, - "step": 17658 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019402726104309005, - "loss": 1.6605, - "step": 17659 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019402692281674896, - "loss": 1.6646, - "step": 17660 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940265845904079, - "loss": 1.5889, - "step": 17661 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019402624636406684, - "loss": 1.6317, - "step": 17662 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019402590813772576, - "loss": 1.6079, - "step": 17663 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940255699113847, - "loss": 1.6271, - "step": 17664 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019402523168504364, - "loss": 1.5082, - "step": 17665 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019402489345870258, - "loss": 1.5833, - "step": 17666 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019402455523236152, - "loss": 1.473, - "step": 17667 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019402421700602044, - "loss": 1.7531, - "step": 17668 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019402387877967938, - "loss": 1.634, - "step": 17669 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019402354055333832, - "loss": 1.5868, - "step": 17670 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019402320232699723, - "loss": 1.5475, - "step": 17671 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019402286410065617, - "loss": 1.6252, - "step": 17672 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019402252587431512, - "loss": 1.6349, - "step": 17673 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019402218764797403, - "loss": 1.5715, - "step": 17674 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019402184942163297, - "loss": 1.7284, - "step": 17675 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940215111952919, - "loss": 1.6894, - "step": 17676 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019402117296895083, - "loss": 1.6085, - "step": 17677 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019402083474260977, - "loss": 1.6175, - "step": 17678 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940204965162687, - "loss": 1.6362, - "step": 17679 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019402015828992762, - "loss": 1.5493, - "step": 17680 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019401982006358656, - "loss": 1.5666, - "step": 17681 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940194818372455, - "loss": 1.5282, - "step": 17682 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019401914361090442, - "loss": 1.6082, - "step": 17683 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019401880538456336, - "loss": 1.7812, - "step": 17684 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940184671582223, - "loss": 1.6176, - "step": 17685 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019401812893188121, - "loss": 1.7539, - "step": 17686 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019401779070554016, - "loss": 1.475, - "step": 17687 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940174524791991, - "loss": 1.6389, - "step": 17688 - }, - { - "epoch": 0.06, - "learning_rate": 0.000194017114252858, - "loss": 1.6995, - "step": 17689 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019401677602651695, - "loss": 1.6291, - "step": 17690 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940164378001759, - "loss": 1.4092, - "step": 17691 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940160995738348, - "loss": 1.5405, - "step": 17692 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019401576134749375, - "loss": 1.7272, - "step": 17693 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940154231211527, - "loss": 1.4997, - "step": 17694 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940150848948116, - "loss": 1.6647, - "step": 17695 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019401474666847055, - "loss": 1.6969, - "step": 17696 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019401440844212949, - "loss": 1.6455, - "step": 17697 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940140702157884, - "loss": 1.564, - "step": 17698 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019401373198944734, - "loss": 1.5625, - "step": 17699 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019401339376310628, - "loss": 1.7446, - "step": 17700 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940130555367652, - "loss": 1.5917, - "step": 17701 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019401271731042414, - "loss": 1.6146, - "step": 17702 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019401237908408308, - "loss": 1.6779, - "step": 17703 - }, - { - "epoch": 0.06, - "learning_rate": 0.000194012040857742, - "loss": 1.6461, - "step": 17704 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019401170263140093, - "loss": 1.5723, - "step": 17705 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019401136440505988, - "loss": 1.5821, - "step": 17706 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940110261787188, - "loss": 1.5254, - "step": 17707 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019401068795237773, - "loss": 1.6009, - "step": 17708 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019401034972603667, - "loss": 1.6694, - "step": 17709 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019401001149969559, - "loss": 1.5525, - "step": 17710 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019400967327335453, - "loss": 1.4145, - "step": 17711 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019400933504701347, - "loss": 1.7036, - "step": 17712 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019400899682067238, - "loss": 1.6431, - "step": 17713 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019400865859433132, - "loss": 1.5201, - "step": 17714 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019400832036799026, - "loss": 1.5364, - "step": 17715 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019400798214164918, - "loss": 1.6291, - "step": 17716 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019400764391530812, - "loss": 1.4671, - "step": 17717 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019400730568896706, - "loss": 1.7594, - "step": 17718 - }, - { - "epoch": 0.06, - "learning_rate": 0.000194006967462626, - "loss": 1.6861, - "step": 17719 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019400662923628492, - "loss": 1.5984, - "step": 17720 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019400629100994388, - "loss": 1.7904, - "step": 17721 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940059527836028, - "loss": 1.6647, - "step": 17722 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019400561455726174, - "loss": 1.5877, - "step": 17723 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019400527633092068, - "loss": 1.5205, - "step": 17724 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940049381045796, - "loss": 1.7036, - "step": 17725 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019400459987823854, - "loss": 1.6242, - "step": 17726 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019400426165189748, - "loss": 1.6689, - "step": 17727 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940039234255564, - "loss": 1.5888, - "step": 17728 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019400358519921533, - "loss": 1.6203, - "step": 17729 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019400324697287427, - "loss": 1.7517, - "step": 17730 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001940029087465332, - "loss": 1.6052, - "step": 17731 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019400257052019213, - "loss": 1.696, - "step": 17732 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019400223229385107, - "loss": 1.4195, - "step": 17733 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019400189406750998, - "loss": 1.6884, - "step": 17734 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019400155584116892, - "loss": 1.5818, - "step": 17735 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019400121761482787, - "loss": 1.5487, - "step": 17736 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019400087938848678, - "loss": 1.6102, - "step": 17737 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019400054116214572, - "loss": 1.6214, - "step": 17738 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019400020293580466, - "loss": 1.5615, - "step": 17739 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019399986470946358, - "loss": 1.6858, - "step": 17740 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019399952648312252, - "loss": 1.4952, - "step": 17741 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019399918825678146, - "loss": 1.7477, - "step": 17742 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019399885003044037, - "loss": 1.6726, - "step": 17743 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939985118040993, - "loss": 1.5119, - "step": 17744 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019399817357775825, - "loss": 1.7345, - "step": 17745 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019399783535141717, - "loss": 1.5768, - "step": 17746 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939974971250761, - "loss": 1.5881, - "step": 17747 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019399715889873505, - "loss": 1.7034, - "step": 17748 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019399682067239396, - "loss": 1.6401, - "step": 17749 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939964824460529, - "loss": 1.5983, - "step": 17750 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019399614421971185, - "loss": 1.5238, - "step": 17751 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019399580599337076, - "loss": 1.5963, - "step": 17752 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939954677670297, - "loss": 1.528, - "step": 17753 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019399512954068864, - "loss": 1.6225, - "step": 17754 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019399479131434756, - "loss": 1.7014, - "step": 17755 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939944530880065, - "loss": 1.5191, - "step": 17756 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019399411486166544, - "loss": 1.5894, - "step": 17757 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019399377663532435, - "loss": 1.659, - "step": 17758 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939934384089833, - "loss": 1.5543, - "step": 17759 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019399310018264224, - "loss": 1.6101, - "step": 17760 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019399276195630115, - "loss": 1.8009, - "step": 17761 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939924237299601, - "loss": 1.5584, - "step": 17762 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019399208550361903, - "loss": 1.533, - "step": 17763 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019399174727727795, - "loss": 1.5264, - "step": 17764 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939914090509369, - "loss": 1.6975, - "step": 17765 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019399107082459583, - "loss": 1.562, - "step": 17766 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019399073259825474, - "loss": 1.4639, - "step": 17767 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019399039437191368, - "loss": 1.5419, - "step": 17768 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019399005614557262, - "loss": 1.4527, - "step": 17769 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019398971791923154, - "loss": 1.6977, - "step": 17770 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019398937969289048, - "loss": 1.425, - "step": 17771 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019398904146654942, - "loss": 1.7159, - "step": 17772 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019398870324020834, - "loss": 1.559, - "step": 17773 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019398836501386728, - "loss": 1.3949, - "step": 17774 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019398802678752622, - "loss": 1.4996, - "step": 17775 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019398768856118513, - "loss": 1.6863, - "step": 17776 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939873503348441, - "loss": 1.4598, - "step": 17777 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019398701210850304, - "loss": 1.702, - "step": 17778 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019398667388216195, - "loss": 1.6441, - "step": 17779 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939863356558209, - "loss": 1.4935, - "step": 17780 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019398599742947984, - "loss": 1.4437, - "step": 17781 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019398565920313875, - "loss": 1.5146, - "step": 17782 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939853209767977, - "loss": 1.6738, - "step": 17783 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019398498275045663, - "loss": 1.6381, - "step": 17784 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019398464452411555, - "loss": 1.7264, - "step": 17785 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939843062977745, - "loss": 1.5492, - "step": 17786 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019398396807143343, - "loss": 1.5563, - "step": 17787 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019398362984509234, - "loss": 1.5517, - "step": 17788 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019398329161875128, - "loss": 1.4056, - "step": 17789 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019398295339241023, - "loss": 1.5551, - "step": 17790 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019398261516606914, - "loss": 1.5647, - "step": 17791 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019398227693972808, - "loss": 1.7279, - "step": 17792 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019398193871338702, - "loss": 1.6227, - "step": 17793 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019398160048704594, - "loss": 1.654, - "step": 17794 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019398126226070488, - "loss": 1.7117, - "step": 17795 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019398092403436382, - "loss": 1.5438, - "step": 17796 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019398058580802273, - "loss": 1.5397, - "step": 17797 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019398024758168167, - "loss": 1.5399, - "step": 17798 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019397990935534062, - "loss": 1.5825, - "step": 17799 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019397957112899953, - "loss": 1.66, - "step": 17800 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019397923290265847, - "loss": 1.6747, - "step": 17801 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939788946763174, - "loss": 1.6614, - "step": 17802 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019397855644997633, - "loss": 1.4655, - "step": 17803 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019397821822363527, - "loss": 1.3461, - "step": 17804 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939778799972942, - "loss": 1.5994, - "step": 17805 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019397754177095312, - "loss": 1.4482, - "step": 17806 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019397720354461206, - "loss": 1.5344, - "step": 17807 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193976865318271, - "loss": 1.6176, - "step": 17808 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019397652709192992, - "loss": 1.5411, - "step": 17809 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019397618886558886, - "loss": 1.5322, - "step": 17810 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939758506392478, - "loss": 1.6115, - "step": 17811 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019397551241290671, - "loss": 1.6721, - "step": 17812 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019397517418656566, - "loss": 1.6535, - "step": 17813 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939748359602246, - "loss": 1.5946, - "step": 17814 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939744977338835, - "loss": 1.5448, - "step": 17815 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019397415950754245, - "loss": 1.5923, - "step": 17816 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939738212812014, - "loss": 1.6178, - "step": 17817 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939734830548603, - "loss": 1.562, - "step": 17818 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019397314482851925, - "loss": 1.6331, - "step": 17819 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939728066021782, - "loss": 1.7346, - "step": 17820 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939724683758371, - "loss": 1.5755, - "step": 17821 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019397213014949604, - "loss": 1.5364, - "step": 17822 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019397179192315499, - "loss": 1.7146, - "step": 17823 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939714536968139, - "loss": 1.6343, - "step": 17824 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019397111547047284, - "loss": 1.5419, - "step": 17825 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019397077724413178, - "loss": 1.6401, - "step": 17826 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939704390177907, - "loss": 1.5436, - "step": 17827 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019397010079144964, - "loss": 1.5005, - "step": 17828 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019396976256510858, - "loss": 1.3596, - "step": 17829 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939694243387675, - "loss": 1.5474, - "step": 17830 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019396908611242643, - "loss": 1.5287, - "step": 17831 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019396874788608537, - "loss": 1.6482, - "step": 17832 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019396840965974432, - "loss": 1.6367, - "step": 17833 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019396807143340326, - "loss": 1.5407, - "step": 17834 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939677332070622, - "loss": 1.5527, - "step": 17835 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939673949807211, - "loss": 1.693, - "step": 17836 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019396705675438005, - "loss": 1.5743, - "step": 17837 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193966718528039, - "loss": 1.4574, - "step": 17838 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939663803016979, - "loss": 1.5294, - "step": 17839 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019396604207535685, - "loss": 1.6502, - "step": 17840 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939657038490158, - "loss": 1.608, - "step": 17841 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939653656226747, - "loss": 1.6692, - "step": 17842 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019396502739633365, - "loss": 1.7553, - "step": 17843 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939646891699926, - "loss": 1.5936, - "step": 17844 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939643509436515, - "loss": 1.5889, - "step": 17845 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019396401271731044, - "loss": 1.6479, - "step": 17846 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019396367449096938, - "loss": 1.4021, - "step": 17847 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939633362646283, - "loss": 1.6471, - "step": 17848 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019396299803828724, - "loss": 1.514, - "step": 17849 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019396265981194618, - "loss": 1.6389, - "step": 17850 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939623215856051, - "loss": 1.5935, - "step": 17851 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019396198335926403, - "loss": 1.6617, - "step": 17852 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019396164513292298, - "loss": 1.5927, - "step": 17853 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939613069065819, - "loss": 1.5646, - "step": 17854 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019396096868024083, - "loss": 1.7331, - "step": 17855 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019396063045389977, - "loss": 1.3999, - "step": 17856 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019396029222755869, - "loss": 1.5582, - "step": 17857 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019395995400121763, - "loss": 1.6178, - "step": 17858 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019395961577487657, - "loss": 1.5292, - "step": 17859 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019395927754853548, - "loss": 1.6064, - "step": 17860 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019395893932219442, - "loss": 1.6037, - "step": 17861 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019395860109585336, - "loss": 1.7974, - "step": 17862 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019395826286951228, - "loss": 1.6497, - "step": 17863 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019395792464317122, - "loss": 1.6638, - "step": 17864 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019395758641683016, - "loss": 1.6006, - "step": 17865 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019395724819048908, - "loss": 1.5509, - "step": 17866 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019395690996414802, - "loss": 1.4873, - "step": 17867 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019395657173780696, - "loss": 1.5038, - "step": 17868 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019395623351146587, - "loss": 1.6556, - "step": 17869 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939558952851248, - "loss": 1.6365, - "step": 17870 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019395555705878375, - "loss": 1.5273, - "step": 17871 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019395521883244267, - "loss": 1.5834, - "step": 17872 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939548806061016, - "loss": 1.7373, - "step": 17873 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019395454237976055, - "loss": 1.677, - "step": 17874 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019395420415341946, - "loss": 1.5082, - "step": 17875 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939538659270784, - "loss": 1.5259, - "step": 17876 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019395352770073735, - "loss": 1.6392, - "step": 17877 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019395318947439626, - "loss": 1.5558, - "step": 17878 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939528512480552, - "loss": 1.6116, - "step": 17879 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019395251302171414, - "loss": 1.6098, - "step": 17880 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019395217479537306, - "loss": 1.5658, - "step": 17881 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193951836569032, - "loss": 1.4675, - "step": 17882 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019395149834269094, - "loss": 1.5364, - "step": 17883 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019395116011634985, - "loss": 1.7583, - "step": 17884 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939508218900088, - "loss": 1.6494, - "step": 17885 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019395048366366774, - "loss": 1.5915, - "step": 17886 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019395014543732665, - "loss": 1.5753, - "step": 17887 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939498072109856, - "loss": 1.5672, - "step": 17888 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019394946898464453, - "loss": 1.5596, - "step": 17889 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019394913075830347, - "loss": 1.5459, - "step": 17890 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019394879253196241, - "loss": 1.5806, - "step": 17891 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019394845430562133, - "loss": 1.567, - "step": 17892 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019394811607928027, - "loss": 1.5248, - "step": 17893 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939477778529392, - "loss": 1.7647, - "step": 17894 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019394743962659812, - "loss": 1.545, - "step": 17895 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019394710140025707, - "loss": 1.5687, - "step": 17896 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193946763173916, - "loss": 1.5428, - "step": 17897 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019394642494757492, - "loss": 1.586, - "step": 17898 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019394608672123386, - "loss": 1.5062, - "step": 17899 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939457484948928, - "loss": 1.5473, - "step": 17900 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019394541026855174, - "loss": 1.7191, - "step": 17901 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019394507204221066, - "loss": 1.6925, - "step": 17902 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939447338158696, - "loss": 1.713, - "step": 17903 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019394439558952854, - "loss": 1.5185, - "step": 17904 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019394405736318745, - "loss": 1.6377, - "step": 17905 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939437191368464, - "loss": 1.5325, - "step": 17906 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019394338091050534, - "loss": 1.5839, - "step": 17907 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019394304268416425, - "loss": 1.6154, - "step": 17908 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939427044578232, - "loss": 1.622, - "step": 17909 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019394236623148213, - "loss": 1.5465, - "step": 17910 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019394202800514105, - "loss": 1.6605, - "step": 17911 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939416897788, - "loss": 1.7291, - "step": 17912 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019394135155245893, - "loss": 1.5805, - "step": 17913 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019394101332611784, - "loss": 1.4775, - "step": 17914 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019394067509977678, - "loss": 1.6248, - "step": 17915 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019394033687343573, - "loss": 1.5851, - "step": 17916 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019393999864709464, - "loss": 1.4972, - "step": 17917 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019393966042075358, - "loss": 1.4262, - "step": 17918 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019393932219441252, - "loss": 1.5463, - "step": 17919 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019393898396807144, - "loss": 1.5719, - "step": 17920 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019393864574173038, - "loss": 1.6305, - "step": 17921 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019393830751538932, - "loss": 1.5357, - "step": 17922 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019393796928904823, - "loss": 1.8155, - "step": 17923 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019393763106270717, - "loss": 1.6233, - "step": 17924 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019393729283636611, - "loss": 1.5025, - "step": 17925 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019393695461002503, - "loss": 1.6174, - "step": 17926 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019393661638368397, - "loss": 1.6064, - "step": 17927 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939362781573429, - "loss": 1.5282, - "step": 17928 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019393593993100182, - "loss": 1.6849, - "step": 17929 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019393560170466077, - "loss": 1.4096, - "step": 17930 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939352634783197, - "loss": 1.5761, - "step": 17931 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019393492525197862, - "loss": 1.707, - "step": 17932 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019393458702563756, - "loss": 1.6437, - "step": 17933 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939342487992965, - "loss": 1.5899, - "step": 17934 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019393391057295542, - "loss": 1.5495, - "step": 17935 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019393357234661436, - "loss": 1.5205, - "step": 17936 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939332341202733, - "loss": 1.5606, - "step": 17937 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019393289589393221, - "loss": 1.5136, - "step": 17938 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019393255766759115, - "loss": 1.6567, - "step": 17939 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939322194412501, - "loss": 1.6753, - "step": 17940 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193931881214909, - "loss": 1.5967, - "step": 17941 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019393154298856795, - "loss": 1.5859, - "step": 17942 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939312047622269, - "loss": 1.6039, - "step": 17943 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939308665358858, - "loss": 1.5271, - "step": 17944 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019393052830954475, - "loss": 1.4941, - "step": 17945 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939301900832037, - "loss": 1.5692, - "step": 17946 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019392985185686263, - "loss": 1.7606, - "step": 17947 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019392951363052157, - "loss": 1.4383, - "step": 17948 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019392917540418048, - "loss": 1.5093, - "step": 17949 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019392883717783943, - "loss": 1.5064, - "step": 17950 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019392849895149837, - "loss": 1.7421, - "step": 17951 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019392816072515728, - "loss": 1.6336, - "step": 17952 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019392782249881622, - "loss": 1.6338, - "step": 17953 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019392748427247516, - "loss": 1.5433, - "step": 17954 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019392714604613408, - "loss": 1.5052, - "step": 17955 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019392680781979302, - "loss": 1.5807, - "step": 17956 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019392646959345196, - "loss": 1.6615, - "step": 17957 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019392613136711087, - "loss": 1.6511, - "step": 17958 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019392579314076982, - "loss": 1.5089, - "step": 17959 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019392545491442876, - "loss": 1.549, - "step": 17960 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019392511668808767, - "loss": 1.5455, - "step": 17961 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939247784617466, - "loss": 1.6534, - "step": 17962 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019392444023540555, - "loss": 1.6514, - "step": 17963 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019392410200906447, - "loss": 1.6476, - "step": 17964 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939237637827234, - "loss": 1.5934, - "step": 17965 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019392342555638235, - "loss": 1.5122, - "step": 17966 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019392308733004126, - "loss": 1.6306, - "step": 17967 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939227491037002, - "loss": 1.5651, - "step": 17968 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019392241087735915, - "loss": 1.4797, - "step": 17969 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019392207265101806, - "loss": 1.6441, - "step": 17970 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193921734424677, - "loss": 1.5599, - "step": 17971 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019392139619833594, - "loss": 1.4908, - "step": 17972 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019392105797199486, - "loss": 1.5154, - "step": 17973 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939207197456538, - "loss": 1.4505, - "step": 17974 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019392038151931274, - "loss": 1.6015, - "step": 17975 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019392004329297165, - "loss": 1.691, - "step": 17976 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939197050666306, - "loss": 1.7125, - "step": 17977 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019391936684028953, - "loss": 1.5828, - "step": 17978 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019391902861394845, - "loss": 1.6826, - "step": 17979 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939186903876074, - "loss": 1.7325, - "step": 17980 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019391835216126633, - "loss": 1.6101, - "step": 17981 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019391801393492524, - "loss": 1.5936, - "step": 17982 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019391767570858419, - "loss": 1.6737, - "step": 17983 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019391733748224313, - "loss": 1.57, - "step": 17984 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019391699925590207, - "loss": 1.7077, - "step": 17985 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019391666102956098, - "loss": 1.5454, - "step": 17986 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019391632280321992, - "loss": 1.6588, - "step": 17987 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019391598457687886, - "loss": 1.4341, - "step": 17988 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019391564635053778, - "loss": 1.6197, - "step": 17989 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019391530812419672, - "loss": 1.7367, - "step": 17990 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019391496989785566, - "loss": 1.6019, - "step": 17991 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019391463167151457, - "loss": 1.7131, - "step": 17992 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019391429344517352, - "loss": 1.4819, - "step": 17993 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019391395521883246, - "loss": 1.7813, - "step": 17994 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019391361699249137, - "loss": 1.5169, - "step": 17995 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939132787661503, - "loss": 1.5997, - "step": 17996 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019391294053980925, - "loss": 1.5216, - "step": 17997 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019391260231346817, - "loss": 1.6485, - "step": 17998 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939122640871271, - "loss": 1.5014, - "step": 17999 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019391192586078605, - "loss": 1.6496, - "step": 18000 - }, - { - "epoch": 0.06, - "eval_loss": 1.3349797175480769, - "mrc_em": 19.0, - "mrc_f1": 30.28095238095238, - "nli_acc": 0.27, - "step": 18000, - "sts_f1": 0.43076923076923074, - "ynat_f1": 0.6547512944571768 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019391158763444496, - "loss": 1.5442, - "step": 18001 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019391124940810393, - "loss": 1.7681, - "step": 18002 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019391091118176285, - "loss": 1.755, - "step": 18003 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939105729554218, - "loss": 1.5564, - "step": 18004 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019391023472908073, - "loss": 1.7157, - "step": 18005 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019390989650273964, - "loss": 1.6283, - "step": 18006 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019390955827639858, - "loss": 1.5422, - "step": 18007 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019390922005005752, - "loss": 1.6238, - "step": 18008 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019390888182371644, - "loss": 1.5878, - "step": 18009 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019390854359737538, - "loss": 1.5797, - "step": 18010 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019390820537103432, - "loss": 1.7267, - "step": 18011 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019390786714469323, - "loss": 1.4404, - "step": 18012 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019390752891835218, - "loss": 1.6793, - "step": 18013 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019390719069201112, - "loss": 1.6018, - "step": 18014 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019390685246567003, - "loss": 1.7375, - "step": 18015 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019390651423932897, - "loss": 1.5472, - "step": 18016 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939061760129879, - "loss": 1.5427, - "step": 18017 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019390583778664683, - "loss": 1.6274, - "step": 18018 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019390549956030577, - "loss": 1.7356, - "step": 18019 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939051613339647, - "loss": 1.4662, - "step": 18020 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019390482310762362, - "loss": 1.523, - "step": 18021 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019390448488128256, - "loss": 1.4645, - "step": 18022 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939041466549415, - "loss": 1.5845, - "step": 18023 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019390380842860042, - "loss": 1.6746, - "step": 18024 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019390347020225936, - "loss": 1.4783, - "step": 18025 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939031319759183, - "loss": 1.5356, - "step": 18026 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019390279374957722, - "loss": 1.4879, - "step": 18027 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019390245552323616, - "loss": 1.6381, - "step": 18028 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939021172968951, - "loss": 1.4975, - "step": 18029 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193901779070554, - "loss": 1.6724, - "step": 18030 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019390144084421295, - "loss": 1.6707, - "step": 18031 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939011026178719, - "loss": 1.5478, - "step": 18032 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939007643915308, - "loss": 1.5045, - "step": 18033 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019390042616518975, - "loss": 1.5801, - "step": 18034 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001939000879388487, - "loss": 1.5082, - "step": 18035 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938997497125076, - "loss": 1.5655, - "step": 18036 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019389941148616655, - "loss": 1.6928, - "step": 18037 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938990732598255, - "loss": 1.4738, - "step": 18038 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938987350334844, - "loss": 1.7056, - "step": 18039 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019389839680714334, - "loss": 1.5413, - "step": 18040 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019389805858080228, - "loss": 1.699, - "step": 18041 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938977203544612, - "loss": 1.6614, - "step": 18042 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019389738212812014, - "loss": 1.4833, - "step": 18043 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019389704390177908, - "loss": 1.7162, - "step": 18044 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193896705675438, - "loss": 1.5479, - "step": 18045 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019389636744909694, - "loss": 1.6574, - "step": 18046 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019389602922275588, - "loss": 1.579, - "step": 18047 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938956909964148, - "loss": 1.4775, - "step": 18048 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019389535277007373, - "loss": 1.4644, - "step": 18049 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019389501454373267, - "loss": 1.4806, - "step": 18050 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938946763173916, - "loss": 1.6528, - "step": 18051 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019389433809105053, - "loss": 1.5698, - "step": 18052 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019389399986470947, - "loss": 1.6056, - "step": 18053 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019389366163836838, - "loss": 1.454, - "step": 18054 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019389332341202732, - "loss": 1.5487, - "step": 18055 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019389298518568627, - "loss": 1.5723, - "step": 18056 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019389264695934518, - "loss": 1.635, - "step": 18057 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019389230873300415, - "loss": 1.6447, - "step": 18058 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938919705066631, - "loss": 1.7487, - "step": 18059 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193891632280322, - "loss": 1.7753, - "step": 18060 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019389129405398094, - "loss": 1.6445, - "step": 18061 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019389095582763989, - "loss": 1.6549, - "step": 18062 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938906176012988, - "loss": 1.643, - "step": 18063 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019389027937495774, - "loss": 1.6386, - "step": 18064 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019388994114861668, - "loss": 1.5789, - "step": 18065 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938896029222756, - "loss": 1.6205, - "step": 18066 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019388926469593454, - "loss": 1.5115, - "step": 18067 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019388892646959348, - "loss": 1.596, - "step": 18068 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938885882432524, - "loss": 1.7098, - "step": 18069 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019388825001691133, - "loss": 1.6694, - "step": 18070 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019388791179057027, - "loss": 1.4984, - "step": 18071 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938875735642292, - "loss": 1.5203, - "step": 18072 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019388723533788813, - "loss": 1.5442, - "step": 18073 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019388689711154707, - "loss": 1.4944, - "step": 18074 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019388655888520598, - "loss": 1.5835, - "step": 18075 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019388622065886493, - "loss": 1.7015, - "step": 18076 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019388588243252387, - "loss": 1.6112, - "step": 18077 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019388554420618278, - "loss": 1.4918, - "step": 18078 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019388520597984172, - "loss": 1.4036, - "step": 18079 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019388486775350066, - "loss": 1.4775, - "step": 18080 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019388452952715958, - "loss": 1.513, - "step": 18081 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019388419130081852, - "loss": 1.4977, - "step": 18082 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019388385307447746, - "loss": 1.6791, - "step": 18083 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019388351484813637, - "loss": 1.423, - "step": 18084 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019388317662179531, - "loss": 1.6813, - "step": 18085 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019388283839545426, - "loss": 1.5074, - "step": 18086 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019388250016911317, - "loss": 1.5607, - "step": 18087 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938821619427721, - "loss": 1.6029, - "step": 18088 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019388182371643105, - "loss": 1.9022, - "step": 18089 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019388148549008997, - "loss": 1.5759, - "step": 18090 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938811472637489, - "loss": 1.4678, - "step": 18091 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019388080903740785, - "loss": 1.4984, - "step": 18092 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019388047081106676, - "loss": 1.7362, - "step": 18093 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938801325847257, - "loss": 1.7931, - "step": 18094 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019387979435838464, - "loss": 1.4194, - "step": 18095 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019387945613204356, - "loss": 1.5695, - "step": 18096 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938791179057025, - "loss": 1.8031, - "step": 18097 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019387877967936144, - "loss": 1.5736, - "step": 18098 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019387844145302035, - "loss": 1.4214, - "step": 18099 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938781032266793, - "loss": 1.5297, - "step": 18100 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019387776500033824, - "loss": 1.5485, - "step": 18101 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019387742677399715, - "loss": 1.6476, - "step": 18102 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938770885476561, - "loss": 1.8248, - "step": 18103 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019387675032131503, - "loss": 1.6603, - "step": 18104 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019387641209497395, - "loss": 1.5818, - "step": 18105 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938760738686329, - "loss": 1.5458, - "step": 18106 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019387573564229183, - "loss": 1.7281, - "step": 18107 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019387539741595074, - "loss": 1.5054, - "step": 18108 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019387505918960968, - "loss": 1.4886, - "step": 18109 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019387472096326863, - "loss": 1.758, - "step": 18110 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019387438273692754, - "loss": 1.4695, - "step": 18111 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019387404451058648, - "loss": 1.6104, - "step": 18112 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019387370628424542, - "loss": 1.5497, - "step": 18113 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019387336805790436, - "loss": 1.6932, - "step": 18114 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938730298315633, - "loss": 1.5678, - "step": 18115 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019387269160522225, - "loss": 1.6222, - "step": 18116 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019387235337888116, - "loss": 1.5375, - "step": 18117 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938720151525401, - "loss": 1.6823, - "step": 18118 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019387167692619904, - "loss": 1.584, - "step": 18119 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019387133869985796, - "loss": 1.5229, - "step": 18120 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938710004735169, - "loss": 1.6872, - "step": 18121 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019387066224717584, - "loss": 1.5895, - "step": 18122 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019387032402083475, - "loss": 1.58, - "step": 18123 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938699857944937, - "loss": 1.5424, - "step": 18124 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019386964756815263, - "loss": 1.6401, - "step": 18125 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019386930934181155, - "loss": 1.4269, - "step": 18126 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938689711154705, - "loss": 1.5553, - "step": 18127 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019386863288912943, - "loss": 1.5045, - "step": 18128 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019386829466278835, - "loss": 1.5798, - "step": 18129 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019386795643644729, - "loss": 1.4989, - "step": 18130 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019386761821010623, - "loss": 1.567, - "step": 18131 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019386727998376514, - "loss": 1.6346, - "step": 18132 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019386694175742408, - "loss": 1.4403, - "step": 18133 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019386660353108302, - "loss": 1.4659, - "step": 18134 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019386626530474194, - "loss": 1.6532, - "step": 18135 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019386592707840088, - "loss": 1.6148, - "step": 18136 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019386558885205982, - "loss": 1.5423, - "step": 18137 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019386525062571873, - "loss": 1.484, - "step": 18138 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019386491239937768, - "loss": 1.6473, - "step": 18139 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019386457417303662, - "loss": 1.7471, - "step": 18140 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019386423594669553, - "loss": 1.7523, - "step": 18141 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019386389772035447, - "loss": 1.6482, - "step": 18142 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938635594940134, - "loss": 1.5656, - "step": 18143 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019386322126767233, - "loss": 1.606, - "step": 18144 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019386288304133127, - "loss": 1.5027, - "step": 18145 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938625448149902, - "loss": 1.5122, - "step": 18146 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019386220658864912, - "loss": 1.5223, - "step": 18147 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019386186836230806, - "loss": 1.7303, - "step": 18148 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193861530135967, - "loss": 1.357, - "step": 18149 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019386119190962592, - "loss": 1.4734, - "step": 18150 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019386085368328486, - "loss": 1.5542, - "step": 18151 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938605154569438, - "loss": 1.5038, - "step": 18152 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019386017723060272, - "loss": 1.5278, - "step": 18153 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019385983900426166, - "loss": 1.7059, - "step": 18154 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938595007779206, - "loss": 1.5556, - "step": 18155 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938591625515795, - "loss": 1.6597, - "step": 18156 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019385882432523845, - "loss": 1.6444, - "step": 18157 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938584860988974, - "loss": 1.5726, - "step": 18158 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938581478725563, - "loss": 1.5323, - "step": 18159 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019385780964621525, - "loss": 1.61, - "step": 18160 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938574714198742, - "loss": 1.6053, - "step": 18161 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938571331935331, - "loss": 1.6022, - "step": 18162 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019385679496719205, - "loss": 1.6731, - "step": 18163 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193856456740851, - "loss": 1.6159, - "step": 18164 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938561185145099, - "loss": 1.5952, - "step": 18165 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019385578028816884, - "loss": 1.4902, - "step": 18166 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019385544206182778, - "loss": 1.6028, - "step": 18167 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938551038354867, - "loss": 1.6821, - "step": 18168 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019385476560914564, - "loss": 1.7038, - "step": 18169 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938544273828046, - "loss": 1.6164, - "step": 18170 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019385408915646352, - "loss": 1.8118, - "step": 18171 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019385375093012246, - "loss": 1.5657, - "step": 18172 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938534127037814, - "loss": 1.532, - "step": 18173 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019385307447744032, - "loss": 1.5797, - "step": 18174 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019385273625109926, - "loss": 1.6467, - "step": 18175 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938523980247582, - "loss": 1.653, - "step": 18176 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938520597984171, - "loss": 1.5438, - "step": 18177 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019385172157207605, - "loss": 1.5114, - "step": 18178 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193851383345735, - "loss": 1.5647, - "step": 18179 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938510451193939, - "loss": 1.485, - "step": 18180 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019385070689305285, - "loss": 1.5185, - "step": 18181 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938503686667118, - "loss": 1.5271, - "step": 18182 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938500304403707, - "loss": 1.5402, - "step": 18183 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019384969221402965, - "loss": 1.5659, - "step": 18184 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938493539876886, - "loss": 1.7185, - "step": 18185 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938490157613475, - "loss": 1.625, - "step": 18186 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019384867753500644, - "loss": 1.6545, - "step": 18187 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019384833930866538, - "loss": 1.5813, - "step": 18188 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938480010823243, - "loss": 1.7566, - "step": 18189 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019384766285598324, - "loss": 1.6111, - "step": 18190 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019384732462964218, - "loss": 1.4982, - "step": 18191 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938469864033011, - "loss": 1.6506, - "step": 18192 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019384664817696004, - "loss": 1.5512, - "step": 18193 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019384630995061898, - "loss": 1.6599, - "step": 18194 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938459717242779, - "loss": 1.6418, - "step": 18195 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019384563349793683, - "loss": 1.5989, - "step": 18196 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019384529527159577, - "loss": 1.5987, - "step": 18197 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938449570452547, - "loss": 1.6678, - "step": 18198 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019384461881891363, - "loss": 1.6734, - "step": 18199 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019384428059257257, - "loss": 1.5727, - "step": 18200 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019384394236623148, - "loss": 1.4664, - "step": 18201 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019384360413989042, - "loss": 1.5015, - "step": 18202 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019384326591354937, - "loss": 1.5481, - "step": 18203 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019384292768720828, - "loss": 1.6026, - "step": 18204 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019384258946086722, - "loss": 1.6205, - "step": 18205 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019384225123452616, - "loss": 1.5768, - "step": 18206 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019384191300818508, - "loss": 1.6867, - "step": 18207 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019384157478184402, - "loss": 1.6803, - "step": 18208 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019384123655550296, - "loss": 1.5822, - "step": 18209 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019384089832916187, - "loss": 1.4656, - "step": 18210 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019384056010282081, - "loss": 1.5992, - "step": 18211 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019384022187647975, - "loss": 1.4456, - "step": 18212 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019383988365013867, - "loss": 1.6319, - "step": 18213 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938395454237976, - "loss": 1.5051, - "step": 18214 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019383920719745655, - "loss": 1.6037, - "step": 18215 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019383886897111547, - "loss": 1.6421, - "step": 18216 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938385307447744, - "loss": 1.5973, - "step": 18217 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019383819251843335, - "loss": 1.6018, - "step": 18218 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019383785429209226, - "loss": 1.5888, - "step": 18219 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938375160657512, - "loss": 1.611, - "step": 18220 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019383717783941014, - "loss": 1.6209, - "step": 18221 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019383683961306906, - "loss": 1.5165, - "step": 18222 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193836501386728, - "loss": 1.5899, - "step": 18223 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019383616316038694, - "loss": 1.8662, - "step": 18224 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019383582493404585, - "loss": 1.7344, - "step": 18225 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938354867077048, - "loss": 1.4696, - "step": 18226 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019383514848136374, - "loss": 1.5099, - "step": 18227 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019383481025502268, - "loss": 1.5956, - "step": 18228 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019383447202868162, - "loss": 1.6153, - "step": 18229 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019383413380234053, - "loss": 1.5754, - "step": 18230 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019383379557599947, - "loss": 1.4506, - "step": 18231 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019383345734965842, - "loss": 1.4907, - "step": 18232 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019383311912331733, - "loss": 1.7216, - "step": 18233 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019383278089697627, - "loss": 1.6172, - "step": 18234 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938324426706352, - "loss": 1.5718, - "step": 18235 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019383210444429413, - "loss": 1.6337, - "step": 18236 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019383176621795307, - "loss": 1.6059, - "step": 18237 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193831427991612, - "loss": 1.7194, - "step": 18238 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019383108976527092, - "loss": 1.6278, - "step": 18239 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019383075153892986, - "loss": 1.5698, - "step": 18240 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938304133125888, - "loss": 1.6743, - "step": 18241 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019383007508624772, - "loss": 1.7422, - "step": 18242 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019382973685990666, - "loss": 1.6243, - "step": 18243 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938293986335656, - "loss": 1.6422, - "step": 18244 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019382906040722451, - "loss": 1.5924, - "step": 18245 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019382872218088346, - "loss": 1.5155, - "step": 18246 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938283839545424, - "loss": 1.6067, - "step": 18247 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938280457282013, - "loss": 1.7819, - "step": 18248 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019382770750186025, - "loss": 1.6261, - "step": 18249 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938273692755192, - "loss": 1.7095, - "step": 18250 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938270310491781, - "loss": 1.7267, - "step": 18251 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019382669282283705, - "loss": 1.4394, - "step": 18252 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193826354596496, - "loss": 1.5776, - "step": 18253 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019382601637015493, - "loss": 1.606, - "step": 18254 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019382567814381384, - "loss": 1.593, - "step": 18255 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019382533991747279, - "loss": 1.4573, - "step": 18256 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019382500169113173, - "loss": 1.566, - "step": 18257 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019382466346479064, - "loss": 1.6016, - "step": 18258 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019382432523844958, - "loss": 1.6694, - "step": 18259 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019382398701210852, - "loss": 1.624, - "step": 18260 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019382364878576744, - "loss": 1.5025, - "step": 18261 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019382331055942638, - "loss": 1.5357, - "step": 18262 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019382297233308532, - "loss": 1.4927, - "step": 18263 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019382263410674423, - "loss": 1.6789, - "step": 18264 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019382229588040317, - "loss": 1.7266, - "step": 18265 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019382195765406212, - "loss": 1.5978, - "step": 18266 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019382161942772103, - "loss": 1.4419, - "step": 18267 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019382128120137997, - "loss": 1.5503, - "step": 18268 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938209429750389, - "loss": 1.7517, - "step": 18269 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019382060474869783, - "loss": 1.5998, - "step": 18270 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019382026652235677, - "loss": 1.7885, - "step": 18271 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938199282960157, - "loss": 1.4774, - "step": 18272 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019381959006967462, - "loss": 1.5069, - "step": 18273 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019381925184333356, - "loss": 1.5038, - "step": 18274 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938189136169925, - "loss": 1.6482, - "step": 18275 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019381857539065142, - "loss": 1.6116, - "step": 18276 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019381823716431036, - "loss": 1.5788, - "step": 18277 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938178989379693, - "loss": 1.6052, - "step": 18278 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019381756071162821, - "loss": 1.6361, - "step": 18279 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019381722248528716, - "loss": 1.5079, - "step": 18280 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938168842589461, - "loss": 1.5143, - "step": 18281 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193816546032605, - "loss": 1.4436, - "step": 18282 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019381620780626398, - "loss": 1.5617, - "step": 18283 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938158695799229, - "loss": 1.5705, - "step": 18284 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019381553135358183, - "loss": 1.5676, - "step": 18285 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019381519312724078, - "loss": 1.419, - "step": 18286 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938148549008997, - "loss": 1.5217, - "step": 18287 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019381451667455863, - "loss": 1.4403, - "step": 18288 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019381417844821757, - "loss": 1.411, - "step": 18289 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019381384022187649, - "loss": 1.6537, - "step": 18290 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019381350199553543, - "loss": 1.6485, - "step": 18291 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019381316376919437, - "loss": 1.6665, - "step": 18292 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019381282554285328, - "loss": 1.6402, - "step": 18293 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019381248731651222, - "loss": 1.563, - "step": 18294 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019381214909017116, - "loss": 1.5282, - "step": 18295 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019381181086383008, - "loss": 1.6406, - "step": 18296 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019381147263748902, - "loss": 1.5706, - "step": 18297 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019381113441114796, - "loss": 1.5676, - "step": 18298 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019381079618480688, - "loss": 1.6612, - "step": 18299 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019381045795846582, - "loss": 1.6169, - "step": 18300 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019381011973212476, - "loss": 1.6511, - "step": 18301 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019380978150578367, - "loss": 1.5395, - "step": 18302 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938094432794426, - "loss": 1.4574, - "step": 18303 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019380910505310155, - "loss": 1.6597, - "step": 18304 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019380876682676047, - "loss": 1.7777, - "step": 18305 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938084286004194, - "loss": 1.6434, - "step": 18306 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019380809037407835, - "loss": 1.5826, - "step": 18307 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019380775214773726, - "loss": 1.5218, - "step": 18308 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938074139213962, - "loss": 1.537, - "step": 18309 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019380707569505515, - "loss": 1.6381, - "step": 18310 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019380673746871406, - "loss": 1.4441, - "step": 18311 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193806399242373, - "loss": 1.5944, - "step": 18312 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019380606101603194, - "loss": 1.4904, - "step": 18313 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019380572278969086, - "loss": 1.4873, - "step": 18314 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938053845633498, - "loss": 1.4485, - "step": 18315 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019380504633700874, - "loss": 1.4525, - "step": 18316 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019380470811066765, - "loss": 1.5921, - "step": 18317 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938043698843266, - "loss": 1.6349, - "step": 18318 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019380403165798554, - "loss": 1.6449, - "step": 18319 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019380369343164445, - "loss": 1.3739, - "step": 18320 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938033552053034, - "loss": 1.5324, - "step": 18321 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019380301697896233, - "loss": 1.4459, - "step": 18322 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019380267875262125, - "loss": 1.4778, - "step": 18323 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001938023405262802, - "loss": 1.5482, - "step": 18324 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019380200229993913, - "loss": 1.6882, - "step": 18325 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019380166407359804, - "loss": 1.5338, - "step": 18326 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019380132584725698, - "loss": 1.6953, - "step": 18327 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019380098762091592, - "loss": 1.5406, - "step": 18328 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019380064939457484, - "loss": 1.6836, - "step": 18329 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019380031116823378, - "loss": 1.5292, - "step": 18330 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019379997294189272, - "loss": 1.7384, - "step": 18331 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019379963471555163, - "loss": 1.5208, - "step": 18332 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019379929648921058, - "loss": 1.4611, - "step": 18333 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019379895826286952, - "loss": 1.4934, - "step": 18334 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019379862003652843, - "loss": 1.5948, - "step": 18335 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019379828181018737, - "loss": 1.5025, - "step": 18336 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937979435838463, - "loss": 1.6103, - "step": 18337 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019379760535750525, - "loss": 1.5446, - "step": 18338 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937972671311642, - "loss": 1.6442, - "step": 18339 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019379692890482314, - "loss": 1.5703, - "step": 18340 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019379659067848205, - "loss": 1.6458, - "step": 18341 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193796252452141, - "loss": 1.4934, - "step": 18342 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019379591422579993, - "loss": 1.6284, - "step": 18343 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019379557599945885, - "loss": 1.4829, - "step": 18344 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937952377731178, - "loss": 1.6885, - "step": 18345 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019379489954677673, - "loss": 1.7158, - "step": 18346 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019379456132043564, - "loss": 1.5473, - "step": 18347 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019379422309409458, - "loss": 1.6046, - "step": 18348 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019379388486775353, - "loss": 1.7485, - "step": 18349 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019379354664141244, - "loss": 1.5011, - "step": 18350 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019379320841507138, - "loss": 1.6496, - "step": 18351 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019379287018873032, - "loss": 1.5886, - "step": 18352 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019379253196238924, - "loss": 1.6681, - "step": 18353 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019379219373604818, - "loss": 1.6789, - "step": 18354 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019379185550970712, - "loss": 1.6467, - "step": 18355 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019379151728336603, - "loss": 1.6252, - "step": 18356 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019379117905702497, - "loss": 1.7811, - "step": 18357 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019379084083068391, - "loss": 1.4864, - "step": 18358 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019379050260434283, - "loss": 1.6257, - "step": 18359 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019379016437800177, - "loss": 1.6953, - "step": 18360 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937898261516607, - "loss": 1.806, - "step": 18361 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019378948792531962, - "loss": 1.6784, - "step": 18362 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019378914969897857, - "loss": 1.5241, - "step": 18363 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937888114726375, - "loss": 1.3958, - "step": 18364 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019378847324629642, - "loss": 1.6451, - "step": 18365 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019378813501995536, - "loss": 1.6251, - "step": 18366 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937877967936143, - "loss": 1.5142, - "step": 18367 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019378745856727322, - "loss": 1.6796, - "step": 18368 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019378712034093216, - "loss": 1.573, - "step": 18369 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937867821145911, - "loss": 1.6435, - "step": 18370 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019378644388825001, - "loss": 1.5233, - "step": 18371 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019378610566190895, - "loss": 1.4947, - "step": 18372 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937857674355679, - "loss": 1.718, - "step": 18373 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937854292092268, - "loss": 1.482, - "step": 18374 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019378509098288575, - "loss": 1.5976, - "step": 18375 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937847527565447, - "loss": 1.498, - "step": 18376 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937844145302036, - "loss": 1.5221, - "step": 18377 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019378407630386255, - "loss": 1.5218, - "step": 18378 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937837380775215, - "loss": 1.6641, - "step": 18379 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937833998511804, - "loss": 1.6124, - "step": 18380 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019378306162483934, - "loss": 1.5972, - "step": 18381 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019378272339849828, - "loss": 1.6126, - "step": 18382 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937823851721572, - "loss": 1.406, - "step": 18383 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019378204694581614, - "loss": 1.6006, - "step": 18384 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019378170871947508, - "loss": 1.6681, - "step": 18385 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193781370493134, - "loss": 1.8635, - "step": 18386 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019378103226679294, - "loss": 1.5926, - "step": 18387 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019378069404045188, - "loss": 1.6065, - "step": 18388 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937803558141108, - "loss": 1.5217, - "step": 18389 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019378001758776973, - "loss": 1.5571, - "step": 18390 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019377967936142867, - "loss": 1.6038, - "step": 18391 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937793411350876, - "loss": 1.5676, - "step": 18392 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019377900290874653, - "loss": 1.7247, - "step": 18393 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019377866468240547, - "loss": 1.585, - "step": 18394 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937783264560644, - "loss": 1.6135, - "step": 18395 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019377798822972335, - "loss": 1.6497, - "step": 18396 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937776500033823, - "loss": 1.6311, - "step": 18397 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937773117770412, - "loss": 1.5925, - "step": 18398 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019377697355070015, - "loss": 1.5815, - "step": 18399 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937766353243591, - "loss": 1.73, - "step": 18400 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193776297098018, - "loss": 1.6539, - "step": 18401 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019377595887167695, - "loss": 1.5327, - "step": 18402 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019377562064533589, - "loss": 1.6515, - "step": 18403 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937752824189948, - "loss": 1.586, - "step": 18404 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019377494419265374, - "loss": 1.5257, - "step": 18405 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019377460596631268, - "loss": 1.4944, - "step": 18406 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937742677399716, - "loss": 1.6963, - "step": 18407 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019377392951363054, - "loss": 1.7446, - "step": 18408 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019377359128728948, - "loss": 1.4833, - "step": 18409 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937732530609484, - "loss": 1.5866, - "step": 18410 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019377291483460733, - "loss": 1.4596, - "step": 18411 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019377257660826628, - "loss": 1.5023, - "step": 18412 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937722383819252, - "loss": 1.5675, - "step": 18413 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019377190015558413, - "loss": 1.5414, - "step": 18414 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019377156192924307, - "loss": 1.7049, - "step": 18415 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019377122370290199, - "loss": 1.6746, - "step": 18416 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019377088547656093, - "loss": 1.6945, - "step": 18417 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019377054725021987, - "loss": 1.6495, - "step": 18418 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019377020902387878, - "loss": 1.7011, - "step": 18419 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019376987079753772, - "loss": 1.5719, - "step": 18420 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019376953257119666, - "loss": 1.6546, - "step": 18421 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019376919434485558, - "loss": 1.551, - "step": 18422 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019376885611851452, - "loss": 1.6207, - "step": 18423 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019376851789217346, - "loss": 1.4418, - "step": 18424 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019376817966583237, - "loss": 1.5709, - "step": 18425 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019376784143949132, - "loss": 1.6111, - "step": 18426 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019376750321315026, - "loss": 1.5226, - "step": 18427 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019376716498680917, - "loss": 1.5736, - "step": 18428 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937668267604681, - "loss": 1.6177, - "step": 18429 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019376648853412705, - "loss": 1.6045, - "step": 18430 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019376615030778597, - "loss": 1.5626, - "step": 18431 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937658120814449, - "loss": 1.443, - "step": 18432 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019376547385510385, - "loss": 1.5036, - "step": 18433 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019376513562876276, - "loss": 1.4326, - "step": 18434 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937647974024217, - "loss": 1.4293, - "step": 18435 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019376445917608065, - "loss": 1.666, - "step": 18436 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019376412094973956, - "loss": 1.7562, - "step": 18437 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937637827233985, - "loss": 1.6379, - "step": 18438 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019376344449705744, - "loss": 1.6737, - "step": 18439 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019376310627071636, - "loss": 1.5515, - "step": 18440 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937627680443753, - "loss": 1.5567, - "step": 18441 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019376242981803424, - "loss": 1.6115, - "step": 18442 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019376209159169315, - "loss": 1.7155, - "step": 18443 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937617533653521, - "loss": 1.4768, - "step": 18444 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019376141513901103, - "loss": 1.5359, - "step": 18445 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019376107691266995, - "loss": 1.5913, - "step": 18446 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937607386863289, - "loss": 1.6707, - "step": 18447 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019376040045998783, - "loss": 1.7248, - "step": 18448 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019376006223364675, - "loss": 1.4648, - "step": 18449 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019375972400730569, - "loss": 1.5589, - "step": 18450 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019375938578096465, - "loss": 1.3988, - "step": 18451 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019375904755462357, - "loss": 1.6266, - "step": 18452 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937587093282825, - "loss": 1.5778, - "step": 18453 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019375837110194145, - "loss": 1.6323, - "step": 18454 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019375803287560036, - "loss": 1.55, - "step": 18455 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937576946492593, - "loss": 1.6157, - "step": 18456 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019375735642291825, - "loss": 1.5843, - "step": 18457 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019375701819657716, - "loss": 1.5909, - "step": 18458 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937566799702361, - "loss": 1.5441, - "step": 18459 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019375634174389504, - "loss": 1.6272, - "step": 18460 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019375600351755396, - "loss": 1.4676, - "step": 18461 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937556652912129, - "loss": 1.6304, - "step": 18462 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019375532706487184, - "loss": 1.6966, - "step": 18463 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019375498883853075, - "loss": 1.6869, - "step": 18464 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937546506121897, - "loss": 1.7885, - "step": 18465 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019375431238584864, - "loss": 1.6458, - "step": 18466 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019375397415950755, - "loss": 1.5486, - "step": 18467 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937536359331665, - "loss": 1.5537, - "step": 18468 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019375329770682543, - "loss": 1.6097, - "step": 18469 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019375295948048435, - "loss": 1.6433, - "step": 18470 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937526212541433, - "loss": 1.5461, - "step": 18471 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019375228302780223, - "loss": 1.6407, - "step": 18472 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019375194480146114, - "loss": 1.5227, - "step": 18473 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019375160657512008, - "loss": 1.5635, - "step": 18474 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019375126834877902, - "loss": 1.5817, - "step": 18475 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019375093012243794, - "loss": 1.6809, - "step": 18476 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019375059189609688, - "loss": 1.5092, - "step": 18477 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019375025366975582, - "loss": 1.646, - "step": 18478 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019374991544341474, - "loss": 1.5121, - "step": 18479 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019374957721707368, - "loss": 1.6186, - "step": 18480 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019374923899073262, - "loss": 1.5374, - "step": 18481 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019374890076439153, - "loss": 1.6681, - "step": 18482 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019374856253805047, - "loss": 1.4726, - "step": 18483 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019374822431170941, - "loss": 1.6055, - "step": 18484 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019374788608536833, - "loss": 1.4103, - "step": 18485 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019374754785902727, - "loss": 1.6577, - "step": 18486 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937472096326862, - "loss": 1.597, - "step": 18487 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019374687140634512, - "loss": 1.5209, - "step": 18488 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019374653318000407, - "loss": 1.5368, - "step": 18489 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193746194953663, - "loss": 1.57, - "step": 18490 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019374585672732192, - "loss": 1.6178, - "step": 18491 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019374551850098086, - "loss": 1.5439, - "step": 18492 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937451802746398, - "loss": 1.6736, - "step": 18493 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019374484204829872, - "loss": 1.3947, - "step": 18494 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019374450382195766, - "loss": 1.5353, - "step": 18495 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937441655956166, - "loss": 1.4223, - "step": 18496 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937438273692755, - "loss": 1.4764, - "step": 18497 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019374348914293445, - "loss": 1.6076, - "step": 18498 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937431509165934, - "loss": 1.5591, - "step": 18499 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937428126902523, - "loss": 1.6893, - "step": 18500 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019374247446391125, - "loss": 1.6054, - "step": 18501 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937421362375702, - "loss": 1.5327, - "step": 18502 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937417980112291, - "loss": 1.5778, - "step": 18503 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019374145978488805, - "loss": 1.5954, - "step": 18504 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193741121558547, - "loss": 1.6103, - "step": 18505 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937407833322059, - "loss": 1.8286, - "step": 18506 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019374044510586487, - "loss": 1.594, - "step": 18507 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019374010687952378, - "loss": 1.4643, - "step": 18508 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019373976865318273, - "loss": 1.5413, - "step": 18509 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019373943042684167, - "loss": 1.6784, - "step": 18510 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019373909220050058, - "loss": 1.5615, - "step": 18511 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019373875397415952, - "loss": 1.5318, - "step": 18512 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019373841574781846, - "loss": 1.5671, - "step": 18513 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019373807752147738, - "loss": 1.5382, - "step": 18514 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019373773929513632, - "loss": 1.6346, - "step": 18515 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019373740106879526, - "loss": 1.4874, - "step": 18516 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019373706284245417, - "loss": 1.5884, - "step": 18517 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019373672461611311, - "loss": 1.56, - "step": 18518 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019373638638977206, - "loss": 1.592, - "step": 18519 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193736048163431, - "loss": 1.5624, - "step": 18520 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937357099370899, - "loss": 1.5536, - "step": 18521 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019373537171074885, - "loss": 1.6216, - "step": 18522 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937350334844078, - "loss": 1.7654, - "step": 18523 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937346952580667, - "loss": 1.5791, - "step": 18524 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019373435703172565, - "loss": 1.6209, - "step": 18525 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937340188053846, - "loss": 1.655, - "step": 18526 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937336805790435, - "loss": 1.4474, - "step": 18527 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019373334235270244, - "loss": 1.7693, - "step": 18528 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019373300412636139, - "loss": 1.5896, - "step": 18529 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937326659000203, - "loss": 1.9211, - "step": 18530 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019373232767367924, - "loss": 1.6609, - "step": 18531 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019373198944733818, - "loss": 1.7175, - "step": 18532 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937316512209971, - "loss": 1.4945, - "step": 18533 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019373131299465604, - "loss": 1.7406, - "step": 18534 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019373097476831498, - "loss": 1.7837, - "step": 18535 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937306365419739, - "loss": 1.6075, - "step": 18536 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019373029831563283, - "loss": 1.566, - "step": 18537 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019372996008929177, - "loss": 1.5824, - "step": 18538 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937296218629507, - "loss": 1.5506, - "step": 18539 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019372928363660963, - "loss": 1.5942, - "step": 18540 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019372894541026857, - "loss": 1.6085, - "step": 18541 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019372860718392748, - "loss": 1.554, - "step": 18542 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019372826895758643, - "loss": 1.4643, - "step": 18543 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019372793073124537, - "loss": 1.6488, - "step": 18544 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019372759250490428, - "loss": 1.5469, - "step": 18545 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019372725427856322, - "loss": 1.6193, - "step": 18546 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019372691605222216, - "loss": 1.6428, - "step": 18547 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019372657782588108, - "loss": 1.4097, - "step": 18548 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019372623959954002, - "loss": 1.6002, - "step": 18549 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019372590137319896, - "loss": 1.5579, - "step": 18550 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019372556314685787, - "loss": 1.6471, - "step": 18551 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019372522492051682, - "loss": 1.5158, - "step": 18552 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019372488669417576, - "loss": 1.5453, - "step": 18553 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019372454846783467, - "loss": 1.6034, - "step": 18554 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937242102414936, - "loss": 1.522, - "step": 18555 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019372387201515255, - "loss": 1.7206, - "step": 18556 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019372353378881147, - "loss": 1.6326, - "step": 18557 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937231955624704, - "loss": 1.5464, - "step": 18558 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019372285733612935, - "loss": 1.5702, - "step": 18559 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019372251910978826, - "loss": 1.6475, - "step": 18560 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937221808834472, - "loss": 1.7456, - "step": 18561 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019372184265710615, - "loss": 1.5545, - "step": 18562 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019372150443076509, - "loss": 1.6024, - "step": 18563 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019372116620442403, - "loss": 1.5898, - "step": 18564 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019372082797808294, - "loss": 1.5688, - "step": 18565 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019372048975174188, - "loss": 1.5913, - "step": 18566 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019372015152540082, - "loss": 1.4944, - "step": 18567 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019371981329905974, - "loss": 1.4975, - "step": 18568 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019371947507271868, - "loss": 1.6131, - "step": 18569 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019371913684637762, - "loss": 1.5406, - "step": 18570 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019371879862003653, - "loss": 1.6469, - "step": 18571 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019371846039369548, - "loss": 1.513, - "step": 18572 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019371812216735442, - "loss": 1.7537, - "step": 18573 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019371778394101333, - "loss": 1.6811, - "step": 18574 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019371744571467227, - "loss": 1.6223, - "step": 18575 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937171074883312, - "loss": 1.6349, - "step": 18576 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019371676926199013, - "loss": 1.5005, - "step": 18577 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019371643103564907, - "loss": 1.4733, - "step": 18578 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193716092809308, - "loss": 1.6319, - "step": 18579 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019371575458296692, - "loss": 1.6598, - "step": 18580 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019371541635662586, - "loss": 1.6084, - "step": 18581 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937150781302848, - "loss": 1.5022, - "step": 18582 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019371473990394372, - "loss": 1.6397, - "step": 18583 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019371440167760266, - "loss": 1.5672, - "step": 18584 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937140634512616, - "loss": 1.6176, - "step": 18585 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019371372522492052, - "loss": 1.5818, - "step": 18586 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019371338699857946, - "loss": 1.5187, - "step": 18587 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937130487722384, - "loss": 1.7615, - "step": 18588 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937127105458973, - "loss": 1.6345, - "step": 18589 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019371237231955625, - "loss": 1.5924, - "step": 18590 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937120340932152, - "loss": 1.587, - "step": 18591 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937116958668741, - "loss": 1.6414, - "step": 18592 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019371135764053305, - "loss": 1.5377, - "step": 18593 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193711019414192, - "loss": 1.6055, - "step": 18594 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937106811878509, - "loss": 1.5604, - "step": 18595 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019371034296150985, - "loss": 1.606, - "step": 18596 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937100047351688, - "loss": 1.4175, - "step": 18597 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937096665088277, - "loss": 1.6193, - "step": 18598 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019370932828248664, - "loss": 1.5069, - "step": 18599 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019370899005614558, - "loss": 1.488, - "step": 18600 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937086518298045, - "loss": 1.6314, - "step": 18601 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019370831360346344, - "loss": 1.5679, - "step": 18602 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019370797537712238, - "loss": 1.4332, - "step": 18603 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937076371507813, - "loss": 1.6916, - "step": 18604 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019370729892444023, - "loss": 1.5877, - "step": 18605 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019370696069809918, - "loss": 1.6919, - "step": 18606 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019370662247175812, - "loss": 1.4954, - "step": 18607 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019370628424541703, - "loss": 1.5619, - "step": 18608 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019370594601907597, - "loss": 1.4241, - "step": 18609 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937056077927349, - "loss": 1.6092, - "step": 18610 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019370526956639383, - "loss": 1.7189, - "step": 18611 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019370493134005277, - "loss": 1.5289, - "step": 18612 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937045931137117, - "loss": 1.5936, - "step": 18613 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019370425488737062, - "loss": 1.4478, - "step": 18614 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019370391666102956, - "loss": 1.5162, - "step": 18615 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937035784346885, - "loss": 1.7209, - "step": 18616 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019370324020834742, - "loss": 1.5175, - "step": 18617 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019370290198200636, - "loss": 1.499, - "step": 18618 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937025637556653, - "loss": 1.5345, - "step": 18619 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019370222552932424, - "loss": 1.5412, - "step": 18620 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019370188730298318, - "loss": 1.6594, - "step": 18621 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937015490766421, - "loss": 1.5947, - "step": 18622 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019370121085030104, - "loss": 1.6403, - "step": 18623 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019370087262395998, - "loss": 1.5817, - "step": 18624 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001937005343976189, - "loss": 1.5271, - "step": 18625 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019370019617127784, - "loss": 1.562, - "step": 18626 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019369985794493678, - "loss": 1.6818, - "step": 18627 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936995197185957, - "loss": 1.6448, - "step": 18628 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019369918149225463, - "loss": 1.5704, - "step": 18629 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019369884326591357, - "loss": 1.6035, - "step": 18630 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936985050395725, - "loss": 1.6787, - "step": 18631 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019369816681323143, - "loss": 1.5603, - "step": 18632 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019369782858689037, - "loss": 1.5831, - "step": 18633 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019369749036054928, - "loss": 1.5726, - "step": 18634 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019369715213420822, - "loss": 1.4955, - "step": 18635 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019369681390786717, - "loss": 1.6195, - "step": 18636 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019369647568152608, - "loss": 1.5622, - "step": 18637 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019369613745518502, - "loss": 1.6056, - "step": 18638 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019369579922884396, - "loss": 1.5405, - "step": 18639 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019369546100250288, - "loss": 1.5686, - "step": 18640 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019369512277616182, - "loss": 1.6055, - "step": 18641 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019369478454982076, - "loss": 1.7472, - "step": 18642 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019369444632347967, - "loss": 1.619, - "step": 18643 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019369410809713861, - "loss": 1.6161, - "step": 18644 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019369376987079755, - "loss": 1.4865, - "step": 18645 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019369343164445647, - "loss": 1.56, - "step": 18646 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936930934181154, - "loss": 1.5541, - "step": 18647 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019369275519177435, - "loss": 1.7193, - "step": 18648 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019369241696543327, - "loss": 1.5161, - "step": 18649 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936920787390922, - "loss": 1.5175, - "step": 18650 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019369174051275115, - "loss": 1.5719, - "step": 18651 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019369140228641006, - "loss": 1.5708, - "step": 18652 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193691064060069, - "loss": 1.6283, - "step": 18653 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019369072583372794, - "loss": 1.5871, - "step": 18654 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019369038760738686, - "loss": 1.6189, - "step": 18655 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936900493810458, - "loss": 1.5131, - "step": 18656 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019368971115470474, - "loss": 1.612, - "step": 18657 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019368937292836365, - "loss": 1.7173, - "step": 18658 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936890347020226, - "loss": 1.694, - "step": 18659 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019368869647568154, - "loss": 1.6369, - "step": 18660 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019368835824934045, - "loss": 1.6165, - "step": 18661 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936880200229994, - "loss": 1.7308, - "step": 18662 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019368768179665833, - "loss": 1.6442, - "step": 18663 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019368734357031725, - "loss": 1.4662, - "step": 18664 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936870053439762, - "loss": 1.3751, - "step": 18665 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019368666711763513, - "loss": 1.6121, - "step": 18666 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019368632889129404, - "loss": 1.6969, - "step": 18667 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019368599066495298, - "loss": 1.6356, - "step": 18668 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019368565243861193, - "loss": 1.7006, - "step": 18669 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019368531421227084, - "loss": 1.5747, - "step": 18670 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019368497598592978, - "loss": 1.7485, - "step": 18671 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019368463775958872, - "loss": 1.514, - "step": 18672 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019368429953324764, - "loss": 1.4744, - "step": 18673 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019368396130690658, - "loss": 1.773, - "step": 18674 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019368362308056552, - "loss": 1.468, - "step": 18675 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019368328485422446, - "loss": 1.5661, - "step": 18676 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936829466278834, - "loss": 1.5938, - "step": 18677 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019368260840154234, - "loss": 1.7172, - "step": 18678 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019368227017520126, - "loss": 1.5936, - "step": 18679 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936819319488602, - "loss": 1.4539, - "step": 18680 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019368159372251914, - "loss": 1.5378, - "step": 18681 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019368125549617805, - "loss": 1.565, - "step": 18682 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193680917269837, - "loss": 1.6921, - "step": 18683 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019368057904349593, - "loss": 1.4291, - "step": 18684 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019368024081715485, - "loss": 1.6608, - "step": 18685 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936799025908138, - "loss": 1.4599, - "step": 18686 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019367956436447273, - "loss": 1.5186, - "step": 18687 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019367922613813164, - "loss": 1.7306, - "step": 18688 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019367888791179059, - "loss": 1.5994, - "step": 18689 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019367854968544953, - "loss": 1.5512, - "step": 18690 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019367821145910844, - "loss": 1.7223, - "step": 18691 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019367787323276738, - "loss": 1.4852, - "step": 18692 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019367753500642632, - "loss": 1.5434, - "step": 18693 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019367719678008524, - "loss": 1.5248, - "step": 18694 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019367685855374418, - "loss": 1.6073, - "step": 18695 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019367652032740312, - "loss": 1.3596, - "step": 18696 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019367618210106203, - "loss": 1.6201, - "step": 18697 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019367584387472097, - "loss": 1.6099, - "step": 18698 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019367550564837992, - "loss": 1.5672, - "step": 18699 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019367516742203883, - "loss": 1.3903, - "step": 18700 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019367482919569777, - "loss": 1.6249, - "step": 18701 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936744909693567, - "loss": 1.5382, - "step": 18702 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019367415274301563, - "loss": 1.7212, - "step": 18703 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019367381451667457, - "loss": 1.554, - "step": 18704 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936734762903335, - "loss": 1.5862, - "step": 18705 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019367313806399242, - "loss": 1.4815, - "step": 18706 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019367279983765136, - "loss": 1.6169, - "step": 18707 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936724616113103, - "loss": 1.6891, - "step": 18708 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019367212338496922, - "loss": 1.6048, - "step": 18709 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019367178515862816, - "loss": 1.584, - "step": 18710 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936714469322871, - "loss": 1.5379, - "step": 18711 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019367110870594601, - "loss": 1.5544, - "step": 18712 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019367077047960496, - "loss": 1.636, - "step": 18713 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936704322532639, - "loss": 1.5797, - "step": 18714 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936700940269228, - "loss": 1.5498, - "step": 18715 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019366975580058175, - "loss": 1.5347, - "step": 18716 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936694175742407, - "loss": 1.6378, - "step": 18717 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936690793478996, - "loss": 1.6587, - "step": 18718 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019366874112155855, - "loss": 1.6717, - "step": 18719 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936684028952175, - "loss": 1.7228, - "step": 18720 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936680646688764, - "loss": 1.5018, - "step": 18721 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019366772644253535, - "loss": 1.7855, - "step": 18722 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019366738821619429, - "loss": 1.6603, - "step": 18723 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936670499898532, - "loss": 1.4925, - "step": 18724 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019366671176351214, - "loss": 1.5773, - "step": 18725 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019366637353717108, - "loss": 1.4866, - "step": 18726 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019366603531083, - "loss": 1.5089, - "step": 18727 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019366569708448894, - "loss": 1.6304, - "step": 18728 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019366535885814788, - "loss": 1.4463, - "step": 18729 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936650206318068, - "loss": 1.6848, - "step": 18730 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019366468240546573, - "loss": 1.5442, - "step": 18731 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936643441791247, - "loss": 1.5643, - "step": 18732 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019366400595278362, - "loss": 1.5721, - "step": 18733 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019366366772644256, - "loss": 1.4715, - "step": 18734 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936633295001015, - "loss": 1.6222, - "step": 18735 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936629912737604, - "loss": 1.5589, - "step": 18736 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019366265304741935, - "loss": 1.4805, - "step": 18737 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936623148210783, - "loss": 1.6413, - "step": 18738 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936619765947372, - "loss": 1.6803, - "step": 18739 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019366163836839615, - "loss": 1.4317, - "step": 18740 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936613001420551, - "loss": 1.6992, - "step": 18741 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193660961915714, - "loss": 1.6135, - "step": 18742 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019366062368937295, - "loss": 1.6071, - "step": 18743 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936602854630319, - "loss": 1.5667, - "step": 18744 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936599472366908, - "loss": 1.568, - "step": 18745 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019365960901034974, - "loss": 1.6915, - "step": 18746 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019365927078400868, - "loss": 1.4315, - "step": 18747 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936589325576676, - "loss": 1.5826, - "step": 18748 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019365859433132654, - "loss": 1.5974, - "step": 18749 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019365825610498548, - "loss": 1.6027, - "step": 18750 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936579178786444, - "loss": 1.5851, - "step": 18751 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019365757965230334, - "loss": 1.6485, - "step": 18752 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019365724142596228, - "loss": 1.5259, - "step": 18753 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936569031996212, - "loss": 1.7412, - "step": 18754 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019365656497328013, - "loss": 1.5747, - "step": 18755 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019365622674693907, - "loss": 1.5626, - "step": 18756 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193655888520598, - "loss": 1.5734, - "step": 18757 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019365555029425693, - "loss": 1.5806, - "step": 18758 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019365521206791587, - "loss": 1.6774, - "step": 18759 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019365487384157478, - "loss": 1.7602, - "step": 18760 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019365453561523372, - "loss": 1.5405, - "step": 18761 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019365419738889267, - "loss": 1.6046, - "step": 18762 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019365385916255158, - "loss": 1.7167, - "step": 18763 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019365352093621052, - "loss": 1.4852, - "step": 18764 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019365318270986946, - "loss": 1.5596, - "step": 18765 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019365284448352838, - "loss": 1.5297, - "step": 18766 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019365250625718732, - "loss": 1.5129, - "step": 18767 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019365216803084626, - "loss": 1.4895, - "step": 18768 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019365182980450517, - "loss": 1.5481, - "step": 18769 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936514915781641, - "loss": 1.5005, - "step": 18770 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019365115335182305, - "loss": 1.6718, - "step": 18771 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019365081512548197, - "loss": 1.5259, - "step": 18772 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936504768991409, - "loss": 1.5463, - "step": 18773 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019365013867279985, - "loss": 1.5347, - "step": 18774 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019364980044645876, - "loss": 1.6862, - "step": 18775 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936494622201177, - "loss": 1.6566, - "step": 18776 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019364912399377665, - "loss": 1.6517, - "step": 18777 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019364878576743556, - "loss": 1.7238, - "step": 18778 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936484475410945, - "loss": 1.4328, - "step": 18779 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019364810931475344, - "loss": 1.4784, - "step": 18780 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019364777108841236, - "loss": 1.533, - "step": 18781 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936474328620713, - "loss": 1.698, - "step": 18782 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019364709463573024, - "loss": 1.6622, - "step": 18783 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019364675640938915, - "loss": 1.408, - "step": 18784 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936464181830481, - "loss": 1.5274, - "step": 18785 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019364607995670704, - "loss": 1.6973, - "step": 18786 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019364574173036595, - "loss": 1.7026, - "step": 18787 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019364540350402492, - "loss": 1.6045, - "step": 18788 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019364506527768386, - "loss": 1.5778, - "step": 18789 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019364472705134277, - "loss": 1.4088, - "step": 18790 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019364438882500171, - "loss": 1.5386, - "step": 18791 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019364405059866066, - "loss": 1.616, - "step": 18792 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019364371237231957, - "loss": 1.5732, - "step": 18793 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936433741459785, - "loss": 1.5528, - "step": 18794 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019364303591963745, - "loss": 1.4609, - "step": 18795 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019364269769329637, - "loss": 1.4997, - "step": 18796 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936423594669553, - "loss": 1.5253, - "step": 18797 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019364202124061425, - "loss": 1.643, - "step": 18798 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019364168301427316, - "loss": 1.418, - "step": 18799 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936413447879321, - "loss": 1.6047, - "step": 18800 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019364100656159104, - "loss": 1.6401, - "step": 18801 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019364066833524996, - "loss": 1.9107, - "step": 18802 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936403301089089, - "loss": 1.5556, - "step": 18803 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019363999188256784, - "loss": 1.6145, - "step": 18804 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019363965365622675, - "loss": 1.534, - "step": 18805 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936393154298857, - "loss": 1.5553, - "step": 18806 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019363897720354464, - "loss": 1.6649, - "step": 18807 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019363863897720355, - "loss": 1.4256, - "step": 18808 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936383007508625, - "loss": 1.5951, - "step": 18809 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019363796252452143, - "loss": 1.5895, - "step": 18810 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019363762429818035, - "loss": 1.4344, - "step": 18811 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936372860718393, - "loss": 1.6213, - "step": 18812 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019363694784549823, - "loss": 1.5761, - "step": 18813 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019363660961915714, - "loss": 1.4228, - "step": 18814 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019363627139281608, - "loss": 1.5589, - "step": 18815 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019363593316647503, - "loss": 1.5609, - "step": 18816 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019363559494013394, - "loss": 1.5543, - "step": 18817 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019363525671379288, - "loss": 1.5195, - "step": 18818 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019363491848745182, - "loss": 1.7011, - "step": 18819 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019363458026111074, - "loss": 1.472, - "step": 18820 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019363424203476968, - "loss": 1.3251, - "step": 18821 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019363390380842862, - "loss": 1.5789, - "step": 18822 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019363356558208753, - "loss": 1.5689, - "step": 18823 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019363322735574647, - "loss": 1.4651, - "step": 18824 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019363288912940542, - "loss": 1.5932, - "step": 18825 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019363255090306433, - "loss": 1.5301, - "step": 18826 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019363221267672327, - "loss": 1.7579, - "step": 18827 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936318744503822, - "loss": 1.5248, - "step": 18828 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019363153622404113, - "loss": 1.4884, - "step": 18829 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019363119799770007, - "loss": 1.7412, - "step": 18830 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193630859771359, - "loss": 1.5012, - "step": 18831 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019363052154501792, - "loss": 1.5684, - "step": 18832 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019363018331867686, - "loss": 1.6418, - "step": 18833 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936298450923358, - "loss": 1.5169, - "step": 18834 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019362950686599472, - "loss": 1.4708, - "step": 18835 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019362916863965366, - "loss": 1.5649, - "step": 18836 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936288304133126, - "loss": 1.6083, - "step": 18837 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019362849218697151, - "loss": 1.7338, - "step": 18838 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019362815396063046, - "loss": 1.6206, - "step": 18839 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936278157342894, - "loss": 1.5406, - "step": 18840 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936274775079483, - "loss": 1.5243, - "step": 18841 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019362713928160725, - "loss": 1.5217, - "step": 18842 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936268010552662, - "loss": 1.5797, - "step": 18843 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019362646282892513, - "loss": 1.6091, - "step": 18844 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019362612460258408, - "loss": 1.5779, - "step": 18845 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193625786376243, - "loss": 1.4732, - "step": 18846 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019362544814990193, - "loss": 1.7058, - "step": 18847 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019362510992356087, - "loss": 1.6161, - "step": 18848 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019362477169721979, - "loss": 1.5739, - "step": 18849 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019362443347087873, - "loss": 1.6494, - "step": 18850 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019362409524453767, - "loss": 1.4852, - "step": 18851 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019362375701819658, - "loss": 1.5824, - "step": 18852 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019362341879185552, - "loss": 1.6951, - "step": 18853 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019362308056551446, - "loss": 1.512, - "step": 18854 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019362274233917338, - "loss": 1.5127, - "step": 18855 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019362240411283232, - "loss": 1.6665, - "step": 18856 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019362206588649126, - "loss": 1.5515, - "step": 18857 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019362172766015017, - "loss": 1.6475, - "step": 18858 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019362138943380912, - "loss": 1.5346, - "step": 18859 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019362105120746806, - "loss": 1.4654, - "step": 18860 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019362071298112697, - "loss": 1.6605, - "step": 18861 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936203747547859, - "loss": 1.649, - "step": 18862 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019362003652844485, - "loss": 1.6152, - "step": 18863 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019361969830210377, - "loss": 1.4579, - "step": 18864 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936193600757627, - "loss": 1.5486, - "step": 18865 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019361902184942165, - "loss": 1.727, - "step": 18866 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019361868362308056, - "loss": 1.4556, - "step": 18867 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936183453967395, - "loss": 1.5758, - "step": 18868 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019361800717039845, - "loss": 1.6501, - "step": 18869 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019361766894405736, - "loss": 1.4809, - "step": 18870 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936173307177163, - "loss": 1.6151, - "step": 18871 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019361699249137524, - "loss": 1.6326, - "step": 18872 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019361665426503418, - "loss": 1.6349, - "step": 18873 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936163160386931, - "loss": 1.6323, - "step": 18874 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019361597781235204, - "loss": 1.5784, - "step": 18875 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019361563958601098, - "loss": 1.6944, - "step": 18876 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936153013596699, - "loss": 1.4214, - "step": 18877 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019361496313332883, - "loss": 1.5883, - "step": 18878 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019361462490698778, - "loss": 1.419, - "step": 18879 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936142866806467, - "loss": 1.6794, - "step": 18880 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019361394845430563, - "loss": 1.6208, - "step": 18881 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019361361022796457, - "loss": 1.6397, - "step": 18882 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019361327200162349, - "loss": 1.7012, - "step": 18883 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019361293377528243, - "loss": 1.4346, - "step": 18884 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019361259554894137, - "loss": 1.5018, - "step": 18885 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019361225732260028, - "loss": 1.5309, - "step": 18886 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019361191909625922, - "loss": 1.6897, - "step": 18887 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019361158086991816, - "loss": 1.5574, - "step": 18888 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019361124264357708, - "loss": 1.5988, - "step": 18889 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019361090441723602, - "loss": 1.502, - "step": 18890 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019361056619089496, - "loss": 1.5728, - "step": 18891 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019361022796455388, - "loss": 1.8392, - "step": 18892 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019360988973821282, - "loss": 1.6398, - "step": 18893 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019360955151187176, - "loss": 1.6886, - "step": 18894 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019360921328553067, - "loss": 1.5949, - "step": 18895 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936088750591896, - "loss": 1.5002, - "step": 18896 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019360853683284855, - "loss": 1.6201, - "step": 18897 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019360819860650747, - "loss": 1.55, - "step": 18898 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936078603801664, - "loss": 1.6444, - "step": 18899 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019360752215382535, - "loss": 1.4887, - "step": 18900 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936071839274843, - "loss": 1.5694, - "step": 18901 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019360684570114323, - "loss": 1.5106, - "step": 18902 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019360650747480215, - "loss": 1.8042, - "step": 18903 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001936061692484611, - "loss": 1.7019, - "step": 18904 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019360583102212003, - "loss": 1.5482, - "step": 18905 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019360549279577894, - "loss": 1.4131, - "step": 18906 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019360515456943788, - "loss": 1.414, - "step": 18907 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019360481634309682, - "loss": 1.6442, - "step": 18908 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019360447811675574, - "loss": 1.6134, - "step": 18909 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019360413989041468, - "loss": 1.6614, - "step": 18910 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019360380166407362, - "loss": 1.5292, - "step": 18911 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019360346343773254, - "loss": 1.5954, - "step": 18912 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019360312521139148, - "loss": 1.4611, - "step": 18913 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019360278698505042, - "loss": 1.702, - "step": 18914 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019360244875870933, - "loss": 1.6135, - "step": 18915 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019360211053236827, - "loss": 1.4724, - "step": 18916 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019360177230602721, - "loss": 1.6656, - "step": 18917 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019360143407968613, - "loss": 1.7436, - "step": 18918 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019360109585334507, - "loss": 1.6218, - "step": 18919 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193600757627004, - "loss": 1.698, - "step": 18920 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019360041940066292, - "loss": 1.55, - "step": 18921 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019360008117432187, - "loss": 1.6275, - "step": 18922 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935997429479808, - "loss": 1.6706, - "step": 18923 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019359940472163972, - "loss": 1.7834, - "step": 18924 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019359906649529866, - "loss": 1.6984, - "step": 18925 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935987282689576, - "loss": 1.6451, - "step": 18926 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019359839004261652, - "loss": 1.4613, - "step": 18927 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019359805181627546, - "loss": 1.5329, - "step": 18928 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935977135899344, - "loss": 1.5089, - "step": 18929 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935973753635933, - "loss": 1.5048, - "step": 18930 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019359703713725225, - "loss": 1.6656, - "step": 18931 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935966989109112, - "loss": 1.5846, - "step": 18932 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935963606845701, - "loss": 1.5241, - "step": 18933 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019359602245822905, - "loss": 1.5033, - "step": 18934 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193595684231888, - "loss": 1.7383, - "step": 18935 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935953460055469, - "loss": 1.5205, - "step": 18936 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019359500777920585, - "loss": 1.4868, - "step": 18937 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935946695528648, - "loss": 1.6509, - "step": 18938 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935943313265237, - "loss": 1.5921, - "step": 18939 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019359399310018264, - "loss": 1.7725, - "step": 18940 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019359365487384158, - "loss": 1.6958, - "step": 18941 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935933166475005, - "loss": 1.6466, - "step": 18942 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019359297842115944, - "loss": 1.5189, - "step": 18943 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019359264019481838, - "loss": 1.6115, - "step": 18944 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935923019684773, - "loss": 1.6704, - "step": 18945 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019359196374213624, - "loss": 1.5127, - "step": 18946 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019359162551579518, - "loss": 1.5888, - "step": 18947 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935912872894541, - "loss": 1.6519, - "step": 18948 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019359094906311303, - "loss": 1.5233, - "step": 18949 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019359061083677197, - "loss": 1.5959, - "step": 18950 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935902726104309, - "loss": 1.6262, - "step": 18951 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019358993438408983, - "loss": 1.5434, - "step": 18952 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019358959615774877, - "loss": 1.4177, - "step": 18953 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019358925793140768, - "loss": 1.6671, - "step": 18954 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019358891970506662, - "loss": 1.6785, - "step": 18955 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019358858147872557, - "loss": 1.655, - "step": 18956 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935882432523845, - "loss": 1.4866, - "step": 18957 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019358790502604345, - "loss": 1.594, - "step": 18958 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935875667997024, - "loss": 1.6029, - "step": 18959 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935872285733613, - "loss": 1.6052, - "step": 18960 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019358689034702024, - "loss": 1.5427, - "step": 18961 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019358655212067919, - "loss": 1.6543, - "step": 18962 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935862138943381, - "loss": 1.6581, - "step": 18963 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019358587566799704, - "loss": 1.6816, - "step": 18964 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019358553744165598, - "loss": 1.3613, - "step": 18965 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935851992153149, - "loss": 1.6014, - "step": 18966 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019358486098897384, - "loss": 1.4881, - "step": 18967 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019358452276263278, - "loss": 1.5963, - "step": 18968 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935841845362917, - "loss": 1.698, - "step": 18969 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019358384630995063, - "loss": 1.5857, - "step": 18970 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019358350808360957, - "loss": 1.6062, - "step": 18971 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935831698572685, - "loss": 1.4068, - "step": 18972 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019358283163092743, - "loss": 1.5469, - "step": 18973 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019358249340458637, - "loss": 1.8005, - "step": 18974 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019358215517824528, - "loss": 1.5043, - "step": 18975 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019358181695190423, - "loss": 1.5061, - "step": 18976 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019358147872556317, - "loss": 1.6805, - "step": 18977 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019358114049922208, - "loss": 1.6484, - "step": 18978 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019358080227288102, - "loss": 1.6729, - "step": 18979 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019358046404653996, - "loss": 1.6046, - "step": 18980 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019358012582019888, - "loss": 1.6014, - "step": 18981 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019357978759385782, - "loss": 1.6284, - "step": 18982 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019357944936751676, - "loss": 1.6317, - "step": 18983 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019357911114117567, - "loss": 1.5707, - "step": 18984 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019357877291483462, - "loss": 1.5731, - "step": 18985 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019357843468849356, - "loss": 1.3483, - "step": 18986 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019357809646215247, - "loss": 1.4619, - "step": 18987 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935777582358114, - "loss": 1.6929, - "step": 18988 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019357742000947035, - "loss": 1.5357, - "step": 18989 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019357708178312927, - "loss": 1.6073, - "step": 18990 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935767435567882, - "loss": 1.5102, - "step": 18991 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019357640533044715, - "loss": 1.5336, - "step": 18992 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019357606710410606, - "loss": 1.5519, - "step": 18993 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193575728877765, - "loss": 1.5826, - "step": 18994 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019357539065142395, - "loss": 1.4652, - "step": 18995 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019357505242508286, - "loss": 1.5338, - "step": 18996 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935747141987418, - "loss": 1.6373, - "step": 18997 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019357437597240074, - "loss": 1.6747, - "step": 18998 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019357403774605966, - "loss": 1.6586, - "step": 18999 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935736995197186, - "loss": 1.5735, - "step": 19000 - }, - { - "epoch": 0.06, - "eval_loss": 1.3721829927884615, - "mrc_em": 16.0, - "mrc_f1": 24.785714285714292, - "nli_acc": 0.1, - "step": 19000, - "sts_f1": 0.6999999999999998, - "ynat_f1": 0.720745714553764 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019357336129337754, - "loss": 1.6278, - "step": 19001 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019357302306703645, - "loss": 1.6447, - "step": 19002 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935726848406954, - "loss": 1.8419, - "step": 19003 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019357234661435433, - "loss": 1.6048, - "step": 19004 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019357200838801325, - "loss": 1.6016, - "step": 19005 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935716701616722, - "loss": 1.5987, - "step": 19006 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019357133193533113, - "loss": 1.53, - "step": 19007 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019357099370899004, - "loss": 1.5809, - "step": 19008 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019357065548264899, - "loss": 1.5828, - "step": 19009 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019357031725630793, - "loss": 1.5876, - "step": 19010 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019356997902996684, - "loss": 1.5645, - "step": 19011 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019356964080362578, - "loss": 1.5874, - "step": 19012 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019356930257728475, - "loss": 1.5775, - "step": 19013 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019356896435094366, - "loss": 1.5103, - "step": 19014 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935686261246026, - "loss": 1.5727, - "step": 19015 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019356828789826155, - "loss": 1.7137, - "step": 19016 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019356794967192046, - "loss": 1.5402, - "step": 19017 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935676114455794, - "loss": 1.6358, - "step": 19018 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019356727321923834, - "loss": 1.5336, - "step": 19019 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019356693499289726, - "loss": 1.5683, - "step": 19020 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935665967665562, - "loss": 1.476, - "step": 19021 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019356625854021514, - "loss": 1.4272, - "step": 19022 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019356592031387405, - "loss": 1.3649, - "step": 19023 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193565582087533, - "loss": 1.4844, - "step": 19024 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019356524386119194, - "loss": 1.5671, - "step": 19025 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019356490563485085, - "loss": 1.7501, - "step": 19026 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935645674085098, - "loss": 1.4988, - "step": 19027 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019356422918216873, - "loss": 1.5751, - "step": 19028 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019356389095582765, - "loss": 1.472, - "step": 19029 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935635527294866, - "loss": 1.6054, - "step": 19030 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019356321450314553, - "loss": 1.673, - "step": 19031 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019356287627680444, - "loss": 1.4586, - "step": 19032 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019356253805046338, - "loss": 1.6114, - "step": 19033 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019356219982412232, - "loss": 1.7222, - "step": 19034 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019356186159778124, - "loss": 1.5578, - "step": 19035 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019356152337144018, - "loss": 1.6183, - "step": 19036 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019356118514509912, - "loss": 1.5787, - "step": 19037 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019356084691875803, - "loss": 1.355, - "step": 19038 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019356050869241698, - "loss": 1.4478, - "step": 19039 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019356017046607592, - "loss": 1.558, - "step": 19040 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019355983223973483, - "loss": 1.5404, - "step": 19041 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019355949401339377, - "loss": 1.6372, - "step": 19042 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935591557870527, - "loss": 1.6909, - "step": 19043 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019355881756071163, - "loss": 1.5235, - "step": 19044 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019355847933437057, - "loss": 1.5637, - "step": 19045 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935581411080295, - "loss": 1.5287, - "step": 19046 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019355780288168842, - "loss": 1.5118, - "step": 19047 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019355746465534736, - "loss": 1.4742, - "step": 19048 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935571264290063, - "loss": 1.5423, - "step": 19049 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019355678820266522, - "loss": 1.6788, - "step": 19050 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019355644997632416, - "loss": 1.5546, - "step": 19051 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935561117499831, - "loss": 1.6092, - "step": 19052 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019355577352364202, - "loss": 1.5955, - "step": 19053 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019355543529730096, - "loss": 1.5466, - "step": 19054 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935550970709599, - "loss": 1.5335, - "step": 19055 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935547588446188, - "loss": 1.7221, - "step": 19056 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019355442061827775, - "loss": 1.6716, - "step": 19057 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935540823919367, - "loss": 1.6942, - "step": 19058 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935537441655956, - "loss": 1.6499, - "step": 19059 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019355340593925455, - "loss": 1.4167, - "step": 19060 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935530677129135, - "loss": 1.481, - "step": 19061 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935527294865724, - "loss": 1.6698, - "step": 19062 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019355239126023135, - "loss": 1.6208, - "step": 19063 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935520530338903, - "loss": 1.5015, - "step": 19064 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935517148075492, - "loss": 1.6088, - "step": 19065 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019355137658120814, - "loss": 1.7243, - "step": 19066 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019355103835486708, - "loss": 1.7369, - "step": 19067 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193550700128526, - "loss": 1.5879, - "step": 19068 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019355036190218497, - "loss": 1.5939, - "step": 19069 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935500236758439, - "loss": 1.5252, - "step": 19070 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019354968544950282, - "loss": 1.6396, - "step": 19071 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019354934722316176, - "loss": 1.6234, - "step": 19072 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935490089968207, - "loss": 1.6927, - "step": 19073 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019354867077047962, - "loss": 1.5081, - "step": 19074 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019354833254413856, - "loss": 1.4882, - "step": 19075 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935479943177975, - "loss": 1.4483, - "step": 19076 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019354765609145641, - "loss": 1.5006, - "step": 19077 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019354731786511535, - "loss": 1.6928, - "step": 19078 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935469796387743, - "loss": 1.5148, - "step": 19079 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935466414124332, - "loss": 1.5224, - "step": 19080 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019354630318609215, - "loss": 1.6584, - "step": 19081 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935459649597511, - "loss": 1.634, - "step": 19082 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019354562673341, - "loss": 1.743, - "step": 19083 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019354528850706895, - "loss": 1.6416, - "step": 19084 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935449502807279, - "loss": 1.523, - "step": 19085 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935446120543868, - "loss": 1.3744, - "step": 19086 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019354427382804574, - "loss": 1.4793, - "step": 19087 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019354393560170469, - "loss": 1.5265, - "step": 19088 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935435973753636, - "loss": 1.4891, - "step": 19089 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019354325914902254, - "loss": 1.6388, - "step": 19090 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019354292092268148, - "loss": 1.6482, - "step": 19091 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935425826963404, - "loss": 1.5969, - "step": 19092 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019354224446999934, - "loss": 1.5376, - "step": 19093 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019354190624365828, - "loss": 1.4462, - "step": 19094 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935415680173172, - "loss": 1.5757, - "step": 19095 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019354122979097613, - "loss": 1.6449, - "step": 19096 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019354089156463507, - "loss": 1.6123, - "step": 19097 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193540553338294, - "loss": 1.4456, - "step": 19098 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019354021511195293, - "loss": 1.5918, - "step": 19099 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019353987688561187, - "loss": 1.4499, - "step": 19100 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019353953865927078, - "loss": 1.507, - "step": 19101 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019353920043292973, - "loss": 1.5318, - "step": 19102 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019353886220658867, - "loss": 1.482, - "step": 19103 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019353852398024758, - "loss": 1.4426, - "step": 19104 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019353818575390652, - "loss": 1.594, - "step": 19105 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019353784752756546, - "loss": 1.5011, - "step": 19106 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019353750930122438, - "loss": 1.5814, - "step": 19107 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019353717107488332, - "loss": 1.6018, - "step": 19108 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019353683284854226, - "loss": 1.5584, - "step": 19109 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019353649462220117, - "loss": 1.5529, - "step": 19110 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019353615639586011, - "loss": 1.5391, - "step": 19111 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019353581816951906, - "loss": 1.5421, - "step": 19112 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019353547994317797, - "loss": 1.6809, - "step": 19113 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935351417168369, - "loss": 1.5749, - "step": 19114 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019353480349049585, - "loss": 1.6212, - "step": 19115 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019353446526415477, - "loss": 1.5738, - "step": 19116 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935341270378137, - "loss": 1.4168, - "step": 19117 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019353378881147265, - "loss": 1.6221, - "step": 19118 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019353345058513156, - "loss": 1.4685, - "step": 19119 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935331123587905, - "loss": 1.516, - "step": 19120 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019353277413244944, - "loss": 1.6995, - "step": 19121 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019353243590610836, - "loss": 1.7604, - "step": 19122 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935320976797673, - "loss": 1.6905, - "step": 19123 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019353175945342624, - "loss": 1.5163, - "step": 19124 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019353142122708518, - "loss": 1.4372, - "step": 19125 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019353108300074412, - "loss": 1.3725, - "step": 19126 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019353074477440304, - "loss": 1.6818, - "step": 19127 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019353040654806198, - "loss": 1.6556, - "step": 19128 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019353006832172092, - "loss": 1.6971, - "step": 19129 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019352973009537983, - "loss": 1.5646, - "step": 19130 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019352939186903877, - "loss": 1.597, - "step": 19131 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019352905364269772, - "loss": 1.6064, - "step": 19132 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019352871541635663, - "loss": 1.426, - "step": 19133 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019352837719001557, - "loss": 1.5956, - "step": 19134 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935280389636745, - "loss": 1.6102, - "step": 19135 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019352770073733343, - "loss": 1.4997, - "step": 19136 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019352736251099237, - "loss": 1.4456, - "step": 19137 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935270242846513, - "loss": 1.5853, - "step": 19138 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019352668605831022, - "loss": 1.659, - "step": 19139 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019352634783196916, - "loss": 1.5898, - "step": 19140 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935260096056281, - "loss": 1.5786, - "step": 19141 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019352567137928705, - "loss": 1.7377, - "step": 19142 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019352533315294596, - "loss": 1.519, - "step": 19143 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935249949266049, - "loss": 1.7504, - "step": 19144 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019352465670026384, - "loss": 1.5373, - "step": 19145 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019352431847392276, - "loss": 1.5009, - "step": 19146 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935239802475817, - "loss": 1.61, - "step": 19147 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019352364202124064, - "loss": 1.7379, - "step": 19148 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019352330379489955, - "loss": 1.4117, - "step": 19149 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935229655685585, - "loss": 1.5792, - "step": 19150 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019352262734221743, - "loss": 1.7814, - "step": 19151 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019352228911587635, - "loss": 1.6543, - "step": 19152 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935219508895353, - "loss": 1.3983, - "step": 19153 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019352161266319423, - "loss": 1.5758, - "step": 19154 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019352127443685315, - "loss": 1.6947, - "step": 19155 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019352093621051209, - "loss": 1.6857, - "step": 19156 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019352059798417103, - "loss": 1.7267, - "step": 19157 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019352025975782994, - "loss": 1.521, - "step": 19158 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019351992153148888, - "loss": 1.6607, - "step": 19159 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019351958330514782, - "loss": 1.3872, - "step": 19160 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019351924507880674, - "loss": 1.5725, - "step": 19161 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019351890685246568, - "loss": 1.5894, - "step": 19162 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019351856862612462, - "loss": 1.5307, - "step": 19163 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019351823039978353, - "loss": 1.6137, - "step": 19164 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019351789217344248, - "loss": 1.5811, - "step": 19165 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019351755394710142, - "loss": 1.4366, - "step": 19166 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019351721572076033, - "loss": 1.5897, - "step": 19167 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019351687749441927, - "loss": 1.6843, - "step": 19168 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935165392680782, - "loss": 1.6244, - "step": 19169 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019351620104173713, - "loss": 1.5076, - "step": 19170 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019351586281539607, - "loss": 1.5788, - "step": 19171 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193515524589055, - "loss": 1.6111, - "step": 19172 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019351518636271392, - "loss": 1.4171, - "step": 19173 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019351484813637286, - "loss": 1.5111, - "step": 19174 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935145099100318, - "loss": 1.537, - "step": 19175 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019351417168369072, - "loss": 1.4798, - "step": 19176 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019351383345734966, - "loss": 1.5969, - "step": 19177 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935134952310086, - "loss": 1.5774, - "step": 19178 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019351315700466752, - "loss": 1.718, - "step": 19179 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019351281877832646, - "loss": 1.639, - "step": 19180 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935124805519854, - "loss": 1.6031, - "step": 19181 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019351214232564434, - "loss": 1.5908, - "step": 19182 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019351180409930328, - "loss": 1.6067, - "step": 19183 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935114658729622, - "loss": 1.4638, - "step": 19184 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019351112764662114, - "loss": 1.5281, - "step": 19185 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019351078942028008, - "loss": 1.6029, - "step": 19186 - }, - { - "epoch": 0.06, - "learning_rate": 0.000193510451193939, - "loss": 1.6245, - "step": 19187 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019351011296759793, - "loss": 1.5112, - "step": 19188 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019350977474125687, - "loss": 1.5357, - "step": 19189 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935094365149158, - "loss": 1.5902, - "step": 19190 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019350909828857473, - "loss": 1.5806, - "step": 19191 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019350876006223367, - "loss": 1.5645, - "step": 19192 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019350842183589258, - "loss": 1.6679, - "step": 19193 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019350808360955152, - "loss": 1.6117, - "step": 19194 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019350774538321047, - "loss": 1.6239, - "step": 19195 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019350740715686938, - "loss": 1.6331, - "step": 19196 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019350706893052832, - "loss": 1.4128, - "step": 19197 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019350673070418726, - "loss": 1.6448, - "step": 19198 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019350639247784618, - "loss": 1.7098, - "step": 19199 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019350605425150512, - "loss": 1.4241, - "step": 19200 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019350571602516406, - "loss": 1.4273, - "step": 19201 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019350537779882297, - "loss": 1.7365, - "step": 19202 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935050395724819, - "loss": 1.4905, - "step": 19203 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019350470134614085, - "loss": 1.5814, - "step": 19204 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019350436311979977, - "loss": 1.596, - "step": 19205 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935040248934587, - "loss": 1.5961, - "step": 19206 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019350368666711765, - "loss": 1.5796, - "step": 19207 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019350334844077656, - "loss": 1.6531, - "step": 19208 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935030102144355, - "loss": 1.5679, - "step": 19209 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019350267198809445, - "loss": 1.5247, - "step": 19210 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019350233376175336, - "loss": 1.5856, - "step": 19211 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935019955354123, - "loss": 1.6534, - "step": 19212 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019350165730907124, - "loss": 1.6257, - "step": 19213 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019350131908273016, - "loss": 1.5176, - "step": 19214 - }, - { - "epoch": 0.06, - "learning_rate": 0.0001935009808563891, - "loss": 1.6678, - "step": 19215 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019350064263004804, - "loss": 1.5055, - "step": 19216 - }, - { - "epoch": 0.06, - "learning_rate": 0.00019350030440370695, - "loss": 1.5704, - "step": 19217 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934999661773659, - "loss": 1.4591, - "step": 19218 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019349962795102484, - "loss": 1.4747, - "step": 19219 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019349928972468375, - "loss": 1.697, - "step": 19220 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934989514983427, - "loss": 1.6068, - "step": 19221 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019349861327200163, - "loss": 1.7851, - "step": 19222 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019349827504566055, - "loss": 1.4519, - "step": 19223 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934979368193195, - "loss": 1.6271, - "step": 19224 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019349759859297843, - "loss": 1.6424, - "step": 19225 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019349726036663737, - "loss": 1.5972, - "step": 19226 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019349692214029628, - "loss": 1.5122, - "step": 19227 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019349658391395522, - "loss": 1.6086, - "step": 19228 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019349624568761417, - "loss": 1.7088, - "step": 19229 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019349590746127308, - "loss": 1.5595, - "step": 19230 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019349556923493202, - "loss": 1.693, - "step": 19231 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019349523100859096, - "loss": 1.7522, - "step": 19232 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019349489278224988, - "loss": 1.7304, - "step": 19233 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019349455455590882, - "loss": 1.6708, - "step": 19234 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019349421632956776, - "loss": 1.738, - "step": 19235 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019349387810322667, - "loss": 1.6179, - "step": 19236 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019349353987688564, - "loss": 1.5644, - "step": 19237 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019349320165054455, - "loss": 1.6727, - "step": 19238 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934928634242035, - "loss": 1.5873, - "step": 19239 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019349252519786244, - "loss": 1.5551, - "step": 19240 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019349218697152135, - "loss": 1.6332, - "step": 19241 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934918487451803, - "loss": 1.5048, - "step": 19242 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019349151051883923, - "loss": 1.5794, - "step": 19243 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019349117229249815, - "loss": 1.461, - "step": 19244 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934908340661571, - "loss": 1.6177, - "step": 19245 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019349049583981603, - "loss": 1.5124, - "step": 19246 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019349015761347494, - "loss": 1.4922, - "step": 19247 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019348981938713389, - "loss": 1.6171, - "step": 19248 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019348948116079283, - "loss": 1.6846, - "step": 19249 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019348914293445174, - "loss": 1.4987, - "step": 19250 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019348880470811068, - "loss": 1.581, - "step": 19251 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019348846648176962, - "loss": 1.6302, - "step": 19252 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019348812825542854, - "loss": 1.352, - "step": 19253 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019348779002908748, - "loss": 1.6283, - "step": 19254 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019348745180274642, - "loss": 1.7122, - "step": 19255 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019348711357640533, - "loss": 1.6515, - "step": 19256 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019348677535006427, - "loss": 1.4188, - "step": 19257 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019348643712372322, - "loss": 1.5787, - "step": 19258 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019348609889738213, - "loss": 1.51, - "step": 19259 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019348576067104107, - "loss": 1.5219, - "step": 19260 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934854224447, - "loss": 1.4994, - "step": 19261 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019348508421835893, - "loss": 1.6002, - "step": 19262 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019348474599201787, - "loss": 1.6823, - "step": 19263 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934844077656768, - "loss": 1.5351, - "step": 19264 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019348406953933572, - "loss": 1.5329, - "step": 19265 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019348373131299466, - "loss": 1.5373, - "step": 19266 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934833930866536, - "loss": 1.6975, - "step": 19267 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019348305486031252, - "loss": 1.5258, - "step": 19268 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019348271663397146, - "loss": 1.6172, - "step": 19269 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934823784076304, - "loss": 1.6992, - "step": 19270 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019348204018128931, - "loss": 1.4811, - "step": 19271 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019348170195494826, - "loss": 1.462, - "step": 19272 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934813637286072, - "loss": 1.5569, - "step": 19273 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934810255022661, - "loss": 1.542, - "step": 19274 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019348068727592505, - "loss": 1.5442, - "step": 19275 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193480349049584, - "loss": 1.5492, - "step": 19276 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934800108232429, - "loss": 1.6432, - "step": 19277 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019347967259690185, - "loss": 1.5487, - "step": 19278 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934793343705608, - "loss": 1.6299, - "step": 19279 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934789961442197, - "loss": 1.5748, - "step": 19280 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019347865791787864, - "loss": 1.612, - "step": 19281 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019347831969153759, - "loss": 1.6074, - "step": 19282 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934779814651965, - "loss": 1.5382, - "step": 19283 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019347764323885544, - "loss": 1.5972, - "step": 19284 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019347730501251438, - "loss": 1.595, - "step": 19285 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934769667861733, - "loss": 1.5524, - "step": 19286 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019347662855983224, - "loss": 1.6586, - "step": 19287 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019347629033349118, - "loss": 1.5131, - "step": 19288 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934759521071501, - "loss": 1.5647, - "step": 19289 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019347561388080903, - "loss": 1.5593, - "step": 19290 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019347527565446797, - "loss": 1.651, - "step": 19291 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934749374281269, - "loss": 1.6887, - "step": 19292 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019347459920178583, - "loss": 1.6185, - "step": 19293 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934742609754448, - "loss": 1.4995, - "step": 19294 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934739227491037, - "loss": 1.6727, - "step": 19295 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019347358452276265, - "loss": 1.4042, - "step": 19296 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934732462964216, - "loss": 1.5753, - "step": 19297 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934729080700805, - "loss": 1.7301, - "step": 19298 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019347256984373945, - "loss": 1.6227, - "step": 19299 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934722316173984, - "loss": 1.3889, - "step": 19300 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934718933910573, - "loss": 1.526, - "step": 19301 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019347155516471625, - "loss": 1.6126, - "step": 19302 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934712169383752, - "loss": 1.4643, - "step": 19303 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934708787120341, - "loss": 1.4885, - "step": 19304 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019347054048569304, - "loss": 1.5637, - "step": 19305 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019347020225935198, - "loss": 1.4683, - "step": 19306 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934698640330109, - "loss": 1.6567, - "step": 19307 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019346952580666984, - "loss": 1.6237, - "step": 19308 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019346918758032878, - "loss": 1.6066, - "step": 19309 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934688493539877, - "loss": 1.5057, - "step": 19310 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019346851112764663, - "loss": 1.59, - "step": 19311 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019346817290130558, - "loss": 1.5233, - "step": 19312 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934678346749645, - "loss": 1.5138, - "step": 19313 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019346749644862343, - "loss": 1.6311, - "step": 19314 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019346715822228237, - "loss": 1.6901, - "step": 19315 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019346681999594129, - "loss": 1.6041, - "step": 19316 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019346648176960023, - "loss": 1.5292, - "step": 19317 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019346614354325917, - "loss": 1.6275, - "step": 19318 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019346580531691808, - "loss": 1.6106, - "step": 19319 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019346546709057702, - "loss": 1.6658, - "step": 19320 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019346512886423596, - "loss": 1.6099, - "step": 19321 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019346479063789488, - "loss": 1.7041, - "step": 19322 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019346445241155382, - "loss": 1.6534, - "step": 19323 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019346411418521276, - "loss": 1.6056, - "step": 19324 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019346377595887168, - "loss": 1.4877, - "step": 19325 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019346343773253062, - "loss": 1.4407, - "step": 19326 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019346309950618956, - "loss": 1.7584, - "step": 19327 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019346276127984847, - "loss": 1.5344, - "step": 19328 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934624230535074, - "loss": 1.4582, - "step": 19329 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019346208482716635, - "loss": 1.6912, - "step": 19330 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019346174660082527, - "loss": 1.8284, - "step": 19331 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934614083744842, - "loss": 1.614, - "step": 19332 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019346107014814315, - "loss": 1.4823, - "step": 19333 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019346073192180206, - "loss": 1.5458, - "step": 19334 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193460393695461, - "loss": 1.699, - "step": 19335 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019346005546911995, - "loss": 1.564, - "step": 19336 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019345971724277886, - "loss": 1.5283, - "step": 19337 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934593790164378, - "loss": 1.5243, - "step": 19338 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019345904079009674, - "loss": 1.6112, - "step": 19339 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019345870256375566, - "loss": 1.6885, - "step": 19340 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934583643374146, - "loss": 1.6089, - "step": 19341 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019345802611107354, - "loss": 1.7665, - "step": 19342 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019345768788473245, - "loss": 1.5954, - "step": 19343 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934573496583914, - "loss": 1.6057, - "step": 19344 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019345701143205034, - "loss": 1.6285, - "step": 19345 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019345667320570925, - "loss": 1.66, - "step": 19346 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934563349793682, - "loss": 1.6186, - "step": 19347 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019345599675302713, - "loss": 1.6477, - "step": 19348 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019345565852668605, - "loss": 1.6981, - "step": 19349 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019345532030034501, - "loss": 1.6015, - "step": 19350 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019345498207400396, - "loss": 1.7712, - "step": 19351 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019345464384766287, - "loss": 1.6695, - "step": 19352 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934543056213218, - "loss": 1.4868, - "step": 19353 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019345396739498075, - "loss": 1.585, - "step": 19354 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019345362916863967, - "loss": 1.5161, - "step": 19355 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934532909422986, - "loss": 1.5276, - "step": 19356 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019345295271595755, - "loss": 1.5633, - "step": 19357 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019345261448961646, - "loss": 1.643, - "step": 19358 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934522762632754, - "loss": 1.4765, - "step": 19359 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019345193803693434, - "loss": 1.5155, - "step": 19360 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019345159981059326, - "loss": 1.4934, - "step": 19361 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934512615842522, - "loss": 1.389, - "step": 19362 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019345092335791114, - "loss": 1.6135, - "step": 19363 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019345058513157005, - "loss": 1.5151, - "step": 19364 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193450246905229, - "loss": 1.562, - "step": 19365 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019344990867888794, - "loss": 1.5092, - "step": 19366 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019344957045254685, - "loss": 1.7882, - "step": 19367 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934492322262058, - "loss": 1.7598, - "step": 19368 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019344889399986473, - "loss": 1.4236, - "step": 19369 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019344855577352365, - "loss": 1.6334, - "step": 19370 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934482175471826, - "loss": 1.5671, - "step": 19371 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019344787932084153, - "loss": 1.5062, - "step": 19372 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019344754109450044, - "loss": 1.4847, - "step": 19373 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019344720286815938, - "loss": 1.5286, - "step": 19374 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019344686464181833, - "loss": 1.648, - "step": 19375 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019344652641547724, - "loss": 1.7721, - "step": 19376 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019344618818913618, - "loss": 1.4793, - "step": 19377 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019344584996279512, - "loss": 1.5936, - "step": 19378 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019344551173645404, - "loss": 1.5997, - "step": 19379 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019344517351011298, - "loss": 1.6422, - "step": 19380 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019344483528377192, - "loss": 1.695, - "step": 19381 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019344449705743083, - "loss": 1.439, - "step": 19382 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019344415883108977, - "loss": 1.3974, - "step": 19383 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019344382060474871, - "loss": 1.5693, - "step": 19384 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019344348237840763, - "loss": 1.6715, - "step": 19385 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019344314415206657, - "loss": 1.417, - "step": 19386 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934428059257255, - "loss": 1.872, - "step": 19387 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019344246769938442, - "loss": 1.6544, - "step": 19388 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019344212947304337, - "loss": 1.6701, - "step": 19389 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934417912467023, - "loss": 1.6531, - "step": 19390 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019344145302036122, - "loss": 1.4952, - "step": 19391 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019344111479402016, - "loss": 1.5658, - "step": 19392 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934407765676791, - "loss": 1.7177, - "step": 19393 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019344043834133802, - "loss": 1.7101, - "step": 19394 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019344010011499696, - "loss": 1.4232, - "step": 19395 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934397618886559, - "loss": 1.4773, - "step": 19396 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019343942366231481, - "loss": 1.6223, - "step": 19397 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019343908543597375, - "loss": 1.4492, - "step": 19398 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934387472096327, - "loss": 1.5657, - "step": 19399 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934384089832916, - "loss": 1.488, - "step": 19400 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019343807075695055, - "loss": 1.5081, - "step": 19401 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934377325306095, - "loss": 1.513, - "step": 19402 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934373943042684, - "loss": 1.5457, - "step": 19403 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019343705607792735, - "loss": 1.691, - "step": 19404 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934367178515863, - "loss": 1.6893, - "step": 19405 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019343637962524523, - "loss": 1.6119, - "step": 19406 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019343604139890417, - "loss": 1.5516, - "step": 19407 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934357031725631, - "loss": 1.5737, - "step": 19408 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019343536494622203, - "loss": 1.6098, - "step": 19409 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019343502671988097, - "loss": 1.6402, - "step": 19410 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934346884935399, - "loss": 1.6791, - "step": 19411 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019343435026719882, - "loss": 1.4685, - "step": 19412 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019343401204085776, - "loss": 1.536, - "step": 19413 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934336738145167, - "loss": 1.6164, - "step": 19414 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019343333558817562, - "loss": 1.6383, - "step": 19415 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019343299736183456, - "loss": 1.4585, - "step": 19416 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934326591354935, - "loss": 1.6415, - "step": 19417 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019343232090915242, - "loss": 1.4148, - "step": 19418 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019343198268281136, - "loss": 1.574, - "step": 19419 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934316444564703, - "loss": 1.5843, - "step": 19420 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934313062301292, - "loss": 1.5803, - "step": 19421 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019343096800378815, - "loss": 1.8019, - "step": 19422 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934306297774471, - "loss": 1.6927, - "step": 19423 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193430291551106, - "loss": 1.61, - "step": 19424 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019342995332476495, - "loss": 1.6085, - "step": 19425 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934296150984239, - "loss": 1.5101, - "step": 19426 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934292768720828, - "loss": 1.6248, - "step": 19427 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019342893864574175, - "loss": 1.757, - "step": 19428 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019342860041940069, - "loss": 1.7202, - "step": 19429 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934282621930596, - "loss": 1.4252, - "step": 19430 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019342792396671854, - "loss": 1.7399, - "step": 19431 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019342758574037748, - "loss": 1.4571, - "step": 19432 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934272475140364, - "loss": 1.5974, - "step": 19433 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019342690928769534, - "loss": 1.5692, - "step": 19434 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019342657106135428, - "loss": 1.5084, - "step": 19435 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934262328350132, - "loss": 1.6914, - "step": 19436 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019342589460867213, - "loss": 1.5395, - "step": 19437 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019342555638233108, - "loss": 1.5601, - "step": 19438 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019342521815599, - "loss": 1.4484, - "step": 19439 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019342487992964893, - "loss": 1.6089, - "step": 19440 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019342454170330787, - "loss": 1.7139, - "step": 19441 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019342420347696679, - "loss": 1.4442, - "step": 19442 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019342386525062573, - "loss": 1.7846, - "step": 19443 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019342352702428467, - "loss": 1.4828, - "step": 19444 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019342318879794358, - "loss": 1.5404, - "step": 19445 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019342285057160252, - "loss": 1.6023, - "step": 19446 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019342251234526146, - "loss": 1.5168, - "step": 19447 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019342217411892038, - "loss": 1.4734, - "step": 19448 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019342183589257932, - "loss": 1.6453, - "step": 19449 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019342149766623826, - "loss": 1.5892, - "step": 19450 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019342115943989717, - "loss": 1.4755, - "step": 19451 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019342082121355612, - "loss": 1.5194, - "step": 19452 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019342048298721506, - "loss": 1.5319, - "step": 19453 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019342014476087397, - "loss": 1.5772, - "step": 19454 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934198065345329, - "loss": 1.8213, - "step": 19455 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019341946830819185, - "loss": 1.7057, - "step": 19456 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019341913008185077, - "loss": 1.555, - "step": 19457 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934187918555097, - "loss": 1.7481, - "step": 19458 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019341845362916865, - "loss": 1.6779, - "step": 19459 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019341811540282756, - "loss": 1.7353, - "step": 19460 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934177771764865, - "loss": 1.5636, - "step": 19461 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019341743895014545, - "loss": 1.6324, - "step": 19462 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934171007238044, - "loss": 1.6456, - "step": 19463 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019341676249746333, - "loss": 1.5496, - "step": 19464 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019341642427112224, - "loss": 1.5409, - "step": 19465 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019341608604478118, - "loss": 1.5859, - "step": 19466 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019341574781844012, - "loss": 1.7177, - "step": 19467 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019341540959209904, - "loss": 1.474, - "step": 19468 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019341507136575798, - "loss": 1.7351, - "step": 19469 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019341473313941692, - "loss": 1.4529, - "step": 19470 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019341439491307583, - "loss": 1.6603, - "step": 19471 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019341405668673478, - "loss": 1.5812, - "step": 19472 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019341371846039372, - "loss": 1.6819, - "step": 19473 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019341338023405263, - "loss": 1.5212, - "step": 19474 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019341304200771157, - "loss": 1.5959, - "step": 19475 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934127037813705, - "loss": 1.5995, - "step": 19476 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019341236555502943, - "loss": 1.4789, - "step": 19477 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019341202732868837, - "loss": 1.6009, - "step": 19478 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934116891023473, - "loss": 1.4346, - "step": 19479 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019341135087600622, - "loss": 1.4521, - "step": 19480 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019341101264966516, - "loss": 1.5318, - "step": 19481 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934106744233241, - "loss": 1.5227, - "step": 19482 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019341033619698302, - "loss": 1.5845, - "step": 19483 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019340999797064196, - "loss": 1.5483, - "step": 19484 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934096597443009, - "loss": 1.4864, - "step": 19485 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019340932151795982, - "loss": 1.5532, - "step": 19486 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019340898329161876, - "loss": 1.4213, - "step": 19487 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934086450652777, - "loss": 1.5392, - "step": 19488 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934083068389366, - "loss": 1.5432, - "step": 19489 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019340796861259555, - "loss": 1.5485, - "step": 19490 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934076303862545, - "loss": 1.5828, - "step": 19491 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019340729215991344, - "loss": 1.6126, - "step": 19492 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019340695393357235, - "loss": 1.6766, - "step": 19493 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934066157072313, - "loss": 1.4994, - "step": 19494 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019340627748089023, - "loss": 1.5161, - "step": 19495 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019340593925454915, - "loss": 1.493, - "step": 19496 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001934056010282081, - "loss": 1.5697, - "step": 19497 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019340526280186703, - "loss": 1.5867, - "step": 19498 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019340492457552594, - "loss": 1.5295, - "step": 19499 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019340458634918488, - "loss": 1.5131, - "step": 19500 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019340424812284382, - "loss": 1.7409, - "step": 19501 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019340390989650274, - "loss": 1.6458, - "step": 19502 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019340357167016168, - "loss": 1.6396, - "step": 19503 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019340323344382062, - "loss": 1.5449, - "step": 19504 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019340289521747954, - "loss": 1.3519, - "step": 19505 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019340255699113848, - "loss": 1.5648, - "step": 19506 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019340221876479742, - "loss": 1.4156, - "step": 19507 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019340188053845633, - "loss": 1.6332, - "step": 19508 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019340154231211527, - "loss": 1.6276, - "step": 19509 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019340120408577421, - "loss": 1.3328, - "step": 19510 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019340086585943313, - "loss": 1.6878, - "step": 19511 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019340052763309207, - "loss": 1.5053, - "step": 19512 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193400189406751, - "loss": 1.3679, - "step": 19513 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019339985118040992, - "loss": 1.5101, - "step": 19514 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019339951295406887, - "loss": 1.6121, - "step": 19515 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933991747277278, - "loss": 1.361, - "step": 19516 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019339883650138672, - "loss": 1.6661, - "step": 19517 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933984982750457, - "loss": 1.5898, - "step": 19518 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933981600487046, - "loss": 1.5574, - "step": 19519 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019339782182236354, - "loss": 1.445, - "step": 19520 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019339748359602249, - "loss": 1.5814, - "step": 19521 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933971453696814, - "loss": 1.7852, - "step": 19522 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019339680714334034, - "loss": 1.405, - "step": 19523 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019339646891699928, - "loss": 1.5918, - "step": 19524 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933961306906582, - "loss": 1.5822, - "step": 19525 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019339579246431714, - "loss": 1.3652, - "step": 19526 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019339545423797608, - "loss": 1.5929, - "step": 19527 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193395116011635, - "loss": 1.4496, - "step": 19528 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019339477778529393, - "loss": 1.5645, - "step": 19529 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019339443955895287, - "loss": 1.5801, - "step": 19530 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933941013326118, - "loss": 1.7145, - "step": 19531 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019339376310627073, - "loss": 1.6098, - "step": 19532 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019339342487992967, - "loss": 1.5733, - "step": 19533 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019339308665358858, - "loss": 1.6248, - "step": 19534 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019339274842724753, - "loss": 1.5795, - "step": 19535 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019339241020090647, - "loss": 1.4324, - "step": 19536 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019339207197456538, - "loss": 1.478, - "step": 19537 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019339173374822432, - "loss": 1.5719, - "step": 19538 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019339139552188326, - "loss": 1.4471, - "step": 19539 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019339105729554218, - "loss": 1.5747, - "step": 19540 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019339071906920112, - "loss": 1.5042, - "step": 19541 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019339038084286006, - "loss": 1.4365, - "step": 19542 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019339004261651897, - "loss": 1.5404, - "step": 19543 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019338970439017791, - "loss": 1.5544, - "step": 19544 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019338936616383686, - "loss": 1.586, - "step": 19545 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019338902793749577, - "loss": 1.6034, - "step": 19546 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933886897111547, - "loss": 1.7102, - "step": 19547 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019338835148481365, - "loss": 1.5715, - "step": 19548 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019338801325847257, - "loss": 1.7811, - "step": 19549 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933876750321315, - "loss": 1.5273, - "step": 19550 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019338733680579045, - "loss": 1.5122, - "step": 19551 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019338699857944936, - "loss": 1.5051, - "step": 19552 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933866603531083, - "loss": 1.6664, - "step": 19553 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019338632212676724, - "loss": 1.6592, - "step": 19554 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019338598390042616, - "loss": 1.5148, - "step": 19555 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933856456740851, - "loss": 1.3591, - "step": 19556 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019338530744774404, - "loss": 1.4837, - "step": 19557 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019338496922140295, - "loss": 1.6841, - "step": 19558 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933846309950619, - "loss": 1.4601, - "step": 19559 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019338429276872084, - "loss": 1.586, - "step": 19560 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019338395454237975, - "loss": 1.6572, - "step": 19561 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933836163160387, - "loss": 1.6525, - "step": 19562 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019338327808969763, - "loss": 1.6679, - "step": 19563 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019338293986335655, - "loss": 1.664, - "step": 19564 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933826016370155, - "loss": 1.4919, - "step": 19565 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019338226341067443, - "loss": 1.4772, - "step": 19566 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019338192518433334, - "loss": 1.5623, - "step": 19567 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019338158695799228, - "loss": 1.5256, - "step": 19568 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019338124873165123, - "loss": 1.6638, - "step": 19569 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019338091050531014, - "loss": 1.6587, - "step": 19570 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019338057227896908, - "loss": 1.5123, - "step": 19571 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019338023405262802, - "loss": 1.7805, - "step": 19572 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019337989582628694, - "loss": 1.5703, - "step": 19573 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933795575999459, - "loss": 1.6408, - "step": 19574 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019337921937360485, - "loss": 1.5453, - "step": 19575 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019337888114726376, - "loss": 1.5269, - "step": 19576 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933785429209227, - "loss": 1.4636, - "step": 19577 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019337820469458164, - "loss": 1.5071, - "step": 19578 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019337786646824056, - "loss": 1.6131, - "step": 19579 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933775282418995, - "loss": 1.704, - "step": 19580 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019337719001555844, - "loss": 1.5433, - "step": 19581 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019337685178921735, - "loss": 1.4708, - "step": 19582 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933765135628763, - "loss": 1.5586, - "step": 19583 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019337617533653523, - "loss": 1.6961, - "step": 19584 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019337583711019415, - "loss": 1.5953, - "step": 19585 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933754988838531, - "loss": 1.6732, - "step": 19586 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019337516065751203, - "loss": 1.596, - "step": 19587 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019337482243117095, - "loss": 1.574, - "step": 19588 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019337448420482989, - "loss": 1.4995, - "step": 19589 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019337414597848883, - "loss": 1.4958, - "step": 19590 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019337380775214774, - "loss": 1.7087, - "step": 19591 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019337346952580668, - "loss": 1.5721, - "step": 19592 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019337313129946562, - "loss": 1.6165, - "step": 19593 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019337279307312454, - "loss": 1.5685, - "step": 19594 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019337245484678348, - "loss": 1.5357, - "step": 19595 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019337211662044242, - "loss": 1.574, - "step": 19596 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019337177839410133, - "loss": 1.5954, - "step": 19597 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019337144016776028, - "loss": 1.5485, - "step": 19598 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019337110194141922, - "loss": 1.4984, - "step": 19599 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019337076371507813, - "loss": 1.6225, - "step": 19600 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019337042548873707, - "loss": 1.8263, - "step": 19601 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193370087262396, - "loss": 1.6224, - "step": 19602 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019336974903605493, - "loss": 1.4382, - "step": 19603 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019336941080971387, - "loss": 1.6695, - "step": 19604 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933690725833728, - "loss": 1.6079, - "step": 19605 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019336873435703172, - "loss": 1.5697, - "step": 19606 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019336839613069066, - "loss": 1.6525, - "step": 19607 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933680579043496, - "loss": 1.5658, - "step": 19608 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019336771967800852, - "loss": 1.4797, - "step": 19609 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019336738145166746, - "loss": 1.6081, - "step": 19610 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933670432253264, - "loss": 1.6131, - "step": 19611 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019336670499898532, - "loss": 1.6668, - "step": 19612 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019336636677264426, - "loss": 1.6365, - "step": 19613 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933660285463032, - "loss": 1.6192, - "step": 19614 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933656903199621, - "loss": 1.5249, - "step": 19615 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019336535209362105, - "loss": 1.4899, - "step": 19616 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019336501386728, - "loss": 1.6603, - "step": 19617 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933646756409389, - "loss": 1.5207, - "step": 19618 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019336433741459785, - "loss": 1.4382, - "step": 19619 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933639991882568, - "loss": 1.464, - "step": 19620 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933636609619157, - "loss": 1.5833, - "step": 19621 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019336332273557465, - "loss": 1.6653, - "step": 19622 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933629845092336, - "loss": 1.5732, - "step": 19623 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933626462828925, - "loss": 1.5147, - "step": 19624 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019336230805655144, - "loss": 1.6269, - "step": 19625 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019336196983021038, - "loss": 1.7473, - "step": 19626 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933616316038693, - "loss": 1.481, - "step": 19627 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019336129337752824, - "loss": 1.5369, - "step": 19628 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019336095515118718, - "loss": 1.6901, - "step": 19629 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933606169248461, - "loss": 1.6203, - "step": 19630 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019336027869850506, - "loss": 1.6485, - "step": 19631 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193359940472164, - "loss": 1.6423, - "step": 19632 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019335960224582292, - "loss": 1.6834, - "step": 19633 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019335926401948186, - "loss": 1.6414, - "step": 19634 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933589257931408, - "loss": 1.4368, - "step": 19635 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933585875667997, - "loss": 1.5432, - "step": 19636 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019335824934045865, - "loss": 1.5651, - "step": 19637 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933579111141176, - "loss": 1.6489, - "step": 19638 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933575728877765, - "loss": 1.5578, - "step": 19639 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019335723466143545, - "loss": 1.5584, - "step": 19640 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933568964350944, - "loss": 1.6541, - "step": 19641 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933565582087533, - "loss": 1.6685, - "step": 19642 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019335621998241225, - "loss": 1.6884, - "step": 19643 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933558817560712, - "loss": 1.5447, - "step": 19644 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933555435297301, - "loss": 1.6526, - "step": 19645 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019335520530338904, - "loss": 1.5721, - "step": 19646 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019335486707704798, - "loss": 1.5512, - "step": 19647 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933545288507069, - "loss": 1.5265, - "step": 19648 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019335419062436584, - "loss": 1.805, - "step": 19649 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019335385239802478, - "loss": 1.6064, - "step": 19650 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933535141716837, - "loss": 1.4511, - "step": 19651 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019335317594534264, - "loss": 1.5923, - "step": 19652 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019335283771900158, - "loss": 1.5574, - "step": 19653 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933524994926605, - "loss": 1.6397, - "step": 19654 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019335216126631943, - "loss": 1.4603, - "step": 19655 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019335182303997837, - "loss": 1.4112, - "step": 19656 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933514848136373, - "loss": 1.5537, - "step": 19657 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019335114658729623, - "loss": 1.6388, - "step": 19658 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019335080836095517, - "loss": 1.5341, - "step": 19659 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019335047013461408, - "loss": 1.6125, - "step": 19660 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019335013190827302, - "loss": 1.5973, - "step": 19661 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019334979368193197, - "loss": 1.6993, - "step": 19662 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019334945545559088, - "loss": 1.4899, - "step": 19663 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019334911722924982, - "loss": 1.5437, - "step": 19664 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019334877900290876, - "loss": 1.5397, - "step": 19665 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019334844077656768, - "loss": 1.4601, - "step": 19666 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019334810255022662, - "loss": 1.7681, - "step": 19667 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019334776432388556, - "loss": 1.6219, - "step": 19668 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019334742609754447, - "loss": 1.4981, - "step": 19669 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019334708787120341, - "loss": 1.6494, - "step": 19670 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019334674964486235, - "loss": 1.5763, - "step": 19671 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019334641141852127, - "loss": 1.6238, - "step": 19672 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933460731921802, - "loss": 1.6634, - "step": 19673 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019334573496583915, - "loss": 1.5656, - "step": 19674 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019334539673949807, - "loss": 1.6589, - "step": 19675 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193345058513157, - "loss": 1.5626, - "step": 19676 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019334472028681595, - "loss": 1.6788, - "step": 19677 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019334438206047486, - "loss": 1.6109, - "step": 19678 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933440438341338, - "loss": 1.5478, - "step": 19679 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019334370560779274, - "loss": 1.5468, - "step": 19680 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019334336738145166, - "loss": 1.3705, - "step": 19681 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933430291551106, - "loss": 1.6519, - "step": 19682 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019334269092876954, - "loss": 1.4706, - "step": 19683 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019334235270242845, - "loss": 1.5876, - "step": 19684 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933420144760874, - "loss": 1.7764, - "step": 19685 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019334167624974634, - "loss": 1.4116, - "step": 19686 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019334133802340528, - "loss": 1.5634, - "step": 19687 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019334099979706422, - "loss": 1.5311, - "step": 19688 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019334066157072316, - "loss": 1.4231, - "step": 19689 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019334032334438207, - "loss": 1.5473, - "step": 19690 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019333998511804102, - "loss": 1.7016, - "step": 19691 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019333964689169996, - "loss": 1.6837, - "step": 19692 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019333930866535887, - "loss": 1.6631, - "step": 19693 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933389704390178, - "loss": 1.7466, - "step": 19694 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019333863221267675, - "loss": 1.6345, - "step": 19695 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019333829398633567, - "loss": 1.4954, - "step": 19696 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933379557599946, - "loss": 1.6086, - "step": 19697 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019333761753365355, - "loss": 1.3831, - "step": 19698 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019333727930731246, - "loss": 1.5738, - "step": 19699 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933369410809714, - "loss": 1.5748, - "step": 19700 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019333660285463035, - "loss": 1.6184, - "step": 19701 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019333626462828926, - "loss": 1.6155, - "step": 19702 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933359264019482, - "loss": 1.6645, - "step": 19703 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019333558817560714, - "loss": 1.6386, - "step": 19704 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019333524994926606, - "loss": 1.6795, - "step": 19705 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193334911722925, - "loss": 1.4273, - "step": 19706 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019333457349658394, - "loss": 1.627, - "step": 19707 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019333423527024285, - "loss": 1.4582, - "step": 19708 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933338970439018, - "loss": 1.6045, - "step": 19709 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019333355881756073, - "loss": 1.6026, - "step": 19710 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019333322059121965, - "loss": 1.687, - "step": 19711 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933328823648786, - "loss": 1.449, - "step": 19712 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019333254413853753, - "loss": 1.5316, - "step": 19713 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019333220591219644, - "loss": 1.4799, - "step": 19714 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019333186768585539, - "loss": 1.5503, - "step": 19715 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019333152945951433, - "loss": 1.5333, - "step": 19716 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019333119123317324, - "loss": 1.5827, - "step": 19717 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019333085300683218, - "loss": 1.4855, - "step": 19718 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019333051478049112, - "loss": 1.6392, - "step": 19719 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019333017655415004, - "loss": 1.5552, - "step": 19720 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019332983832780898, - "loss": 1.6186, - "step": 19721 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019332950010146792, - "loss": 1.631, - "step": 19722 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019332916187512683, - "loss": 1.6269, - "step": 19723 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019332882364878577, - "loss": 1.5439, - "step": 19724 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019332848542244472, - "loss": 1.6544, - "step": 19725 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019332814719610363, - "loss": 1.4795, - "step": 19726 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019332780896976257, - "loss": 1.444, - "step": 19727 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933274707434215, - "loss": 1.5967, - "step": 19728 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019332713251708043, - "loss": 1.6536, - "step": 19729 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019332679429073937, - "loss": 1.4958, - "step": 19730 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933264560643983, - "loss": 1.7379, - "step": 19731 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019332611783805722, - "loss": 1.6996, - "step": 19732 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019332577961171616, - "loss": 1.5926, - "step": 19733 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933254413853751, - "loss": 1.6181, - "step": 19734 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019332510315903402, - "loss": 1.6516, - "step": 19735 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019332476493269296, - "loss": 1.6426, - "step": 19736 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933244267063519, - "loss": 1.6411, - "step": 19737 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019332408848001082, - "loss": 1.4878, - "step": 19738 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019332375025366976, - "loss": 1.5577, - "step": 19739 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933234120273287, - "loss": 1.7981, - "step": 19740 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933230738009876, - "loss": 1.5532, - "step": 19741 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019332273557464655, - "loss": 1.6584, - "step": 19742 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933223973483055, - "loss": 1.6896, - "step": 19743 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019332205912196443, - "loss": 1.7177, - "step": 19744 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019332172089562338, - "loss": 1.6009, - "step": 19745 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933213826692823, - "loss": 1.56, - "step": 19746 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019332104444294123, - "loss": 1.6203, - "step": 19747 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019332070621660017, - "loss": 1.6109, - "step": 19748 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019332036799025909, - "loss": 1.6068, - "step": 19749 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019332002976391803, - "loss": 1.5572, - "step": 19750 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019331969153757697, - "loss": 1.5426, - "step": 19751 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019331935331123588, - "loss": 1.57, - "step": 19752 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019331901508489482, - "loss": 1.6137, - "step": 19753 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019331867685855376, - "loss": 1.7191, - "step": 19754 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019331833863221268, - "loss": 1.5095, - "step": 19755 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019331800040587162, - "loss": 1.646, - "step": 19756 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019331766217953056, - "loss": 1.3904, - "step": 19757 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019331732395318948, - "loss": 1.5955, - "step": 19758 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019331698572684842, - "loss": 1.5944, - "step": 19759 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019331664750050736, - "loss": 1.5583, - "step": 19760 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933163092741663, - "loss": 1.5414, - "step": 19761 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933159710478252, - "loss": 1.5348, - "step": 19762 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019331563282148415, - "loss": 1.5725, - "step": 19763 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933152945951431, - "loss": 1.556, - "step": 19764 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193314956368802, - "loss": 1.6678, - "step": 19765 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019331461814246095, - "loss": 1.5685, - "step": 19766 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933142799161199, - "loss": 1.6359, - "step": 19767 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933139416897788, - "loss": 1.6219, - "step": 19768 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019331360346343775, - "loss": 1.4026, - "step": 19769 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933132652370967, - "loss": 1.4875, - "step": 19770 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933129270107556, - "loss": 1.5755, - "step": 19771 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019331258878441454, - "loss": 1.6273, - "step": 19772 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019331225055807348, - "loss": 1.5322, - "step": 19773 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933119123317324, - "loss": 1.4372, - "step": 19774 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019331157410539134, - "loss": 1.5877, - "step": 19775 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019331123587905028, - "loss": 1.496, - "step": 19776 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933108976527092, - "loss": 1.4655, - "step": 19777 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019331055942636814, - "loss": 1.6392, - "step": 19778 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019331022120002708, - "loss": 1.6229, - "step": 19779 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193309882973686, - "loss": 1.5624, - "step": 19780 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019330954474734493, - "loss": 1.4982, - "step": 19781 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019330920652100387, - "loss": 1.5521, - "step": 19782 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933088682946628, - "loss": 1.5815, - "step": 19783 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019330853006832173, - "loss": 1.6275, - "step": 19784 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019330819184198067, - "loss": 1.6225, - "step": 19785 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019330785361563958, - "loss": 1.5095, - "step": 19786 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019330751538929852, - "loss": 1.6581, - "step": 19787 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019330717716295747, - "loss": 1.6618, - "step": 19788 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019330683893661638, - "loss": 1.5576, - "step": 19789 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019330650071027532, - "loss": 1.6347, - "step": 19790 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019330616248393426, - "loss": 1.5223, - "step": 19791 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019330582425759318, - "loss": 1.5258, - "step": 19792 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019330548603125212, - "loss": 1.6082, - "step": 19793 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019330514780491106, - "loss": 1.6779, - "step": 19794 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019330480957856997, - "loss": 1.6564, - "step": 19795 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933044713522289, - "loss": 1.3944, - "step": 19796 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019330413312588785, - "loss": 1.4482, - "step": 19797 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019330379489954677, - "loss": 1.7079, - "step": 19798 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019330345667320574, - "loss": 1.4575, - "step": 19799 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019330311844686465, - "loss": 1.6145, - "step": 19800 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933027802205236, - "loss": 1.7559, - "step": 19801 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019330244199418253, - "loss": 1.5812, - "step": 19802 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019330210376784145, - "loss": 1.5374, - "step": 19803 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001933017655415004, - "loss": 1.6273, - "step": 19804 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019330142731515933, - "loss": 1.4905, - "step": 19805 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019330108908881824, - "loss": 1.614, - "step": 19806 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019330075086247718, - "loss": 1.5452, - "step": 19807 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019330041263613613, - "loss": 1.4713, - "step": 19808 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019330007440979504, - "loss": 1.5091, - "step": 19809 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019329973618345398, - "loss": 1.5225, - "step": 19810 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019329939795711292, - "loss": 1.5667, - "step": 19811 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019329905973077184, - "loss": 1.6156, - "step": 19812 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019329872150443078, - "loss": 1.7394, - "step": 19813 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019329838327808972, - "loss": 1.6089, - "step": 19814 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019329804505174863, - "loss": 1.5277, - "step": 19815 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019329770682540757, - "loss": 1.5415, - "step": 19816 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019329736859906651, - "loss": 1.6544, - "step": 19817 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019329703037272543, - "loss": 1.6841, - "step": 19818 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019329669214638437, - "loss": 1.5197, - "step": 19819 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932963539200433, - "loss": 1.6124, - "step": 19820 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019329601569370222, - "loss": 1.6285, - "step": 19821 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019329567746736117, - "loss": 1.6523, - "step": 19822 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932953392410201, - "loss": 1.3774, - "step": 19823 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019329500101467902, - "loss": 1.5226, - "step": 19824 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019329466278833796, - "loss": 1.5743, - "step": 19825 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932943245619969, - "loss": 1.5135, - "step": 19826 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019329398633565582, - "loss": 1.7115, - "step": 19827 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019329364810931476, - "loss": 1.626, - "step": 19828 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932933098829737, - "loss": 1.6266, - "step": 19829 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019329297165663261, - "loss": 1.656, - "step": 19830 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019329263343029155, - "loss": 1.6243, - "step": 19831 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932922952039505, - "loss": 1.5844, - "step": 19832 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932919569776094, - "loss": 1.6027, - "step": 19833 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019329161875126835, - "loss": 1.6543, - "step": 19834 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932912805249273, - "loss": 1.5982, - "step": 19835 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932909422985862, - "loss": 1.5799, - "step": 19836 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019329060407224515, - "loss": 1.4157, - "step": 19837 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932902658459041, - "loss": 1.3188, - "step": 19838 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193289927619563, - "loss": 1.6233, - "step": 19839 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019328958939322194, - "loss": 1.6041, - "step": 19840 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019328925116688089, - "loss": 1.7251, - "step": 19841 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932889129405398, - "loss": 1.724, - "step": 19842 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019328857471419874, - "loss": 1.4156, - "step": 19843 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019328823648785768, - "loss": 1.6049, - "step": 19844 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019328789826151662, - "loss": 1.562, - "step": 19845 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019328756003517554, - "loss": 1.6066, - "step": 19846 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019328722180883448, - "loss": 1.7545, - "step": 19847 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019328688358249342, - "loss": 1.7072, - "step": 19848 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019328654535615233, - "loss": 1.7226, - "step": 19849 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019328620712981127, - "loss": 1.4671, - "step": 19850 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019328586890347022, - "loss": 1.6759, - "step": 19851 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019328553067712913, - "loss": 1.4389, - "step": 19852 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019328519245078807, - "loss": 1.5417, - "step": 19853 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193284854224447, - "loss": 1.6065, - "step": 19854 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019328451599810595, - "loss": 1.551, - "step": 19855 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932841777717649, - "loss": 1.7162, - "step": 19856 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932838395454238, - "loss": 1.544, - "step": 19857 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019328350131908275, - "loss": 1.588, - "step": 19858 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932831630927417, - "loss": 1.5651, - "step": 19859 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932828248664006, - "loss": 1.5937, - "step": 19860 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019328248664005955, - "loss": 1.5092, - "step": 19861 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019328214841371849, - "loss": 1.6155, - "step": 19862 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932818101873774, - "loss": 1.5977, - "step": 19863 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019328147196103634, - "loss": 1.5847, - "step": 19864 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019328113373469528, - "loss": 1.7895, - "step": 19865 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932807955083542, - "loss": 1.7378, - "step": 19866 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019328045728201314, - "loss": 1.5578, - "step": 19867 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019328011905567208, - "loss": 1.6471, - "step": 19868 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193279780829331, - "loss": 1.609, - "step": 19869 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019327944260298993, - "loss": 1.5477, - "step": 19870 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019327910437664888, - "loss": 1.6684, - "step": 19871 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932787661503078, - "loss": 1.5049, - "step": 19872 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019327842792396673, - "loss": 1.5591, - "step": 19873 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019327808969762567, - "loss": 1.5624, - "step": 19874 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019327775147128459, - "loss": 1.5948, - "step": 19875 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019327741324494353, - "loss": 1.5111, - "step": 19876 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019327707501860247, - "loss": 1.5638, - "step": 19877 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019327673679226138, - "loss": 1.6682, - "step": 19878 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019327639856592032, - "loss": 1.5796, - "step": 19879 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019327606033957926, - "loss": 1.516, - "step": 19880 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019327572211323818, - "loss": 1.6509, - "step": 19881 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019327538388689712, - "loss": 1.5325, - "step": 19882 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019327504566055606, - "loss": 1.578, - "step": 19883 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019327470743421497, - "loss": 1.5932, - "step": 19884 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019327436920787392, - "loss": 1.3689, - "step": 19885 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019327403098153286, - "loss": 1.7411, - "step": 19886 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019327369275519177, - "loss": 1.5921, - "step": 19887 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932733545288507, - "loss": 1.5099, - "step": 19888 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019327301630250965, - "loss": 1.6201, - "step": 19889 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019327267807616857, - "loss": 1.6284, - "step": 19890 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932723398498275, - "loss": 1.5352, - "step": 19891 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019327200162348645, - "loss": 1.6183, - "step": 19892 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019327166339714536, - "loss": 1.5979, - "step": 19893 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932713251708043, - "loss": 1.6939, - "step": 19894 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019327098694446325, - "loss": 1.5336, - "step": 19895 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019327064871812216, - "loss": 1.6157, - "step": 19896 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932703104917811, - "loss": 1.3882, - "step": 19897 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019326997226544004, - "loss": 1.7541, - "step": 19898 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019326963403909896, - "loss": 1.4814, - "step": 19899 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932692958127579, - "loss": 1.4658, - "step": 19900 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019326895758641684, - "loss": 1.4445, - "step": 19901 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019326861936007575, - "loss": 1.4601, - "step": 19902 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932682811337347, - "loss": 1.3458, - "step": 19903 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019326794290739363, - "loss": 1.6165, - "step": 19904 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019326760468105255, - "loss": 1.7015, - "step": 19905 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932672664547115, - "loss": 1.6801, - "step": 19906 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019326692822837043, - "loss": 1.5795, - "step": 19907 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019326659000202935, - "loss": 1.4322, - "step": 19908 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019326625177568829, - "loss": 1.548, - "step": 19909 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019326591354934723, - "loss": 1.5876, - "step": 19910 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019326557532300617, - "loss": 1.645, - "step": 19911 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932652370966651, - "loss": 1.5733, - "step": 19912 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019326489887032405, - "loss": 1.651, - "step": 19913 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019326456064398296, - "loss": 1.3176, - "step": 19914 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932642224176419, - "loss": 1.6715, - "step": 19915 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019326388419130085, - "loss": 1.4637, - "step": 19916 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019326354596495976, - "loss": 1.5543, - "step": 19917 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932632077386187, - "loss": 1.4502, - "step": 19918 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019326286951227764, - "loss": 1.7305, - "step": 19919 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019326253128593656, - "loss": 1.6697, - "step": 19920 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932621930595955, - "loss": 1.5502, - "step": 19921 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019326185483325444, - "loss": 1.5154, - "step": 19922 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019326151660691335, - "loss": 1.4632, - "step": 19923 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932611783805723, - "loss": 1.5297, - "step": 19924 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019326084015423124, - "loss": 1.6708, - "step": 19925 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019326050192789015, - "loss": 1.4915, - "step": 19926 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932601637015491, - "loss": 1.5215, - "step": 19927 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019325982547520803, - "loss": 1.5636, - "step": 19928 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019325948724886695, - "loss": 1.6069, - "step": 19929 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932591490225259, - "loss": 1.4819, - "step": 19930 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019325881079618483, - "loss": 1.3489, - "step": 19931 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019325847256984374, - "loss": 1.5191, - "step": 19932 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019325813434350268, - "loss": 1.5585, - "step": 19933 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019325779611716162, - "loss": 1.6786, - "step": 19934 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019325745789082054, - "loss": 1.7188, - "step": 19935 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019325711966447948, - "loss": 1.657, - "step": 19936 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019325678143813842, - "loss": 1.6348, - "step": 19937 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019325644321179734, - "loss": 1.7421, - "step": 19938 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019325610498545628, - "loss": 1.6851, - "step": 19939 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019325576675911522, - "loss": 1.5442, - "step": 19940 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019325542853277413, - "loss": 1.6168, - "step": 19941 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019325509030643307, - "loss": 1.6172, - "step": 19942 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019325475208009201, - "loss": 1.5777, - "step": 19943 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019325441385375093, - "loss": 1.6709, - "step": 19944 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019325407562740987, - "loss": 1.748, - "step": 19945 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932537374010688, - "loss": 1.531, - "step": 19946 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019325339917472772, - "loss": 1.613, - "step": 19947 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019325306094838667, - "loss": 1.4752, - "step": 19948 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932527227220456, - "loss": 1.5285, - "step": 19949 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019325238449570452, - "loss": 1.4822, - "step": 19950 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019325204626936346, - "loss": 1.3842, - "step": 19951 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932517080430224, - "loss": 1.6613, - "step": 19952 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019325136981668132, - "loss": 1.4939, - "step": 19953 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019325103159034026, - "loss": 1.4657, - "step": 19954 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932506933639992, - "loss": 1.6649, - "step": 19955 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932503551376581, - "loss": 1.5762, - "step": 19956 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019325001691131705, - "loss": 1.4, - "step": 19957 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193249678684976, - "loss": 1.4833, - "step": 19958 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932493404586349, - "loss": 1.5048, - "step": 19959 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019324900223229385, - "loss": 1.558, - "step": 19960 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932486640059528, - "loss": 1.6993, - "step": 19961 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932483257796117, - "loss": 1.578, - "step": 19962 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019324798755327065, - "loss": 1.4324, - "step": 19963 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932476493269296, - "loss": 1.6292, - "step": 19964 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932473111005885, - "loss": 1.5981, - "step": 19965 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019324697287424744, - "loss": 1.6024, - "step": 19966 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019324663464790638, - "loss": 1.6094, - "step": 19967 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019324629642156533, - "loss": 1.5444, - "step": 19968 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019324595819522427, - "loss": 1.8024, - "step": 19969 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932456199688832, - "loss": 1.5421, - "step": 19970 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019324528174254212, - "loss": 1.7761, - "step": 19971 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019324494351620106, - "loss": 1.7572, - "step": 19972 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019324460528986, - "loss": 1.4763, - "step": 19973 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019324426706351892, - "loss": 1.5365, - "step": 19974 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019324392883717786, - "loss": 1.4669, - "step": 19975 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932435906108368, - "loss": 1.5724, - "step": 19976 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019324325238449571, - "loss": 1.4021, - "step": 19977 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019324291415815466, - "loss": 1.5277, - "step": 19978 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932425759318136, - "loss": 1.679, - "step": 19979 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932422377054725, - "loss": 1.3536, - "step": 19980 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019324189947913145, - "loss": 1.5752, - "step": 19981 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932415612527904, - "loss": 1.5699, - "step": 19982 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932412230264493, - "loss": 1.5544, - "step": 19983 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019324088480010825, - "loss": 1.548, - "step": 19984 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932405465737672, - "loss": 1.7146, - "step": 19985 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932402083474261, - "loss": 1.4335, - "step": 19986 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019323987012108504, - "loss": 1.6349, - "step": 19987 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019323953189474399, - "loss": 1.6037, - "step": 19988 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932391936684029, - "loss": 1.6013, - "step": 19989 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019323885544206184, - "loss": 1.4304, - "step": 19990 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019323851721572078, - "loss": 1.5684, - "step": 19991 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932381789893797, - "loss": 1.6417, - "step": 19992 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019323784076303864, - "loss": 1.7894, - "step": 19993 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019323750253669758, - "loss": 1.563, - "step": 19994 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932371643103565, - "loss": 1.5125, - "step": 19995 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019323682608401543, - "loss": 1.6311, - "step": 19996 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019323648785767437, - "loss": 1.6896, - "step": 19997 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932361496313333, - "loss": 1.6038, - "step": 19998 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019323581140499223, - "loss": 1.6815, - "step": 19999 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019323547317865117, - "loss": 1.7366, - "step": 20000 - }, - { - "epoch": 0.07, - "eval_loss": 1.5273061899038463, - "mrc_em": 26.0, - "mrc_f1": 33.019047619047626, - "nli_acc": 0.03, - "step": 20000, - "sts_f1": 0.6846846846846848, - "ynat_f1": 0.6904304627669371 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019323513495231008, - "loss": 1.6481, - "step": 20001 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019323479672596903, - "loss": 1.6622, - "step": 20002 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019323445849962797, - "loss": 1.5449, - "step": 20003 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019323412027328688, - "loss": 1.5995, - "step": 20004 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019323378204694582, - "loss": 1.5543, - "step": 20005 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019323344382060476, - "loss": 1.7, - "step": 20006 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019323310559426368, - "loss": 1.493, - "step": 20007 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019323276736792262, - "loss": 1.7049, - "step": 20008 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019323242914158156, - "loss": 1.3741, - "step": 20009 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019323209091524047, - "loss": 1.568, - "step": 20010 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019323175268889942, - "loss": 1.5452, - "step": 20011 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019323141446255836, - "loss": 1.4466, - "step": 20012 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019323107623621727, - "loss": 1.3845, - "step": 20013 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932307380098762, - "loss": 1.5564, - "step": 20014 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019323039978353515, - "loss": 1.4019, - "step": 20015 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019323006155719407, - "loss": 1.5556, - "step": 20016 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193229723330853, - "loss": 1.5618, - "step": 20017 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019322938510451195, - "loss": 1.664, - "step": 20018 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019322904687817086, - "loss": 1.6105, - "step": 20019 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932287086518298, - "loss": 1.4957, - "step": 20020 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019322837042548875, - "loss": 1.516, - "step": 20021 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019322803219914766, - "loss": 1.5234, - "step": 20022 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932276939728066, - "loss": 1.3911, - "step": 20023 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019322735574646554, - "loss": 1.598, - "step": 20024 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019322701752012448, - "loss": 1.6448, - "step": 20025 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019322667929378342, - "loss": 1.6078, - "step": 20026 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019322634106744236, - "loss": 1.5146, - "step": 20027 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019322600284110128, - "loss": 1.5611, - "step": 20028 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019322566461476022, - "loss": 1.6804, - "step": 20029 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019322532638841916, - "loss": 1.622, - "step": 20030 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019322498816207808, - "loss": 1.5402, - "step": 20031 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019322464993573702, - "loss": 1.429, - "step": 20032 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019322431170939596, - "loss": 1.5874, - "step": 20033 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019322397348305487, - "loss": 1.6453, - "step": 20034 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932236352567138, - "loss": 1.7223, - "step": 20035 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019322329703037275, - "loss": 1.5125, - "step": 20036 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019322295880403167, - "loss": 1.5838, - "step": 20037 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932226205776906, - "loss": 1.5189, - "step": 20038 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019322228235134955, - "loss": 1.751, - "step": 20039 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019322194412500846, - "loss": 1.5708, - "step": 20040 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932216058986674, - "loss": 1.5221, - "step": 20041 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019322126767232635, - "loss": 1.3623, - "step": 20042 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019322092944598526, - "loss": 1.5153, - "step": 20043 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932205912196442, - "loss": 1.7103, - "step": 20044 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019322025299330314, - "loss": 1.6157, - "step": 20045 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019321991476696206, - "loss": 1.628, - "step": 20046 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193219576540621, - "loss": 1.6567, - "step": 20047 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019321923831427994, - "loss": 1.5429, - "step": 20048 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019321890008793885, - "loss": 1.6568, - "step": 20049 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932185618615978, - "loss": 1.5826, - "step": 20050 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019321822363525674, - "loss": 1.4084, - "step": 20051 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019321788540891565, - "loss": 1.5417, - "step": 20052 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932175471825746, - "loss": 1.6739, - "step": 20053 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019321720895623353, - "loss": 1.6138, - "step": 20054 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019321687072989245, - "loss": 1.5426, - "step": 20055 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932165325035514, - "loss": 1.3765, - "step": 20056 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019321619427721033, - "loss": 1.7042, - "step": 20057 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019321585605086924, - "loss": 1.5931, - "step": 20058 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019321551782452818, - "loss": 1.6113, - "step": 20059 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019321517959818712, - "loss": 1.7275, - "step": 20060 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019321484137184604, - "loss": 1.4641, - "step": 20061 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019321450314550498, - "loss": 1.4486, - "step": 20062 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019321416491916392, - "loss": 1.6005, - "step": 20063 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019321382669282283, - "loss": 1.6946, - "step": 20064 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019321348846648178, - "loss": 1.6189, - "step": 20065 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019321315024014072, - "loss": 1.8467, - "step": 20066 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019321281201379963, - "loss": 1.5974, - "step": 20067 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019321247378745857, - "loss": 1.541, - "step": 20068 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932121355611175, - "loss": 1.4982, - "step": 20069 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019321179733477643, - "loss": 1.5444, - "step": 20070 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019321145910843537, - "loss": 1.4741, - "step": 20071 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932111208820943, - "loss": 1.4896, - "step": 20072 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019321078265575322, - "loss": 1.8052, - "step": 20073 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019321044442941216, - "loss": 1.4588, - "step": 20074 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932101062030711, - "loss": 1.6001, - "step": 20075 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019320976797673002, - "loss": 1.4894, - "step": 20076 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019320942975038896, - "loss": 1.6097, - "step": 20077 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932090915240479, - "loss": 1.5768, - "step": 20078 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019320875329770682, - "loss": 1.7087, - "step": 20079 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019320841507136578, - "loss": 1.4602, - "step": 20080 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932080768450247, - "loss": 1.6424, - "step": 20081 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019320773861868364, - "loss": 1.6137, - "step": 20082 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019320740039234258, - "loss": 1.4671, - "step": 20083 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932070621660015, - "loss": 1.322, - "step": 20084 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019320672393966044, - "loss": 1.3676, - "step": 20085 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019320638571331938, - "loss": 1.6758, - "step": 20086 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932060474869783, - "loss": 1.6103, - "step": 20087 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019320570926063723, - "loss": 1.7019, - "step": 20088 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019320537103429617, - "loss": 1.5691, - "step": 20089 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001932050328079551, - "loss": 1.6637, - "step": 20090 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019320469458161403, - "loss": 1.6505, - "step": 20091 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019320435635527297, - "loss": 1.605, - "step": 20092 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019320401812893188, - "loss": 1.5357, - "step": 20093 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019320367990259082, - "loss": 1.5446, - "step": 20094 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019320334167624977, - "loss": 1.5256, - "step": 20095 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019320300344990868, - "loss": 1.4859, - "step": 20096 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019320266522356762, - "loss": 1.4853, - "step": 20097 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019320232699722656, - "loss": 1.5804, - "step": 20098 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019320198877088548, - "loss": 1.4734, - "step": 20099 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019320165054454442, - "loss": 1.5509, - "step": 20100 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019320131231820336, - "loss": 1.6486, - "step": 20101 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019320097409186227, - "loss": 1.4898, - "step": 20102 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019320063586552121, - "loss": 1.5729, - "step": 20103 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019320029763918015, - "loss": 1.6257, - "step": 20104 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019319995941283907, - "loss": 1.6502, - "step": 20105 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193199621186498, - "loss": 1.5851, - "step": 20106 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019319928296015695, - "loss": 1.5936, - "step": 20107 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019319894473381587, - "loss": 1.4386, - "step": 20108 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931986065074748, - "loss": 1.573, - "step": 20109 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019319826828113375, - "loss": 1.7126, - "step": 20110 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019319793005479266, - "loss": 1.6218, - "step": 20111 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931975918284516, - "loss": 1.6395, - "step": 20112 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019319725360211054, - "loss": 1.6942, - "step": 20113 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019319691537576949, - "loss": 1.577, - "step": 20114 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931965771494284, - "loss": 1.6387, - "step": 20115 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019319623892308734, - "loss": 1.6612, - "step": 20116 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019319590069674628, - "loss": 1.6898, - "step": 20117 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931955624704052, - "loss": 1.5646, - "step": 20118 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019319522424406414, - "loss": 1.6586, - "step": 20119 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019319488601772308, - "loss": 1.5643, - "step": 20120 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193194547791382, - "loss": 1.6762, - "step": 20121 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019319420956504093, - "loss": 1.5187, - "step": 20122 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019319387133869987, - "loss": 1.5658, - "step": 20123 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931935331123588, - "loss": 1.5388, - "step": 20124 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019319319488601773, - "loss": 1.4372, - "step": 20125 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019319285665967667, - "loss": 1.6511, - "step": 20126 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019319251843333558, - "loss": 1.4868, - "step": 20127 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019319218020699453, - "loss": 1.5204, - "step": 20128 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019319184198065347, - "loss": 1.5741, - "step": 20129 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019319150375431238, - "loss": 1.614, - "step": 20130 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019319116552797132, - "loss": 1.6623, - "step": 20131 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019319082730163026, - "loss": 1.4133, - "step": 20132 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019319048907528918, - "loss": 1.5031, - "step": 20133 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019319015084894812, - "loss": 1.6296, - "step": 20134 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019318981262260706, - "loss": 1.5887, - "step": 20135 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193189474396266, - "loss": 1.586, - "step": 20136 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019318913616992494, - "loss": 1.663, - "step": 20137 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019318879794358386, - "loss": 1.6302, - "step": 20138 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931884597172428, - "loss": 1.7405, - "step": 20139 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019318812149090174, - "loss": 1.5738, - "step": 20140 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019318778326456065, - "loss": 1.6882, - "step": 20141 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931874450382196, - "loss": 1.6568, - "step": 20142 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019318710681187853, - "loss": 1.5756, - "step": 20143 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019318676858553745, - "loss": 1.7426, - "step": 20144 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931864303591964, - "loss": 1.5944, - "step": 20145 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019318609213285533, - "loss": 1.6479, - "step": 20146 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019318575390651424, - "loss": 1.5421, - "step": 20147 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019318541568017319, - "loss": 1.5511, - "step": 20148 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019318507745383213, - "loss": 1.7221, - "step": 20149 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019318473922749104, - "loss": 1.5451, - "step": 20150 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019318440100114998, - "loss": 1.6013, - "step": 20151 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019318406277480892, - "loss": 1.565, - "step": 20152 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019318372454846784, - "loss": 1.5253, - "step": 20153 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019318338632212678, - "loss": 1.5661, - "step": 20154 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019318304809578572, - "loss": 1.4505, - "step": 20155 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019318270986944463, - "loss": 1.4589, - "step": 20156 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019318237164310357, - "loss": 1.565, - "step": 20157 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019318203341676252, - "loss": 1.6347, - "step": 20158 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019318169519042143, - "loss": 1.7556, - "step": 20159 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019318135696408037, - "loss": 1.59, - "step": 20160 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931810187377393, - "loss": 1.7788, - "step": 20161 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019318068051139823, - "loss": 1.6344, - "step": 20162 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019318034228505717, - "loss": 1.5172, - "step": 20163 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931800040587161, - "loss": 1.5533, - "step": 20164 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019317966583237502, - "loss": 1.6834, - "step": 20165 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019317932760603396, - "loss": 1.5826, - "step": 20166 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931789893796929, - "loss": 1.3857, - "step": 20167 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019317865115335182, - "loss": 1.6095, - "step": 20168 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019317831292701076, - "loss": 1.4144, - "step": 20169 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931779747006697, - "loss": 1.5761, - "step": 20170 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019317763647432862, - "loss": 1.4955, - "step": 20171 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019317729824798756, - "loss": 1.5906, - "step": 20172 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931769600216465, - "loss": 1.5346, - "step": 20173 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931766217953054, - "loss": 1.7253, - "step": 20174 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019317628356896435, - "loss": 1.5074, - "step": 20175 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931759453426233, - "loss": 1.3834, - "step": 20176 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931756071162822, - "loss": 1.5935, - "step": 20177 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019317526888994115, - "loss": 1.5546, - "step": 20178 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931749306636001, - "loss": 1.5896, - "step": 20179 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193174592437259, - "loss": 1.7345, - "step": 20180 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019317425421091795, - "loss": 1.482, - "step": 20181 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019317391598457689, - "loss": 1.7626, - "step": 20182 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931735777582358, - "loss": 1.413, - "step": 20183 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019317323953189474, - "loss": 1.5401, - "step": 20184 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019317290130555368, - "loss": 1.325, - "step": 20185 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931725630792126, - "loss": 1.4544, - "step": 20186 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019317222485287154, - "loss": 1.5617, - "step": 20187 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019317188662653048, - "loss": 1.68, - "step": 20188 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931715484001894, - "loss": 1.6558, - "step": 20189 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019317121017384833, - "loss": 1.5653, - "step": 20190 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019317087194750728, - "loss": 1.7342, - "step": 20191 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019317053372116622, - "loss": 1.5855, - "step": 20192 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019317019549482516, - "loss": 1.506, - "step": 20193 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931698572684841, - "loss": 1.5156, - "step": 20194 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193169519042143, - "loss": 1.5828, - "step": 20195 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019316918081580195, - "loss": 1.6078, - "step": 20196 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931688425894609, - "loss": 1.5081, - "step": 20197 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931685043631198, - "loss": 1.6033, - "step": 20198 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019316816613677875, - "loss": 1.5483, - "step": 20199 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931678279104377, - "loss": 1.636, - "step": 20200 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931674896840966, - "loss": 1.6639, - "step": 20201 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019316715145775555, - "loss": 1.635, - "step": 20202 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931668132314145, - "loss": 1.5869, - "step": 20203 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931664750050734, - "loss": 1.7358, - "step": 20204 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019316613677873234, - "loss": 1.5235, - "step": 20205 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019316579855239128, - "loss": 1.7285, - "step": 20206 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931654603260502, - "loss": 1.6835, - "step": 20207 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019316512209970914, - "loss": 1.7316, - "step": 20208 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019316478387336808, - "loss": 1.6351, - "step": 20209 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193164445647027, - "loss": 1.3965, - "step": 20210 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019316410742068594, - "loss": 1.5855, - "step": 20211 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019316376919434488, - "loss": 1.6159, - "step": 20212 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931634309680038, - "loss": 1.5535, - "step": 20213 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019316309274166273, - "loss": 1.6243, - "step": 20214 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019316275451532167, - "loss": 1.4909, - "step": 20215 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931624162889806, - "loss": 1.6408, - "step": 20216 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019316207806263953, - "loss": 1.5591, - "step": 20217 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019316173983629847, - "loss": 1.5539, - "step": 20218 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019316140160995738, - "loss": 1.5487, - "step": 20219 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019316106338361632, - "loss": 1.4557, - "step": 20220 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019316072515727527, - "loss": 1.5659, - "step": 20221 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019316038693093418, - "loss": 1.6921, - "step": 20222 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019316004870459312, - "loss": 1.5227, - "step": 20223 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019315971047825206, - "loss": 1.5718, - "step": 20224 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019315937225191098, - "loss": 1.4464, - "step": 20225 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019315903402556992, - "loss": 1.6739, - "step": 20226 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019315869579922886, - "loss": 1.5812, - "step": 20227 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019315835757288777, - "loss": 1.6105, - "step": 20228 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931580193465467, - "loss": 1.6129, - "step": 20229 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019315768112020565, - "loss": 1.6439, - "step": 20230 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019315734289386457, - "loss": 1.5334, - "step": 20231 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931570046675235, - "loss": 1.6413, - "step": 20232 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019315666644118245, - "loss": 1.4272, - "step": 20233 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019315632821484136, - "loss": 1.4889, - "step": 20234 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931559899885003, - "loss": 1.5951, - "step": 20235 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019315565176215925, - "loss": 1.5824, - "step": 20236 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019315531353581816, - "loss": 1.5447, - "step": 20237 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931549753094771, - "loss": 1.7793, - "step": 20238 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019315463708313604, - "loss": 1.6759, - "step": 20239 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019315429885679496, - "loss": 1.4576, - "step": 20240 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931539606304539, - "loss": 1.4782, - "step": 20241 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019315362240411284, - "loss": 1.6355, - "step": 20242 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019315328417777175, - "loss": 1.5259, - "step": 20243 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931529459514307, - "loss": 1.7141, - "step": 20244 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019315260772508964, - "loss": 1.7044, - "step": 20245 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019315226949874855, - "loss": 1.5436, - "step": 20246 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931519312724075, - "loss": 1.5246, - "step": 20247 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019315159304606646, - "loss": 1.5929, - "step": 20248 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019315125481972537, - "loss": 1.594, - "step": 20249 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019315091659338431, - "loss": 1.7992, - "step": 20250 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019315057836704326, - "loss": 1.5033, - "step": 20251 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019315024014070217, - "loss": 1.5319, - "step": 20252 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931499019143611, - "loss": 1.6187, - "step": 20253 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019314956368802005, - "loss": 1.5964, - "step": 20254 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019314922546167897, - "loss": 1.6034, - "step": 20255 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931488872353379, - "loss": 1.5112, - "step": 20256 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019314854900899685, - "loss": 1.4722, - "step": 20257 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019314821078265576, - "loss": 1.6896, - "step": 20258 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931478725563147, - "loss": 1.7025, - "step": 20259 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019314753432997364, - "loss": 1.4378, - "step": 20260 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019314719610363256, - "loss": 1.4815, - "step": 20261 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931468578772915, - "loss": 1.442, - "step": 20262 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019314651965095044, - "loss": 1.4148, - "step": 20263 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019314618142460935, - "loss": 1.5461, - "step": 20264 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931458431982683, - "loss": 1.6005, - "step": 20265 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019314550497192724, - "loss": 1.5532, - "step": 20266 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019314516674558615, - "loss": 1.731, - "step": 20267 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931448285192451, - "loss": 1.5838, - "step": 20268 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019314449029290403, - "loss": 1.4843, - "step": 20269 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019314415206656295, - "loss": 1.6892, - "step": 20270 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931438138402219, - "loss": 1.4434, - "step": 20271 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019314347561388083, - "loss": 1.4399, - "step": 20272 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019314313738753974, - "loss": 1.8124, - "step": 20273 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019314279916119869, - "loss": 1.5753, - "step": 20274 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019314246093485763, - "loss": 1.466, - "step": 20275 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019314212270851654, - "loss": 1.6135, - "step": 20276 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019314178448217548, - "loss": 1.512, - "step": 20277 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019314144625583442, - "loss": 1.5318, - "step": 20278 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019314110802949334, - "loss": 1.649, - "step": 20279 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019314076980315228, - "loss": 1.5924, - "step": 20280 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019314043157681122, - "loss": 1.5239, - "step": 20281 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019314009335047013, - "loss": 1.6383, - "step": 20282 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019313975512412907, - "loss": 1.569, - "step": 20283 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019313941689778802, - "loss": 1.6981, - "step": 20284 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019313907867144693, - "loss": 1.5831, - "step": 20285 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019313874044510587, - "loss": 1.4798, - "step": 20286 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931384022187648, - "loss": 1.4944, - "step": 20287 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019313806399242373, - "loss": 1.714, - "step": 20288 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019313772576608267, - "loss": 1.7926, - "step": 20289 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931373875397416, - "loss": 1.5433, - "step": 20290 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019313704931340052, - "loss": 1.4668, - "step": 20291 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019313671108705946, - "loss": 1.5573, - "step": 20292 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931363728607184, - "loss": 1.6932, - "step": 20293 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019313603463437732, - "loss": 1.6817, - "step": 20294 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019313569640803626, - "loss": 1.5407, - "step": 20295 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931353581816952, - "loss": 1.5149, - "step": 20296 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019313501995535411, - "loss": 1.5106, - "step": 20297 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019313468172901306, - "loss": 1.5943, - "step": 20298 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193134343502672, - "loss": 1.5229, - "step": 20299 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931340052763309, - "loss": 1.6245, - "step": 20300 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019313366704998985, - "loss": 1.6744, - "step": 20301 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931333288236488, - "loss": 1.4152, - "step": 20302 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931329905973077, - "loss": 1.5545, - "step": 20303 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019313265237096665, - "loss": 1.6964, - "step": 20304 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019313231414462562, - "loss": 1.5076, - "step": 20305 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019313197591828453, - "loss": 1.4803, - "step": 20306 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019313163769194347, - "loss": 1.5803, - "step": 20307 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931312994656024, - "loss": 1.6163, - "step": 20308 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019313096123926133, - "loss": 1.6492, - "step": 20309 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019313062301292027, - "loss": 1.5922, - "step": 20310 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931302847865792, - "loss": 1.7795, - "step": 20311 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019312994656023812, - "loss": 1.6212, - "step": 20312 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019312960833389706, - "loss": 1.6019, - "step": 20313 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193129270107556, - "loss": 1.6282, - "step": 20314 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019312893188121492, - "loss": 1.6319, - "step": 20315 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019312859365487386, - "loss": 1.5038, - "step": 20316 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931282554285328, - "loss": 1.6245, - "step": 20317 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019312791720219172, - "loss": 1.7304, - "step": 20318 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019312757897585066, - "loss": 1.5845, - "step": 20319 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931272407495096, - "loss": 1.7003, - "step": 20320 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931269025231685, - "loss": 1.6301, - "step": 20321 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019312656429682745, - "loss": 1.5116, - "step": 20322 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931262260704864, - "loss": 1.5031, - "step": 20323 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931258878441453, - "loss": 1.5446, - "step": 20324 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019312554961780425, - "loss": 1.5668, - "step": 20325 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931252113914632, - "loss": 1.4892, - "step": 20326 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931248731651221, - "loss": 1.7119, - "step": 20327 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019312453493878105, - "loss": 1.6997, - "step": 20328 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019312419671244, - "loss": 1.4795, - "step": 20329 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931238584860989, - "loss": 1.4958, - "step": 20330 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019312352025975784, - "loss": 1.6184, - "step": 20331 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019312318203341678, - "loss": 1.5811, - "step": 20332 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931228438070757, - "loss": 1.7281, - "step": 20333 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019312250558073464, - "loss": 1.5747, - "step": 20334 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019312216735439358, - "loss": 1.7861, - "step": 20335 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931218291280525, - "loss": 1.4908, - "step": 20336 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019312149090171143, - "loss": 1.5556, - "step": 20337 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019312115267537038, - "loss": 1.7694, - "step": 20338 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931208144490293, - "loss": 1.5695, - "step": 20339 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019312047622268823, - "loss": 1.6016, - "step": 20340 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019312013799634717, - "loss": 1.7037, - "step": 20341 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019311979977000609, - "loss": 1.5156, - "step": 20342 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019311946154366503, - "loss": 1.5389, - "step": 20343 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019311912331732397, - "loss": 1.5229, - "step": 20344 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019311878509098288, - "loss": 1.5525, - "step": 20345 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019311844686464182, - "loss": 1.6464, - "step": 20346 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019311810863830076, - "loss": 1.6448, - "step": 20347 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019311777041195968, - "loss": 1.5878, - "step": 20348 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019311743218561862, - "loss": 1.648, - "step": 20349 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019311709395927756, - "loss": 1.586, - "step": 20350 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019311675573293648, - "loss": 1.464, - "step": 20351 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019311641750659542, - "loss": 1.6589, - "step": 20352 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019311607928025436, - "loss": 1.5126, - "step": 20353 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019311574105391327, - "loss": 1.4583, - "step": 20354 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931154028275722, - "loss": 1.6168, - "step": 20355 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019311506460123115, - "loss": 1.5683, - "step": 20356 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019311472637489007, - "loss": 1.6313, - "step": 20357 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193114388148549, - "loss": 1.4776, - "step": 20358 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019311404992220795, - "loss": 1.5925, - "step": 20359 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019311371169586686, - "loss": 1.4257, - "step": 20360 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019311337346952583, - "loss": 1.6165, - "step": 20361 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019311303524318475, - "loss": 1.5523, - "step": 20362 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931126970168437, - "loss": 1.6223, - "step": 20363 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019311235879050263, - "loss": 1.5325, - "step": 20364 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019311202056416154, - "loss": 1.6128, - "step": 20365 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019311168233782048, - "loss": 1.6078, - "step": 20366 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019311134411147942, - "loss": 1.5082, - "step": 20367 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019311100588513834, - "loss": 1.4911, - "step": 20368 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019311066765879728, - "loss": 1.5664, - "step": 20369 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019311032943245622, - "loss": 1.6288, - "step": 20370 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019310999120611514, - "loss": 1.5843, - "step": 20371 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019310965297977408, - "loss": 1.603, - "step": 20372 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019310931475343302, - "loss": 1.577, - "step": 20373 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019310897652709193, - "loss": 1.634, - "step": 20374 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019310863830075087, - "loss": 1.5959, - "step": 20375 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019310830007440981, - "loss": 1.4318, - "step": 20376 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019310796184806873, - "loss": 1.6229, - "step": 20377 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019310762362172767, - "loss": 1.4739, - "step": 20378 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931072853953866, - "loss": 1.4391, - "step": 20379 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019310694716904555, - "loss": 1.5997, - "step": 20380 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019310660894270447, - "loss": 1.5288, - "step": 20381 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931062707163634, - "loss": 1.654, - "step": 20382 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019310593249002235, - "loss": 1.6257, - "step": 20383 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019310559426368126, - "loss": 1.6166, - "step": 20384 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931052560373402, - "loss": 1.644, - "step": 20385 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019310491781099914, - "loss": 1.5669, - "step": 20386 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019310457958465806, - "loss": 1.3981, - "step": 20387 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193104241358317, - "loss": 1.5473, - "step": 20388 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019310390313197594, - "loss": 1.4792, - "step": 20389 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019310356490563485, - "loss": 1.639, - "step": 20390 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931032266792938, - "loss": 1.5624, - "step": 20391 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019310288845295274, - "loss": 1.4897, - "step": 20392 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019310255022661165, - "loss": 1.6938, - "step": 20393 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931022120002706, - "loss": 1.4494, - "step": 20394 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019310187377392953, - "loss": 1.5482, - "step": 20395 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019310153554758845, - "loss": 1.606, - "step": 20396 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001931011973212474, - "loss": 1.4722, - "step": 20397 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019310085909490633, - "loss": 1.8186, - "step": 20398 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019310052086856524, - "loss": 1.5669, - "step": 20399 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019310018264222418, - "loss": 1.5419, - "step": 20400 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019309984441588313, - "loss": 1.5131, - "step": 20401 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019309950618954204, - "loss": 1.5774, - "step": 20402 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019309916796320098, - "loss": 1.469, - "step": 20403 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019309882973685992, - "loss": 1.5704, - "step": 20404 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019309849151051884, - "loss": 1.4476, - "step": 20405 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019309815328417778, - "loss": 1.561, - "step": 20406 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019309781505783672, - "loss": 1.5444, - "step": 20407 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019309747683149563, - "loss": 1.5069, - "step": 20408 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019309713860515457, - "loss": 1.619, - "step": 20409 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019309680037881351, - "loss": 1.5783, - "step": 20410 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019309646215247243, - "loss": 1.4946, - "step": 20411 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019309612392613137, - "loss": 1.6219, - "step": 20412 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930957856997903, - "loss": 1.4229, - "step": 20413 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019309544747344922, - "loss": 1.5944, - "step": 20414 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019309510924710817, - "loss": 1.5165, - "step": 20415 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930947710207671, - "loss": 1.5928, - "step": 20416 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019309443279442605, - "loss": 1.5946, - "step": 20417 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193094094568085, - "loss": 1.8197, - "step": 20418 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930937563417439, - "loss": 1.6669, - "step": 20419 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019309341811540284, - "loss": 1.4866, - "step": 20420 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019309307988906179, - "loss": 1.7374, - "step": 20421 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930927416627207, - "loss": 1.6357, - "step": 20422 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019309240343637964, - "loss": 1.6699, - "step": 20423 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019309206521003858, - "loss": 1.4346, - "step": 20424 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930917269836975, - "loss": 1.5388, - "step": 20425 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019309138875735644, - "loss": 1.5835, - "step": 20426 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019309105053101538, - "loss": 1.541, - "step": 20427 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930907123046743, - "loss": 1.4519, - "step": 20428 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019309037407833323, - "loss": 1.6377, - "step": 20429 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019309003585199217, - "loss": 1.7433, - "step": 20430 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930896976256511, - "loss": 1.4485, - "step": 20431 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019308935939931003, - "loss": 1.5429, - "step": 20432 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019308902117296897, - "loss": 1.6965, - "step": 20433 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019308868294662789, - "loss": 1.7072, - "step": 20434 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019308834472028683, - "loss": 1.5324, - "step": 20435 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019308800649394577, - "loss": 1.5529, - "step": 20436 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019308766826760468, - "loss": 1.6017, - "step": 20437 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019308733004126362, - "loss": 1.5879, - "step": 20438 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019308699181492256, - "loss": 1.6584, - "step": 20439 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019308665358858148, - "loss": 1.4822, - "step": 20440 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019308631536224042, - "loss": 1.5068, - "step": 20441 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019308597713589936, - "loss": 1.5674, - "step": 20442 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019308563890955827, - "loss": 1.5911, - "step": 20443 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019308530068321722, - "loss": 1.4027, - "step": 20444 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019308496245687616, - "loss": 1.7018, - "step": 20445 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019308462423053507, - "loss": 1.4988, - "step": 20446 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193084286004194, - "loss": 1.6256, - "step": 20447 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019308394777785295, - "loss": 1.6903, - "step": 20448 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019308360955151187, - "loss": 1.7577, - "step": 20449 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930832713251708, - "loss": 1.7232, - "step": 20450 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019308293309882975, - "loss": 1.7495, - "step": 20451 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019308259487248866, - "loss": 1.6277, - "step": 20452 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930822566461476, - "loss": 1.5458, - "step": 20453 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019308191841980655, - "loss": 1.4749, - "step": 20454 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019308158019346546, - "loss": 1.6257, - "step": 20455 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930812419671244, - "loss": 1.6735, - "step": 20456 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019308090374078334, - "loss": 1.4469, - "step": 20457 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019308056551444226, - "loss": 1.5864, - "step": 20458 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930802272881012, - "loss": 1.5382, - "step": 20459 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019307988906176014, - "loss": 1.5039, - "step": 20460 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019307955083541905, - "loss": 1.6449, - "step": 20461 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193079212609078, - "loss": 1.5528, - "step": 20462 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019307887438273693, - "loss": 1.6161, - "step": 20463 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019307853615639585, - "loss": 1.524, - "step": 20464 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930781979300548, - "loss": 1.6724, - "step": 20465 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019307785970371373, - "loss": 1.6888, - "step": 20466 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019307752147737267, - "loss": 1.5543, - "step": 20467 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019307718325103159, - "loss": 1.5271, - "step": 20468 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019307684502469053, - "loss": 1.6914, - "step": 20469 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019307650679834947, - "loss": 1.5678, - "step": 20470 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019307616857200838, - "loss": 1.5517, - "step": 20471 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019307583034566732, - "loss": 1.499, - "step": 20472 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019307549211932626, - "loss": 1.5444, - "step": 20473 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930751538929852, - "loss": 1.5842, - "step": 20474 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019307481566664415, - "loss": 1.6082, - "step": 20475 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019307447744030306, - "loss": 1.6279, - "step": 20476 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193074139213962, - "loss": 1.593, - "step": 20477 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019307380098762094, - "loss": 1.4677, - "step": 20478 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019307346276127986, - "loss": 1.6242, - "step": 20479 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930731245349388, - "loss": 1.5915, - "step": 20480 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019307278630859774, - "loss": 1.5594, - "step": 20481 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019307244808225665, - "loss": 1.6338, - "step": 20482 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930721098559156, - "loss": 1.5024, - "step": 20483 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019307177162957454, - "loss": 1.4419, - "step": 20484 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019307143340323345, - "loss": 1.6732, - "step": 20485 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930710951768924, - "loss": 1.5198, - "step": 20486 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019307075695055133, - "loss": 1.5336, - "step": 20487 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019307041872421025, - "loss": 1.522, - "step": 20488 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930700804978692, - "loss": 1.6175, - "step": 20489 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019306974227152813, - "loss": 1.5348, - "step": 20490 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019306940404518704, - "loss": 1.4665, - "step": 20491 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019306906581884598, - "loss": 1.5277, - "step": 20492 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019306872759250492, - "loss": 1.5678, - "step": 20493 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019306838936616384, - "loss": 1.556, - "step": 20494 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019306805113982278, - "loss": 1.4929, - "step": 20495 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019306771291348172, - "loss": 1.6598, - "step": 20496 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019306737468714063, - "loss": 1.6375, - "step": 20497 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019306703646079958, - "loss": 1.5276, - "step": 20498 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019306669823445852, - "loss": 1.623, - "step": 20499 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019306636000811743, - "loss": 1.5005, - "step": 20500 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019306602178177637, - "loss": 1.4702, - "step": 20501 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930656835554353, - "loss": 1.4926, - "step": 20502 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019306534532909423, - "loss": 1.6644, - "step": 20503 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019306500710275317, - "loss": 1.6919, - "step": 20504 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930646688764121, - "loss": 1.5577, - "step": 20505 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019306433065007102, - "loss": 1.5931, - "step": 20506 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019306399242372996, - "loss": 1.4418, - "step": 20507 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930636541973889, - "loss": 1.5798, - "step": 20508 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019306331597104782, - "loss": 1.5113, - "step": 20509 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019306297774470676, - "loss": 1.6439, - "step": 20510 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930626395183657, - "loss": 1.512, - "step": 20511 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019306230129202462, - "loss": 1.7129, - "step": 20512 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019306196306568356, - "loss": 1.2699, - "step": 20513 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930616248393425, - "loss": 1.4184, - "step": 20514 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930612866130014, - "loss": 1.4959, - "step": 20515 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019306094838666035, - "loss": 1.5756, - "step": 20516 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930606101603193, - "loss": 1.6924, - "step": 20517 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930602719339782, - "loss": 1.5495, - "step": 20518 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019305993370763715, - "loss": 1.7967, - "step": 20519 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930595954812961, - "loss": 1.6342, - "step": 20520 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193059257254955, - "loss": 1.6029, - "step": 20521 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019305891902861395, - "loss": 1.5753, - "step": 20522 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930585808022729, - "loss": 1.5203, - "step": 20523 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930582425759318, - "loss": 1.6422, - "step": 20524 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019305790434959074, - "loss": 1.5998, - "step": 20525 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019305756612324968, - "loss": 1.4664, - "step": 20526 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930572278969086, - "loss": 1.6173, - "step": 20527 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019305688967056754, - "loss": 1.6623, - "step": 20528 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930565514442265, - "loss": 1.5423, - "step": 20529 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019305621321788542, - "loss": 1.4963, - "step": 20530 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019305587499154436, - "loss": 1.5642, - "step": 20531 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930555367652033, - "loss": 1.4746, - "step": 20532 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019305519853886222, - "loss": 1.652, - "step": 20533 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019305486031252116, - "loss": 1.6365, - "step": 20534 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930545220861801, - "loss": 1.5907, - "step": 20535 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019305418385983901, - "loss": 1.5896, - "step": 20536 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019305384563349796, - "loss": 1.7114, - "step": 20537 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930535074071569, - "loss": 1.6561, - "step": 20538 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930531691808158, - "loss": 1.4541, - "step": 20539 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019305283095447475, - "loss": 1.532, - "step": 20540 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930524927281337, - "loss": 1.5899, - "step": 20541 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930521545017926, - "loss": 1.581, - "step": 20542 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019305181627545155, - "loss": 1.3541, - "step": 20543 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930514780491105, - "loss": 1.6279, - "step": 20544 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930511398227694, - "loss": 1.7273, - "step": 20545 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019305080159642834, - "loss": 1.6301, - "step": 20546 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019305046337008729, - "loss": 1.6439, - "step": 20547 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930501251437462, - "loss": 1.5739, - "step": 20548 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019304978691740514, - "loss": 1.5045, - "step": 20549 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019304944869106408, - "loss": 1.4947, - "step": 20550 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193049110464723, - "loss": 1.6694, - "step": 20551 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019304877223838194, - "loss": 1.5028, - "step": 20552 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019304843401204088, - "loss": 1.5001, - "step": 20553 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930480957856998, - "loss": 1.6173, - "step": 20554 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019304775755935873, - "loss": 1.6091, - "step": 20555 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019304741933301767, - "loss": 1.7292, - "step": 20556 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930470811066766, - "loss": 1.6155, - "step": 20557 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019304674288033553, - "loss": 1.6213, - "step": 20558 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019304640465399447, - "loss": 1.5573, - "step": 20559 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019304606642765338, - "loss": 1.7349, - "step": 20560 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019304572820131233, - "loss": 1.4912, - "step": 20561 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019304538997497127, - "loss": 1.5658, - "step": 20562 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019304505174863018, - "loss": 1.4875, - "step": 20563 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019304471352228912, - "loss": 1.6141, - "step": 20564 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019304437529594806, - "loss": 1.5324, - "step": 20565 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019304403706960698, - "loss": 1.4766, - "step": 20566 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019304369884326592, - "loss": 1.4904, - "step": 20567 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019304336061692486, - "loss": 1.5844, - "step": 20568 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019304302239058377, - "loss": 1.5758, - "step": 20569 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019304268416424271, - "loss": 1.5858, - "step": 20570 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019304234593790166, - "loss": 1.6851, - "step": 20571 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019304200771156057, - "loss": 1.6643, - "step": 20572 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930416694852195, - "loss": 1.6801, - "step": 20573 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019304133125887845, - "loss": 1.4472, - "step": 20574 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019304099303253737, - "loss": 1.5969, - "step": 20575 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930406548061963, - "loss": 1.594, - "step": 20576 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019304031657985525, - "loss": 1.5955, - "step": 20577 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019303997835351416, - "loss": 1.4116, - "step": 20578 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930396401271731, - "loss": 1.6312, - "step": 20579 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019303930190083204, - "loss": 1.6392, - "step": 20580 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019303896367449096, - "loss": 1.5732, - "step": 20581 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930386254481499, - "loss": 1.4089, - "step": 20582 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019303828722180884, - "loss": 1.486, - "step": 20583 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019303794899546775, - "loss": 1.5478, - "step": 20584 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019303761076912672, - "loss": 1.4052, - "step": 20585 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019303727254278566, - "loss": 1.6461, - "step": 20586 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019303693431644458, - "loss": 1.4391, - "step": 20587 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019303659609010352, - "loss": 1.5876, - "step": 20588 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019303625786376246, - "loss": 1.5539, - "step": 20589 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019303591963742137, - "loss": 1.7035, - "step": 20590 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019303558141108032, - "loss": 1.4393, - "step": 20591 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019303524318473926, - "loss": 1.7432, - "step": 20592 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019303490495839817, - "loss": 1.4792, - "step": 20593 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930345667320571, - "loss": 1.57, - "step": 20594 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019303422850571605, - "loss": 1.5723, - "step": 20595 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019303389027937497, - "loss": 1.4985, - "step": 20596 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930335520530339, - "loss": 1.5085, - "step": 20597 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019303321382669285, - "loss": 1.6408, - "step": 20598 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019303287560035176, - "loss": 1.6291, - "step": 20599 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930325373740107, - "loss": 1.5181, - "step": 20600 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019303219914766965, - "loss": 1.5589, - "step": 20601 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019303186092132856, - "loss": 1.4942, - "step": 20602 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930315226949875, - "loss": 1.7133, - "step": 20603 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019303118446864644, - "loss": 1.5038, - "step": 20604 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019303084624230536, - "loss": 1.6542, - "step": 20605 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930305080159643, - "loss": 1.4602, - "step": 20606 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019303016978962324, - "loss": 1.5427, - "step": 20607 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019302983156328215, - "loss": 1.5354, - "step": 20608 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930294933369411, - "loss": 1.4237, - "step": 20609 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019302915511060003, - "loss": 1.6903, - "step": 20610 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019302881688425895, - "loss": 1.6051, - "step": 20611 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930284786579179, - "loss": 1.6292, - "step": 20612 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019302814043157683, - "loss": 1.6501, - "step": 20613 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019302780220523575, - "loss": 1.5883, - "step": 20614 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019302746397889469, - "loss": 1.632, - "step": 20615 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019302712575255363, - "loss": 1.7227, - "step": 20616 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019302678752621254, - "loss": 1.5611, - "step": 20617 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019302644929987148, - "loss": 1.6762, - "step": 20618 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019302611107353042, - "loss": 1.5427, - "step": 20619 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019302577284718934, - "loss": 1.5911, - "step": 20620 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019302543462084828, - "loss": 1.6048, - "step": 20621 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019302509639450722, - "loss": 1.5882, - "step": 20622 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019302475816816613, - "loss": 1.5202, - "step": 20623 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019302441994182508, - "loss": 1.5986, - "step": 20624 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019302408171548402, - "loss": 1.5504, - "step": 20625 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019302374348914293, - "loss": 1.6029, - "step": 20626 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019302340526280187, - "loss": 1.4735, - "step": 20627 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930230670364608, - "loss": 1.5578, - "step": 20628 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019302272881011973, - "loss": 1.5802, - "step": 20629 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019302239058377867, - "loss": 1.6341, - "step": 20630 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930220523574376, - "loss": 1.7021, - "step": 20631 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019302171413109652, - "loss": 1.6611, - "step": 20632 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019302137590475546, - "loss": 1.4972, - "step": 20633 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930210376784144, - "loss": 1.5543, - "step": 20634 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019302069945207332, - "loss": 1.5368, - "step": 20635 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019302036122573226, - "loss": 1.6079, - "step": 20636 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930200229993912, - "loss": 1.5222, - "step": 20637 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019301968477305012, - "loss": 1.4751, - "step": 20638 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019301934654670906, - "loss": 1.4848, - "step": 20639 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193019008320368, - "loss": 1.7488, - "step": 20640 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930186700940269, - "loss": 1.6312, - "step": 20641 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019301833186768588, - "loss": 1.592, - "step": 20642 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930179936413448, - "loss": 1.632, - "step": 20643 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019301765541500374, - "loss": 1.4988, - "step": 20644 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019301731718866268, - "loss": 1.6038, - "step": 20645 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930169789623216, - "loss": 1.4938, - "step": 20646 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019301664073598053, - "loss": 1.5168, - "step": 20647 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019301630250963947, - "loss": 1.5418, - "step": 20648 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019301596428329841, - "loss": 1.6936, - "step": 20649 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019301562605695733, - "loss": 1.6061, - "step": 20650 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019301528783061627, - "loss": 1.6567, - "step": 20651 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930149496042752, - "loss": 1.729, - "step": 20652 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019301461137793412, - "loss": 1.6097, - "step": 20653 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019301427315159307, - "loss": 1.5315, - "step": 20654 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193013934925252, - "loss": 1.5422, - "step": 20655 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019301359669891092, - "loss": 1.6226, - "step": 20656 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019301325847256986, - "loss": 1.7185, - "step": 20657 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930129202462288, - "loss": 1.6353, - "step": 20658 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019301258201988772, - "loss": 1.608, - "step": 20659 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019301224379354666, - "loss": 1.5914, - "step": 20660 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930119055672056, - "loss": 1.6504, - "step": 20661 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930115673408645, - "loss": 1.631, - "step": 20662 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019301122911452345, - "loss": 1.771, - "step": 20663 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930108908881824, - "loss": 1.6241, - "step": 20664 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930105526618413, - "loss": 1.4126, - "step": 20665 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019301021443550025, - "loss": 1.5635, - "step": 20666 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930098762091592, - "loss": 1.6041, - "step": 20667 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930095379828181, - "loss": 1.5379, - "step": 20668 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019300919975647705, - "loss": 1.703, - "step": 20669 - }, - { - "epoch": 0.07, - "learning_rate": 0.000193008861530136, - "loss": 1.6251, - "step": 20670 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930085233037949, - "loss": 1.5233, - "step": 20671 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019300818507745384, - "loss": 1.6843, - "step": 20672 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019300784685111278, - "loss": 1.5244, - "step": 20673 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930075086247717, - "loss": 1.7005, - "step": 20674 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019300717039843064, - "loss": 1.5094, - "step": 20675 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019300683217208958, - "loss": 1.3508, - "step": 20676 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930064939457485, - "loss": 1.6597, - "step": 20677 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019300615571940744, - "loss": 1.7343, - "step": 20678 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019300581749306638, - "loss": 1.6321, - "step": 20679 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930054792667253, - "loss": 1.5915, - "step": 20680 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019300514104038423, - "loss": 1.7487, - "step": 20681 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019300480281404317, - "loss": 1.5226, - "step": 20682 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001930044645877021, - "loss": 1.585, - "step": 20683 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019300412636136103, - "loss": 1.6839, - "step": 20684 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019300378813501997, - "loss": 1.5754, - "step": 20685 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019300344990867888, - "loss": 1.727, - "step": 20686 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019300311168233782, - "loss": 1.555, - "step": 20687 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019300277345599677, - "loss": 1.556, - "step": 20688 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019300243522965568, - "loss": 1.6291, - "step": 20689 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019300209700331462, - "loss": 1.5332, - "step": 20690 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019300175877697356, - "loss": 1.5817, - "step": 20691 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019300142055063248, - "loss": 1.5431, - "step": 20692 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019300108232429142, - "loss": 1.6859, - "step": 20693 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019300074409795036, - "loss": 1.671, - "step": 20694 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019300040587160927, - "loss": 1.3847, - "step": 20695 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019300006764526821, - "loss": 1.4709, - "step": 20696 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019299972941892715, - "loss": 1.7155, - "step": 20697 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929993911925861, - "loss": 1.4707, - "step": 20698 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019299905296624504, - "loss": 1.671, - "step": 20699 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019299871473990395, - "loss": 1.6121, - "step": 20700 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929983765135629, - "loss": 1.5712, - "step": 20701 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019299803828722183, - "loss": 1.6661, - "step": 20702 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019299770006088075, - "loss": 1.5155, - "step": 20703 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929973618345397, - "loss": 1.552, - "step": 20704 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019299702360819863, - "loss": 1.8815, - "step": 20705 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019299668538185754, - "loss": 1.5308, - "step": 20706 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019299634715551649, - "loss": 1.4965, - "step": 20707 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019299600892917543, - "loss": 1.6163, - "step": 20708 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019299567070283434, - "loss": 1.6907, - "step": 20709 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019299533247649328, - "loss": 1.6399, - "step": 20710 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019299499425015222, - "loss": 1.539, - "step": 20711 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019299465602381114, - "loss": 1.5609, - "step": 20712 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019299431779747008, - "loss": 1.5025, - "step": 20713 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019299397957112902, - "loss": 1.534, - "step": 20714 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019299364134478793, - "loss": 1.5664, - "step": 20715 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019299330311844687, - "loss": 1.7927, - "step": 20716 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019299296489210582, - "loss": 1.6601, - "step": 20717 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019299262666576473, - "loss": 1.6451, - "step": 20718 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019299228843942367, - "loss": 1.6171, - "step": 20719 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929919502130826, - "loss": 1.7025, - "step": 20720 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019299161198674153, - "loss": 1.5642, - "step": 20721 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019299127376040047, - "loss": 1.5386, - "step": 20722 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929909355340594, - "loss": 1.6143, - "step": 20723 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019299059730771832, - "loss": 1.5141, - "step": 20724 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019299025908137726, - "loss": 1.7591, - "step": 20725 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929899208550362, - "loss": 1.5292, - "step": 20726 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019298958262869512, - "loss": 1.6686, - "step": 20727 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019298924440235406, - "loss": 1.4124, - "step": 20728 - }, - { - "epoch": 0.07, - "learning_rate": 0.000192988906176013, - "loss": 1.6305, - "step": 20729 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019298856794967191, - "loss": 1.4493, - "step": 20730 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019298822972333086, - "loss": 1.6681, - "step": 20731 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929878914969898, - "loss": 1.5496, - "step": 20732 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019298755327064874, - "loss": 1.569, - "step": 20733 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019298721504430765, - "loss": 1.6434, - "step": 20734 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929868768179666, - "loss": 1.5015, - "step": 20735 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019298653859162553, - "loss": 1.4929, - "step": 20736 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019298620036528445, - "loss": 1.5808, - "step": 20737 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929858621389434, - "loss": 1.6924, - "step": 20738 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019298552391260233, - "loss": 1.7987, - "step": 20739 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019298518568626124, - "loss": 1.5355, - "step": 20740 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019298484745992019, - "loss": 1.4057, - "step": 20741 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019298450923357913, - "loss": 1.5423, - "step": 20742 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019298417100723804, - "loss": 1.4516, - "step": 20743 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019298383278089698, - "loss": 1.6376, - "step": 20744 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019298349455455592, - "loss": 1.4431, - "step": 20745 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019298315632821484, - "loss": 1.6477, - "step": 20746 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019298281810187378, - "loss": 1.5026, - "step": 20747 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019298247987553272, - "loss": 1.5681, - "step": 20748 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019298214164919163, - "loss": 1.5446, - "step": 20749 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019298180342285057, - "loss": 1.6042, - "step": 20750 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019298146519650952, - "loss": 1.5703, - "step": 20751 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019298112697016843, - "loss": 1.5275, - "step": 20752 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019298078874382737, - "loss": 1.4436, - "step": 20753 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929804505174863, - "loss": 1.5749, - "step": 20754 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019298011229114525, - "loss": 1.3802, - "step": 20755 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929797740648042, - "loss": 1.6571, - "step": 20756 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929794358384631, - "loss": 1.6424, - "step": 20757 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019297909761212205, - "loss": 1.6088, - "step": 20758 - }, - { - "epoch": 0.07, - "learning_rate": 0.000192978759385781, - "loss": 1.5173, - "step": 20759 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929784211594399, - "loss": 1.5059, - "step": 20760 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019297808293309885, - "loss": 1.6488, - "step": 20761 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929777447067578, - "loss": 1.5639, - "step": 20762 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929774064804167, - "loss": 1.6143, - "step": 20763 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019297706825407564, - "loss": 1.6011, - "step": 20764 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019297673002773458, - "loss": 1.4279, - "step": 20765 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929763918013935, - "loss": 1.7078, - "step": 20766 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019297605357505244, - "loss": 1.5732, - "step": 20767 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019297571534871138, - "loss": 1.7529, - "step": 20768 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929753771223703, - "loss": 1.6868, - "step": 20769 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019297503889602923, - "loss": 1.5063, - "step": 20770 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019297470066968818, - "loss": 1.634, - "step": 20771 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929743624433471, - "loss": 1.6801, - "step": 20772 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019297402421700603, - "loss": 1.5316, - "step": 20773 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019297368599066497, - "loss": 1.4881, - "step": 20774 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019297334776432389, - "loss": 1.8187, - "step": 20775 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019297300953798283, - "loss": 1.3424, - "step": 20776 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019297267131164177, - "loss": 1.6306, - "step": 20777 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019297233308530068, - "loss": 1.6038, - "step": 20778 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019297199485895962, - "loss": 1.5816, - "step": 20779 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019297165663261856, - "loss": 1.4593, - "step": 20780 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019297131840627748, - "loss": 1.6351, - "step": 20781 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019297098017993642, - "loss": 1.6581, - "step": 20782 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019297064195359536, - "loss": 1.6868, - "step": 20783 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019297030372725428, - "loss": 1.5896, - "step": 20784 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019296996550091322, - "loss": 1.5505, - "step": 20785 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019296962727457216, - "loss": 1.7595, - "step": 20786 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019296928904823107, - "loss": 1.7701, - "step": 20787 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019296895082189, - "loss": 1.6202, - "step": 20788 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019296861259554895, - "loss": 1.5616, - "step": 20789 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019296827436920787, - "loss": 1.5635, - "step": 20790 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929679361428668, - "loss": 1.6212, - "step": 20791 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019296759791652575, - "loss": 1.5791, - "step": 20792 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019296725969018466, - "loss": 1.563, - "step": 20793 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929669214638436, - "loss": 1.636, - "step": 20794 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019296658323750255, - "loss": 1.6499, - "step": 20795 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019296624501116146, - "loss": 1.4833, - "step": 20796 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929659067848204, - "loss": 1.6512, - "step": 20797 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019296556855847934, - "loss": 1.6122, - "step": 20798 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019296523033213826, - "loss": 1.5984, - "step": 20799 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929648921057972, - "loss": 1.5829, - "step": 20800 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019296455387945614, - "loss": 1.595, - "step": 20801 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019296421565311505, - "loss": 1.6468, - "step": 20802 - }, - { - "epoch": 0.07, - "learning_rate": 0.000192963877426774, - "loss": 1.5845, - "step": 20803 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019296353920043294, - "loss": 1.5585, - "step": 20804 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019296320097409185, - "loss": 1.5315, - "step": 20805 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929628627477508, - "loss": 1.5564, - "step": 20806 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019296252452140973, - "loss": 1.6825, - "step": 20807 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019296218629506865, - "loss": 1.5497, - "step": 20808 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929618480687276, - "loss": 1.4462, - "step": 20809 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019296150984238656, - "loss": 1.5848, - "step": 20810 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019296117161604547, - "loss": 1.6537, - "step": 20811 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929608333897044, - "loss": 1.6465, - "step": 20812 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019296049516336335, - "loss": 1.4571, - "step": 20813 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019296015693702227, - "loss": 1.6744, - "step": 20814 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929598187106812, - "loss": 1.5288, - "step": 20815 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019295948048434015, - "loss": 1.5171, - "step": 20816 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019295914225799906, - "loss": 1.3765, - "step": 20817 - }, - { - "epoch": 0.07, - "learning_rate": 0.000192958804031658, - "loss": 1.4222, - "step": 20818 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019295846580531694, - "loss": 1.422, - "step": 20819 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019295812757897586, - "loss": 1.5938, - "step": 20820 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929577893526348, - "loss": 1.4425, - "step": 20821 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019295745112629374, - "loss": 1.5859, - "step": 20822 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019295711289995265, - "loss": 1.618, - "step": 20823 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929567746736116, - "loss": 1.8383, - "step": 20824 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019295643644727054, - "loss": 1.5281, - "step": 20825 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019295609822092945, - "loss": 1.5687, - "step": 20826 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929557599945884, - "loss": 1.5638, - "step": 20827 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019295542176824733, - "loss": 1.5069, - "step": 20828 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019295508354190625, - "loss": 1.6197, - "step": 20829 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929547453155652, - "loss": 1.5718, - "step": 20830 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019295440708922413, - "loss": 1.6931, - "step": 20831 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019295406886288304, - "loss": 1.6131, - "step": 20832 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019295373063654198, - "loss": 1.6459, - "step": 20833 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019295339241020093, - "loss": 1.5237, - "step": 20834 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019295305418385984, - "loss": 1.5866, - "step": 20835 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019295271595751878, - "loss": 1.5111, - "step": 20836 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019295237773117772, - "loss": 1.6159, - "step": 20837 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019295203950483664, - "loss": 1.4977, - "step": 20838 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019295170127849558, - "loss": 1.5259, - "step": 20839 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019295136305215452, - "loss": 1.4366, - "step": 20840 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019295102482581343, - "loss": 1.538, - "step": 20841 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019295068659947237, - "loss": 1.7263, - "step": 20842 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019295034837313131, - "loss": 1.6852, - "step": 20843 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019295001014679023, - "loss": 1.5938, - "step": 20844 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019294967192044917, - "loss": 1.5243, - "step": 20845 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929493336941081, - "loss": 1.6128, - "step": 20846 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019294899546776702, - "loss": 1.629, - "step": 20847 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019294865724142597, - "loss": 1.5597, - "step": 20848 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929483190150849, - "loss": 1.4249, - "step": 20849 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019294798078874382, - "loss": 1.4774, - "step": 20850 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019294764256240276, - "loss": 1.6214, - "step": 20851 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929473043360617, - "loss": 1.5109, - "step": 20852 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019294696610972062, - "loss": 1.5423, - "step": 20853 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019294662788337956, - "loss": 1.5889, - "step": 20854 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929462896570385, - "loss": 1.6115, - "step": 20855 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019294595143069741, - "loss": 1.5672, - "step": 20856 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019294561320435635, - "loss": 1.6994, - "step": 20857 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929452749780153, - "loss": 1.4656, - "step": 20858 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929449367516742, - "loss": 1.5092, - "step": 20859 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019294459852533315, - "loss": 1.7212, - "step": 20860 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929442602989921, - "loss": 1.4261, - "step": 20861 - }, - { - "epoch": 0.07, - "learning_rate": 0.000192943922072651, - "loss": 1.5771, - "step": 20862 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019294358384630995, - "loss": 1.4514, - "step": 20863 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929432456199689, - "loss": 1.6395, - "step": 20864 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929429073936278, - "loss": 1.5652, - "step": 20865 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019294256916728677, - "loss": 1.5569, - "step": 20866 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929422309409457, - "loss": 1.7962, - "step": 20867 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019294189271460463, - "loss": 1.8107, - "step": 20868 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019294155448826357, - "loss": 1.6366, - "step": 20869 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929412162619225, - "loss": 1.5094, - "step": 20870 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019294087803558142, - "loss": 1.5371, - "step": 20871 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019294053980924036, - "loss": 1.8446, - "step": 20872 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929402015828993, - "loss": 1.6938, - "step": 20873 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019293986335655822, - "loss": 1.61, - "step": 20874 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019293952513021716, - "loss": 1.5502, - "step": 20875 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929391869038761, - "loss": 1.6696, - "step": 20876 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019293884867753502, - "loss": 1.6929, - "step": 20877 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019293851045119396, - "loss": 1.5396, - "step": 20878 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929381722248529, - "loss": 1.7047, - "step": 20879 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929378339985118, - "loss": 1.6591, - "step": 20880 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019293749577217075, - "loss": 1.5039, - "step": 20881 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929371575458297, - "loss": 1.4762, - "step": 20882 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929368193194886, - "loss": 1.578, - "step": 20883 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019293648109314755, - "loss": 1.5334, - "step": 20884 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929361428668065, - "loss": 1.6587, - "step": 20885 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929358046404654, - "loss": 1.5946, - "step": 20886 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019293546641412435, - "loss": 1.5193, - "step": 20887 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019293512818778329, - "loss": 1.4593, - "step": 20888 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929347899614422, - "loss": 1.446, - "step": 20889 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019293445173510114, - "loss": 1.4608, - "step": 20890 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019293411350876008, - "loss": 1.4692, - "step": 20891 - }, - { - "epoch": 0.07, - "learning_rate": 0.000192933775282419, - "loss": 1.6616, - "step": 20892 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019293343705607794, - "loss": 1.6402, - "step": 20893 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019293309882973688, - "loss": 1.5856, - "step": 20894 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929327606033958, - "loss": 1.4886, - "step": 20895 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019293242237705473, - "loss": 1.685, - "step": 20896 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019293208415071368, - "loss": 1.5385, - "step": 20897 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929317459243726, - "loss": 1.4168, - "step": 20898 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019293140769803153, - "loss": 1.358, - "step": 20899 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019293106947169047, - "loss": 1.5926, - "step": 20900 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019293073124534939, - "loss": 1.4641, - "step": 20901 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019293039301900833, - "loss": 1.5363, - "step": 20902 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019293005479266727, - "loss": 1.552, - "step": 20903 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019292971656632618, - "loss": 1.501, - "step": 20904 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019292937833998512, - "loss": 1.5908, - "step": 20905 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019292904011364406, - "loss": 1.5615, - "step": 20906 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019292870188730298, - "loss": 1.6245, - "step": 20907 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019292836366096192, - "loss": 1.5598, - "step": 20908 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019292802543462086, - "loss": 1.6093, - "step": 20909 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019292768720827977, - "loss": 1.6438, - "step": 20910 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019292734898193872, - "loss": 1.461, - "step": 20911 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019292701075559766, - "loss": 1.6393, - "step": 20912 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019292667252925657, - "loss": 1.5578, - "step": 20913 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929263343029155, - "loss": 1.576, - "step": 20914 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019292599607657445, - "loss": 1.423, - "step": 20915 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019292565785023337, - "loss": 1.656, - "step": 20916 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929253196238923, - "loss": 1.4457, - "step": 20917 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019292498139755125, - "loss": 1.6404, - "step": 20918 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019292464317121016, - "loss": 1.5951, - "step": 20919 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929243049448691, - "loss": 1.494, - "step": 20920 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019292396671852805, - "loss": 1.6018, - "step": 20921 - }, - { - "epoch": 0.07, - "learning_rate": 0.000192923628492187, - "loss": 1.559, - "step": 20922 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019292329026584593, - "loss": 1.6022, - "step": 20923 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019292295203950487, - "loss": 1.5048, - "step": 20924 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019292261381316378, - "loss": 1.4543, - "step": 20925 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019292227558682272, - "loss": 1.6522, - "step": 20926 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019292193736048167, - "loss": 1.8323, - "step": 20927 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019292159913414058, - "loss": 1.3972, - "step": 20928 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019292126090779952, - "loss": 1.6169, - "step": 20929 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019292092268145846, - "loss": 1.6831, - "step": 20930 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019292058445511738, - "loss": 1.5142, - "step": 20931 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019292024622877632, - "loss": 1.4292, - "step": 20932 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019291990800243526, - "loss": 1.5099, - "step": 20933 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019291956977609417, - "loss": 1.4831, - "step": 20934 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929192315497531, - "loss": 1.5453, - "step": 20935 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019291889332341205, - "loss": 1.5631, - "step": 20936 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019291855509707097, - "loss": 1.6488, - "step": 20937 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929182168707299, - "loss": 1.5752, - "step": 20938 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019291787864438885, - "loss": 1.4928, - "step": 20939 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019291754041804776, - "loss": 1.6097, - "step": 20940 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929172021917067, - "loss": 1.5196, - "step": 20941 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019291686396536565, - "loss": 1.6157, - "step": 20942 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019291652573902456, - "loss": 1.6735, - "step": 20943 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929161875126835, - "loss": 1.3957, - "step": 20944 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019291584928634244, - "loss": 1.6298, - "step": 20945 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019291551106000136, - "loss": 1.6295, - "step": 20946 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929151728336603, - "loss": 1.5913, - "step": 20947 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019291483460731924, - "loss": 1.4734, - "step": 20948 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019291449638097815, - "loss": 1.5726, - "step": 20949 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929141581546371, - "loss": 1.6016, - "step": 20950 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019291381992829604, - "loss": 1.6178, - "step": 20951 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019291348170195495, - "loss": 1.6675, - "step": 20952 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929131434756139, - "loss": 1.5677, - "step": 20953 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019291280524927283, - "loss": 1.579, - "step": 20954 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019291246702293175, - "loss": 1.5015, - "step": 20955 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929121287965907, - "loss": 1.5054, - "step": 20956 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019291179057024963, - "loss": 1.6123, - "step": 20957 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019291145234390854, - "loss": 1.6896, - "step": 20958 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019291111411756748, - "loss": 1.7144, - "step": 20959 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019291077589122642, - "loss": 1.5912, - "step": 20960 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019291043766488534, - "loss": 1.6911, - "step": 20961 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019291009943854428, - "loss": 1.5157, - "step": 20962 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019290976121220322, - "loss": 1.652, - "step": 20963 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019290942298586214, - "loss": 1.6071, - "step": 20964 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019290908475952108, - "loss": 1.492, - "step": 20965 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019290874653318002, - "loss": 1.4577, - "step": 20966 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019290840830683893, - "loss": 1.3757, - "step": 20967 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019290807008049787, - "loss": 1.7029, - "step": 20968 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019290773185415681, - "loss": 1.5219, - "step": 20969 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019290739362781573, - "loss": 1.5575, - "step": 20970 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019290705540147467, - "loss": 1.5603, - "step": 20971 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929067171751336, - "loss": 1.5094, - "step": 20972 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019290637894879252, - "loss": 1.5492, - "step": 20973 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019290604072245147, - "loss": 1.8434, - "step": 20974 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929057024961104, - "loss": 1.4792, - "step": 20975 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019290536426976932, - "loss": 1.5671, - "step": 20976 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019290502604342826, - "loss": 1.5085, - "step": 20977 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929046878170872, - "loss": 1.7062, - "step": 20978 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019290434959074614, - "loss": 1.5521, - "step": 20979 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019290401136440509, - "loss": 1.462, - "step": 20980 - }, - { - "epoch": 0.07, - "learning_rate": 0.000192903673138064, - "loss": 1.5344, - "step": 20981 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019290333491172294, - "loss": 1.355, - "step": 20982 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019290299668538188, - "loss": 1.543, - "step": 20983 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929026584590408, - "loss": 1.6126, - "step": 20984 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019290232023269974, - "loss": 1.3718, - "step": 20985 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019290198200635868, - "loss": 1.6218, - "step": 20986 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929016437800176, - "loss": 1.6194, - "step": 20987 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019290130555367653, - "loss": 1.6794, - "step": 20988 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019290096732733547, - "loss": 1.5485, - "step": 20989 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001929006291009944, - "loss": 1.5577, - "step": 20990 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019290029087465333, - "loss": 1.6438, - "step": 20991 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019289995264831227, - "loss": 1.6528, - "step": 20992 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019289961442197118, - "loss": 1.5235, - "step": 20993 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019289927619563013, - "loss": 1.7035, - "step": 20994 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019289893796928907, - "loss": 1.606, - "step": 20995 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019289859974294798, - "loss": 1.6894, - "step": 20996 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019289826151660692, - "loss": 1.5006, - "step": 20997 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019289792329026586, - "loss": 1.5684, - "step": 20998 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019289758506392478, - "loss": 1.5316, - "step": 20999 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019289724683758372, - "loss": 1.5155, - "step": 21000 - }, - { - "epoch": 0.07, - "eval_loss": 1.333251953125, - "mrc_em": 20.0, - "mrc_f1": 27.285714285714292, - "nli_acc": 0.13, - "step": 21000, - "sts_f1": 0.3548387096774194, - "ynat_f1": 0.5368883196332983 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019289690861124266, - "loss": 1.4022, - "step": 21001 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928965703849016, - "loss": 1.3607, - "step": 21002 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019289623215856051, - "loss": 1.567, - "step": 21003 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019289589393221946, - "loss": 1.6192, - "step": 21004 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928955557058784, - "loss": 1.6121, - "step": 21005 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928952174795373, - "loss": 1.6117, - "step": 21006 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019289487925319625, - "loss": 1.4649, - "step": 21007 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928945410268552, - "loss": 1.6578, - "step": 21008 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928942028005141, - "loss": 1.5264, - "step": 21009 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019289386457417305, - "loss": 1.4699, - "step": 21010 - }, - { - "epoch": 0.07, - "learning_rate": 0.000192893526347832, - "loss": 1.5695, - "step": 21011 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928931881214909, - "loss": 1.5018, - "step": 21012 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019289284989514984, - "loss": 1.5427, - "step": 21013 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019289251166880879, - "loss": 1.5028, - "step": 21014 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928921734424677, - "loss": 1.6181, - "step": 21015 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019289183521612664, - "loss": 1.6894, - "step": 21016 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019289149698978558, - "loss": 1.5027, - "step": 21017 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928911587634445, - "loss": 1.6264, - "step": 21018 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019289082053710344, - "loss": 1.7705, - "step": 21019 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019289048231076238, - "loss": 1.6284, - "step": 21020 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928901440844213, - "loss": 1.6058, - "step": 21021 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019288980585808023, - "loss": 1.4946, - "step": 21022 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019288946763173917, - "loss": 1.504, - "step": 21023 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928891294053981, - "loss": 1.5731, - "step": 21024 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019288879117905703, - "loss": 1.6846, - "step": 21025 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019288845295271597, - "loss": 1.5164, - "step": 21026 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019288811472637488, - "loss": 1.6063, - "step": 21027 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019288777650003383, - "loss": 1.4423, - "step": 21028 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019288743827369277, - "loss": 1.7973, - "step": 21029 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019288710004735168, - "loss": 1.5276, - "step": 21030 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019288676182101062, - "loss": 1.9279, - "step": 21031 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019288642359466956, - "loss": 1.4269, - "step": 21032 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019288608536832848, - "loss": 1.6303, - "step": 21033 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019288574714198742, - "loss": 1.4934, - "step": 21034 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019288540891564636, - "loss": 1.7096, - "step": 21035 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928850706893053, - "loss": 1.5634, - "step": 21036 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019288473246296424, - "loss": 1.6958, - "step": 21037 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019288439423662316, - "loss": 1.6105, - "step": 21038 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928840560102821, - "loss": 1.5653, - "step": 21039 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019288371778394104, - "loss": 1.4892, - "step": 21040 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019288337955759995, - "loss": 1.4391, - "step": 21041 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928830413312589, - "loss": 1.6533, - "step": 21042 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019288270310491783, - "loss": 1.4731, - "step": 21043 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019288236487857675, - "loss": 1.6673, - "step": 21044 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928820266522357, - "loss": 1.3631, - "step": 21045 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019288168842589463, - "loss": 1.7901, - "step": 21046 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019288135019955355, - "loss": 1.5032, - "step": 21047 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019288101197321249, - "loss": 1.5578, - "step": 21048 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019288067374687143, - "loss": 1.5389, - "step": 21049 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019288033552053034, - "loss": 1.4939, - "step": 21050 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019287999729418928, - "loss": 1.7613, - "step": 21051 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019287965906784822, - "loss": 1.6973, - "step": 21052 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019287932084150714, - "loss": 1.4772, - "step": 21053 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019287898261516608, - "loss": 1.64, - "step": 21054 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019287864438882502, - "loss": 1.5516, - "step": 21055 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019287830616248393, - "loss": 1.7326, - "step": 21056 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019287796793614288, - "loss": 1.782, - "step": 21057 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019287762970980182, - "loss": 1.4236, - "step": 21058 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019287729148346073, - "loss": 1.479, - "step": 21059 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019287695325711967, - "loss": 1.6269, - "step": 21060 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928766150307786, - "loss": 1.6994, - "step": 21061 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019287627680443753, - "loss": 1.5399, - "step": 21062 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019287593857809647, - "loss": 1.5616, - "step": 21063 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928756003517554, - "loss": 1.5621, - "step": 21064 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019287526212541432, - "loss": 1.5035, - "step": 21065 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019287492389907326, - "loss": 1.617, - "step": 21066 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928745856727322, - "loss": 1.5732, - "step": 21067 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019287424744639112, - "loss": 1.5615, - "step": 21068 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019287390922005006, - "loss": 1.7118, - "step": 21069 - }, - { - "epoch": 0.07, - "learning_rate": 0.000192873570993709, - "loss": 1.485, - "step": 21070 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019287323276736792, - "loss": 1.6286, - "step": 21071 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019287289454102686, - "loss": 1.5154, - "step": 21072 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928725563146858, - "loss": 1.675, - "step": 21073 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928722180883447, - "loss": 1.7584, - "step": 21074 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019287187986200365, - "loss": 1.4242, - "step": 21075 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928715416356626, - "loss": 1.478, - "step": 21076 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928712034093215, - "loss": 1.6304, - "step": 21077 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019287086518298045, - "loss": 1.6036, - "step": 21078 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928705269566394, - "loss": 1.5419, - "step": 21079 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928701887302983, - "loss": 1.5725, - "step": 21080 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019286985050395725, - "loss": 1.3299, - "step": 21081 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928695122776162, - "loss": 1.6004, - "step": 21082 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928691740512751, - "loss": 1.779, - "step": 21083 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019286883582493404, - "loss": 1.8441, - "step": 21084 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019286849759859298, - "loss": 1.472, - "step": 21085 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019286815937225192, - "loss": 1.657, - "step": 21086 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019286782114591084, - "loss": 1.699, - "step": 21087 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019286748291956978, - "loss": 1.579, - "step": 21088 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019286714469322872, - "loss": 1.6885, - "step": 21089 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019286680646688763, - "loss": 1.5159, - "step": 21090 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928664682405466, - "loss": 1.4479, - "step": 21091 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019286613001420552, - "loss": 1.4644, - "step": 21092 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019286579178786446, - "loss": 1.6905, - "step": 21093 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928654535615234, - "loss": 1.5713, - "step": 21094 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928651153351823, - "loss": 1.42, - "step": 21095 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019286477710884125, - "loss": 1.628, - "step": 21096 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928644388825002, - "loss": 1.6289, - "step": 21097 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928641006561591, - "loss": 1.6638, - "step": 21098 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019286376242981805, - "loss": 1.5972, - "step": 21099 - }, - { - "epoch": 0.07, - "learning_rate": 0.000192863424203477, - "loss": 1.5208, - "step": 21100 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928630859771359, - "loss": 1.6256, - "step": 21101 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019286274775079485, - "loss": 1.6836, - "step": 21102 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928624095244538, - "loss": 1.5587, - "step": 21103 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928620712981127, - "loss": 1.5508, - "step": 21104 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019286173307177164, - "loss": 1.6127, - "step": 21105 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019286139484543058, - "loss": 1.5194, - "step": 21106 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928610566190895, - "loss": 1.4981, - "step": 21107 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019286071839274844, - "loss": 1.5462, - "step": 21108 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019286038016640738, - "loss": 1.6052, - "step": 21109 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928600419400663, - "loss": 1.4548, - "step": 21110 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019285970371372524, - "loss": 1.6592, - "step": 21111 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019285936548738418, - "loss": 1.5598, - "step": 21112 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928590272610431, - "loss": 1.528, - "step": 21113 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019285868903470203, - "loss": 1.4086, - "step": 21114 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019285835080836097, - "loss": 1.4417, - "step": 21115 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928580125820199, - "loss": 1.6458, - "step": 21116 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019285767435567883, - "loss": 1.4524, - "step": 21117 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019285733612933777, - "loss": 1.5472, - "step": 21118 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019285699790299668, - "loss": 1.4501, - "step": 21119 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019285665967665562, - "loss": 1.3923, - "step": 21120 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019285632145031457, - "loss": 1.537, - "step": 21121 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019285598322397348, - "loss": 1.6141, - "step": 21122 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019285564499763242, - "loss": 1.516, - "step": 21123 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019285530677129136, - "loss": 1.3904, - "step": 21124 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019285496854495028, - "loss": 1.6931, - "step": 21125 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019285463031860922, - "loss": 1.5429, - "step": 21126 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019285429209226816, - "loss": 1.6198, - "step": 21127 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019285395386592707, - "loss": 1.7602, - "step": 21128 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019285361563958601, - "loss": 1.5797, - "step": 21129 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019285327741324496, - "loss": 1.6314, - "step": 21130 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019285293918690387, - "loss": 1.4995, - "step": 21131 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928526009605628, - "loss": 1.6228, - "step": 21132 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019285226273422175, - "loss": 1.5356, - "step": 21133 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019285192450788067, - "loss": 1.6625, - "step": 21134 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928515862815396, - "loss": 1.6995, - "step": 21135 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019285124805519855, - "loss": 1.5639, - "step": 21136 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019285090982885746, - "loss": 1.6018, - "step": 21137 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928505716025164, - "loss": 1.57, - "step": 21138 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019285023337617534, - "loss": 1.6061, - "step": 21139 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019284989514983426, - "loss": 1.6002, - "step": 21140 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928495569234932, - "loss": 1.7432, - "step": 21141 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019284921869715214, - "loss": 1.5396, - "step": 21142 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019284888047081105, - "loss": 1.6157, - "step": 21143 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019284854224447, - "loss": 1.6085, - "step": 21144 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019284820401812894, - "loss": 1.5796, - "step": 21145 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019284786579178785, - "loss": 1.5444, - "step": 21146 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019284752756544682, - "loss": 1.6921, - "step": 21147 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019284718933910576, - "loss": 1.6707, - "step": 21148 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019284685111276467, - "loss": 1.6166, - "step": 21149 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019284651288642362, - "loss": 1.5519, - "step": 21150 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019284617466008256, - "loss": 1.5397, - "step": 21151 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019284583643374147, - "loss": 1.4942, - "step": 21152 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928454982074004, - "loss": 1.7237, - "step": 21153 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019284515998105935, - "loss": 1.6011, - "step": 21154 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019284482175471827, - "loss": 1.4604, - "step": 21155 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928444835283772, - "loss": 1.7078, - "step": 21156 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019284414530203615, - "loss": 1.5818, - "step": 21157 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019284380707569506, - "loss": 1.5865, - "step": 21158 - }, - { - "epoch": 0.07, - "learning_rate": 0.000192843468849354, - "loss": 1.6125, - "step": 21159 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019284313062301295, - "loss": 1.6332, - "step": 21160 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019284279239667186, - "loss": 1.6863, - "step": 21161 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928424541703308, - "loss": 1.3858, - "step": 21162 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019284211594398974, - "loss": 1.456, - "step": 21163 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019284177771764866, - "loss": 1.4738, - "step": 21164 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928414394913076, - "loss": 1.5064, - "step": 21165 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019284110126496654, - "loss": 1.5442, - "step": 21166 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019284076303862545, - "loss": 1.4887, - "step": 21167 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928404248122844, - "loss": 1.6155, - "step": 21168 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019284008658594333, - "loss": 1.6127, - "step": 21169 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019283974835960225, - "loss": 1.6081, - "step": 21170 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928394101332612, - "loss": 1.5347, - "step": 21171 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019283907190692013, - "loss": 1.5244, - "step": 21172 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019283873368057904, - "loss": 1.5073, - "step": 21173 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019283839545423799, - "loss": 1.6751, - "step": 21174 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019283805722789693, - "loss": 1.7625, - "step": 21175 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019283771900155584, - "loss": 1.7065, - "step": 21176 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019283738077521478, - "loss": 1.6481, - "step": 21177 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019283704254887372, - "loss": 1.407, - "step": 21178 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019283670432253264, - "loss": 1.6079, - "step": 21179 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019283636609619158, - "loss": 1.4891, - "step": 21180 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019283602786985052, - "loss": 1.6815, - "step": 21181 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019283568964350943, - "loss": 1.6164, - "step": 21182 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019283535141716837, - "loss": 1.5308, - "step": 21183 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019283501319082732, - "loss": 1.4885, - "step": 21184 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019283467496448623, - "loss": 1.467, - "step": 21185 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019283433673814517, - "loss": 1.6813, - "step": 21186 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928339985118041, - "loss": 1.618, - "step": 21187 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019283366028546303, - "loss": 1.5626, - "step": 21188 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019283332205912197, - "loss": 1.5282, - "step": 21189 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928329838327809, - "loss": 1.5797, - "step": 21190 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019283264560643982, - "loss": 1.7127, - "step": 21191 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019283230738009876, - "loss": 1.6834, - "step": 21192 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928319691537577, - "loss": 1.5087, - "step": 21193 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019283163092741662, - "loss": 1.6948, - "step": 21194 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019283129270107556, - "loss": 1.5393, - "step": 21195 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928309544747345, - "loss": 1.5177, - "step": 21196 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019283061624839342, - "loss": 1.575, - "step": 21197 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019283027802205236, - "loss": 1.4327, - "step": 21198 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928299397957113, - "loss": 1.5108, - "step": 21199 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928296015693702, - "loss": 1.5832, - "step": 21200 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019282926334302915, - "loss": 1.6669, - "step": 21201 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928289251166881, - "loss": 1.5256, - "step": 21202 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019282858689034703, - "loss": 1.5843, - "step": 21203 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019282824866400598, - "loss": 1.7782, - "step": 21204 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019282791043766492, - "loss": 1.6342, - "step": 21205 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019282757221132383, - "loss": 1.4927, - "step": 21206 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019282723398498277, - "loss": 1.5787, - "step": 21207 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928268957586417, - "loss": 1.5361, - "step": 21208 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019282655753230063, - "loss": 1.6097, - "step": 21209 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019282621930595957, - "loss": 1.7793, - "step": 21210 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928258810796185, - "loss": 1.6128, - "step": 21211 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019282554285327742, - "loss": 1.5032, - "step": 21212 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019282520462693636, - "loss": 1.5071, - "step": 21213 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928248664005953, - "loss": 1.3683, - "step": 21214 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019282452817425422, - "loss": 1.6386, - "step": 21215 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019282418994791316, - "loss": 1.3902, - "step": 21216 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928238517215721, - "loss": 1.6088, - "step": 21217 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019282351349523102, - "loss": 1.4432, - "step": 21218 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019282317526888996, - "loss": 1.4952, - "step": 21219 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928228370425489, - "loss": 1.511, - "step": 21220 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928224988162078, - "loss": 1.4731, - "step": 21221 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019282216058986675, - "loss": 1.4963, - "step": 21222 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928218223635257, - "loss": 1.6176, - "step": 21223 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928214841371846, - "loss": 1.4428, - "step": 21224 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019282114591084355, - "loss": 1.5384, - "step": 21225 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928208076845025, - "loss": 1.5127, - "step": 21226 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928204694581614, - "loss": 1.6747, - "step": 21227 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019282013123182035, - "loss": 1.4858, - "step": 21228 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928197930054793, - "loss": 1.6175, - "step": 21229 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928194547791382, - "loss": 1.4568, - "step": 21230 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019281911655279714, - "loss": 1.5771, - "step": 21231 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019281877832645608, - "loss": 1.6545, - "step": 21232 - }, - { - "epoch": 0.07, - "learning_rate": 0.000192818440100115, - "loss": 1.6286, - "step": 21233 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019281810187377394, - "loss": 1.5139, - "step": 21234 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019281776364743288, - "loss": 1.6086, - "step": 21235 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928174254210918, - "loss": 1.6589, - "step": 21236 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019281708719475074, - "loss": 1.5812, - "step": 21237 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019281674896840968, - "loss": 1.4196, - "step": 21238 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928164107420686, - "loss": 1.6161, - "step": 21239 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019281607251572753, - "loss": 1.3834, - "step": 21240 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019281573428938647, - "loss": 1.513, - "step": 21241 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928153960630454, - "loss": 1.6889, - "step": 21242 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019281505783670433, - "loss": 1.5529, - "step": 21243 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019281471961036327, - "loss": 1.5678, - "step": 21244 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019281438138402218, - "loss": 1.5157, - "step": 21245 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019281404315768112, - "loss": 1.7772, - "step": 21246 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019281370493134007, - "loss": 1.4814, - "step": 21247 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019281336670499898, - "loss": 1.644, - "step": 21248 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019281302847865792, - "loss": 1.4813, - "step": 21249 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019281269025231686, - "loss": 1.4335, - "step": 21250 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019281235202597578, - "loss": 1.5327, - "step": 21251 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019281201379963472, - "loss": 1.5499, - "step": 21252 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019281167557329366, - "loss": 1.5911, - "step": 21253 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019281133734695257, - "loss": 1.5003, - "step": 21254 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928109991206115, - "loss": 1.7711, - "step": 21255 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019281066089427045, - "loss": 1.5528, - "step": 21256 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019281032266792937, - "loss": 1.5913, - "step": 21257 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928099844415883, - "loss": 1.6035, - "step": 21258 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019280964621524725, - "loss": 1.6131, - "step": 21259 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928093079889062, - "loss": 1.3987, - "step": 21260 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019280896976256513, - "loss": 1.4851, - "step": 21261 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019280863153622405, - "loss": 1.6253, - "step": 21262 - }, - { - "epoch": 0.07, - "learning_rate": 0.000192808293309883, - "loss": 1.316, - "step": 21263 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019280795508354193, - "loss": 1.6647, - "step": 21264 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019280761685720084, - "loss": 1.5082, - "step": 21265 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019280727863085978, - "loss": 1.7043, - "step": 21266 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019280694040451873, - "loss": 1.6195, - "step": 21267 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019280660217817767, - "loss": 1.6233, - "step": 21268 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019280626395183658, - "loss": 1.5701, - "step": 21269 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019280592572549552, - "loss": 1.6094, - "step": 21270 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019280558749915446, - "loss": 1.6083, - "step": 21271 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019280524927281338, - "loss": 1.6807, - "step": 21272 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019280491104647232, - "loss": 1.4681, - "step": 21273 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019280457282013126, - "loss": 1.5611, - "step": 21274 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019280423459379017, - "loss": 1.5806, - "step": 21275 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019280389636744911, - "loss": 1.46, - "step": 21276 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019280355814110806, - "loss": 1.4601, - "step": 21277 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019280321991476697, - "loss": 1.6256, - "step": 21278 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928028816884259, - "loss": 1.3586, - "step": 21279 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019280254346208485, - "loss": 1.5908, - "step": 21280 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019280220523574377, - "loss": 1.4585, - "step": 21281 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928018670094027, - "loss": 1.5402, - "step": 21282 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019280152878306165, - "loss": 1.6637, - "step": 21283 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019280119055672056, - "loss": 1.604, - "step": 21284 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001928008523303795, - "loss": 1.6242, - "step": 21285 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019280051410403844, - "loss": 1.5938, - "step": 21286 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019280017587769736, - "loss": 1.5168, - "step": 21287 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927998376513563, - "loss": 1.3787, - "step": 21288 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019279949942501524, - "loss": 1.4875, - "step": 21289 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019279916119867415, - "loss": 1.6114, - "step": 21290 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927988229723331, - "loss": 1.5461, - "step": 21291 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019279848474599204, - "loss": 1.4049, - "step": 21292 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019279814651965095, - "loss": 1.5957, - "step": 21293 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927978082933099, - "loss": 1.6334, - "step": 21294 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019279747006696883, - "loss": 1.5744, - "step": 21295 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019279713184062775, - "loss": 1.7633, - "step": 21296 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927967936142867, - "loss": 1.478, - "step": 21297 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019279645538794563, - "loss": 1.5928, - "step": 21298 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019279611716160454, - "loss": 1.6838, - "step": 21299 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019279577893526349, - "loss": 1.5078, - "step": 21300 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019279544070892243, - "loss": 1.5122, - "step": 21301 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019279510248258134, - "loss": 1.5411, - "step": 21302 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019279476425624028, - "loss": 1.6269, - "step": 21303 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019279442602989922, - "loss": 1.714, - "step": 21304 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019279408780355814, - "loss": 1.5603, - "step": 21305 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019279374957721708, - "loss": 1.5778, - "step": 21306 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019279341135087602, - "loss": 1.5791, - "step": 21307 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019279307312453493, - "loss": 1.5248, - "step": 21308 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019279273489819387, - "loss": 1.5101, - "step": 21309 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019279239667185282, - "loss": 1.7069, - "step": 21310 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019279205844551173, - "loss": 1.6244, - "step": 21311 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019279172021917067, - "loss": 1.6392, - "step": 21312 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927913819928296, - "loss": 1.5406, - "step": 21313 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019279104376648853, - "loss": 1.4831, - "step": 21314 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019279070554014747, - "loss": 1.6386, - "step": 21315 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927903673138064, - "loss": 1.6352, - "step": 21316 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019279002908746535, - "loss": 1.6736, - "step": 21317 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927896908611243, - "loss": 1.5048, - "step": 21318 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927893526347832, - "loss": 1.6105, - "step": 21319 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019278901440844215, - "loss": 1.6638, - "step": 21320 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019278867618210109, - "loss": 1.3807, - "step": 21321 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019278833795576, - "loss": 1.3321, - "step": 21322 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019278799972941894, - "loss": 1.5421, - "step": 21323 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019278766150307788, - "loss": 1.547, - "step": 21324 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927873232767368, - "loss": 1.5445, - "step": 21325 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019278698505039574, - "loss": 1.5749, - "step": 21326 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019278664682405468, - "loss": 1.5512, - "step": 21327 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927863085977136, - "loss": 1.58, - "step": 21328 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019278597037137253, - "loss": 1.5008, - "step": 21329 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019278563214503148, - "loss": 1.558, - "step": 21330 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927852939186904, - "loss": 1.6698, - "step": 21331 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019278495569234933, - "loss": 1.648, - "step": 21332 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019278461746600827, - "loss": 1.6681, - "step": 21333 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019278427923966719, - "loss": 1.5295, - "step": 21334 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019278394101332613, - "loss": 1.5925, - "step": 21335 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019278360278698507, - "loss": 1.5186, - "step": 21336 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019278326456064398, - "loss": 1.5441, - "step": 21337 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019278292633430292, - "loss": 1.5663, - "step": 21338 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019278258810796186, - "loss": 1.6107, - "step": 21339 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019278224988162078, - "loss": 1.6532, - "step": 21340 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019278191165527972, - "loss": 1.491, - "step": 21341 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019278157342893866, - "loss": 1.5277, - "step": 21342 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019278123520259757, - "loss": 1.5241, - "step": 21343 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019278089697625652, - "loss": 1.497, - "step": 21344 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019278055874991546, - "loss": 1.5242, - "step": 21345 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019278022052357437, - "loss": 1.5294, - "step": 21346 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927798822972333, - "loss": 1.4981, - "step": 21347 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019277954407089225, - "loss": 1.5188, - "step": 21348 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019277920584455117, - "loss": 1.5675, - "step": 21349 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927788676182101, - "loss": 1.4326, - "step": 21350 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019277852939186905, - "loss": 1.6039, - "step": 21351 - }, - { - "epoch": 0.07, - "learning_rate": 0.000192778191165528, - "loss": 1.4786, - "step": 21352 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927778529391869, - "loss": 1.6588, - "step": 21353 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019277751471284585, - "loss": 1.7033, - "step": 21354 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927771764865048, - "loss": 1.5268, - "step": 21355 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927768382601637, - "loss": 1.833, - "step": 21356 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019277650003382264, - "loss": 1.6754, - "step": 21357 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019277616180748158, - "loss": 1.4932, - "step": 21358 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927758235811405, - "loss": 1.5451, - "step": 21359 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019277548535479944, - "loss": 1.4929, - "step": 21360 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019277514712845838, - "loss": 1.6102, - "step": 21361 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927748089021173, - "loss": 1.5048, - "step": 21362 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019277447067577623, - "loss": 1.7226, - "step": 21363 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019277413244943518, - "loss": 1.4116, - "step": 21364 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927737942230941, - "loss": 1.4297, - "step": 21365 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019277345599675303, - "loss": 1.5393, - "step": 21366 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019277311777041197, - "loss": 1.4615, - "step": 21367 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019277277954407089, - "loss": 1.6091, - "step": 21368 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019277244131772983, - "loss": 1.5223, - "step": 21369 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019277210309138877, - "loss": 1.7056, - "step": 21370 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019277176486504768, - "loss": 1.6162, - "step": 21371 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019277142663870665, - "loss": 1.6533, - "step": 21372 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019277108841236556, - "loss": 1.6541, - "step": 21373 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927707501860245, - "loss": 1.6786, - "step": 21374 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019277041195968345, - "loss": 1.557, - "step": 21375 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019277007373334236, - "loss": 1.5665, - "step": 21376 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927697355070013, - "loss": 1.5123, - "step": 21377 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019276939728066024, - "loss": 1.5184, - "step": 21378 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019276905905431916, - "loss": 1.5512, - "step": 21379 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927687208279781, - "loss": 1.4073, - "step": 21380 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019276838260163704, - "loss": 1.4894, - "step": 21381 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019276804437529595, - "loss": 1.5179, - "step": 21382 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927677061489549, - "loss": 1.6333, - "step": 21383 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019276736792261384, - "loss": 1.4711, - "step": 21384 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019276702969627275, - "loss": 1.5723, - "step": 21385 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927666914699317, - "loss": 1.5411, - "step": 21386 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019276635324359063, - "loss": 1.5143, - "step": 21387 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019276601501724955, - "loss": 1.7241, - "step": 21388 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927656767909085, - "loss": 1.5783, - "step": 21389 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019276533856456743, - "loss": 1.5495, - "step": 21390 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019276500033822634, - "loss": 1.5868, - "step": 21391 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019276466211188528, - "loss": 1.4963, - "step": 21392 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019276432388554422, - "loss": 1.5636, - "step": 21393 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019276398565920314, - "loss": 1.5378, - "step": 21394 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019276364743286208, - "loss": 1.5661, - "step": 21395 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019276330920652102, - "loss": 1.6419, - "step": 21396 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019276297098017994, - "loss": 1.4145, - "step": 21397 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019276263275383888, - "loss": 1.6032, - "step": 21398 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019276229452749782, - "loss": 1.5803, - "step": 21399 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019276195630115673, - "loss": 1.4272, - "step": 21400 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019276161807481567, - "loss": 1.5756, - "step": 21401 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019276127984847461, - "loss": 1.5434, - "step": 21402 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019276094162213353, - "loss": 1.7647, - "step": 21403 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019276060339579247, - "loss": 1.4087, - "step": 21404 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927602651694514, - "loss": 1.5865, - "step": 21405 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019275992694311032, - "loss": 1.6296, - "step": 21406 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019275958871676927, - "loss": 1.6519, - "step": 21407 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927592504904282, - "loss": 1.5439, - "step": 21408 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019275891226408712, - "loss": 1.6459, - "step": 21409 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019275857403774606, - "loss": 1.5939, - "step": 21410 - }, - { - "epoch": 0.07, - "learning_rate": 0.000192758235811405, - "loss": 1.5338, - "step": 21411 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019275789758506392, - "loss": 1.4588, - "step": 21412 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019275755935872286, - "loss": 1.6254, - "step": 21413 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927572211323818, - "loss": 1.6605, - "step": 21414 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927568829060407, - "loss": 1.6113, - "step": 21415 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019275654467969965, - "loss": 1.4241, - "step": 21416 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927562064533586, - "loss": 1.6583, - "step": 21417 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927558682270175, - "loss": 1.6641, - "step": 21418 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019275553000067645, - "loss": 1.5074, - "step": 21419 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927551917743354, - "loss": 1.6371, - "step": 21420 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927548535479943, - "loss": 1.622, - "step": 21421 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019275451532165325, - "loss": 1.5755, - "step": 21422 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927541770953122, - "loss": 1.6157, - "step": 21423 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927538388689711, - "loss": 1.5726, - "step": 21424 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019275350064263004, - "loss": 1.555, - "step": 21425 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019275316241628898, - "loss": 1.4456, - "step": 21426 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927528241899479, - "loss": 1.7982, - "step": 21427 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019275248596360687, - "loss": 1.6015, - "step": 21428 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927521477372658, - "loss": 1.5728, - "step": 21429 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019275180951092472, - "loss": 1.574, - "step": 21430 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019275147128458366, - "loss": 1.3316, - "step": 21431 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927511330582426, - "loss": 1.5285, - "step": 21432 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019275079483190152, - "loss": 1.4842, - "step": 21433 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019275045660556046, - "loss": 1.5709, - "step": 21434 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927501183792194, - "loss": 1.5182, - "step": 21435 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019274978015287831, - "loss": 1.5035, - "step": 21436 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019274944192653726, - "loss": 1.6593, - "step": 21437 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927491037001962, - "loss": 1.451, - "step": 21438 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927487654738551, - "loss": 1.4339, - "step": 21439 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019274842724751405, - "loss": 1.6516, - "step": 21440 - }, - { - "epoch": 0.07, - "learning_rate": 0.000192748089021173, - "loss": 1.5891, - "step": 21441 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927477507948319, - "loss": 1.6378, - "step": 21442 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019274741256849085, - "loss": 1.5609, - "step": 21443 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927470743421498, - "loss": 1.5099, - "step": 21444 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927467361158087, - "loss": 1.7723, - "step": 21445 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019274639788946764, - "loss": 1.5695, - "step": 21446 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019274605966312659, - "loss": 1.5995, - "step": 21447 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927457214367855, - "loss": 1.4167, - "step": 21448 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019274538321044444, - "loss": 1.4549, - "step": 21449 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019274504498410338, - "loss": 1.621, - "step": 21450 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927447067577623, - "loss": 1.5116, - "step": 21451 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019274436853142124, - "loss": 1.5476, - "step": 21452 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019274403030508018, - "loss": 1.6859, - "step": 21453 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927436920787391, - "loss": 1.5212, - "step": 21454 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019274335385239803, - "loss": 1.6233, - "step": 21455 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019274301562605697, - "loss": 1.6295, - "step": 21456 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927426773997159, - "loss": 1.3799, - "step": 21457 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019274233917337483, - "loss": 1.5156, - "step": 21458 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019274200094703377, - "loss": 1.6493, - "step": 21459 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019274166272069269, - "loss": 1.5028, - "step": 21460 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019274132449435163, - "loss": 1.5342, - "step": 21461 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019274098626801057, - "loss": 1.5107, - "step": 21462 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019274064804166948, - "loss": 1.3587, - "step": 21463 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019274030981532842, - "loss": 1.5382, - "step": 21464 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019273997158898736, - "loss": 1.4359, - "step": 21465 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019273963336264628, - "loss": 1.5838, - "step": 21466 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019273929513630522, - "loss": 1.5724, - "step": 21467 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019273895690996416, - "loss": 1.7299, - "step": 21468 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019273861868362307, - "loss": 1.6472, - "step": 21469 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019273828045728202, - "loss": 1.6582, - "step": 21470 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019273794223094096, - "loss": 1.5627, - "step": 21471 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019273760400459987, - "loss": 1.5869, - "step": 21472 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927372657782588, - "loss": 1.602, - "step": 21473 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019273692755191775, - "loss": 1.5909, - "step": 21474 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019273658932557667, - "loss": 1.52, - "step": 21475 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927362510992356, - "loss": 1.4155, - "step": 21476 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019273591287289455, - "loss": 1.475, - "step": 21477 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019273557464655346, - "loss": 1.5689, - "step": 21478 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927352364202124, - "loss": 1.6733, - "step": 21479 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019273489819387135, - "loss": 1.5811, - "step": 21480 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019273455996753026, - "loss": 1.5285, - "step": 21481 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927342217411892, - "loss": 1.6706, - "step": 21482 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019273388351484814, - "loss": 1.6916, - "step": 21483 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019273354528850708, - "loss": 1.595, - "step": 21484 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019273320706216602, - "loss": 1.4951, - "step": 21485 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019273286883582496, - "loss": 1.517, - "step": 21486 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019273253060948388, - "loss": 1.6359, - "step": 21487 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019273219238314282, - "loss": 1.3065, - "step": 21488 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019273185415680176, - "loss": 1.5096, - "step": 21489 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019273151593046068, - "loss": 1.5498, - "step": 21490 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019273117770411962, - "loss": 1.6013, - "step": 21491 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019273083947777856, - "loss": 1.4998, - "step": 21492 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019273050125143747, - "loss": 1.6284, - "step": 21493 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927301630250964, - "loss": 1.5412, - "step": 21494 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019272982479875535, - "loss": 1.384, - "step": 21495 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019272948657241427, - "loss": 1.5108, - "step": 21496 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927291483460732, - "loss": 1.4375, - "step": 21497 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019272881011973215, - "loss": 1.6789, - "step": 21498 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019272847189339106, - "loss": 1.5566, - "step": 21499 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019272813366705, - "loss": 1.5948, - "step": 21500 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019272779544070895, - "loss": 1.6505, - "step": 21501 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019272745721436786, - "loss": 1.672, - "step": 21502 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927271189880268, - "loss": 1.6472, - "step": 21503 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019272678076168574, - "loss": 1.4924, - "step": 21504 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019272644253534466, - "loss": 1.464, - "step": 21505 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927261043090036, - "loss": 1.5407, - "step": 21506 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019272576608266254, - "loss": 1.6026, - "step": 21507 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019272542785632145, - "loss": 1.4902, - "step": 21508 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927250896299804, - "loss": 1.4187, - "step": 21509 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019272475140363934, - "loss": 1.5834, - "step": 21510 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019272441317729825, - "loss": 1.61, - "step": 21511 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927240749509572, - "loss": 1.5867, - "step": 21512 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019272373672461613, - "loss": 1.5739, - "step": 21513 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019272339849827505, - "loss": 1.7557, - "step": 21514 - }, - { - "epoch": 0.07, - "learning_rate": 0.000192723060271934, - "loss": 1.7063, - "step": 21515 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019272272204559293, - "loss": 1.6164, - "step": 21516 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019272238381925184, - "loss": 1.6343, - "step": 21517 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019272204559291078, - "loss": 1.5901, - "step": 21518 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019272170736656972, - "loss": 1.5658, - "step": 21519 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019272136914022864, - "loss": 1.5743, - "step": 21520 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019272103091388758, - "loss": 1.4976, - "step": 21521 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019272069268754652, - "loss": 1.7125, - "step": 21522 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019272035446120543, - "loss": 1.4838, - "step": 21523 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019272001623486438, - "loss": 1.5295, - "step": 21524 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019271967800852332, - "loss": 1.6907, - "step": 21525 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019271933978218223, - "loss": 1.4231, - "step": 21526 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019271900155584117, - "loss": 1.6043, - "step": 21527 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927186633295001, - "loss": 1.5956, - "step": 21528 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019271832510315903, - "loss": 1.5247, - "step": 21529 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019271798687681797, - "loss": 1.5373, - "step": 21530 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927176486504769, - "loss": 1.4406, - "step": 21531 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019271731042413582, - "loss": 1.6452, - "step": 21532 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019271697219779476, - "loss": 1.5885, - "step": 21533 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927166339714537, - "loss": 1.6879, - "step": 21534 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019271629574511262, - "loss": 1.5619, - "step": 21535 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019271595751877156, - "loss": 1.5736, - "step": 21536 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927156192924305, - "loss": 1.6409, - "step": 21537 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019271528106608942, - "loss": 1.5455, - "step": 21538 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019271494283974836, - "loss": 1.6993, - "step": 21539 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019271460461340733, - "loss": 1.4881, - "step": 21540 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019271426638706624, - "loss": 1.7014, - "step": 21541 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019271392816072518, - "loss": 1.514, - "step": 21542 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019271358993438412, - "loss": 1.4522, - "step": 21543 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019271325170804304, - "loss": 1.5642, - "step": 21544 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019271291348170198, - "loss": 1.589, - "step": 21545 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019271257525536092, - "loss": 1.5824, - "step": 21546 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019271223702901983, - "loss": 1.5378, - "step": 21547 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019271189880267877, - "loss": 1.5965, - "step": 21548 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019271156057633771, - "loss": 1.5584, - "step": 21549 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019271122234999663, - "loss": 1.5393, - "step": 21550 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019271088412365557, - "loss": 1.3665, - "step": 21551 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927105458973145, - "loss": 1.5682, - "step": 21552 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019271020767097342, - "loss": 1.5131, - "step": 21553 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019270986944463237, - "loss": 1.5587, - "step": 21554 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927095312182913, - "loss": 1.6305, - "step": 21555 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019270919299195022, - "loss": 1.5055, - "step": 21556 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019270885476560916, - "loss": 1.6451, - "step": 21557 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927085165392681, - "loss": 1.5159, - "step": 21558 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019270817831292702, - "loss": 1.3233, - "step": 21559 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019270784008658596, - "loss": 1.6693, - "step": 21560 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927075018602449, - "loss": 1.4212, - "step": 21561 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019270716363390381, - "loss": 1.7599, - "step": 21562 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019270682540756276, - "loss": 1.557, - "step": 21563 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927064871812217, - "loss": 1.5267, - "step": 21564 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927061489548806, - "loss": 1.6089, - "step": 21565 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019270581072853955, - "loss": 1.5926, - "step": 21566 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927054725021985, - "loss": 1.5008, - "step": 21567 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927051342758574, - "loss": 1.6869, - "step": 21568 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019270479604951635, - "loss": 1.6395, - "step": 21569 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927044578231753, - "loss": 1.5967, - "step": 21570 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927041195968342, - "loss": 1.7019, - "step": 21571 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019270378137049314, - "loss": 1.5799, - "step": 21572 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019270344314415209, - "loss": 1.5296, - "step": 21573 - }, - { - "epoch": 0.07, - "learning_rate": 0.000192703104917811, - "loss": 1.5667, - "step": 21574 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019270276669146994, - "loss": 1.6286, - "step": 21575 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019270242846512888, - "loss": 1.4219, - "step": 21576 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927020902387878, - "loss": 1.5063, - "step": 21577 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019270175201244674, - "loss": 1.429, - "step": 21578 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019270141378610568, - "loss": 1.4552, - "step": 21579 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927010755597646, - "loss": 1.4816, - "step": 21580 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019270073733342353, - "loss": 1.6009, - "step": 21581 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019270039910708247, - "loss": 1.5754, - "step": 21582 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001927000608807414, - "loss": 1.6023, - "step": 21583 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019269972265440033, - "loss": 1.5201, - "step": 21584 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019269938442805927, - "loss": 1.4817, - "step": 21585 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019269904620171818, - "loss": 1.5211, - "step": 21586 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019269870797537713, - "loss": 1.5797, - "step": 21587 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019269836974903607, - "loss": 1.6393, - "step": 21588 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019269803152269498, - "loss": 1.5069, - "step": 21589 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019269769329635392, - "loss": 1.5585, - "step": 21590 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019269735507001286, - "loss": 1.4653, - "step": 21591 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019269701684367178, - "loss": 1.6082, - "step": 21592 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019269667861733072, - "loss": 1.6531, - "step": 21593 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019269634039098966, - "loss": 1.5863, - "step": 21594 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019269600216464857, - "loss": 1.7081, - "step": 21595 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019269566393830754, - "loss": 1.6439, - "step": 21596 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019269532571196646, - "loss": 1.4327, - "step": 21597 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926949874856254, - "loss": 1.5251, - "step": 21598 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019269464925928434, - "loss": 1.5397, - "step": 21599 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019269431103294325, - "loss": 1.5664, - "step": 21600 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926939728066022, - "loss": 1.6967, - "step": 21601 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019269363458026113, - "loss": 1.5551, - "step": 21602 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019269329635392005, - "loss": 1.676, - "step": 21603 - }, - { - "epoch": 0.07, - "learning_rate": 0.000192692958127579, - "loss": 1.5573, - "step": 21604 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019269261990123793, - "loss": 1.4631, - "step": 21605 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019269228167489684, - "loss": 1.464, - "step": 21606 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019269194344855579, - "loss": 1.6186, - "step": 21607 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019269160522221473, - "loss": 1.4273, - "step": 21608 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019269126699587364, - "loss": 1.4706, - "step": 21609 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019269092876953258, - "loss": 1.5046, - "step": 21610 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019269059054319152, - "loss": 1.5329, - "step": 21611 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019269025231685044, - "loss": 1.64, - "step": 21612 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019268991409050938, - "loss": 1.502, - "step": 21613 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019268957586416832, - "loss": 1.6919, - "step": 21614 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019268923763782723, - "loss": 1.6006, - "step": 21615 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019268889941148617, - "loss": 1.6672, - "step": 21616 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019268856118514512, - "loss": 1.6279, - "step": 21617 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019268822295880403, - "loss": 1.5049, - "step": 21618 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019268788473246297, - "loss": 1.6306, - "step": 21619 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926875465061219, - "loss": 1.3612, - "step": 21620 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019268720827978085, - "loss": 1.6172, - "step": 21621 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019268687005343977, - "loss": 1.3977, - "step": 21622 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926865318270987, - "loss": 1.621, - "step": 21623 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019268619360075765, - "loss": 1.5082, - "step": 21624 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019268585537441656, - "loss": 1.5446, - "step": 21625 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926855171480755, - "loss": 1.5931, - "step": 21626 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019268517892173445, - "loss": 1.5009, - "step": 21627 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019268484069539336, - "loss": 1.447, - "step": 21628 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926845024690523, - "loss": 1.7604, - "step": 21629 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019268416424271124, - "loss": 1.5713, - "step": 21630 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019268382601637016, - "loss": 1.5824, - "step": 21631 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926834877900291, - "loss": 1.5853, - "step": 21632 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019268314956368804, - "loss": 1.576, - "step": 21633 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019268281133734695, - "loss": 1.5362, - "step": 21634 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926824731110059, - "loss": 1.6681, - "step": 21635 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019268213488466483, - "loss": 1.8454, - "step": 21636 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019268179665832375, - "loss": 1.6542, - "step": 21637 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926814584319827, - "loss": 1.4932, - "step": 21638 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019268112020564163, - "loss": 1.529, - "step": 21639 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019268078197930055, - "loss": 1.6925, - "step": 21640 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019268044375295949, - "loss": 1.5532, - "step": 21641 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019268010552661843, - "loss": 1.6817, - "step": 21642 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019267976730027734, - "loss": 1.5613, - "step": 21643 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019267942907393628, - "loss": 1.7386, - "step": 21644 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019267909084759522, - "loss": 1.6366, - "step": 21645 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019267875262125414, - "loss": 1.4505, - "step": 21646 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019267841439491308, - "loss": 1.5334, - "step": 21647 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019267807616857202, - "loss": 1.4849, - "step": 21648 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019267773794223093, - "loss": 1.4098, - "step": 21649 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019267739971588988, - "loss": 1.5491, - "step": 21650 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019267706148954882, - "loss": 1.5915, - "step": 21651 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019267672326320773, - "loss": 1.6091, - "step": 21652 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926763850368667, - "loss": 1.636, - "step": 21653 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926760468105256, - "loss": 1.5566, - "step": 21654 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019267570858418455, - "loss": 1.5364, - "step": 21655 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926753703578435, - "loss": 1.7235, - "step": 21656 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926750321315024, - "loss": 1.5624, - "step": 21657 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019267469390516135, - "loss": 1.5257, - "step": 21658 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926743556788203, - "loss": 1.4871, - "step": 21659 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926740174524792, - "loss": 1.5428, - "step": 21660 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019267367922613815, - "loss": 1.5427, - "step": 21661 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926733409997971, - "loss": 1.7468, - "step": 21662 - }, - { - "epoch": 0.07, - "learning_rate": 0.000192673002773456, - "loss": 1.6285, - "step": 21663 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019267266454711494, - "loss": 1.5817, - "step": 21664 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019267232632077388, - "loss": 1.6671, - "step": 21665 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926719880944328, - "loss": 1.6089, - "step": 21666 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019267164986809174, - "loss": 1.6357, - "step": 21667 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019267131164175068, - "loss": 1.7452, - "step": 21668 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926709734154096, - "loss": 1.4121, - "step": 21669 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019267063518906854, - "loss": 1.528, - "step": 21670 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019267029696272748, - "loss": 1.4982, - "step": 21671 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926699587363864, - "loss": 1.7427, - "step": 21672 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019266962051004533, - "loss": 1.7232, - "step": 21673 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019266928228370427, - "loss": 1.5094, - "step": 21674 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926689440573632, - "loss": 1.7271, - "step": 21675 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019266860583102213, - "loss": 1.6582, - "step": 21676 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019266826760468107, - "loss": 1.4186, - "step": 21677 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019266792937833998, - "loss": 1.5695, - "step": 21678 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019266759115199892, - "loss": 1.4548, - "step": 21679 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019266725292565787, - "loss": 1.5303, - "step": 21680 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019266691469931678, - "loss": 1.5982, - "step": 21681 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019266657647297572, - "loss": 1.4743, - "step": 21682 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019266623824663466, - "loss": 1.6593, - "step": 21683 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019266590002029358, - "loss": 1.6202, - "step": 21684 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019266556179395252, - "loss": 1.4232, - "step": 21685 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019266522356761146, - "loss": 1.5738, - "step": 21686 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019266488534127037, - "loss": 1.4717, - "step": 21687 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926645471149293, - "loss": 1.5973, - "step": 21688 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019266420888858825, - "loss": 1.5473, - "step": 21689 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019266387066224717, - "loss": 1.8069, - "step": 21690 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926635324359061, - "loss": 1.5159, - "step": 21691 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019266319420956505, - "loss": 1.5159, - "step": 21692 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019266285598322396, - "loss": 1.7389, - "step": 21693 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926625177568829, - "loss": 1.5386, - "step": 21694 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019266217953054185, - "loss": 1.5147, - "step": 21695 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019266184130420076, - "loss": 1.5889, - "step": 21696 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926615030778597, - "loss": 1.6759, - "step": 21697 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019266116485151864, - "loss": 1.5082, - "step": 21698 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019266082662517756, - "loss": 1.5487, - "step": 21699 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926604883988365, - "loss": 1.5174, - "step": 21700 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019266015017249544, - "loss": 1.6849, - "step": 21701 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019265981194615435, - "loss": 1.7425, - "step": 21702 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926594737198133, - "loss": 1.5839, - "step": 21703 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019265913549347224, - "loss": 1.4688, - "step": 21704 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019265879726713118, - "loss": 1.5082, - "step": 21705 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926584590407901, - "loss": 1.545, - "step": 21706 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019265812081444903, - "loss": 1.6006, - "step": 21707 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019265778258810797, - "loss": 1.5361, - "step": 21708 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019265744436176691, - "loss": 1.5393, - "step": 21709 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019265710613542586, - "loss": 1.5527, - "step": 21710 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019265676790908477, - "loss": 1.4736, - "step": 21711 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926564296827437, - "loss": 1.5872, - "step": 21712 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019265609145640265, - "loss": 1.3924, - "step": 21713 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019265575323006157, - "loss": 1.6262, - "step": 21714 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926554150037205, - "loss": 1.5547, - "step": 21715 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019265507677737945, - "loss": 1.6184, - "step": 21716 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019265473855103836, - "loss": 1.5748, - "step": 21717 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926544003246973, - "loss": 1.5434, - "step": 21718 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019265406209835624, - "loss": 1.5674, - "step": 21719 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019265372387201516, - "loss": 1.6121, - "step": 21720 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926533856456741, - "loss": 1.4758, - "step": 21721 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019265304741933304, - "loss": 1.6085, - "step": 21722 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019265270919299195, - "loss": 1.4398, - "step": 21723 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926523709666509, - "loss": 1.3979, - "step": 21724 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019265203274030984, - "loss": 1.619, - "step": 21725 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019265169451396875, - "loss": 1.596, - "step": 21726 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926513562876277, - "loss": 1.598, - "step": 21727 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019265101806128663, - "loss": 1.4582, - "step": 21728 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019265067983494555, - "loss": 1.4841, - "step": 21729 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926503416086045, - "loss": 1.5706, - "step": 21730 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019265000338226343, - "loss": 1.6545, - "step": 21731 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019264966515592234, - "loss": 1.6636, - "step": 21732 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019264932692958129, - "loss": 1.5292, - "step": 21733 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019264898870324023, - "loss": 1.4414, - "step": 21734 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019264865047689914, - "loss": 1.442, - "step": 21735 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019264831225055808, - "loss": 1.5417, - "step": 21736 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019264797402421702, - "loss": 1.6021, - "step": 21737 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019264763579787594, - "loss": 1.5835, - "step": 21738 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019264729757153488, - "loss": 1.612, - "step": 21739 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019264695934519382, - "loss": 1.7123, - "step": 21740 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019264662111885273, - "loss": 1.4975, - "step": 21741 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019264628289251167, - "loss": 1.5873, - "step": 21742 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019264594466617062, - "loss": 1.609, - "step": 21743 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019264560643982953, - "loss": 1.5729, - "step": 21744 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019264526821348847, - "loss": 1.5174, - "step": 21745 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926449299871474, - "loss": 1.6327, - "step": 21746 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019264459176080633, - "loss": 1.6889, - "step": 21747 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019264425353446527, - "loss": 1.6, - "step": 21748 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926439153081242, - "loss": 1.6698, - "step": 21749 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019264357708178312, - "loss": 1.501, - "step": 21750 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019264323885544206, - "loss": 1.5095, - "step": 21751 - }, - { - "epoch": 0.07, - "learning_rate": 0.000192642900629101, - "loss": 1.6299, - "step": 21752 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019264256240275992, - "loss": 1.8269, - "step": 21753 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019264222417641886, - "loss": 1.5531, - "step": 21754 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926418859500778, - "loss": 1.6457, - "step": 21755 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019264154772373671, - "loss": 1.5868, - "step": 21756 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019264120949739566, - "loss": 1.3283, - "step": 21757 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926408712710546, - "loss": 1.4971, - "step": 21758 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926405330447135, - "loss": 1.5711, - "step": 21759 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019264019481837245, - "loss": 1.6171, - "step": 21760 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926398565920314, - "loss": 1.5725, - "step": 21761 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926395183656903, - "loss": 1.5975, - "step": 21762 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019263918013934925, - "loss": 1.6233, - "step": 21763 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926388419130082, - "loss": 1.6711, - "step": 21764 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019263850368666713, - "loss": 1.659, - "step": 21765 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019263816546032607, - "loss": 1.6122, - "step": 21766 - }, - { - "epoch": 0.07, - "learning_rate": 0.000192637827233985, - "loss": 1.5152, - "step": 21767 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019263748900764393, - "loss": 1.8156, - "step": 21768 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019263715078130287, - "loss": 1.7792, - "step": 21769 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926368125549618, - "loss": 1.574, - "step": 21770 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019263647432862072, - "loss": 1.6538, - "step": 21771 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019263613610227966, - "loss": 1.5908, - "step": 21772 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926357978759386, - "loss": 1.508, - "step": 21773 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019263545964959752, - "loss": 1.4564, - "step": 21774 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019263512142325646, - "loss": 1.7223, - "step": 21775 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926347831969154, - "loss": 1.5357, - "step": 21776 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019263444497057432, - "loss": 1.6015, - "step": 21777 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019263410674423326, - "loss": 1.6115, - "step": 21778 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926337685178922, - "loss": 1.5755, - "step": 21779 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926334302915511, - "loss": 1.4845, - "step": 21780 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019263309206521005, - "loss": 1.5919, - "step": 21781 - }, - { - "epoch": 0.07, - "learning_rate": 0.000192632753838869, - "loss": 1.556, - "step": 21782 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926324156125279, - "loss": 1.6698, - "step": 21783 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019263207738618685, - "loss": 1.6428, - "step": 21784 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926317391598458, - "loss": 1.3427, - "step": 21785 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926314009335047, - "loss": 1.4967, - "step": 21786 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019263106270716365, - "loss": 1.6293, - "step": 21787 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926307244808226, - "loss": 1.6268, - "step": 21788 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926303862544815, - "loss": 1.6294, - "step": 21789 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019263004802814044, - "loss": 1.532, - "step": 21790 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019262970980179938, - "loss": 1.594, - "step": 21791 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926293715754583, - "loss": 1.4915, - "step": 21792 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019262903334911724, - "loss": 1.7014, - "step": 21793 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019262869512277618, - "loss": 1.519, - "step": 21794 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926283568964351, - "loss": 1.4958, - "step": 21795 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019262801867009403, - "loss": 1.6984, - "step": 21796 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019262768044375298, - "loss": 1.6985, - "step": 21797 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926273422174119, - "loss": 1.7999, - "step": 21798 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019262700399107083, - "loss": 1.6683, - "step": 21799 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019262666576472977, - "loss": 1.4355, - "step": 21800 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019262632753838869, - "loss": 1.5099, - "step": 21801 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019262598931204763, - "loss": 1.3883, - "step": 21802 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019262565108570657, - "loss": 1.6718, - "step": 21803 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019262531285936548, - "loss": 1.7122, - "step": 21804 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019262497463302442, - "loss": 1.5329, - "step": 21805 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019262463640668336, - "loss": 1.5792, - "step": 21806 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019262429818034228, - "loss": 1.4033, - "step": 21807 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019262395995400122, - "loss": 1.8306, - "step": 21808 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019262362172766016, - "loss": 1.4909, - "step": 21809 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019262328350131908, - "loss": 1.5423, - "step": 21810 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019262294527497802, - "loss": 1.4595, - "step": 21811 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019262260704863696, - "loss": 1.4233, - "step": 21812 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019262226882229587, - "loss": 1.5519, - "step": 21813 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926219305959548, - "loss": 1.6969, - "step": 21814 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019262159236961375, - "loss": 1.5862, - "step": 21815 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019262125414327267, - "loss": 1.3817, - "step": 21816 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926209159169316, - "loss": 1.5305, - "step": 21817 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019262057769059055, - "loss": 1.5877, - "step": 21818 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019262023946424946, - "loss": 1.3057, - "step": 21819 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926199012379084, - "loss": 1.6314, - "step": 21820 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019261956301156737, - "loss": 1.7028, - "step": 21821 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926192247852263, - "loss": 1.4374, - "step": 21822 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019261888655888523, - "loss": 1.4741, - "step": 21823 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019261854833254417, - "loss": 1.493, - "step": 21824 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019261821010620308, - "loss": 1.4875, - "step": 21825 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019261787187986203, - "loss": 1.4524, - "step": 21826 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019261753365352097, - "loss": 1.6475, - "step": 21827 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019261719542717988, - "loss": 1.4384, - "step": 21828 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019261685720083882, - "loss": 1.6007, - "step": 21829 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019261651897449776, - "loss": 1.5678, - "step": 21830 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019261618074815668, - "loss": 1.5926, - "step": 21831 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019261584252181562, - "loss": 1.5018, - "step": 21832 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019261550429547456, - "loss": 1.6067, - "step": 21833 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019261516606913347, - "loss": 1.6474, - "step": 21834 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019261482784279241, - "loss": 1.4901, - "step": 21835 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019261448961645136, - "loss": 1.6449, - "step": 21836 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019261415139011027, - "loss": 1.5272, - "step": 21837 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926138131637692, - "loss": 1.6536, - "step": 21838 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019261347493742815, - "loss": 1.5479, - "step": 21839 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019261313671108707, - "loss": 1.5944, - "step": 21840 - }, - { - "epoch": 0.07, - "learning_rate": 0.000192612798484746, - "loss": 1.5415, - "step": 21841 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019261246025840495, - "loss": 1.6448, - "step": 21842 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019261212203206386, - "loss": 1.7101, - "step": 21843 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926117838057228, - "loss": 1.4308, - "step": 21844 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019261144557938174, - "loss": 1.8243, - "step": 21845 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019261110735304066, - "loss": 1.4075, - "step": 21846 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926107691266996, - "loss": 1.5029, - "step": 21847 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019261043090035854, - "loss": 1.6607, - "step": 21848 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019261009267401745, - "loss": 1.7158, - "step": 21849 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926097544476764, - "loss": 1.4292, - "step": 21850 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019260941622133534, - "loss": 1.8344, - "step": 21851 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019260907799499425, - "loss": 1.5772, - "step": 21852 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926087397686532, - "loss": 1.5662, - "step": 21853 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019260840154231213, - "loss": 1.4724, - "step": 21854 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019260806331597105, - "loss": 1.6289, - "step": 21855 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019260772508963, - "loss": 1.5036, - "step": 21856 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019260738686328893, - "loss": 1.6176, - "step": 21857 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019260704863694784, - "loss": 1.5191, - "step": 21858 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019260671041060678, - "loss": 1.529, - "step": 21859 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019260637218426573, - "loss": 1.5509, - "step": 21860 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019260603395792464, - "loss": 1.6195, - "step": 21861 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019260569573158358, - "loss": 1.6239, - "step": 21862 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019260535750524252, - "loss": 1.611, - "step": 21863 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019260501927890144, - "loss": 1.6157, - "step": 21864 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019260468105256038, - "loss": 1.5705, - "step": 21865 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019260434282621932, - "loss": 1.4894, - "step": 21866 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019260400459987823, - "loss": 1.5904, - "step": 21867 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019260366637353717, - "loss": 1.503, - "step": 21868 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019260332814719611, - "loss": 1.4955, - "step": 21869 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019260298992085503, - "loss": 1.5321, - "step": 21870 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019260265169451397, - "loss": 1.6087, - "step": 21871 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926023134681729, - "loss": 1.516, - "step": 21872 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019260197524183182, - "loss": 1.5386, - "step": 21873 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019260163701549077, - "loss": 1.6234, - "step": 21874 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926012987891497, - "loss": 1.6919, - "step": 21875 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019260096056280862, - "loss": 1.471, - "step": 21876 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926006223364676, - "loss": 1.6053, - "step": 21877 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001926002841101265, - "loss": 1.5965, - "step": 21878 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019259994588378544, - "loss": 1.5639, - "step": 21879 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019259960765744439, - "loss": 1.8342, - "step": 21880 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925992694311033, - "loss": 1.5421, - "step": 21881 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019259893120476224, - "loss": 1.7001, - "step": 21882 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019259859297842118, - "loss": 1.6229, - "step": 21883 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925982547520801, - "loss": 1.7161, - "step": 21884 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019259791652573904, - "loss": 1.6094, - "step": 21885 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019259757829939798, - "loss": 1.6305, - "step": 21886 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019259724007305692, - "loss": 1.5631, - "step": 21887 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019259690184671583, - "loss": 1.4967, - "step": 21888 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019259656362037477, - "loss": 1.5081, - "step": 21889 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019259622539403372, - "loss": 1.5457, - "step": 21890 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019259588716769263, - "loss": 1.575, - "step": 21891 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019259554894135157, - "loss": 1.6309, - "step": 21892 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925952107150105, - "loss": 1.6421, - "step": 21893 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019259487248866943, - "loss": 1.6022, - "step": 21894 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019259453426232837, - "loss": 1.6551, - "step": 21895 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925941960359873, - "loss": 1.5774, - "step": 21896 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019259385780964622, - "loss": 1.5958, - "step": 21897 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019259351958330516, - "loss": 1.6434, - "step": 21898 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925931813569641, - "loss": 1.5348, - "step": 21899 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019259284313062302, - "loss": 1.4353, - "step": 21900 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019259250490428196, - "loss": 1.5972, - "step": 21901 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925921666779409, - "loss": 1.5401, - "step": 21902 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019259182845159982, - "loss": 1.4897, - "step": 21903 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019259149022525876, - "loss": 1.6639, - "step": 21904 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925911519989177, - "loss": 1.6717, - "step": 21905 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925908137725766, - "loss": 1.4512, - "step": 21906 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019259047554623555, - "loss": 1.3737, - "step": 21907 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925901373198945, - "loss": 1.2827, - "step": 21908 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925897990935534, - "loss": 1.6161, - "step": 21909 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019258946086721235, - "loss": 1.5416, - "step": 21910 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925891226408713, - "loss": 1.6691, - "step": 21911 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925887844145302, - "loss": 1.719, - "step": 21912 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019258844618818915, - "loss": 1.642, - "step": 21913 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019258810796184809, - "loss": 1.7052, - "step": 21914 - }, - { - "epoch": 0.07, - "learning_rate": 0.000192587769735507, - "loss": 1.5914, - "step": 21915 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019258743150916594, - "loss": 1.5171, - "step": 21916 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019258709328282488, - "loss": 1.7599, - "step": 21917 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925867550564838, - "loss": 1.5139, - "step": 21918 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019258641683014274, - "loss": 1.5259, - "step": 21919 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019258607860380168, - "loss": 1.6167, - "step": 21920 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925857403774606, - "loss": 1.4826, - "step": 21921 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019258540215111953, - "loss": 1.664, - "step": 21922 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019258506392477848, - "loss": 1.67, - "step": 21923 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925847256984374, - "loss": 1.5133, - "step": 21924 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019258438747209633, - "loss": 1.7419, - "step": 21925 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019258404924575527, - "loss": 1.5462, - "step": 21926 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019258371101941419, - "loss": 1.5024, - "step": 21927 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019258337279307313, - "loss": 1.6134, - "step": 21928 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019258303456673207, - "loss": 1.4854, - "step": 21929 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019258269634039098, - "loss": 1.5038, - "step": 21930 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019258235811404992, - "loss": 1.5923, - "step": 21931 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019258201988770886, - "loss": 1.4861, - "step": 21932 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925816816613678, - "loss": 1.5581, - "step": 21933 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019258134343502675, - "loss": 1.5899, - "step": 21934 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019258100520868566, - "loss": 1.4868, - "step": 21935 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925806669823446, - "loss": 1.5596, - "step": 21936 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019258032875600354, - "loss": 1.5267, - "step": 21937 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019257999052966246, - "loss": 1.5838, - "step": 21938 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925796523033214, - "loss": 1.8117, - "step": 21939 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019257931407698034, - "loss": 1.5168, - "step": 21940 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019257897585063925, - "loss": 1.6032, - "step": 21941 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925786376242982, - "loss": 1.5165, - "step": 21942 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019257829939795714, - "loss": 1.624, - "step": 21943 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019257796117161605, - "loss": 1.6391, - "step": 21944 - }, - { - "epoch": 0.07, - "learning_rate": 0.000192577622945275, - "loss": 1.7189, - "step": 21945 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019257728471893393, - "loss": 1.4678, - "step": 21946 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019257694649259285, - "loss": 1.7934, - "step": 21947 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925766082662518, - "loss": 1.5365, - "step": 21948 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019257627003991073, - "loss": 1.4343, - "step": 21949 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019257593181356964, - "loss": 1.5517, - "step": 21950 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019257559358722858, - "loss": 1.6304, - "step": 21951 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019257525536088752, - "loss": 1.4296, - "step": 21952 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019257491713454644, - "loss": 1.7143, - "step": 21953 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019257457890820538, - "loss": 1.469, - "step": 21954 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019257424068186432, - "loss": 1.6355, - "step": 21955 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019257390245552323, - "loss": 1.5784, - "step": 21956 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019257356422918218, - "loss": 1.5613, - "step": 21957 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019257322600284112, - "loss": 1.5085, - "step": 21958 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019257288777650003, - "loss": 1.6024, - "step": 21959 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019257254955015897, - "loss": 1.7176, - "step": 21960 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925722113238179, - "loss": 1.5449, - "step": 21961 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019257187309747683, - "loss": 1.7137, - "step": 21962 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019257153487113577, - "loss": 1.5163, - "step": 21963 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925711966447947, - "loss": 1.5301, - "step": 21964 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019257085841845362, - "loss": 1.5344, - "step": 21965 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019257052019211256, - "loss": 1.5972, - "step": 21966 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925701819657715, - "loss": 1.6089, - "step": 21967 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019256984373943042, - "loss": 1.6364, - "step": 21968 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019256950551308936, - "loss": 1.5653, - "step": 21969 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925691672867483, - "loss": 1.5934, - "step": 21970 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019256882906040722, - "loss": 1.7052, - "step": 21971 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019256849083406616, - "loss": 1.6291, - "step": 21972 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925681526077251, - "loss": 1.5118, - "step": 21973 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019256781438138404, - "loss": 1.5851, - "step": 21974 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019256747615504295, - "loss": 1.6818, - "step": 21975 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925671379287019, - "loss": 1.6773, - "step": 21976 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019256679970236084, - "loss": 1.663, - "step": 21977 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019256646147601975, - "loss": 1.4813, - "step": 21978 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925661232496787, - "loss": 1.5154, - "step": 21979 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019256578502333763, - "loss": 1.5282, - "step": 21980 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019256544679699655, - "loss": 1.4912, - "step": 21981 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925651085706555, - "loss": 1.6343, - "step": 21982 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019256477034431443, - "loss": 1.405, - "step": 21983 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019256443211797334, - "loss": 1.4822, - "step": 21984 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019256409389163228, - "loss": 1.529, - "step": 21985 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019256375566529122, - "loss": 1.5359, - "step": 21986 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019256341743895014, - "loss": 1.6581, - "step": 21987 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019256307921260908, - "loss": 1.657, - "step": 21988 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019256274098626802, - "loss": 1.5332, - "step": 21989 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019256240275992696, - "loss": 1.4648, - "step": 21990 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925620645335859, - "loss": 1.8973, - "step": 21991 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019256172630724482, - "loss": 1.4335, - "step": 21992 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019256138808090376, - "loss": 1.4947, - "step": 21993 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925610498545627, - "loss": 1.613, - "step": 21994 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019256071162822161, - "loss": 1.6023, - "step": 21995 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019256037340188056, - "loss": 1.5579, - "step": 21996 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925600351755395, - "loss": 1.7155, - "step": 21997 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925596969491984, - "loss": 1.5785, - "step": 21998 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019255935872285735, - "loss": 1.5348, - "step": 21999 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925590204965163, - "loss": 1.3942, - "step": 22000 - }, - { - "epoch": 0.07, - "eval_loss": 1.4208420973557692, - "mrc_em": 24.0, - "mrc_f1": 32.82380952380953, - "nli_acc": 0.07, - "step": 22000, - "sts_f1": 0.627906976744186, - "ynat_f1": 0.684992784992785 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925586822701752, - "loss": 1.5129, - "step": 22001 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019255834404383415, - "loss": 1.6431, - "step": 22002 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925580058174931, - "loss": 1.6345, - "step": 22003 - }, - { - "epoch": 0.07, - "learning_rate": 0.000192557667591152, - "loss": 1.6593, - "step": 22004 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019255732936481094, - "loss": 1.5937, - "step": 22005 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019255699113846989, - "loss": 1.7461, - "step": 22006 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925566529121288, - "loss": 1.4316, - "step": 22007 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019255631468578774, - "loss": 1.5242, - "step": 22008 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019255597645944668, - "loss": 1.6016, - "step": 22009 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925556382331056, - "loss": 1.4307, - "step": 22010 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019255530000676454, - "loss": 1.4666, - "step": 22011 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019255496178042348, - "loss": 1.626, - "step": 22012 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925546235540824, - "loss": 1.7275, - "step": 22013 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019255428532774133, - "loss": 1.7634, - "step": 22014 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019255394710140027, - "loss": 1.6618, - "step": 22015 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925536088750592, - "loss": 1.4867, - "step": 22016 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019255327064871813, - "loss": 1.4663, - "step": 22017 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019255293242237707, - "loss": 1.4102, - "step": 22018 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019255259419603598, - "loss": 1.3414, - "step": 22019 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019255225596969493, - "loss": 1.5175, - "step": 22020 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019255191774335387, - "loss": 1.6389, - "step": 22021 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019255157951701278, - "loss": 1.6917, - "step": 22022 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019255124129067172, - "loss": 1.6451, - "step": 22023 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019255090306433066, - "loss": 1.5582, - "step": 22024 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019255056483798958, - "loss": 1.7501, - "step": 22025 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019255022661164852, - "loss": 1.5457, - "step": 22026 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019254988838530746, - "loss": 1.6208, - "step": 22027 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019254955015896637, - "loss": 1.6633, - "step": 22028 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019254921193262531, - "loss": 1.6358, - "step": 22029 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019254887370628426, - "loss": 1.6462, - "step": 22030 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019254853547994317, - "loss": 1.6652, - "step": 22031 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925481972536021, - "loss": 1.6302, - "step": 22032 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019254785902726105, - "loss": 1.5744, - "step": 22033 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019254752080091997, - "loss": 1.6119, - "step": 22034 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925471825745789, - "loss": 1.4866, - "step": 22035 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019254684434823785, - "loss": 1.6496, - "step": 22036 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019254650612189676, - "loss": 1.4942, - "step": 22037 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925461678955557, - "loss": 1.5198, - "step": 22038 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019254582966921464, - "loss": 1.4215, - "step": 22039 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019254549144287356, - "loss": 1.7347, - "step": 22040 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925451532165325, - "loss": 1.625, - "step": 22041 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019254481499019144, - "loss": 1.6428, - "step": 22042 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019254447676385035, - "loss": 1.4454, - "step": 22043 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925441385375093, - "loss": 1.6679, - "step": 22044 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019254380031116824, - "loss": 1.5371, - "step": 22045 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019254346208482718, - "loss": 1.5125, - "step": 22046 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019254312385848612, - "loss": 1.5943, - "step": 22047 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019254278563214506, - "loss": 1.5745, - "step": 22048 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019254244740580397, - "loss": 1.4939, - "step": 22049 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019254210917946292, - "loss": 1.5677, - "step": 22050 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019254177095312186, - "loss": 1.5647, - "step": 22051 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019254143272678077, - "loss": 1.6189, - "step": 22052 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925410945004397, - "loss": 1.5891, - "step": 22053 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019254075627409865, - "loss": 1.467, - "step": 22054 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019254041804775757, - "loss": 1.5999, - "step": 22055 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925400798214165, - "loss": 1.5021, - "step": 22056 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019253974159507545, - "loss": 1.5078, - "step": 22057 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019253940336873436, - "loss": 1.5603, - "step": 22058 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925390651423933, - "loss": 1.4888, - "step": 22059 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019253872691605225, - "loss": 1.4942, - "step": 22060 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019253838868971116, - "loss": 1.5207, - "step": 22061 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925380504633701, - "loss": 1.5847, - "step": 22062 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019253771223702904, - "loss": 1.514, - "step": 22063 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019253737401068796, - "loss": 1.434, - "step": 22064 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925370357843469, - "loss": 1.69, - "step": 22065 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019253669755800584, - "loss": 1.5723, - "step": 22066 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019253635933166475, - "loss": 1.6815, - "step": 22067 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925360211053237, - "loss": 1.6958, - "step": 22068 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019253568287898263, - "loss": 1.4684, - "step": 22069 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019253534465264155, - "loss": 1.7691, - "step": 22070 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925350064263005, - "loss": 1.6185, - "step": 22071 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019253466819995943, - "loss": 1.6431, - "step": 22072 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019253432997361835, - "loss": 1.5309, - "step": 22073 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019253399174727729, - "loss": 1.4137, - "step": 22074 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019253365352093623, - "loss": 1.4084, - "step": 22075 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019253331529459514, - "loss": 1.5173, - "step": 22076 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019253297706825408, - "loss": 1.7109, - "step": 22077 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019253263884191302, - "loss": 1.54, - "step": 22078 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019253230061557194, - "loss": 1.5728, - "step": 22079 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019253196238923088, - "loss": 1.7494, - "step": 22080 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019253162416288982, - "loss": 1.5565, - "step": 22081 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019253128593654873, - "loss": 1.5129, - "step": 22082 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019253094771020768, - "loss": 1.7896, - "step": 22083 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019253060948386662, - "loss": 1.689, - "step": 22084 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019253027125752553, - "loss": 1.4547, - "step": 22085 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019252993303118447, - "loss": 1.5806, - "step": 22086 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925295948048434, - "loss": 1.5217, - "step": 22087 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019252925657850233, - "loss": 1.5609, - "step": 22088 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019252891835216127, - "loss": 1.5882, - "step": 22089 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925285801258202, - "loss": 1.481, - "step": 22090 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019252824189947912, - "loss": 1.6667, - "step": 22091 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019252790367313806, - "loss": 1.7441, - "step": 22092 - }, - { - "epoch": 0.07, - "learning_rate": 0.000192527565446797, - "loss": 1.5971, - "step": 22093 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019252722722045592, - "loss": 1.514, - "step": 22094 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019252688899411486, - "loss": 1.4038, - "step": 22095 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925265507677738, - "loss": 1.691, - "step": 22096 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019252621254143272, - "loss": 1.5542, - "step": 22097 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019252587431509166, - "loss": 1.5541, - "step": 22098 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925255360887506, - "loss": 1.5351, - "step": 22099 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925251978624095, - "loss": 1.4931, - "step": 22100 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019252485963606845, - "loss": 1.4919, - "step": 22101 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019252452140972742, - "loss": 1.7217, - "step": 22102 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019252418318338634, - "loss": 1.758, - "step": 22103 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019252384495704528, - "loss": 1.5765, - "step": 22104 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019252350673070422, - "loss": 1.5452, - "step": 22105 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019252316850436313, - "loss": 1.6813, - "step": 22106 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019252283027802207, - "loss": 1.4524, - "step": 22107 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019252249205168101, - "loss": 1.7227, - "step": 22108 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019252215382533993, - "loss": 1.5227, - "step": 22109 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019252181559899887, - "loss": 1.5115, - "step": 22110 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925214773726578, - "loss": 1.3807, - "step": 22111 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019252113914631672, - "loss": 1.5453, - "step": 22112 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019252080091997567, - "loss": 1.551, - "step": 22113 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925204626936346, - "loss": 1.5626, - "step": 22114 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019252012446729352, - "loss": 1.6028, - "step": 22115 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019251978624095246, - "loss": 1.3953, - "step": 22116 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925194480146114, - "loss": 1.5747, - "step": 22117 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019251910978827032, - "loss": 1.5671, - "step": 22118 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019251877156192926, - "loss": 1.5233, - "step": 22119 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925184333355882, - "loss": 1.5365, - "step": 22120 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925180951092471, - "loss": 1.5504, - "step": 22121 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019251775688290605, - "loss": 1.4747, - "step": 22122 - }, - { - "epoch": 0.07, - "learning_rate": 0.000192517418656565, - "loss": 1.7565, - "step": 22123 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925170804302239, - "loss": 1.7717, - "step": 22124 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019251674220388285, - "loss": 1.3567, - "step": 22125 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925164039775418, - "loss": 1.482, - "step": 22126 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925160657512007, - "loss": 1.64, - "step": 22127 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019251572752485965, - "loss": 1.6711, - "step": 22128 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925153892985186, - "loss": 1.5421, - "step": 22129 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925150510721775, - "loss": 1.4507, - "step": 22130 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019251471284583644, - "loss": 1.5118, - "step": 22131 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019251437461949538, - "loss": 1.4776, - "step": 22132 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925140363931543, - "loss": 1.6787, - "step": 22133 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019251369816681324, - "loss": 1.4825, - "step": 22134 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019251335994047218, - "loss": 1.5498, - "step": 22135 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925130217141311, - "loss": 1.6467, - "step": 22136 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019251268348779004, - "loss": 1.517, - "step": 22137 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019251234526144898, - "loss": 1.6112, - "step": 22138 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925120070351079, - "loss": 1.6919, - "step": 22139 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019251166880876683, - "loss": 1.6407, - "step": 22140 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019251133058242577, - "loss": 1.7044, - "step": 22141 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925109923560847, - "loss": 1.5421, - "step": 22142 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019251065412974363, - "loss": 1.5866, - "step": 22143 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019251031590340257, - "loss": 1.7111, - "step": 22144 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019250997767706148, - "loss": 1.5202, - "step": 22145 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019250963945072042, - "loss": 1.4941, - "step": 22146 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019250930122437937, - "loss": 1.547, - "step": 22147 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019250896299803828, - "loss": 1.7403, - "step": 22148 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019250862477169722, - "loss": 1.4405, - "step": 22149 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019250828654535616, - "loss": 1.4926, - "step": 22150 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019250794831901508, - "loss": 1.507, - "step": 22151 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019250761009267402, - "loss": 1.5581, - "step": 22152 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019250727186633296, - "loss": 1.4096, - "step": 22153 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019250693363999187, - "loss": 1.4603, - "step": 22154 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019250659541365081, - "loss": 1.6383, - "step": 22155 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019250625718730976, - "loss": 1.8122, - "step": 22156 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019250591896096867, - "loss": 1.4986, - "step": 22157 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019250558073462764, - "loss": 1.6376, - "step": 22158 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019250524250828658, - "loss": 1.543, - "step": 22159 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925049042819455, - "loss": 1.6233, - "step": 22160 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019250456605560443, - "loss": 1.6753, - "step": 22161 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019250422782926337, - "loss": 1.592, - "step": 22162 - }, - { - "epoch": 0.07, - "learning_rate": 0.0001925038896029223, - "loss": 1.6882, - "step": 22163 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019250355137658123, - "loss": 1.565, - "step": 22164 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019250321315024017, - "loss": 1.5947, - "step": 22165 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019250287492389909, - "loss": 1.5904, - "step": 22166 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019250253669755803, - "loss": 1.5997, - "step": 22167 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019250219847121697, - "loss": 1.655, - "step": 22168 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019250186024487588, - "loss": 1.5222, - "step": 22169 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019250152201853482, - "loss": 1.4485, - "step": 22170 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019250118379219376, - "loss": 1.5142, - "step": 22171 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019250084556585268, - "loss": 1.5423, - "step": 22172 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019250050733951162, - "loss": 1.5458, - "step": 22173 - }, - { - "epoch": 0.07, - "learning_rate": 0.00019250016911317056, - "loss": 1.644, - "step": 22174 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019249983088682947, - "loss": 1.7588, - "step": 22175 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019249949266048842, - "loss": 1.6503, - "step": 22176 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019249915443414736, - "loss": 1.6472, - "step": 22177 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019249881620780627, - "loss": 1.8237, - "step": 22178 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924984779814652, - "loss": 1.5597, - "step": 22179 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019249813975512415, - "loss": 1.4993, - "step": 22180 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019249780152878307, - "loss": 1.6494, - "step": 22181 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192497463302442, - "loss": 1.5432, - "step": 22182 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019249712507610095, - "loss": 1.5513, - "step": 22183 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019249678684975986, - "loss": 1.6201, - "step": 22184 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924964486234188, - "loss": 1.7387, - "step": 22185 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019249611039707775, - "loss": 1.5415, - "step": 22186 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019249577217073666, - "loss": 1.5861, - "step": 22187 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924954339443956, - "loss": 1.6338, - "step": 22188 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019249509571805454, - "loss": 1.5768, - "step": 22189 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019249475749171346, - "loss": 1.6142, - "step": 22190 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924944192653724, - "loss": 1.4616, - "step": 22191 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019249408103903134, - "loss": 1.5316, - "step": 22192 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019249374281269025, - "loss": 1.5818, - "step": 22193 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924934045863492, - "loss": 1.4863, - "step": 22194 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019249306636000813, - "loss": 1.5867, - "step": 22195 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019249272813366705, - "loss": 1.5753, - "step": 22196 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192492389907326, - "loss": 1.4734, - "step": 22197 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019249205168098493, - "loss": 1.5148, - "step": 22198 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019249171345464384, - "loss": 1.4485, - "step": 22199 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019249137522830279, - "loss": 1.6097, - "step": 22200 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019249103700196173, - "loss": 1.5591, - "step": 22201 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019249069877562064, - "loss": 1.5187, - "step": 22202 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019249036054927958, - "loss": 1.6292, - "step": 22203 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019249002232293852, - "loss": 1.4033, - "step": 22204 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019248968409659744, - "loss": 1.4934, - "step": 22205 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019248934587025638, - "loss": 1.7608, - "step": 22206 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019248900764391532, - "loss": 1.6882, - "step": 22207 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019248866941757423, - "loss": 1.5935, - "step": 22208 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019248833119123317, - "loss": 1.4792, - "step": 22209 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019248799296489212, - "loss": 1.6023, - "step": 22210 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019248765473855103, - "loss": 1.478, - "step": 22211 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019248731651220997, - "loss": 1.6457, - "step": 22212 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924869782858689, - "loss": 1.5498, - "step": 22213 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019248664005952785, - "loss": 1.5257, - "step": 22214 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924863018331868, - "loss": 1.5358, - "step": 22215 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924859636068457, - "loss": 1.5925, - "step": 22216 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019248562538050465, - "loss": 1.509, - "step": 22217 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924852871541636, - "loss": 1.5807, - "step": 22218 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924849489278225, - "loss": 1.5918, - "step": 22219 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019248461070148145, - "loss": 1.6271, - "step": 22220 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924842724751404, - "loss": 1.4912, - "step": 22221 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924839342487993, - "loss": 1.676, - "step": 22222 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019248359602245824, - "loss": 1.5281, - "step": 22223 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019248325779611718, - "loss": 1.6051, - "step": 22224 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924829195697761, - "loss": 1.6084, - "step": 22225 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019248258134343504, - "loss": 1.4348, - "step": 22226 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019248224311709398, - "loss": 1.5355, - "step": 22227 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924819048907529, - "loss": 1.6486, - "step": 22228 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019248156666441183, - "loss": 1.6121, - "step": 22229 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019248122843807078, - "loss": 1.558, - "step": 22230 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924808902117297, - "loss": 1.4978, - "step": 22231 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019248055198538863, - "loss": 1.6418, - "step": 22232 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019248021375904757, - "loss": 1.7281, - "step": 22233 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019247987553270649, - "loss": 1.7029, - "step": 22234 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019247953730636543, - "loss": 1.5391, - "step": 22235 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019247919908002437, - "loss": 1.3762, - "step": 22236 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019247886085368328, - "loss": 1.7049, - "step": 22237 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019247852262734222, - "loss": 1.8265, - "step": 22238 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019247818440100116, - "loss": 1.5453, - "step": 22239 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924778461746601, - "loss": 1.5499, - "step": 22240 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019247750794831902, - "loss": 1.5107, - "step": 22241 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019247716972197796, - "loss": 1.6044, - "step": 22242 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924768314956369, - "loss": 1.58, - "step": 22243 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019247649326929582, - "loss": 1.7341, - "step": 22244 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019247615504295476, - "loss": 1.6355, - "step": 22245 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924758168166137, - "loss": 1.5132, - "step": 22246 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924754785902726, - "loss": 1.5332, - "step": 22247 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019247514036393155, - "loss": 1.6306, - "step": 22248 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924748021375905, - "loss": 1.5849, - "step": 22249 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924744639112494, - "loss": 1.5095, - "step": 22250 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019247412568490835, - "loss": 1.4943, - "step": 22251 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924737874585673, - "loss": 1.5929, - "step": 22252 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924734492322262, - "loss": 1.5286, - "step": 22253 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019247311100588515, - "loss": 1.5429, - "step": 22254 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924727727795441, - "loss": 1.4707, - "step": 22255 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192472434553203, - "loss": 1.6022, - "step": 22256 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019247209632686194, - "loss": 1.6776, - "step": 22257 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019247175810052088, - "loss": 1.5013, - "step": 22258 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924714198741798, - "loss": 1.5343, - "step": 22259 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019247108164783874, - "loss": 1.7355, - "step": 22260 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019247074342149768, - "loss": 1.4784, - "step": 22261 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924704051951566, - "loss": 1.4723, - "step": 22262 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019247006696881554, - "loss": 1.5348, - "step": 22263 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019246972874247448, - "loss": 1.4905, - "step": 22264 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924693905161334, - "loss": 1.6505, - "step": 22265 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019246905228979233, - "loss": 1.3814, - "step": 22266 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019246871406345127, - "loss": 1.4394, - "step": 22267 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924683758371102, - "loss": 1.5797, - "step": 22268 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019246803761076913, - "loss": 1.514, - "step": 22269 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019246769938442807, - "loss": 1.6369, - "step": 22270 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192467361158087, - "loss": 1.5987, - "step": 22271 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019246702293174595, - "loss": 1.4999, - "step": 22272 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019246668470540487, - "loss": 1.6163, - "step": 22273 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924663464790638, - "loss": 1.5676, - "step": 22274 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019246600825272275, - "loss": 1.6234, - "step": 22275 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019246567002638166, - "loss": 1.6212, - "step": 22276 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924653318000406, - "loss": 1.6028, - "step": 22277 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019246499357369954, - "loss": 1.5579, - "step": 22278 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019246465534735846, - "loss": 1.4509, - "step": 22279 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924643171210174, - "loss": 1.5411, - "step": 22280 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019246397889467634, - "loss": 1.6744, - "step": 22281 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019246364066833525, - "loss": 1.473, - "step": 22282 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924633024419942, - "loss": 1.6645, - "step": 22283 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019246296421565314, - "loss": 1.5955, - "step": 22284 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019246262598931205, - "loss": 1.4682, - "step": 22285 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192462287762971, - "loss": 1.6845, - "step": 22286 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019246194953662993, - "loss": 1.5294, - "step": 22287 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019246161131028885, - "loss": 1.4915, - "step": 22288 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924612730839478, - "loss": 1.8109, - "step": 22289 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019246093485760673, - "loss": 1.814, - "step": 22290 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019246059663126564, - "loss": 1.6975, - "step": 22291 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019246025840492458, - "loss": 1.5307, - "step": 22292 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019245992017858353, - "loss": 1.6944, - "step": 22293 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019245958195224244, - "loss": 1.5662, - "step": 22294 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019245924372590138, - "loss": 1.5108, - "step": 22295 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019245890549956032, - "loss": 1.561, - "step": 22296 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019245856727321924, - "loss": 1.6192, - "step": 22297 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019245822904687818, - "loss": 1.6696, - "step": 22298 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019245789082053712, - "loss": 1.4002, - "step": 22299 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019245755259419603, - "loss": 1.525, - "step": 22300 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019245721436785497, - "loss": 1.6775, - "step": 22301 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019245687614151391, - "loss": 1.6141, - "step": 22302 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019245653791517283, - "loss": 1.4752, - "step": 22303 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019245619968883177, - "loss": 1.5933, - "step": 22304 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924558614624907, - "loss": 1.5206, - "step": 22305 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019245552323614962, - "loss": 1.4777, - "step": 22306 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019245518500980857, - "loss": 1.4921, - "step": 22307 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924548467834675, - "loss": 1.6334, - "step": 22308 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019245450855712642, - "loss": 1.5957, - "step": 22309 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019245417033078536, - "loss": 1.4306, - "step": 22310 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924538321044443, - "loss": 1.696, - "step": 22311 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019245349387810322, - "loss": 1.4929, - "step": 22312 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019245315565176216, - "loss": 1.6575, - "step": 22313 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924528174254211, - "loss": 1.5828, - "step": 22314 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019245247919908001, - "loss": 1.6868, - "step": 22315 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019245214097273895, - "loss": 1.4341, - "step": 22316 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924518027463979, - "loss": 1.6092, - "step": 22317 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924514645200568, - "loss": 1.6121, - "step": 22318 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019245112629371575, - "loss": 1.5058, - "step": 22319 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924507880673747, - "loss": 1.6424, - "step": 22320 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924504498410336, - "loss": 1.6517, - "step": 22321 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019245011161469255, - "loss": 1.6408, - "step": 22322 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924497733883515, - "loss": 1.477, - "step": 22323 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019244943516201043, - "loss": 1.559, - "step": 22324 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019244909693566934, - "loss": 1.4749, - "step": 22325 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019244875870932829, - "loss": 1.6373, - "step": 22326 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019244842048298723, - "loss": 1.6265, - "step": 22327 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019244808225664617, - "loss": 1.5066, - "step": 22328 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924477440303051, - "loss": 1.7317, - "step": 22329 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019244740580396402, - "loss": 1.7097, - "step": 22330 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019244706757762296, - "loss": 1.5295, - "step": 22331 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924467293512819, - "loss": 1.5774, - "step": 22332 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019244639112494082, - "loss": 1.702, - "step": 22333 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019244605289859976, - "loss": 1.6257, - "step": 22334 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924457146722587, - "loss": 1.6032, - "step": 22335 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019244537644591762, - "loss": 1.6255, - "step": 22336 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019244503821957656, - "loss": 1.5466, - "step": 22337 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924446999932355, - "loss": 1.5867, - "step": 22338 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924443617668944, - "loss": 1.5716, - "step": 22339 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019244402354055335, - "loss": 1.6367, - "step": 22340 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924436853142123, - "loss": 1.7217, - "step": 22341 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924433470878712, - "loss": 1.5813, - "step": 22342 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019244300886153015, - "loss": 1.3885, - "step": 22343 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924426706351891, - "loss": 1.5169, - "step": 22344 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192442332408848, - "loss": 1.6364, - "step": 22345 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019244199418250695, - "loss": 1.6213, - "step": 22346 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019244165595616589, - "loss": 1.6298, - "step": 22347 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924413177298248, - "loss": 1.6924, - "step": 22348 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019244097950348374, - "loss": 1.5218, - "step": 22349 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019244064127714268, - "loss": 1.4844, - "step": 22350 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924403030508016, - "loss": 1.3911, - "step": 22351 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019243996482446054, - "loss": 1.472, - "step": 22352 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019243962659811948, - "loss": 1.5387, - "step": 22353 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924392883717784, - "loss": 1.4355, - "step": 22354 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019243895014543733, - "loss": 1.6047, - "step": 22355 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019243861191909628, - "loss": 1.4172, - "step": 22356 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924382736927552, - "loss": 1.644, - "step": 22357 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019243793546641413, - "loss": 1.6599, - "step": 22358 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019243759724007307, - "loss": 1.5126, - "step": 22359 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019243725901373199, - "loss": 1.6077, - "step": 22360 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019243692078739093, - "loss": 1.7377, - "step": 22361 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019243658256104987, - "loss": 1.5785, - "step": 22362 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019243624433470878, - "loss": 1.5298, - "step": 22363 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019243590610836772, - "loss": 1.5017, - "step": 22364 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019243556788202666, - "loss": 1.4661, - "step": 22365 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019243522965568558, - "loss": 1.7721, - "step": 22366 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019243489142934452, - "loss": 1.5674, - "step": 22367 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019243455320300346, - "loss": 1.5617, - "step": 22368 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019243421497666237, - "loss": 1.524, - "step": 22369 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019243387675032132, - "loss": 1.6308, - "step": 22370 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019243353852398026, - "loss": 1.6141, - "step": 22371 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019243320029763917, - "loss": 1.5471, - "step": 22372 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924328620712981, - "loss": 1.5796, - "step": 22373 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019243252384495705, - "loss": 1.6205, - "step": 22374 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019243218561861597, - "loss": 1.5477, - "step": 22375 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924318473922749, - "loss": 1.3948, - "step": 22376 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019243150916593385, - "loss": 1.506, - "step": 22377 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019243117093959276, - "loss": 1.5308, - "step": 22378 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924308327132517, - "loss": 1.7209, - "step": 22379 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019243049448691065, - "loss": 1.612, - "step": 22380 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019243015626056956, - "loss": 1.4676, - "step": 22381 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924298180342285, - "loss": 1.5148, - "step": 22382 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019242947980788747, - "loss": 1.5726, - "step": 22383 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019242914158154638, - "loss": 1.552, - "step": 22384 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019242880335520532, - "loss": 1.534, - "step": 22385 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019242846512886427, - "loss": 1.5219, - "step": 22386 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019242812690252318, - "loss": 1.5413, - "step": 22387 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019242778867618212, - "loss": 1.4619, - "step": 22388 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019242745044984106, - "loss": 1.6342, - "step": 22389 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019242711222349998, - "loss": 1.6827, - "step": 22390 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019242677399715892, - "loss": 1.5742, - "step": 22391 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019242643577081786, - "loss": 1.6103, - "step": 22392 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019242609754447677, - "loss": 1.5526, - "step": 22393 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924257593181357, - "loss": 1.6319, - "step": 22394 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019242542109179465, - "loss": 1.3409, - "step": 22395 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019242508286545357, - "loss": 1.6914, - "step": 22396 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924247446391125, - "loss": 1.6256, - "step": 22397 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019242440641277145, - "loss": 1.5288, - "step": 22398 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019242406818643036, - "loss": 1.5034, - "step": 22399 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924237299600893, - "loss": 1.6262, - "step": 22400 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019242339173374825, - "loss": 1.5722, - "step": 22401 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019242305350740716, - "loss": 1.6411, - "step": 22402 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924227152810661, - "loss": 1.6503, - "step": 22403 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019242237705472504, - "loss": 1.5047, - "step": 22404 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019242203882838396, - "loss": 1.5614, - "step": 22405 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924217006020429, - "loss": 1.5401, - "step": 22406 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019242136237570184, - "loss": 1.761, - "step": 22407 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019242102414936075, - "loss": 1.3942, - "step": 22408 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924206859230197, - "loss": 1.4454, - "step": 22409 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019242034769667864, - "loss": 1.7344, - "step": 22410 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019242000947033755, - "loss": 1.5231, - "step": 22411 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924196712439965, - "loss": 1.3836, - "step": 22412 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019241933301765543, - "loss": 1.6153, - "step": 22413 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019241899479131435, - "loss": 1.5971, - "step": 22414 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924186565649733, - "loss": 1.7525, - "step": 22415 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019241831833863223, - "loss": 1.5765, - "step": 22416 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019241798011229114, - "loss": 1.6, - "step": 22417 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019241764188595008, - "loss": 1.4514, - "step": 22418 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019241730365960902, - "loss": 1.6398, - "step": 22419 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019241696543326794, - "loss": 1.5726, - "step": 22420 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019241662720692688, - "loss": 1.5835, - "step": 22421 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019241628898058582, - "loss": 1.5662, - "step": 22422 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019241595075424474, - "loss": 1.5562, - "step": 22423 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019241561252790368, - "loss": 1.4363, - "step": 22424 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019241527430156262, - "loss": 1.6719, - "step": 22425 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019241493607522153, - "loss": 1.628, - "step": 22426 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019241459784888047, - "loss": 1.506, - "step": 22427 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019241425962253941, - "loss": 1.4576, - "step": 22428 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019241392139619833, - "loss": 1.5989, - "step": 22429 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019241358316985727, - "loss": 1.5786, - "step": 22430 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924132449435162, - "loss": 1.5104, - "step": 22431 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019241290671717512, - "loss": 1.5339, - "step": 22432 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019241256849083407, - "loss": 1.5494, - "step": 22433 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192412230264493, - "loss": 1.596, - "step": 22434 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019241189203815192, - "loss": 1.4186, - "step": 22435 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019241155381181086, - "loss": 1.6919, - "step": 22436 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924112155854698, - "loss": 1.6458, - "step": 22437 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019241087735912872, - "loss": 1.7022, - "step": 22438 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019241053913278769, - "loss": 1.4464, - "step": 22439 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019241020090644663, - "loss": 1.5982, - "step": 22440 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019240986268010554, - "loss": 1.621, - "step": 22441 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019240952445376448, - "loss": 1.4335, - "step": 22442 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019240918622742342, - "loss": 1.5854, - "step": 22443 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019240884800108234, - "loss": 1.6009, - "step": 22444 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019240850977474128, - "loss": 1.6355, - "step": 22445 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019240817154840022, - "loss": 1.6603, - "step": 22446 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019240783332205913, - "loss": 1.4907, - "step": 22447 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019240749509571807, - "loss": 1.6209, - "step": 22448 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019240715686937702, - "loss": 1.5996, - "step": 22449 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019240681864303593, - "loss": 1.6736, - "step": 22450 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019240648041669487, - "loss": 1.4387, - "step": 22451 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924061421903538, - "loss": 1.6843, - "step": 22452 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019240580396401273, - "loss": 1.6767, - "step": 22453 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019240546573767167, - "loss": 1.5689, - "step": 22454 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924051275113306, - "loss": 1.4664, - "step": 22455 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019240478928498952, - "loss": 1.7171, - "step": 22456 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019240445105864846, - "loss": 1.7079, - "step": 22457 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924041128323074, - "loss": 1.4257, - "step": 22458 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019240377460596632, - "loss": 1.5055, - "step": 22459 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019240343637962526, - "loss": 1.6024, - "step": 22460 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924030981532842, - "loss": 1.4551, - "step": 22461 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019240275992694311, - "loss": 1.4797, - "step": 22462 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019240242170060206, - "loss": 1.4863, - "step": 22463 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192402083474261, - "loss": 1.3717, - "step": 22464 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924017452479199, - "loss": 1.5286, - "step": 22465 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019240140702157885, - "loss": 1.5716, - "step": 22466 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924010687952378, - "loss": 1.6806, - "step": 22467 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924007305688967, - "loss": 1.4332, - "step": 22468 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019240039234255565, - "loss": 1.6484, - "step": 22469 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001924000541162146, - "loss": 1.5208, - "step": 22470 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923997158898735, - "loss": 1.5642, - "step": 22471 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019239937766353244, - "loss": 1.4976, - "step": 22472 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019239903943719139, - "loss": 1.7133, - "step": 22473 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923987012108503, - "loss": 1.4602, - "step": 22474 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019239836298450924, - "loss": 1.5699, - "step": 22475 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019239802475816818, - "loss": 1.4828, - "step": 22476 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923976865318271, - "loss": 1.3921, - "step": 22477 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019239734830548604, - "loss": 1.6263, - "step": 22478 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019239701007914498, - "loss": 1.5387, - "step": 22479 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923966718528039, - "loss": 1.5562, - "step": 22480 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019239633362646283, - "loss": 1.6044, - "step": 22481 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019239599540012177, - "loss": 1.7275, - "step": 22482 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923956571737807, - "loss": 1.72, - "step": 22483 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019239531894743963, - "loss": 1.5883, - "step": 22484 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019239498072109857, - "loss": 1.5004, - "step": 22485 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019239464249475749, - "loss": 1.3851, - "step": 22486 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019239430426841643, - "loss": 1.6681, - "step": 22487 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019239396604207537, - "loss": 1.6111, - "step": 22488 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019239362781573428, - "loss": 1.5806, - "step": 22489 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019239328958939322, - "loss": 1.6644, - "step": 22490 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019239295136305216, - "loss": 1.4567, - "step": 22491 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019239261313671108, - "loss": 1.7348, - "step": 22492 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019239227491037002, - "loss": 1.497, - "step": 22493 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019239193668402896, - "loss": 1.6297, - "step": 22494 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923915984576879, - "loss": 1.7925, - "step": 22495 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019239126023134684, - "loss": 1.6356, - "step": 22496 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019239092200500576, - "loss": 1.6157, - "step": 22497 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923905837786647, - "loss": 1.4913, - "step": 22498 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019239024555232364, - "loss": 1.6226, - "step": 22499 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019238990732598255, - "loss": 1.7057, - "step": 22500 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923895690996415, - "loss": 1.5354, - "step": 22501 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019238923087330043, - "loss": 1.628, - "step": 22502 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019238889264695935, - "loss": 1.4514, - "step": 22503 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923885544206183, - "loss": 1.5689, - "step": 22504 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019238821619427723, - "loss": 1.5673, - "step": 22505 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019238787796793615, - "loss": 1.6304, - "step": 22506 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019238753974159509, - "loss": 1.4877, - "step": 22507 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019238720151525403, - "loss": 1.5308, - "step": 22508 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019238686328891297, - "loss": 1.6872, - "step": 22509 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019238652506257188, - "loss": 1.6398, - "step": 22510 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019238618683623082, - "loss": 1.52, - "step": 22511 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019238584860988976, - "loss": 1.7233, - "step": 22512 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019238551038354868, - "loss": 1.5671, - "step": 22513 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019238517215720762, - "loss": 1.5471, - "step": 22514 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019238483393086656, - "loss": 1.5674, - "step": 22515 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019238449570452548, - "loss": 1.6046, - "step": 22516 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019238415747818442, - "loss": 1.5822, - "step": 22517 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019238381925184336, - "loss": 1.5786, - "step": 22518 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019238348102550227, - "loss": 1.4464, - "step": 22519 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923831427991612, - "loss": 1.4463, - "step": 22520 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019238280457282015, - "loss": 1.636, - "step": 22521 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019238246634647907, - "loss": 1.5724, - "step": 22522 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192382128120138, - "loss": 1.4415, - "step": 22523 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019238178989379695, - "loss": 1.6001, - "step": 22524 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019238145166745586, - "loss": 1.6159, - "step": 22525 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923811134411148, - "loss": 1.5089, - "step": 22526 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019238077521477375, - "loss": 1.7033, - "step": 22527 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019238043698843266, - "loss": 1.4333, - "step": 22528 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923800987620916, - "loss": 1.5771, - "step": 22529 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019237976053575054, - "loss": 1.6504, - "step": 22530 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019237942230940946, - "loss": 1.5522, - "step": 22531 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923790840830684, - "loss": 1.808, - "step": 22532 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019237874585672734, - "loss": 1.4375, - "step": 22533 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019237840763038625, - "loss": 1.6314, - "step": 22534 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923780694040452, - "loss": 1.5939, - "step": 22535 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019237773117770414, - "loss": 1.3726, - "step": 22536 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019237739295136305, - "loss": 1.5306, - "step": 22537 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192377054725022, - "loss": 1.6689, - "step": 22538 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019237671649868093, - "loss": 1.5648, - "step": 22539 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019237637827233985, - "loss": 1.5283, - "step": 22540 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923760400459988, - "loss": 1.5373, - "step": 22541 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019237570181965773, - "loss": 1.6484, - "step": 22542 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019237536359331664, - "loss": 1.5795, - "step": 22543 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019237502536697558, - "loss": 1.5396, - "step": 22544 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019237468714063452, - "loss": 1.6461, - "step": 22545 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019237434891429344, - "loss": 1.5681, - "step": 22546 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019237401068795238, - "loss": 1.5768, - "step": 22547 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019237367246161132, - "loss": 1.4716, - "step": 22548 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019237333423527023, - "loss": 1.5269, - "step": 22549 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019237299600892918, - "loss": 1.6172, - "step": 22550 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019237265778258812, - "loss": 1.4508, - "step": 22551 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019237231955624706, - "loss": 1.5435, - "step": 22552 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192371981329906, - "loss": 1.7441, - "step": 22553 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923716431035649, - "loss": 1.7157, - "step": 22554 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019237130487722385, - "loss": 1.4599, - "step": 22555 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923709666508828, - "loss": 1.6265, - "step": 22556 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923706284245417, - "loss": 1.5433, - "step": 22557 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019237029019820065, - "loss": 1.55, - "step": 22558 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923699519718596, - "loss": 1.6465, - "step": 22559 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923696137455185, - "loss": 1.4306, - "step": 22560 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019236927551917745, - "loss": 1.5449, - "step": 22561 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923689372928364, - "loss": 1.533, - "step": 22562 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923685990664953, - "loss": 1.5638, - "step": 22563 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019236826084015424, - "loss": 1.647, - "step": 22564 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019236792261381318, - "loss": 1.6324, - "step": 22565 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923675843874721, - "loss": 1.4994, - "step": 22566 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019236724616113104, - "loss": 1.4761, - "step": 22567 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019236690793478998, - "loss": 1.7287, - "step": 22568 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923665697084489, - "loss": 1.402, - "step": 22569 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019236623148210784, - "loss": 1.5903, - "step": 22570 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019236589325576678, - "loss": 1.5987, - "step": 22571 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923655550294257, - "loss": 1.5898, - "step": 22572 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019236521680308463, - "loss": 1.5065, - "step": 22573 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019236487857674357, - "loss": 1.664, - "step": 22574 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923645403504025, - "loss": 1.6012, - "step": 22575 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019236420212406143, - "loss": 1.4875, - "step": 22576 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019236386389772037, - "loss": 1.4344, - "step": 22577 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019236352567137928, - "loss": 1.6318, - "step": 22578 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019236318744503822, - "loss": 1.4851, - "step": 22579 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019236284921869717, - "loss": 1.3997, - "step": 22580 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019236251099235608, - "loss": 1.5727, - "step": 22581 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019236217276601502, - "loss": 1.5064, - "step": 22582 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019236183453967396, - "loss": 1.5371, - "step": 22583 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019236149631333288, - "loss": 1.5742, - "step": 22584 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019236115808699182, - "loss": 1.4899, - "step": 22585 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019236081986065076, - "loss": 1.5999, - "step": 22586 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019236048163430967, - "loss": 1.5333, - "step": 22587 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019236014340796861, - "loss": 1.6315, - "step": 22588 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019235980518162756, - "loss": 1.6236, - "step": 22589 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019235946695528647, - "loss": 1.506, - "step": 22590 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923591287289454, - "loss": 1.7072, - "step": 22591 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019235879050260435, - "loss": 1.5456, - "step": 22592 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923584522762633, - "loss": 1.5271, - "step": 22593 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923581140499222, - "loss": 1.5073, - "step": 22594 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019235777582358115, - "loss": 1.4891, - "step": 22595 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923574375972401, - "loss": 1.5276, - "step": 22596 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192357099370899, - "loss": 1.565, - "step": 22597 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019235676114455794, - "loss": 1.7622, - "step": 22598 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019235642291821689, - "loss": 1.427, - "step": 22599 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923560846918758, - "loss": 1.3798, - "step": 22600 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019235574646553474, - "loss": 1.5613, - "step": 22601 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019235540823919368, - "loss": 1.5751, - "step": 22602 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923550700128526, - "loss": 1.5964, - "step": 22603 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019235473178651154, - "loss": 1.4564, - "step": 22604 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019235439356017048, - "loss": 1.42, - "step": 22605 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923540553338294, - "loss": 1.5571, - "step": 22606 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019235371710748836, - "loss": 1.5311, - "step": 22607 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019235337888114727, - "loss": 1.5331, - "step": 22608 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019235304065480622, - "loss": 1.5984, - "step": 22609 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019235270242846516, - "loss": 1.4831, - "step": 22610 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019235236420212407, - "loss": 1.6454, - "step": 22611 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192352025975783, - "loss": 1.5829, - "step": 22612 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019235168774944195, - "loss": 1.5074, - "step": 22613 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019235134952310087, - "loss": 1.6029, - "step": 22614 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923510112967598, - "loss": 1.6789, - "step": 22615 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019235067307041875, - "loss": 1.4562, - "step": 22616 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019235033484407766, - "loss": 1.5572, - "step": 22617 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923499966177366, - "loss": 1.4969, - "step": 22618 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019234965839139555, - "loss": 1.5699, - "step": 22619 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019234932016505446, - "loss": 1.4555, - "step": 22620 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923489819387134, - "loss": 1.5232, - "step": 22621 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019234864371237234, - "loss": 1.5081, - "step": 22622 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019234830548603126, - "loss": 1.4522, - "step": 22623 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923479672596902, - "loss": 1.624, - "step": 22624 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019234762903334914, - "loss": 1.6091, - "step": 22625 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019234729080700805, - "loss": 1.7022, - "step": 22626 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192346952580667, - "loss": 1.7035, - "step": 22627 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019234661435432593, - "loss": 1.3818, - "step": 22628 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019234627612798485, - "loss": 1.5298, - "step": 22629 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923459379016438, - "loss": 1.6515, - "step": 22630 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019234559967530273, - "loss": 1.5522, - "step": 22631 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019234526144896164, - "loss": 1.5212, - "step": 22632 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019234492322262059, - "loss": 1.5576, - "step": 22633 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019234458499627953, - "loss": 1.6103, - "step": 22634 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019234424676993844, - "loss": 1.78, - "step": 22635 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019234390854359738, - "loss": 1.5942, - "step": 22636 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019234357031725632, - "loss": 1.6233, - "step": 22637 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019234323209091524, - "loss": 1.7766, - "step": 22638 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019234289386457418, - "loss": 1.6609, - "step": 22639 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019234255563823312, - "loss": 1.6364, - "step": 22640 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019234221741189203, - "loss": 1.6472, - "step": 22641 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019234187918555097, - "loss": 1.5921, - "step": 22642 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019234154095920992, - "loss": 1.6879, - "step": 22643 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019234120273286883, - "loss": 1.8267, - "step": 22644 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019234086450652777, - "loss": 1.5015, - "step": 22645 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923405262801867, - "loss": 1.5665, - "step": 22646 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019234018805384563, - "loss": 1.5096, - "step": 22647 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019233984982750457, - "loss": 1.7215, - "step": 22648 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923395116011635, - "loss": 1.553, - "step": 22649 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019233917337482242, - "loss": 1.7175, - "step": 22650 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019233883514848136, - "loss": 1.6104, - "step": 22651 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923384969221403, - "loss": 1.5689, - "step": 22652 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019233815869579922, - "loss": 1.6276, - "step": 22653 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019233782046945816, - "loss": 1.6012, - "step": 22654 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923374822431171, - "loss": 1.6796, - "step": 22655 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019233714401677602, - "loss": 1.6789, - "step": 22656 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019233680579043496, - "loss": 1.4423, - "step": 22657 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923364675640939, - "loss": 1.4931, - "step": 22658 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923361293377528, - "loss": 1.5391, - "step": 22659 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019233579111141175, - "loss": 1.665, - "step": 22660 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923354528850707, - "loss": 1.6586, - "step": 22661 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923351146587296, - "loss": 1.5518, - "step": 22662 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019233477643238855, - "loss": 1.4379, - "step": 22663 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019233443820604752, - "loss": 1.6168, - "step": 22664 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019233409997970643, - "loss": 1.6767, - "step": 22665 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019233376175336537, - "loss": 1.6687, - "step": 22666 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923334235270243, - "loss": 1.6405, - "step": 22667 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019233308530068323, - "loss": 1.6116, - "step": 22668 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019233274707434217, - "loss": 1.7198, - "step": 22669 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923324088480011, - "loss": 1.5648, - "step": 22670 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019233207062166002, - "loss": 1.4458, - "step": 22671 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019233173239531896, - "loss": 1.4257, - "step": 22672 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923313941689779, - "loss": 1.6901, - "step": 22673 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019233105594263682, - "loss": 1.6562, - "step": 22674 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019233071771629576, - "loss": 1.5408, - "step": 22675 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923303794899547, - "loss": 1.4495, - "step": 22676 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019233004126361362, - "loss": 1.4767, - "step": 22677 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019232970303727256, - "loss": 1.7581, - "step": 22678 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923293648109315, - "loss": 1.6057, - "step": 22679 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923290265845904, - "loss": 1.667, - "step": 22680 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019232868835824935, - "loss": 1.591, - "step": 22681 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923283501319083, - "loss": 1.5951, - "step": 22682 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923280119055672, - "loss": 1.3286, - "step": 22683 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019232767367922615, - "loss": 1.5869, - "step": 22684 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923273354528851, - "loss": 1.6356, - "step": 22685 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192326997226544, - "loss": 1.4919, - "step": 22686 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019232665900020295, - "loss": 1.6833, - "step": 22687 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923263207738619, - "loss": 1.5005, - "step": 22688 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923259825475208, - "loss": 1.597, - "step": 22689 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019232564432117974, - "loss": 1.5322, - "step": 22690 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019232530609483868, - "loss": 1.5448, - "step": 22691 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923249678684976, - "loss": 1.7062, - "step": 22692 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019232462964215654, - "loss": 1.5774, - "step": 22693 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019232429141581548, - "loss": 1.703, - "step": 22694 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923239531894744, - "loss": 1.4814, - "step": 22695 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019232361496313334, - "loss": 1.5147, - "step": 22696 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019232327673679228, - "loss": 1.5919, - "step": 22697 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923229385104512, - "loss": 1.5817, - "step": 22698 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019232260028411013, - "loss": 1.5716, - "step": 22699 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019232226205776907, - "loss": 1.6238, - "step": 22700 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192321923831428, - "loss": 1.5264, - "step": 22701 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019232158560508693, - "loss": 1.6236, - "step": 22702 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019232124737874587, - "loss": 1.5416, - "step": 22703 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019232090915240478, - "loss": 1.6836, - "step": 22704 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019232057092606372, - "loss": 1.4242, - "step": 22705 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019232023269972267, - "loss": 1.5774, - "step": 22706 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019231989447338158, - "loss": 1.6471, - "step": 22707 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019231955624704052, - "loss": 1.5281, - "step": 22708 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019231921802069946, - "loss": 1.5091, - "step": 22709 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019231887979435838, - "loss": 1.5914, - "step": 22710 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019231854156801732, - "loss": 1.6554, - "step": 22711 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019231820334167626, - "loss": 1.4045, - "step": 22712 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019231786511533517, - "loss": 1.4514, - "step": 22713 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923175268889941, - "loss": 1.4972, - "step": 22714 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019231718866265305, - "loss": 1.6022, - "step": 22715 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019231685043631197, - "loss": 1.4632, - "step": 22716 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923165122099709, - "loss": 1.8091, - "step": 22717 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019231617398362985, - "loss": 1.5665, - "step": 22718 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019231583575728876, - "loss": 1.4311, - "step": 22719 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019231549753094773, - "loss": 1.5538, - "step": 22720 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019231515930460667, - "loss": 1.7084, - "step": 22721 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923148210782656, - "loss": 1.6023, - "step": 22722 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019231448285192453, - "loss": 1.5668, - "step": 22723 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019231414462558347, - "loss": 1.3416, - "step": 22724 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019231380639924238, - "loss": 1.4175, - "step": 22725 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019231346817290133, - "loss": 1.4488, - "step": 22726 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019231312994656027, - "loss": 1.6214, - "step": 22727 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019231279172021918, - "loss": 1.431, - "step": 22728 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019231245349387812, - "loss": 1.4297, - "step": 22729 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019231211526753706, - "loss": 1.478, - "step": 22730 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019231177704119598, - "loss": 1.6311, - "step": 22731 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019231143881485492, - "loss": 1.5021, - "step": 22732 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019231110058851386, - "loss": 1.6515, - "step": 22733 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019231076236217277, - "loss": 1.6862, - "step": 22734 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019231042413583171, - "loss": 1.8793, - "step": 22735 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019231008590949066, - "loss": 1.6057, - "step": 22736 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019230974768314957, - "loss": 1.5557, - "step": 22737 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923094094568085, - "loss": 1.6716, - "step": 22738 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019230907123046745, - "loss": 1.6621, - "step": 22739 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019230873300412637, - "loss": 1.6846, - "step": 22740 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923083947777853, - "loss": 1.4906, - "step": 22741 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019230805655144425, - "loss": 1.5991, - "step": 22742 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019230771832510316, - "loss": 1.5992, - "step": 22743 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923073800987621, - "loss": 1.7065, - "step": 22744 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019230704187242104, - "loss": 1.4424, - "step": 22745 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019230670364607996, - "loss": 1.5884, - "step": 22746 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923063654197389, - "loss": 1.725, - "step": 22747 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019230602719339784, - "loss": 1.5232, - "step": 22748 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019230568896705676, - "loss": 1.5162, - "step": 22749 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923053507407157, - "loss": 1.6785, - "step": 22750 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019230501251437464, - "loss": 1.5726, - "step": 22751 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019230467428803355, - "loss": 1.5838, - "step": 22752 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923043360616925, - "loss": 1.5929, - "step": 22753 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019230399783535143, - "loss": 1.6279, - "step": 22754 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019230365960901035, - "loss": 1.5687, - "step": 22755 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001923033213826693, - "loss": 1.5941, - "step": 22756 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019230298315632823, - "loss": 1.5299, - "step": 22757 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019230264492998714, - "loss": 1.4588, - "step": 22758 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019230230670364609, - "loss": 1.4469, - "step": 22759 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019230196847730503, - "loss": 1.5797, - "step": 22760 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019230163025096394, - "loss": 1.6467, - "step": 22761 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019230129202462288, - "loss": 1.4914, - "step": 22762 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019230095379828182, - "loss": 1.576, - "step": 22763 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019230061557194074, - "loss": 1.6303, - "step": 22764 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019230027734559968, - "loss": 1.4446, - "step": 22765 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019229993911925862, - "loss": 1.6741, - "step": 22766 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019229960089291753, - "loss": 1.7326, - "step": 22767 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019229926266657647, - "loss": 1.48, - "step": 22768 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019229892444023542, - "loss": 1.5809, - "step": 22769 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019229858621389433, - "loss": 1.4681, - "step": 22770 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019229824798755327, - "loss": 1.6309, - "step": 22771 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922979097612122, - "loss": 1.4478, - "step": 22772 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019229757153487113, - "loss": 1.5584, - "step": 22773 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019229723330853007, - "loss": 1.624, - "step": 22774 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192296895082189, - "loss": 1.6199, - "step": 22775 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019229655685584795, - "loss": 1.7237, - "step": 22776 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922962186295069, - "loss": 1.525, - "step": 22777 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019229588040316583, - "loss": 1.7593, - "step": 22778 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019229554217682475, - "loss": 1.5692, - "step": 22779 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019229520395048369, - "loss": 1.686, - "step": 22780 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019229486572414263, - "loss": 1.5889, - "step": 22781 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019229452749780154, - "loss": 1.5719, - "step": 22782 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019229418927146048, - "loss": 1.593, - "step": 22783 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019229385104511942, - "loss": 1.5371, - "step": 22784 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019229351281877834, - "loss": 1.6033, - "step": 22785 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019229317459243728, - "loss": 1.7407, - "step": 22786 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019229283636609622, - "loss": 1.572, - "step": 22787 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019229249813975513, - "loss": 1.4153, - "step": 22788 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019229215991341408, - "loss": 1.4248, - "step": 22789 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019229182168707302, - "loss": 1.4429, - "step": 22790 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019229148346073193, - "loss": 1.6529, - "step": 22791 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019229114523439087, - "loss": 1.6572, - "step": 22792 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922908070080498, - "loss": 1.5606, - "step": 22793 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019229046878170873, - "loss": 1.7735, - "step": 22794 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019229013055536767, - "loss": 1.6156, - "step": 22795 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922897923290266, - "loss": 1.7282, - "step": 22796 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019228945410268552, - "loss": 1.538, - "step": 22797 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019228911587634446, - "loss": 1.5778, - "step": 22798 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922887776500034, - "loss": 1.3944, - "step": 22799 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019228843942366232, - "loss": 1.6232, - "step": 22800 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019228810119732126, - "loss": 1.4612, - "step": 22801 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922877629709802, - "loss": 1.5833, - "step": 22802 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019228742474463912, - "loss": 1.5696, - "step": 22803 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019228708651829806, - "loss": 1.4976, - "step": 22804 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192286748291957, - "loss": 1.6171, - "step": 22805 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922864100656159, - "loss": 1.6719, - "step": 22806 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019228607183927485, - "loss": 1.5631, - "step": 22807 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922857336129338, - "loss": 1.6352, - "step": 22808 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922853953865927, - "loss": 1.587, - "step": 22809 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019228505716025165, - "loss": 1.6914, - "step": 22810 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922847189339106, - "loss": 1.6481, - "step": 22811 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922843807075695, - "loss": 1.4487, - "step": 22812 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019228404248122845, - "loss": 1.4683, - "step": 22813 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922837042548874, - "loss": 1.4099, - "step": 22814 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922833660285463, - "loss": 1.5378, - "step": 22815 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019228302780220524, - "loss": 1.5406, - "step": 22816 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019228268957586418, - "loss": 1.4402, - "step": 22817 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922823513495231, - "loss": 1.5556, - "step": 22818 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019228201312318204, - "loss": 1.764, - "step": 22819 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019228167489684098, - "loss": 1.4527, - "step": 22820 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922813366704999, - "loss": 1.7222, - "step": 22821 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019228099844415883, - "loss": 1.5394, - "step": 22822 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019228066021781778, - "loss": 1.5472, - "step": 22823 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922803219914767, - "loss": 1.6025, - "step": 22824 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019227998376513563, - "loss": 1.4896, - "step": 22825 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019227964553879457, - "loss": 1.546, - "step": 22826 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019227930731245349, - "loss": 1.5324, - "step": 22827 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019227896908611243, - "loss": 1.6231, - "step": 22828 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019227863085977137, - "loss": 1.6027, - "step": 22829 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019227829263343028, - "loss": 1.5945, - "step": 22830 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019227795440708922, - "loss": 1.6795, - "step": 22831 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019227761618074816, - "loss": 1.6137, - "step": 22832 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922772779544071, - "loss": 1.4378, - "step": 22833 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019227693972806605, - "loss": 1.5355, - "step": 22834 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019227660150172496, - "loss": 1.653, - "step": 22835 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922762632753839, - "loss": 1.6663, - "step": 22836 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019227592504904284, - "loss": 1.3988, - "step": 22837 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019227558682270176, - "loss": 1.5387, - "step": 22838 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922752485963607, - "loss": 1.5894, - "step": 22839 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019227491037001964, - "loss": 1.3976, - "step": 22840 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019227457214367855, - "loss": 1.4959, - "step": 22841 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922742339173375, - "loss": 1.5679, - "step": 22842 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019227389569099644, - "loss": 1.6392, - "step": 22843 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019227355746465535, - "loss": 1.5221, - "step": 22844 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922732192383143, - "loss": 1.5088, - "step": 22845 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019227288101197323, - "loss": 1.6582, - "step": 22846 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019227254278563215, - "loss": 1.4531, - "step": 22847 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922722045592911, - "loss": 1.5501, - "step": 22848 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019227186633295003, - "loss": 1.5924, - "step": 22849 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019227152810660894, - "loss": 1.5214, - "step": 22850 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019227118988026788, - "loss": 1.5239, - "step": 22851 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019227085165392683, - "loss": 1.5857, - "step": 22852 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019227051342758574, - "loss": 1.4729, - "step": 22853 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019227017520124468, - "loss": 1.6373, - "step": 22854 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019226983697490362, - "loss": 1.5106, - "step": 22855 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019226949874856254, - "loss": 1.6202, - "step": 22856 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019226916052222148, - "loss": 1.5547, - "step": 22857 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019226882229588042, - "loss": 1.6127, - "step": 22858 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019226848406953933, - "loss": 1.6457, - "step": 22859 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019226814584319827, - "loss": 1.6176, - "step": 22860 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019226780761685721, - "loss": 1.574, - "step": 22861 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019226746939051616, - "loss": 1.6403, - "step": 22862 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019226713116417507, - "loss": 1.4695, - "step": 22863 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192266792937834, - "loss": 1.5646, - "step": 22864 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019226645471149295, - "loss": 1.3998, - "step": 22865 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019226611648515187, - "loss": 1.5365, - "step": 22866 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922657782588108, - "loss": 1.5008, - "step": 22867 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019226544003246975, - "loss": 1.4657, - "step": 22868 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019226510180612866, - "loss": 1.634, - "step": 22869 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922647635797876, - "loss": 1.5008, - "step": 22870 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019226442535344654, - "loss": 1.6885, - "step": 22871 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019226408712710546, - "loss": 1.6349, - "step": 22872 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922637489007644, - "loss": 1.6276, - "step": 22873 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019226341067442334, - "loss": 1.4146, - "step": 22874 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019226307244808225, - "loss": 1.5188, - "step": 22875 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922627342217412, - "loss": 1.9302, - "step": 22876 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019226239599540014, - "loss": 1.6024, - "step": 22877 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019226205776905905, - "loss": 1.5979, - "step": 22878 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192261719542718, - "loss": 1.4241, - "step": 22879 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019226138131637693, - "loss": 1.5384, - "step": 22880 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019226104309003585, - "loss": 1.6929, - "step": 22881 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922607048636948, - "loss": 1.6831, - "step": 22882 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019226036663735373, - "loss": 1.3531, - "step": 22883 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019226002841101264, - "loss": 1.5913, - "step": 22884 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019225969018467158, - "loss": 1.5939, - "step": 22885 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019225935195833053, - "loss": 1.8802, - "step": 22886 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019225901373198944, - "loss": 1.6058, - "step": 22887 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922586755056484, - "loss": 1.4995, - "step": 22888 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019225833727930732, - "loss": 1.6723, - "step": 22889 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019225799905296626, - "loss": 1.8056, - "step": 22890 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922576608266252, - "loss": 1.5515, - "step": 22891 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019225732260028412, - "loss": 1.5909, - "step": 22892 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019225698437394306, - "loss": 1.7496, - "step": 22893 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192256646147602, - "loss": 1.5108, - "step": 22894 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019225630792126091, - "loss": 1.6072, - "step": 22895 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019225596969491986, - "loss": 1.6203, - "step": 22896 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922556314685788, - "loss": 1.5806, - "step": 22897 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922552932422377, - "loss": 1.5262, - "step": 22898 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019225495501589665, - "loss": 1.4702, - "step": 22899 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922546167895556, - "loss": 1.5167, - "step": 22900 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922542785632145, - "loss": 1.6086, - "step": 22901 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019225394033687345, - "loss": 1.4323, - "step": 22902 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922536021105324, - "loss": 1.5344, - "step": 22903 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922532638841913, - "loss": 1.3661, - "step": 22904 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019225292565785024, - "loss": 1.5044, - "step": 22905 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019225258743150919, - "loss": 1.6784, - "step": 22906 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922522492051681, - "loss": 1.5166, - "step": 22907 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019225191097882704, - "loss": 1.4153, - "step": 22908 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019225157275248598, - "loss": 1.5918, - "step": 22909 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922512345261449, - "loss": 1.5776, - "step": 22910 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019225089629980384, - "loss": 1.5917, - "step": 22911 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019225055807346278, - "loss": 1.5292, - "step": 22912 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922502198471217, - "loss": 1.7132, - "step": 22913 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019224988162078063, - "loss": 1.4292, - "step": 22914 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019224954339443957, - "loss": 1.5256, - "step": 22915 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922492051680985, - "loss": 1.7138, - "step": 22916 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019224886694175743, - "loss": 1.6402, - "step": 22917 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019224852871541637, - "loss": 1.453, - "step": 22918 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019224819048907529, - "loss": 1.5138, - "step": 22919 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019224785226273423, - "loss": 1.4773, - "step": 22920 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019224751403639317, - "loss": 1.4299, - "step": 22921 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019224717581005208, - "loss": 1.5942, - "step": 22922 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019224683758371102, - "loss": 1.5512, - "step": 22923 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019224649935736996, - "loss": 1.6876, - "step": 22924 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019224616113102888, - "loss": 1.6334, - "step": 22925 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019224582290468782, - "loss": 1.5133, - "step": 22926 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019224548467834676, - "loss": 1.6405, - "step": 22927 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019224514645200567, - "loss": 1.4134, - "step": 22928 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019224480822566462, - "loss": 1.3732, - "step": 22929 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019224446999932356, - "loss": 1.4414, - "step": 22930 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019224413177298247, - "loss": 1.5623, - "step": 22931 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922437935466414, - "loss": 1.4602, - "step": 22932 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019224345532030035, - "loss": 1.7302, - "step": 22933 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019224311709395927, - "loss": 1.4946, - "step": 22934 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922427788676182, - "loss": 1.5907, - "step": 22935 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019224244064127715, - "loss": 1.6509, - "step": 22936 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019224210241493606, - "loss": 1.5326, - "step": 22937 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192241764188595, - "loss": 1.6409, - "step": 22938 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019224142596225395, - "loss": 1.5529, - "step": 22939 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019224108773591286, - "loss": 1.7127, - "step": 22940 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922407495095718, - "loss": 1.556, - "step": 22941 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019224041128323074, - "loss": 1.7206, - "step": 22942 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019224007305688966, - "loss": 1.5364, - "step": 22943 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019223973483054862, - "loss": 1.6295, - "step": 22944 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019223939660420756, - "loss": 1.4851, - "step": 22945 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019223905837786648, - "loss": 1.5879, - "step": 22946 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019223872015152542, - "loss": 1.7146, - "step": 22947 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019223838192518436, - "loss": 1.4577, - "step": 22948 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019223804369884328, - "loss": 1.5638, - "step": 22949 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019223770547250222, - "loss": 1.6206, - "step": 22950 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019223736724616116, - "loss": 1.3908, - "step": 22951 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019223702901982007, - "loss": 1.5434, - "step": 22952 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192236690793479, - "loss": 1.4894, - "step": 22953 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019223635256713795, - "loss": 1.5398, - "step": 22954 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019223601434079687, - "loss": 1.4725, - "step": 22955 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922356761144558, - "loss": 1.5168, - "step": 22956 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019223533788811475, - "loss": 1.5116, - "step": 22957 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019223499966177366, - "loss": 1.7078, - "step": 22958 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922346614354326, - "loss": 1.6542, - "step": 22959 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019223432320909155, - "loss": 1.5389, - "step": 22960 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019223398498275046, - "loss": 1.5315, - "step": 22961 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922336467564094, - "loss": 1.5625, - "step": 22962 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019223330853006834, - "loss": 1.5954, - "step": 22963 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019223297030372726, - "loss": 1.5978, - "step": 22964 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922326320773862, - "loss": 1.6912, - "step": 22965 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019223229385104514, - "loss": 1.5308, - "step": 22966 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019223195562470405, - "loss": 1.5386, - "step": 22967 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192231617398363, - "loss": 1.3944, - "step": 22968 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019223127917202194, - "loss": 1.6426, - "step": 22969 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019223094094568085, - "loss": 1.6467, - "step": 22970 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922306027193398, - "loss": 1.6744, - "step": 22971 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019223026449299873, - "loss": 1.4735, - "step": 22972 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019222992626665765, - "loss": 1.664, - "step": 22973 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922295880403166, - "loss": 1.6157, - "step": 22974 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019222924981397553, - "loss": 1.5811, - "step": 22975 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019222891158763444, - "loss": 1.6018, - "step": 22976 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019222857336129338, - "loss": 1.4649, - "step": 22977 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019222823513495232, - "loss": 1.5548, - "step": 22978 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019222789690861124, - "loss": 1.4824, - "step": 22979 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019222755868227018, - "loss": 1.553, - "step": 22980 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019222722045592912, - "loss": 1.51, - "step": 22981 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019222688222958803, - "loss": 1.5467, - "step": 22982 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019222654400324698, - "loss": 1.3586, - "step": 22983 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019222620577690592, - "loss": 1.5381, - "step": 22984 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019222586755056483, - "loss": 1.6004, - "step": 22985 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019222552932422377, - "loss": 1.5899, - "step": 22986 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922251910978827, - "loss": 1.6794, - "step": 22987 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019222485287154163, - "loss": 1.5102, - "step": 22988 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019222451464520057, - "loss": 1.564, - "step": 22989 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922241764188595, - "loss": 1.5299, - "step": 22990 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019222383819251842, - "loss": 1.5201, - "step": 22991 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019222349996617736, - "loss": 1.5465, - "step": 22992 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922231617398363, - "loss": 1.3715, - "step": 22993 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019222282351349522, - "loss": 1.4994, - "step": 22994 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019222248528715416, - "loss": 1.5898, - "step": 22995 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922221470608131, - "loss": 1.5714, - "step": 22996 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019222180883447202, - "loss": 1.6203, - "step": 22997 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019222147060813096, - "loss": 1.5695, - "step": 22998 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922211323817899, - "loss": 1.5355, - "step": 22999 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922207941554488, - "loss": 1.6115, - "step": 23000 - }, - { - "epoch": 0.08, - "eval_loss": 1.3640512319711537, - "mrc_em": 22.0, - "mrc_f1": 30.9, - "nli_acc": 0.11, - "step": 23000, - "sts_f1": 0.6766917293233083, - "ynat_f1": 0.6243784963378064 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019222045592910778, - "loss": 1.6353, - "step": 23001 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019222011770276672, - "loss": 1.4297, - "step": 23002 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019221977947642564, - "loss": 1.6503, - "step": 23003 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019221944125008458, - "loss": 1.4654, - "step": 23004 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019221910302374352, - "loss": 1.6265, - "step": 23005 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019221876479740243, - "loss": 1.5844, - "step": 23006 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019221842657106137, - "loss": 1.658, - "step": 23007 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019221808834472031, - "loss": 1.6109, - "step": 23008 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019221775011837923, - "loss": 1.4081, - "step": 23009 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019221741189203817, - "loss": 1.3244, - "step": 23010 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922170736656971, - "loss": 1.4981, - "step": 23011 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019221673543935602, - "loss": 1.4494, - "step": 23012 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019221639721301497, - "loss": 1.517, - "step": 23013 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922160589866739, - "loss": 1.5488, - "step": 23014 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019221572076033282, - "loss": 1.5579, - "step": 23015 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019221538253399176, - "loss": 1.6052, - "step": 23016 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922150443076507, - "loss": 1.6946, - "step": 23017 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019221470608130962, - "loss": 1.561, - "step": 23018 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019221436785496856, - "loss": 1.5894, - "step": 23019 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922140296286275, - "loss": 1.549, - "step": 23020 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019221369140228641, - "loss": 1.5067, - "step": 23021 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019221335317594536, - "loss": 1.6482, - "step": 23022 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922130149496043, - "loss": 1.5145, - "step": 23023 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922126767232632, - "loss": 1.7821, - "step": 23024 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019221233849692215, - "loss": 1.5912, - "step": 23025 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922120002705811, - "loss": 1.5826, - "step": 23026 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019221166204424, - "loss": 1.5454, - "step": 23027 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019221132381789895, - "loss": 1.6077, - "step": 23028 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922109855915579, - "loss": 1.5579, - "step": 23029 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922106473652168, - "loss": 1.5394, - "step": 23030 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019221030913887574, - "loss": 1.4933, - "step": 23031 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019220997091253469, - "loss": 1.4145, - "step": 23032 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922096326861936, - "loss": 1.4783, - "step": 23033 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019220929445985254, - "loss": 1.485, - "step": 23034 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019220895623351148, - "loss": 1.6829, - "step": 23035 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922086180071704, - "loss": 1.5486, - "step": 23036 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019220827978082934, - "loss": 1.5785, - "step": 23037 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019220794155448828, - "loss": 1.4568, - "step": 23038 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922076033281472, - "loss": 1.5455, - "step": 23039 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019220726510180613, - "loss": 1.697, - "step": 23040 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019220692687546507, - "loss": 1.4307, - "step": 23041 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192206588649124, - "loss": 1.4544, - "step": 23042 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019220625042278293, - "loss": 1.677, - "step": 23043 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019220591219644187, - "loss": 1.6001, - "step": 23044 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019220557397010078, - "loss": 1.6849, - "step": 23045 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019220523574375973, - "loss": 1.5057, - "step": 23046 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019220489751741867, - "loss": 1.4605, - "step": 23047 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019220455929107758, - "loss": 1.4994, - "step": 23048 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019220422106473652, - "loss": 1.4679, - "step": 23049 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019220388283839546, - "loss": 1.6905, - "step": 23050 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019220354461205438, - "loss": 1.4248, - "step": 23051 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019220320638571332, - "loss": 1.6982, - "step": 23052 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019220286815937226, - "loss": 1.6435, - "step": 23053 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019220252993303117, - "loss": 1.6121, - "step": 23054 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019220219170669011, - "loss": 1.6088, - "step": 23055 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019220185348034906, - "loss": 1.5481, - "step": 23056 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192201515254008, - "loss": 1.4467, - "step": 23057 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019220117702766694, - "loss": 1.6969, - "step": 23058 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019220083880132588, - "loss": 1.64, - "step": 23059 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001922005005749848, - "loss": 1.5031, - "step": 23060 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019220016234864373, - "loss": 1.6574, - "step": 23061 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019219982412230268, - "loss": 1.5207, - "step": 23062 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921994858959616, - "loss": 1.4635, - "step": 23063 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019219914766962053, - "loss": 1.5957, - "step": 23064 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019219880944327947, - "loss": 1.3999, - "step": 23065 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019219847121693839, - "loss": 1.45, - "step": 23066 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019219813299059733, - "loss": 1.5497, - "step": 23067 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019219779476425627, - "loss": 1.5319, - "step": 23068 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019219745653791518, - "loss": 1.6472, - "step": 23069 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019219711831157412, - "loss": 1.5171, - "step": 23070 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019219678008523306, - "loss": 1.4335, - "step": 23071 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019219644185889198, - "loss": 1.5183, - "step": 23072 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019219610363255092, - "loss": 1.6057, - "step": 23073 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019219576540620986, - "loss": 1.6834, - "step": 23074 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019219542717986877, - "loss": 1.5047, - "step": 23075 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019219508895352772, - "loss": 1.4231, - "step": 23076 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019219475072718666, - "loss": 1.6513, - "step": 23077 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019219441250084557, - "loss": 1.4725, - "step": 23078 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921940742745045, - "loss": 1.5343, - "step": 23079 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019219373604816345, - "loss": 1.5111, - "step": 23080 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019219339782182237, - "loss": 1.6581, - "step": 23081 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921930595954813, - "loss": 1.4885, - "step": 23082 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019219272136914025, - "loss": 1.5128, - "step": 23083 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019219238314279916, - "loss": 1.5899, - "step": 23084 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921920449164581, - "loss": 1.513, - "step": 23085 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019219170669011705, - "loss": 1.3903, - "step": 23086 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019219136846377596, - "loss": 1.6369, - "step": 23087 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921910302374349, - "loss": 1.6761, - "step": 23088 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019219069201109384, - "loss": 1.5966, - "step": 23089 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019219035378475276, - "loss": 1.6202, - "step": 23090 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921900155584117, - "loss": 1.5589, - "step": 23091 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019218967733207064, - "loss": 1.5151, - "step": 23092 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019218933910572955, - "loss": 1.7109, - "step": 23093 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921890008793885, - "loss": 1.5379, - "step": 23094 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019218866265304743, - "loss": 1.6408, - "step": 23095 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019218832442670635, - "loss": 1.6151, - "step": 23096 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921879862003653, - "loss": 1.6474, - "step": 23097 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019218764797402423, - "loss": 1.5159, - "step": 23098 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019218730974768315, - "loss": 1.7084, - "step": 23099 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019218697152134209, - "loss": 1.6104, - "step": 23100 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019218663329500103, - "loss": 1.5334, - "step": 23101 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019218629506865994, - "loss": 1.5438, - "step": 23102 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019218595684231888, - "loss": 1.6919, - "step": 23103 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019218561861597782, - "loss": 1.5608, - "step": 23104 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019218528038963674, - "loss": 1.4589, - "step": 23105 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019218494216329568, - "loss": 1.4403, - "step": 23106 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019218460393695462, - "loss": 1.5676, - "step": 23107 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019218426571061353, - "loss": 1.4637, - "step": 23108 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019218392748427248, - "loss": 1.5217, - "step": 23109 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019218358925793142, - "loss": 1.6897, - "step": 23110 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019218325103159033, - "loss": 1.7182, - "step": 23111 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019218291280524927, - "loss": 1.6666, - "step": 23112 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921825745789082, - "loss": 1.5052, - "step": 23113 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019218223635256715, - "loss": 1.4389, - "step": 23114 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921818981262261, - "loss": 1.6331, - "step": 23115 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192181559899885, - "loss": 1.67, - "step": 23116 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019218122167354395, - "loss": 1.4347, - "step": 23117 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921808834472029, - "loss": 1.6507, - "step": 23118 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921805452208618, - "loss": 1.441, - "step": 23119 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019218020699452075, - "loss": 1.4727, - "step": 23120 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921798687681797, - "loss": 1.6378, - "step": 23121 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921795305418386, - "loss": 1.6272, - "step": 23122 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019217919231549754, - "loss": 1.6347, - "step": 23123 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019217885408915648, - "loss": 1.695, - "step": 23124 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921785158628154, - "loss": 1.5082, - "step": 23125 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019217817763647434, - "loss": 1.3847, - "step": 23126 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019217783941013328, - "loss": 1.6193, - "step": 23127 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019217750118379222, - "loss": 1.6381, - "step": 23128 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019217716295745114, - "loss": 1.6083, - "step": 23129 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019217682473111008, - "loss": 1.6499, - "step": 23130 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019217648650476902, - "loss": 1.6613, - "step": 23131 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019217614827842793, - "loss": 1.6456, - "step": 23132 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019217581005208687, - "loss": 1.6636, - "step": 23133 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019217547182574581, - "loss": 1.3848, - "step": 23134 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019217513359940473, - "loss": 1.5854, - "step": 23135 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019217479537306367, - "loss": 1.5825, - "step": 23136 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921744571467226, - "loss": 1.4252, - "step": 23137 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019217411892038152, - "loss": 1.4995, - "step": 23138 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019217378069404047, - "loss": 1.5479, - "step": 23139 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921734424676994, - "loss": 1.5381, - "step": 23140 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019217310424135832, - "loss": 1.5572, - "step": 23141 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019217276601501726, - "loss": 1.7422, - "step": 23142 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921724277886762, - "loss": 1.5035, - "step": 23143 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019217208956233512, - "loss": 1.4457, - "step": 23144 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019217175133599406, - "loss": 1.5657, - "step": 23145 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192171413109653, - "loss": 1.5083, - "step": 23146 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921710748833119, - "loss": 1.386, - "step": 23147 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019217073665697085, - "loss": 1.5321, - "step": 23148 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921703984306298, - "loss": 1.5584, - "step": 23149 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921700602042887, - "loss": 1.5267, - "step": 23150 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019216972197794765, - "loss": 1.6122, - "step": 23151 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921693837516066, - "loss": 1.424, - "step": 23152 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921690455252655, - "loss": 1.4806, - "step": 23153 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019216870729892445, - "loss": 1.6537, - "step": 23154 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921683690725834, - "loss": 1.2378, - "step": 23155 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921680308462423, - "loss": 1.3411, - "step": 23156 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019216769261990124, - "loss": 1.5842, - "step": 23157 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019216735439356018, - "loss": 1.48, - "step": 23158 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921670161672191, - "loss": 1.5815, - "step": 23159 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019216667794087804, - "loss": 1.5955, - "step": 23160 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019216633971453698, - "loss": 1.6488, - "step": 23161 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921660014881959, - "loss": 1.655, - "step": 23162 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019216566326185484, - "loss": 1.682, - "step": 23163 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019216532503551378, - "loss": 1.4383, - "step": 23164 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921649868091727, - "loss": 1.4333, - "step": 23165 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019216464858283163, - "loss": 1.4998, - "step": 23166 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019216431035649057, - "loss": 1.669, - "step": 23167 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921639721301495, - "loss": 1.6483, - "step": 23168 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019216363390380846, - "loss": 1.4829, - "step": 23169 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019216329567746737, - "loss": 1.5793, - "step": 23170 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921629574511263, - "loss": 1.6374, - "step": 23171 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019216261922478525, - "loss": 1.6587, - "step": 23172 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019216228099844417, - "loss": 1.5796, - "step": 23173 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921619427721031, - "loss": 1.5192, - "step": 23174 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019216160454576205, - "loss": 1.7152, - "step": 23175 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019216126631942096, - "loss": 1.6301, - "step": 23176 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921609280930799, - "loss": 1.6245, - "step": 23177 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019216058986673884, - "loss": 1.637, - "step": 23178 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019216025164039776, - "loss": 1.5899, - "step": 23179 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921599134140567, - "loss": 1.4034, - "step": 23180 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019215957518771564, - "loss": 1.5838, - "step": 23181 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019215923696137456, - "loss": 1.6643, - "step": 23182 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921588987350335, - "loss": 1.6111, - "step": 23183 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019215856050869244, - "loss": 1.4122, - "step": 23184 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019215822228235135, - "loss": 1.5486, - "step": 23185 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921578840560103, - "loss": 1.6042, - "step": 23186 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019215754582966923, - "loss": 1.5867, - "step": 23187 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019215720760332815, - "loss": 1.5683, - "step": 23188 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921568693769871, - "loss": 1.5745, - "step": 23189 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019215653115064603, - "loss": 1.6303, - "step": 23190 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019215619292430494, - "loss": 1.4678, - "step": 23191 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019215585469796389, - "loss": 1.5613, - "step": 23192 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019215551647162283, - "loss": 1.6526, - "step": 23193 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019215517824528174, - "loss": 1.6562, - "step": 23194 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019215484001894068, - "loss": 1.4912, - "step": 23195 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019215450179259962, - "loss": 1.5153, - "step": 23196 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019215416356625854, - "loss": 1.6127, - "step": 23197 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019215382533991748, - "loss": 1.6134, - "step": 23198 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019215348711357642, - "loss": 1.6448, - "step": 23199 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019215314888723533, - "loss": 1.4537, - "step": 23200 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019215281066089427, - "loss": 1.6258, - "step": 23201 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019215247243455322, - "loss": 1.6311, - "step": 23202 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019215213420821213, - "loss": 1.5392, - "step": 23203 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019215179598187107, - "loss": 1.5945, - "step": 23204 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019215145775553, - "loss": 1.7458, - "step": 23205 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019215111952918893, - "loss": 1.3969, - "step": 23206 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019215078130284787, - "loss": 1.5206, - "step": 23207 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921504430765068, - "loss": 1.6578, - "step": 23208 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019215010485016572, - "loss": 1.5113, - "step": 23209 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019214976662382466, - "loss": 1.446, - "step": 23210 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921494283974836, - "loss": 1.4246, - "step": 23211 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019214909017114255, - "loss": 1.5794, - "step": 23212 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019214875194480146, - "loss": 1.59, - "step": 23213 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921484137184604, - "loss": 1.55, - "step": 23214 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019214807549211934, - "loss": 1.5916, - "step": 23215 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019214773726577826, - "loss": 1.3514, - "step": 23216 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921473990394372, - "loss": 1.3684, - "step": 23217 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019214706081309614, - "loss": 1.544, - "step": 23218 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019214672258675505, - "loss": 1.6265, - "step": 23219 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192146384360414, - "loss": 1.4945, - "step": 23220 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019214604613407293, - "loss": 1.4079, - "step": 23221 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019214570790773185, - "loss": 1.6002, - "step": 23222 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921453696813908, - "loss": 1.5479, - "step": 23223 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019214503145504973, - "loss": 1.4577, - "step": 23224 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019214469322870867, - "loss": 1.4415, - "step": 23225 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921443550023676, - "loss": 1.5155, - "step": 23226 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019214401677602653, - "loss": 1.739, - "step": 23227 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019214367854968547, - "loss": 1.5857, - "step": 23228 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921433403233444, - "loss": 1.7624, - "step": 23229 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019214300209700332, - "loss": 1.488, - "step": 23230 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019214266387066226, - "loss": 1.8454, - "step": 23231 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921423256443212, - "loss": 1.4969, - "step": 23232 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019214198741798012, - "loss": 1.5008, - "step": 23233 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019214164919163906, - "loss": 1.5616, - "step": 23234 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192141310965298, - "loss": 1.5835, - "step": 23235 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019214097273895692, - "loss": 1.5381, - "step": 23236 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019214063451261586, - "loss": 1.5635, - "step": 23237 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921402962862748, - "loss": 1.5734, - "step": 23238 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921399580599337, - "loss": 1.4138, - "step": 23239 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019213961983359265, - "loss": 1.6238, - "step": 23240 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921392816072516, - "loss": 1.5423, - "step": 23241 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921389433809105, - "loss": 1.4694, - "step": 23242 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019213860515456945, - "loss": 1.5525, - "step": 23243 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921382669282284, - "loss": 1.5825, - "step": 23244 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921379287018873, - "loss": 1.4942, - "step": 23245 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019213759047554625, - "loss": 1.5583, - "step": 23246 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921372522492052, - "loss": 1.6822, - "step": 23247 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921369140228641, - "loss": 1.751, - "step": 23248 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019213657579652304, - "loss": 1.5355, - "step": 23249 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019213623757018198, - "loss": 1.6968, - "step": 23250 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921358993438409, - "loss": 1.3802, - "step": 23251 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019213556111749984, - "loss": 1.6654, - "step": 23252 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019213522289115878, - "loss": 1.5247, - "step": 23253 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921348846648177, - "loss": 1.6683, - "step": 23254 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019213454643847663, - "loss": 1.4789, - "step": 23255 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019213420821213558, - "loss": 1.404, - "step": 23256 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921338699857945, - "loss": 1.6968, - "step": 23257 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019213353175945343, - "loss": 1.5309, - "step": 23258 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019213319353311237, - "loss": 1.5494, - "step": 23259 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019213285530677129, - "loss": 1.5999, - "step": 23260 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019213251708043023, - "loss": 1.4853, - "step": 23261 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019213217885408917, - "loss": 1.5627, - "step": 23262 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019213184062774808, - "loss": 1.6548, - "step": 23263 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019213150240140702, - "loss": 1.683, - "step": 23264 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019213116417506596, - "loss": 1.4325, - "step": 23265 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019213082594872488, - "loss": 1.7076, - "step": 23266 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019213048772238382, - "loss": 1.6172, - "step": 23267 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019213014949604276, - "loss": 1.5364, - "step": 23268 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019212981126970168, - "loss": 1.7672, - "step": 23269 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019212947304336062, - "loss": 1.476, - "step": 23270 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019212913481701956, - "loss": 1.6351, - "step": 23271 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019212879659067847, - "loss": 1.7231, - "step": 23272 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921284583643374, - "loss": 1.6708, - "step": 23273 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019212812013799635, - "loss": 1.6041, - "step": 23274 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019212778191165527, - "loss": 1.5779, - "step": 23275 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921274436853142, - "loss": 1.5171, - "step": 23276 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019212710545897315, - "loss": 1.6099, - "step": 23277 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019212676723263206, - "loss": 1.5972, - "step": 23278 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192126429006291, - "loss": 1.4254, - "step": 23279 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019212609077994995, - "loss": 1.6729, - "step": 23280 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921257525536089, - "loss": 1.537, - "step": 23281 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019212541432726783, - "loss": 1.6861, - "step": 23282 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019212507610092677, - "loss": 1.6175, - "step": 23283 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019212473787458568, - "loss": 1.5151, - "step": 23284 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019212439964824463, - "loss": 1.5792, - "step": 23285 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019212406142190357, - "loss": 1.6026, - "step": 23286 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019212372319556248, - "loss": 1.5514, - "step": 23287 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019212338496922142, - "loss": 1.5715, - "step": 23288 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019212304674288036, - "loss": 1.5871, - "step": 23289 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019212270851653928, - "loss": 1.5431, - "step": 23290 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019212237029019822, - "loss": 1.5351, - "step": 23291 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019212203206385716, - "loss": 1.7914, - "step": 23292 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019212169383751607, - "loss": 1.4444, - "step": 23293 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019212135561117501, - "loss": 1.7534, - "step": 23294 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019212101738483396, - "loss": 1.6326, - "step": 23295 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019212067915849287, - "loss": 1.5356, - "step": 23296 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921203409321518, - "loss": 1.5372, - "step": 23297 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019212000270581075, - "loss": 1.5882, - "step": 23298 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019211966447946967, - "loss": 1.591, - "step": 23299 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921193262531286, - "loss": 1.5156, - "step": 23300 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019211898802678755, - "loss": 1.5412, - "step": 23301 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019211864980044646, - "loss": 1.579, - "step": 23302 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921183115741054, - "loss": 1.5552, - "step": 23303 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019211797334776434, - "loss": 1.5734, - "step": 23304 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019211763512142326, - "loss": 1.6161, - "step": 23305 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921172968950822, - "loss": 1.5405, - "step": 23306 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019211695866874114, - "loss": 1.343, - "step": 23307 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019211662044240005, - "loss": 1.5715, - "step": 23308 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192116282216059, - "loss": 1.518, - "step": 23309 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019211594398971794, - "loss": 1.5591, - "step": 23310 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019211560576337685, - "loss": 1.5999, - "step": 23311 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921152675370358, - "loss": 1.4356, - "step": 23312 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019211492931069473, - "loss": 1.6276, - "step": 23313 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019211459108435365, - "loss": 1.4778, - "step": 23314 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921142528580126, - "loss": 1.6416, - "step": 23315 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019211391463167153, - "loss": 1.4528, - "step": 23316 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019211357640533044, - "loss": 1.5455, - "step": 23317 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019211323817898938, - "loss": 1.5262, - "step": 23318 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019211289995264833, - "loss": 1.568, - "step": 23319 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019211256172630724, - "loss": 1.3665, - "step": 23320 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019211222349996618, - "loss": 1.6761, - "step": 23321 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019211188527362512, - "loss": 1.4658, - "step": 23322 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019211154704728404, - "loss": 1.59, - "step": 23323 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019211120882094298, - "loss": 1.5343, - "step": 23324 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019211087059460192, - "loss": 1.4522, - "step": 23325 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019211053236826083, - "loss": 1.493, - "step": 23326 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019211019414191977, - "loss": 1.6406, - "step": 23327 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019210985591557871, - "loss": 1.5534, - "step": 23328 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019210951768923763, - "loss": 1.6779, - "step": 23329 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019210917946289657, - "loss": 1.4745, - "step": 23330 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921088412365555, - "loss": 1.5713, - "step": 23331 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019210850301021442, - "loss": 1.6211, - "step": 23332 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019210816478387337, - "loss": 1.3901, - "step": 23333 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921078265575323, - "loss": 1.5118, - "step": 23334 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019210748833119122, - "loss": 1.712, - "step": 23335 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019210715010485016, - "loss": 1.6992, - "step": 23336 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921068118785091, - "loss": 1.5209, - "step": 23337 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019210647365216804, - "loss": 1.6696, - "step": 23338 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019210613542582699, - "loss": 1.5033, - "step": 23339 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019210579719948593, - "loss": 1.6691, - "step": 23340 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019210545897314484, - "loss": 1.6045, - "step": 23341 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019210512074680378, - "loss": 1.6404, - "step": 23342 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019210478252046272, - "loss": 1.4781, - "step": 23343 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019210444429412164, - "loss": 1.5536, - "step": 23344 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019210410606778058, - "loss": 1.7993, - "step": 23345 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019210376784143952, - "loss": 1.3963, - "step": 23346 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019210342961509843, - "loss": 1.5033, - "step": 23347 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019210309138875737, - "loss": 1.5959, - "step": 23348 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019210275316241632, - "loss": 1.4692, - "step": 23349 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019210241493607523, - "loss": 1.4974, - "step": 23350 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019210207670973417, - "loss": 1.5698, - "step": 23351 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921017384833931, - "loss": 1.6566, - "step": 23352 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019210140025705203, - "loss": 1.593, - "step": 23353 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019210106203071097, - "loss": 1.5987, - "step": 23354 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001921007238043699, - "loss": 1.7002, - "step": 23355 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019210038557802882, - "loss": 1.5005, - "step": 23356 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019210004735168776, - "loss": 1.5959, - "step": 23357 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920997091253467, - "loss": 1.5483, - "step": 23358 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019209937089900562, - "loss": 1.6409, - "step": 23359 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019209903267266456, - "loss": 1.5947, - "step": 23360 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920986944463235, - "loss": 1.6364, - "step": 23361 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019209835621998242, - "loss": 1.7054, - "step": 23362 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019209801799364136, - "loss": 1.6291, - "step": 23363 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920976797673003, - "loss": 1.5309, - "step": 23364 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920973415409592, - "loss": 1.6276, - "step": 23365 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019209700331461815, - "loss": 1.7313, - "step": 23366 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920966650882771, - "loss": 1.6177, - "step": 23367 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192096326861936, - "loss": 1.4203, - "step": 23368 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019209598863559495, - "loss": 1.6725, - "step": 23369 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920956504092539, - "loss": 1.4875, - "step": 23370 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920953121829128, - "loss": 1.7214, - "step": 23371 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019209497395657175, - "loss": 1.5877, - "step": 23372 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019209463573023069, - "loss": 1.605, - "step": 23373 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920942975038896, - "loss": 1.5032, - "step": 23374 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019209395927754854, - "loss": 1.6294, - "step": 23375 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019209362105120748, - "loss": 1.6334, - "step": 23376 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920932828248664, - "loss": 1.6063, - "step": 23377 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019209294459852534, - "loss": 1.5873, - "step": 23378 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019209260637218428, - "loss": 1.3783, - "step": 23379 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920922681458432, - "loss": 1.6565, - "step": 23380 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019209192991950213, - "loss": 1.391, - "step": 23381 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019209159169316108, - "loss": 1.495, - "step": 23382 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019209125346682, - "loss": 1.7886, - "step": 23383 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019209091524047893, - "loss": 1.5875, - "step": 23384 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019209057701413787, - "loss": 1.5607, - "step": 23385 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019209023878779679, - "loss": 1.6268, - "step": 23386 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019208990056145573, - "loss": 1.4067, - "step": 23387 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019208956233511467, - "loss": 1.4928, - "step": 23388 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019208922410877358, - "loss": 1.4303, - "step": 23389 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019208888588243252, - "loss": 1.4372, - "step": 23390 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019208854765609146, - "loss": 1.6223, - "step": 23391 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019208820942975038, - "loss": 1.4956, - "step": 23392 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019208787120340932, - "loss": 1.6654, - "step": 23393 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019208753297706826, - "loss": 1.6909, - "step": 23394 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920871947507272, - "loss": 1.6702, - "step": 23395 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019208685652438614, - "loss": 1.6166, - "step": 23396 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019208651829804508, - "loss": 1.7355, - "step": 23397 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192086180071704, - "loss": 1.5732, - "step": 23398 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019208584184536294, - "loss": 1.5746, - "step": 23399 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019208550361902188, - "loss": 1.4341, - "step": 23400 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920851653926808, - "loss": 1.6552, - "step": 23401 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019208482716633974, - "loss": 1.4477, - "step": 23402 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019208448893999868, - "loss": 1.4155, - "step": 23403 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920841507136576, - "loss": 1.4798, - "step": 23404 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019208381248731653, - "loss": 1.7575, - "step": 23405 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019208347426097547, - "loss": 1.6807, - "step": 23406 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920831360346344, - "loss": 1.537, - "step": 23407 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019208279780829333, - "loss": 1.5677, - "step": 23408 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019208245958195227, - "loss": 1.6461, - "step": 23409 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019208212135561118, - "loss": 1.56, - "step": 23410 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019208178312927012, - "loss": 1.5584, - "step": 23411 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019208144490292907, - "loss": 1.5127, - "step": 23412 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019208110667658798, - "loss": 1.3946, - "step": 23413 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019208076845024692, - "loss": 1.412, - "step": 23414 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019208043022390586, - "loss": 1.449, - "step": 23415 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019208009199756478, - "loss": 1.4588, - "step": 23416 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019207975377122372, - "loss": 1.6259, - "step": 23417 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019207941554488266, - "loss": 1.4841, - "step": 23418 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019207907731854157, - "loss": 1.6017, - "step": 23419 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920787390922005, - "loss": 1.5735, - "step": 23420 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019207840086585945, - "loss": 1.6106, - "step": 23421 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019207806263951837, - "loss": 1.5171, - "step": 23422 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920777244131773, - "loss": 1.5769, - "step": 23423 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019207738618683625, - "loss": 1.6289, - "step": 23424 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019207704796049516, - "loss": 1.5732, - "step": 23425 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920767097341541, - "loss": 1.5427, - "step": 23426 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019207637150781305, - "loss": 1.7156, - "step": 23427 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019207603328147196, - "loss": 1.4804, - "step": 23428 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920756950551309, - "loss": 1.4903, - "step": 23429 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019207535682878984, - "loss": 1.5345, - "step": 23430 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019207501860244876, - "loss": 1.745, - "step": 23431 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920746803761077, - "loss": 1.4838, - "step": 23432 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019207434214976664, - "loss": 1.6199, - "step": 23433 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019207400392342555, - "loss": 1.5477, - "step": 23434 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920736656970845, - "loss": 1.6713, - "step": 23435 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019207332747074344, - "loss": 1.6216, - "step": 23436 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019207298924440235, - "loss": 1.4859, - "step": 23437 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920726510180613, - "loss": 1.6226, - "step": 23438 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019207231279172023, - "loss": 1.42, - "step": 23439 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019207197456537915, - "loss": 1.5477, - "step": 23440 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920716363390381, - "loss": 1.7406, - "step": 23441 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019207129811269703, - "loss": 1.5396, - "step": 23442 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019207095988635594, - "loss": 1.625, - "step": 23443 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019207062166001488, - "loss": 1.5793, - "step": 23444 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019207028343367383, - "loss": 1.6577, - "step": 23445 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019206994520733274, - "loss": 1.6512, - "step": 23446 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019206960698099168, - "loss": 1.6823, - "step": 23447 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019206926875465062, - "loss": 1.6356, - "step": 23448 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019206893052830954, - "loss": 1.7385, - "step": 23449 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920685923019685, - "loss": 1.4796, - "step": 23450 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019206825407562742, - "loss": 1.7145, - "step": 23451 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019206791584928636, - "loss": 1.6346, - "step": 23452 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920675776229453, - "loss": 1.5265, - "step": 23453 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019206723939660421, - "loss": 1.7662, - "step": 23454 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019206690117026316, - "loss": 1.6603, - "step": 23455 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920665629439221, - "loss": 1.5442, - "step": 23456 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192066224717581, - "loss": 1.4895, - "step": 23457 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019206588649123995, - "loss": 1.5736, - "step": 23458 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920655482648989, - "loss": 1.3556, - "step": 23459 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920652100385578, - "loss": 1.548, - "step": 23460 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019206487181221675, - "loss": 1.582, - "step": 23461 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920645335858757, - "loss": 1.5884, - "step": 23462 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920641953595346, - "loss": 1.6295, - "step": 23463 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019206385713319354, - "loss": 1.5917, - "step": 23464 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019206351890685249, - "loss": 1.601, - "step": 23465 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920631806805114, - "loss": 1.5453, - "step": 23466 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019206284245417034, - "loss": 1.4945, - "step": 23467 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019206250422782928, - "loss": 1.5443, - "step": 23468 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920621660014882, - "loss": 1.5857, - "step": 23469 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019206182777514714, - "loss": 1.4021, - "step": 23470 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019206148954880608, - "loss": 1.4813, - "step": 23471 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192061151322465, - "loss": 1.5343, - "step": 23472 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019206081309612393, - "loss": 1.457, - "step": 23473 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019206047486978287, - "loss": 1.6047, - "step": 23474 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920601366434418, - "loss": 1.4287, - "step": 23475 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019205979841710073, - "loss": 1.4758, - "step": 23476 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019205946019075967, - "loss": 1.5184, - "step": 23477 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019205912196441858, - "loss": 1.4981, - "step": 23478 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019205878373807753, - "loss": 1.6389, - "step": 23479 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019205844551173647, - "loss": 1.5726, - "step": 23480 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920581072853954, - "loss": 1.6329, - "step": 23481 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019205776905905432, - "loss": 1.5129, - "step": 23482 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019205743083271326, - "loss": 1.5921, - "step": 23483 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920570926063722, - "loss": 1.4188, - "step": 23484 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019205675438003112, - "loss": 1.6251, - "step": 23485 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019205641615369006, - "loss": 1.612, - "step": 23486 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192056077927349, - "loss": 1.5111, - "step": 23487 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019205573970100791, - "loss": 1.5392, - "step": 23488 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019205540147466686, - "loss": 1.3305, - "step": 23489 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920550632483258, - "loss": 1.538, - "step": 23490 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920547250219847, - "loss": 1.4637, - "step": 23491 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019205438679564365, - "loss": 1.6332, - "step": 23492 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920540485693026, - "loss": 1.5483, - "step": 23493 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920537103429615, - "loss": 1.5348, - "step": 23494 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019205337211662045, - "loss": 1.5993, - "step": 23495 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920530338902794, - "loss": 1.5917, - "step": 23496 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920526956639383, - "loss": 1.511, - "step": 23497 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019205235743759724, - "loss": 1.5398, - "step": 23498 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019205201921125619, - "loss": 1.434, - "step": 23499 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920516809849151, - "loss": 1.7516, - "step": 23500 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019205134275857404, - "loss": 1.4984, - "step": 23501 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019205100453223298, - "loss": 1.5699, - "step": 23502 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920506663058919, - "loss": 1.5013, - "step": 23503 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019205032807955084, - "loss": 1.4691, - "step": 23504 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019204998985320978, - "loss": 1.3432, - "step": 23505 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019204965162686872, - "loss": 1.5062, - "step": 23506 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019204931340052766, - "loss": 1.6289, - "step": 23507 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019204897517418657, - "loss": 1.5747, - "step": 23508 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019204863694784552, - "loss": 1.5001, - "step": 23509 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019204829872150446, - "loss": 1.4367, - "step": 23510 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019204796049516337, - "loss": 1.4108, - "step": 23511 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920476222688223, - "loss": 1.4285, - "step": 23512 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019204728404248125, - "loss": 1.4906, - "step": 23513 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019204694581614017, - "loss": 1.6742, - "step": 23514 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920466075897991, - "loss": 1.5761, - "step": 23515 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019204626936345805, - "loss": 1.6216, - "step": 23516 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019204593113711696, - "loss": 1.6805, - "step": 23517 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920455929107759, - "loss": 1.6186, - "step": 23518 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019204525468443485, - "loss": 1.6487, - "step": 23519 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019204491645809376, - "loss": 1.6071, - "step": 23520 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920445782317527, - "loss": 1.4916, - "step": 23521 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019204424000541164, - "loss": 1.5421, - "step": 23522 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019204390177907056, - "loss": 1.6081, - "step": 23523 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920435635527295, - "loss": 1.6274, - "step": 23524 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019204322532638844, - "loss": 1.483, - "step": 23525 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019204288710004735, - "loss": 1.6879, - "step": 23526 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920425488737063, - "loss": 1.5593, - "step": 23527 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019204221064736523, - "loss": 1.5566, - "step": 23528 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019204187242102415, - "loss": 1.4883, - "step": 23529 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920415341946831, - "loss": 1.5519, - "step": 23530 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019204119596834203, - "loss": 1.5608, - "step": 23531 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019204085774200095, - "loss": 1.4319, - "step": 23532 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019204051951565989, - "loss": 1.5136, - "step": 23533 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019204018128931883, - "loss": 1.5378, - "step": 23534 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019203984306297774, - "loss": 1.4933, - "step": 23535 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019203950483663668, - "loss": 1.5569, - "step": 23536 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019203916661029562, - "loss": 1.6364, - "step": 23537 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019203882838395454, - "loss": 1.4412, - "step": 23538 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019203849015761348, - "loss": 1.4721, - "step": 23539 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019203815193127242, - "loss": 1.6318, - "step": 23540 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019203781370493133, - "loss": 1.5136, - "step": 23541 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019203747547859028, - "loss": 1.5885, - "step": 23542 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019203713725224922, - "loss": 1.6402, - "step": 23543 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019203679902590813, - "loss": 1.4984, - "step": 23544 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019203646079956707, - "loss": 1.5128, - "step": 23545 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192036122573226, - "loss": 1.6054, - "step": 23546 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019203578434688493, - "loss": 1.5977, - "step": 23547 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019203544612054387, - "loss": 1.6085, - "step": 23548 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920351078942028, - "loss": 1.7965, - "step": 23549 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019203476966786172, - "loss": 1.5927, - "step": 23550 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019203443144152066, - "loss": 1.6554, - "step": 23551 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920340932151796, - "loss": 1.4281, - "step": 23552 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019203375498883852, - "loss": 1.5683, - "step": 23553 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019203341676249746, - "loss": 1.5809, - "step": 23554 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920330785361564, - "loss": 1.6663, - "step": 23555 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019203274030981532, - "loss": 1.5476, - "step": 23556 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019203240208347426, - "loss": 1.4512, - "step": 23557 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920320638571332, - "loss": 1.5587, - "step": 23558 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920317256307921, - "loss": 1.5155, - "step": 23559 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019203138740445105, - "loss": 1.5471, - "step": 23560 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019203104917811, - "loss": 1.586, - "step": 23561 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019203071095176894, - "loss": 1.6161, - "step": 23562 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019203037272542788, - "loss": 1.5438, - "step": 23563 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019203003449908682, - "loss": 1.688, - "step": 23564 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019202969627274573, - "loss": 1.6165, - "step": 23565 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019202935804640467, - "loss": 1.5063, - "step": 23566 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019202901982006361, - "loss": 1.5744, - "step": 23567 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019202868159372253, - "loss": 1.4231, - "step": 23568 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019202834336738147, - "loss": 1.4315, - "step": 23569 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920280051410404, - "loss": 1.4664, - "step": 23570 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019202766691469932, - "loss": 1.6548, - "step": 23571 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019202732868835827, - "loss": 1.6261, - "step": 23572 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920269904620172, - "loss": 1.6791, - "step": 23573 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019202665223567612, - "loss": 1.5832, - "step": 23574 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019202631400933506, - "loss": 1.5961, - "step": 23575 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192025975782994, - "loss": 1.4509, - "step": 23576 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019202563755665292, - "loss": 1.5453, - "step": 23577 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019202529933031186, - "loss": 1.5185, - "step": 23578 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920249611039708, - "loss": 1.6883, - "step": 23579 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920246228776297, - "loss": 1.3683, - "step": 23580 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019202428465128865, - "loss": 1.4414, - "step": 23581 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920239464249476, - "loss": 1.562, - "step": 23582 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920236081986065, - "loss": 1.6219, - "step": 23583 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019202326997226545, - "loss": 1.6124, - "step": 23584 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920229317459244, - "loss": 1.5541, - "step": 23585 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920225935195833, - "loss": 1.6647, - "step": 23586 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019202225529324225, - "loss": 1.5524, - "step": 23587 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920219170669012, - "loss": 1.5201, - "step": 23588 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920215788405601, - "loss": 1.5194, - "step": 23589 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019202124061421904, - "loss": 1.5654, - "step": 23590 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019202090238787798, - "loss": 1.4554, - "step": 23591 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920205641615369, - "loss": 1.5878, - "step": 23592 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019202022593519584, - "loss": 1.6298, - "step": 23593 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019201988770885478, - "loss": 1.6871, - "step": 23594 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920195494825137, - "loss": 1.5505, - "step": 23595 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019201921125617264, - "loss": 1.4818, - "step": 23596 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019201887302983158, - "loss": 1.4745, - "step": 23597 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920185348034905, - "loss": 1.5014, - "step": 23598 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019201819657714943, - "loss": 1.605, - "step": 23599 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019201785835080837, - "loss": 1.537, - "step": 23600 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920175201244673, - "loss": 1.3834, - "step": 23601 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019201718189812623, - "loss": 1.4963, - "step": 23602 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019201684367178517, - "loss": 1.6019, - "step": 23603 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019201650544544408, - "loss": 1.5495, - "step": 23604 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019201616721910302, - "loss": 1.6641, - "step": 23605 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019201582899276197, - "loss": 1.5773, - "step": 23606 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019201549076642088, - "loss": 1.5814, - "step": 23607 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019201515254007982, - "loss": 1.6354, - "step": 23608 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019201481431373876, - "loss": 1.5445, - "step": 23609 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019201447608739768, - "loss": 1.5905, - "step": 23610 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019201413786105662, - "loss": 1.6083, - "step": 23611 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019201379963471556, - "loss": 1.5319, - "step": 23612 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019201346140837447, - "loss": 1.5016, - "step": 23613 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019201312318203341, - "loss": 1.6372, - "step": 23614 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019201278495569236, - "loss": 1.3898, - "step": 23615 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019201244672935127, - "loss": 1.5639, - "step": 23616 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920121085030102, - "loss": 1.5185, - "step": 23617 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019201177027666918, - "loss": 1.5861, - "step": 23618 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920114320503281, - "loss": 1.7712, - "step": 23619 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019201109382398703, - "loss": 1.8132, - "step": 23620 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019201075559764597, - "loss": 1.5158, - "step": 23621 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920104173713049, - "loss": 1.5619, - "step": 23622 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019201007914496383, - "loss": 1.5771, - "step": 23623 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019200974091862277, - "loss": 1.5054, - "step": 23624 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019200940269228169, - "loss": 1.5161, - "step": 23625 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019200906446594063, - "loss": 1.6595, - "step": 23626 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019200872623959957, - "loss": 1.6452, - "step": 23627 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019200838801325848, - "loss": 1.603, - "step": 23628 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019200804978691742, - "loss": 1.6547, - "step": 23629 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019200771156057636, - "loss": 1.6631, - "step": 23630 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019200737333423528, - "loss": 1.4305, - "step": 23631 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019200703510789422, - "loss": 1.5468, - "step": 23632 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019200669688155316, - "loss": 1.4855, - "step": 23633 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019200635865521207, - "loss": 1.4506, - "step": 23634 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019200602042887102, - "loss": 1.5626, - "step": 23635 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019200568220252996, - "loss": 1.5877, - "step": 23636 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019200534397618887, - "loss": 1.4593, - "step": 23637 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920050057498478, - "loss": 1.587, - "step": 23638 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019200466752350675, - "loss": 1.5, - "step": 23639 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019200432929716567, - "loss": 1.5673, - "step": 23640 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920039910708246, - "loss": 1.528, - "step": 23641 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019200365284448355, - "loss": 1.5347, - "step": 23642 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019200331461814246, - "loss": 1.5388, - "step": 23643 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920029763918014, - "loss": 1.6323, - "step": 23644 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019200263816546035, - "loss": 1.5566, - "step": 23645 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019200229993911926, - "loss": 1.4976, - "step": 23646 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001920019617127782, - "loss": 1.6586, - "step": 23647 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019200162348643714, - "loss": 1.6433, - "step": 23648 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019200128526009606, - "loss": 1.6225, - "step": 23649 - }, - { - "epoch": 0.08, - "learning_rate": 0.000192000947033755, - "loss": 1.4968, - "step": 23650 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019200060880741394, - "loss": 1.6107, - "step": 23651 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019200027058107285, - "loss": 1.667, - "step": 23652 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919999323547318, - "loss": 1.6594, - "step": 23653 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019199959412839073, - "loss": 1.4006, - "step": 23654 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019199925590204965, - "loss": 1.5708, - "step": 23655 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919989176757086, - "loss": 1.5551, - "step": 23656 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019199857944936753, - "loss": 1.7695, - "step": 23657 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019199824122302644, - "loss": 1.5792, - "step": 23658 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019199790299668539, - "loss": 1.5627, - "step": 23659 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019199756477034433, - "loss": 1.6397, - "step": 23660 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019199722654400324, - "loss": 1.4375, - "step": 23661 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019199688831766218, - "loss": 1.6622, - "step": 23662 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019199655009132112, - "loss": 1.4882, - "step": 23663 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019199621186498004, - "loss": 1.4388, - "step": 23664 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019199587363863898, - "loss": 1.5919, - "step": 23665 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019199553541229792, - "loss": 1.4501, - "step": 23666 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019199519718595683, - "loss": 1.5217, - "step": 23667 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019199485895961577, - "loss": 1.6382, - "step": 23668 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019199452073327472, - "loss": 1.5558, - "step": 23669 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019199418250693363, - "loss": 1.5874, - "step": 23670 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019199384428059257, - "loss": 1.5319, - "step": 23671 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919935060542515, - "loss": 1.4614, - "step": 23672 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019199316782791043, - "loss": 1.5299, - "step": 23673 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019199282960156937, - "loss": 1.5818, - "step": 23674 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019199249137522834, - "loss": 1.4551, - "step": 23675 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019199215314888725, - "loss": 1.4359, - "step": 23676 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919918149225462, - "loss": 1.6434, - "step": 23677 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019199147669620513, - "loss": 1.5207, - "step": 23678 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019199113846986405, - "loss": 1.4952, - "step": 23679 - }, - { - "epoch": 0.08, - "learning_rate": 0.000191990800243523, - "loss": 1.529, - "step": 23680 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019199046201718193, - "loss": 1.754, - "step": 23681 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019199012379084084, - "loss": 1.4159, - "step": 23682 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019198978556449978, - "loss": 1.5515, - "step": 23683 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019198944733815872, - "loss": 1.466, - "step": 23684 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019198910911181764, - "loss": 1.6762, - "step": 23685 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019198877088547658, - "loss": 1.4602, - "step": 23686 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019198843265913552, - "loss": 1.562, - "step": 23687 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019198809443279443, - "loss": 1.5706, - "step": 23688 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019198775620645338, - "loss": 1.5671, - "step": 23689 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019198741798011232, - "loss": 1.4471, - "step": 23690 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019198707975377123, - "loss": 1.4827, - "step": 23691 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019198674152743017, - "loss": 1.4789, - "step": 23692 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919864033010891, - "loss": 1.535, - "step": 23693 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019198606507474803, - "loss": 1.5536, - "step": 23694 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019198572684840697, - "loss": 1.6039, - "step": 23695 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919853886220659, - "loss": 1.5517, - "step": 23696 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019198505039572482, - "loss": 1.543, - "step": 23697 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019198471216938376, - "loss": 1.5341, - "step": 23698 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919843739430427, - "loss": 1.5365, - "step": 23699 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019198403571670162, - "loss": 1.4106, - "step": 23700 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019198369749036056, - "loss": 1.5222, - "step": 23701 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919833592640195, - "loss": 1.3964, - "step": 23702 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019198302103767842, - "loss": 1.5307, - "step": 23703 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019198268281133736, - "loss": 1.5681, - "step": 23704 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919823445849963, - "loss": 1.6217, - "step": 23705 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919820063586552, - "loss": 1.5786, - "step": 23706 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019198166813231415, - "loss": 1.6289, - "step": 23707 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919813299059731, - "loss": 1.3059, - "step": 23708 - }, - { - "epoch": 0.08, - "learning_rate": 0.000191980991679632, - "loss": 1.4474, - "step": 23709 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019198065345329095, - "loss": 1.5259, - "step": 23710 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919803152269499, - "loss": 1.6565, - "step": 23711 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919799770006088, - "loss": 1.6961, - "step": 23712 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019197963877426775, - "loss": 1.4658, - "step": 23713 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919793005479267, - "loss": 1.5359, - "step": 23714 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919789623215856, - "loss": 1.5884, - "step": 23715 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019197862409524454, - "loss": 1.5279, - "step": 23716 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019197828586890348, - "loss": 1.6057, - "step": 23717 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919779476425624, - "loss": 1.697, - "step": 23718 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019197760941622134, - "loss": 1.5168, - "step": 23719 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019197727118988028, - "loss": 1.5765, - "step": 23720 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919769329635392, - "loss": 1.5119, - "step": 23721 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019197659473719814, - "loss": 1.6301, - "step": 23722 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019197625651085708, - "loss": 1.6141, - "step": 23723 - }, - { - "epoch": 0.08, - "learning_rate": 0.000191975918284516, - "loss": 1.5277, - "step": 23724 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019197558005817493, - "loss": 1.5834, - "step": 23725 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019197524183183387, - "loss": 1.5373, - "step": 23726 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919749036054928, - "loss": 1.6043, - "step": 23727 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019197456537915173, - "loss": 1.4978, - "step": 23728 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019197422715281067, - "loss": 1.7107, - "step": 23729 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019197388892646958, - "loss": 1.5541, - "step": 23730 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019197355070012855, - "loss": 1.6107, - "step": 23731 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019197321247378747, - "loss": 1.6176, - "step": 23732 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919728742474464, - "loss": 1.5766, - "step": 23733 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019197253602110535, - "loss": 1.5621, - "step": 23734 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019197219779476426, - "loss": 1.557, - "step": 23735 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919718595684232, - "loss": 1.6486, - "step": 23736 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019197152134208214, - "loss": 1.5736, - "step": 23737 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019197118311574106, - "loss": 1.6321, - "step": 23738 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919708448894, - "loss": 1.3977, - "step": 23739 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019197050666305894, - "loss": 1.5971, - "step": 23740 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019197016843671785, - "loss": 1.4737, - "step": 23741 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919698302103768, - "loss": 1.4828, - "step": 23742 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019196949198403574, - "loss": 1.5982, - "step": 23743 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019196915375769465, - "loss": 1.4661, - "step": 23744 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919688155313536, - "loss": 1.4815, - "step": 23745 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019196847730501253, - "loss": 1.4367, - "step": 23746 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019196813907867147, - "loss": 1.5326, - "step": 23747 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919678008523304, - "loss": 1.3823, - "step": 23748 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019196746262598933, - "loss": 1.4893, - "step": 23749 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019196712439964827, - "loss": 1.4704, - "step": 23750 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019196678617330718, - "loss": 1.5749, - "step": 23751 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019196644794696613, - "loss": 1.5757, - "step": 23752 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019196610972062507, - "loss": 1.6119, - "step": 23753 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019196577149428398, - "loss": 1.4338, - "step": 23754 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019196543326794292, - "loss": 1.6106, - "step": 23755 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019196509504160186, - "loss": 1.6544, - "step": 23756 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019196475681526078, - "loss": 1.6044, - "step": 23757 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019196441858891972, - "loss": 1.6456, - "step": 23758 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019196408036257866, - "loss": 1.5656, - "step": 23759 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019196374213623757, - "loss": 1.5694, - "step": 23760 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019196340390989651, - "loss": 1.4504, - "step": 23761 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019196306568355546, - "loss": 1.4051, - "step": 23762 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019196272745721437, - "loss": 1.5214, - "step": 23763 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919623892308733, - "loss": 1.4472, - "step": 23764 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019196205100453225, - "loss": 1.5925, - "step": 23765 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019196171277819117, - "loss": 1.5841, - "step": 23766 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919613745518501, - "loss": 1.6662, - "step": 23767 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019196103632550905, - "loss": 1.6747, - "step": 23768 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019196069809916796, - "loss": 1.6768, - "step": 23769 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919603598728269, - "loss": 1.4135, - "step": 23770 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019196002164648584, - "loss": 1.5809, - "step": 23771 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019195968342014476, - "loss": 1.5817, - "step": 23772 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919593451938037, - "loss": 1.4632, - "step": 23773 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019195900696746264, - "loss": 1.491, - "step": 23774 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019195866874112156, - "loss": 1.7026, - "step": 23775 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919583305147805, - "loss": 1.6298, - "step": 23776 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019195799228843944, - "loss": 1.5648, - "step": 23777 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019195765406209835, - "loss": 1.8032, - "step": 23778 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919573158357573, - "loss": 1.5665, - "step": 23779 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019195697760941623, - "loss": 1.4843, - "step": 23780 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019195663938307515, - "loss": 1.5227, - "step": 23781 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919563011567341, - "loss": 1.6082, - "step": 23782 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019195596293039303, - "loss": 1.4679, - "step": 23783 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019195562470405194, - "loss": 1.6002, - "step": 23784 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019195528647771089, - "loss": 1.4553, - "step": 23785 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019195494825136983, - "loss": 1.5552, - "step": 23786 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019195461002502877, - "loss": 1.5198, - "step": 23787 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919542717986877, - "loss": 1.6092, - "step": 23788 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019195393357234662, - "loss": 1.6794, - "step": 23789 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019195359534600556, - "loss": 1.6743, - "step": 23790 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919532571196645, - "loss": 1.6092, - "step": 23791 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019195291889332342, - "loss": 1.4407, - "step": 23792 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019195258066698236, - "loss": 1.5112, - "step": 23793 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919522424406413, - "loss": 1.524, - "step": 23794 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019195190421430022, - "loss": 1.5148, - "step": 23795 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019195156598795916, - "loss": 1.5684, - "step": 23796 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919512277616181, - "loss": 1.5471, - "step": 23797 - }, - { - "epoch": 0.08, - "learning_rate": 0.000191950889535277, - "loss": 1.459, - "step": 23798 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019195055130893595, - "loss": 1.5507, - "step": 23799 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919502130825949, - "loss": 1.6723, - "step": 23800 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919498748562538, - "loss": 1.4772, - "step": 23801 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019194953662991275, - "loss": 1.6335, - "step": 23802 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919491984035717, - "loss": 1.699, - "step": 23803 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919488601772306, - "loss": 1.5853, - "step": 23804 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019194852195088955, - "loss": 1.6714, - "step": 23805 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019194818372454849, - "loss": 1.3812, - "step": 23806 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919478454982074, - "loss": 1.5459, - "step": 23807 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019194750727186634, - "loss": 1.5228, - "step": 23808 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019194716904552528, - "loss": 1.435, - "step": 23809 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919468308191842, - "loss": 1.5183, - "step": 23810 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019194649259284314, - "loss": 1.5258, - "step": 23811 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019194615436650208, - "loss": 1.552, - "step": 23812 - }, - { - "epoch": 0.08, - "learning_rate": 0.000191945816140161, - "loss": 1.7141, - "step": 23813 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019194547791381993, - "loss": 1.6371, - "step": 23814 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019194513968747888, - "loss": 1.6853, - "step": 23815 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919448014611378, - "loss": 1.494, - "step": 23816 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019194446323479673, - "loss": 1.5487, - "step": 23817 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019194412500845567, - "loss": 1.6165, - "step": 23818 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019194378678211459, - "loss": 1.7498, - "step": 23819 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019194344855577353, - "loss": 1.5338, - "step": 23820 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019194311032943247, - "loss": 1.5043, - "step": 23821 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019194277210309138, - "loss": 1.5202, - "step": 23822 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019194243387675032, - "loss": 1.6679, - "step": 23823 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019194209565040926, - "loss": 1.5083, - "step": 23824 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019194175742406818, - "loss": 1.5525, - "step": 23825 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019194141919772712, - "loss": 1.6171, - "step": 23826 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019194108097138606, - "loss": 1.6611, - "step": 23827 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019194074274504497, - "loss": 1.6698, - "step": 23828 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019194040451870392, - "loss": 1.4229, - "step": 23829 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019194006629236286, - "loss": 1.5511, - "step": 23830 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019193972806602177, - "loss": 1.4219, - "step": 23831 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919393898396807, - "loss": 1.5971, - "step": 23832 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019193905161333965, - "loss": 1.5021, - "step": 23833 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919387133869986, - "loss": 1.4603, - "step": 23834 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919383751606575, - "loss": 1.576, - "step": 23835 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019193803693431645, - "loss": 1.5623, - "step": 23836 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919376987079754, - "loss": 1.5689, - "step": 23837 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919373604816343, - "loss": 1.5963, - "step": 23838 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019193702225529325, - "loss": 1.5457, - "step": 23839 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919366840289522, - "loss": 1.5271, - "step": 23840 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919363458026111, - "loss": 1.3913, - "step": 23841 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019193600757627004, - "loss": 1.7464, - "step": 23842 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019193566934992898, - "loss": 1.5836, - "step": 23843 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019193533112358792, - "loss": 1.4446, - "step": 23844 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019193499289724687, - "loss": 1.5895, - "step": 23845 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019193465467090578, - "loss": 1.6035, - "step": 23846 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019193431644456472, - "loss": 1.6679, - "step": 23847 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019193397821822366, - "loss": 1.5066, - "step": 23848 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019193363999188258, - "loss": 1.568, - "step": 23849 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019193330176554152, - "loss": 1.5206, - "step": 23850 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019193296353920046, - "loss": 1.4843, - "step": 23851 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019193262531285937, - "loss": 1.5383, - "step": 23852 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919322870865183, - "loss": 1.7443, - "step": 23853 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019193194886017725, - "loss": 1.4745, - "step": 23854 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019193161063383617, - "loss": 1.3953, - "step": 23855 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919312724074951, - "loss": 1.5875, - "step": 23856 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019193093418115405, - "loss": 1.782, - "step": 23857 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019193059595481296, - "loss": 1.5724, - "step": 23858 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919302577284719, - "loss": 1.433, - "step": 23859 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019192991950213085, - "loss": 1.6133, - "step": 23860 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019192958127578976, - "loss": 1.4987, - "step": 23861 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919292430494487, - "loss": 1.565, - "step": 23862 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019192890482310764, - "loss": 1.6445, - "step": 23863 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019192856659676656, - "loss": 1.531, - "step": 23864 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919282283704255, - "loss": 1.6391, - "step": 23865 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019192789014408444, - "loss": 1.5571, - "step": 23866 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019192755191774335, - "loss": 1.6106, - "step": 23867 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919272136914023, - "loss": 1.5662, - "step": 23868 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019192687546506124, - "loss": 1.5678, - "step": 23869 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019192653723872015, - "loss": 1.6164, - "step": 23870 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919261990123791, - "loss": 1.4091, - "step": 23871 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019192586078603803, - "loss": 1.581, - "step": 23872 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019192552255969695, - "loss": 1.4873, - "step": 23873 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919251843333559, - "loss": 1.5083, - "step": 23874 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019192484610701483, - "loss": 1.394, - "step": 23875 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019192450788067374, - "loss": 1.3606, - "step": 23876 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019192416965433268, - "loss": 1.6064, - "step": 23877 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019192383142799163, - "loss": 1.6535, - "step": 23878 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019192349320165054, - "loss": 1.5404, - "step": 23879 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019192315497530948, - "loss": 1.5664, - "step": 23880 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019192281674896842, - "loss": 1.6084, - "step": 23881 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019192247852262734, - "loss": 1.6112, - "step": 23882 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019192214029628628, - "loss": 1.4923, - "step": 23883 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019192180206994522, - "loss": 1.6424, - "step": 23884 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019192146384360413, - "loss": 1.5048, - "step": 23885 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019192112561726307, - "loss": 1.6359, - "step": 23886 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019192078739092201, - "loss": 1.5064, - "step": 23887 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019192044916458093, - "loss": 1.6379, - "step": 23888 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019192011093823987, - "loss": 1.5457, - "step": 23889 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919197727118988, - "loss": 1.5504, - "step": 23890 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019191943448555772, - "loss": 1.6081, - "step": 23891 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019191909625921667, - "loss": 1.4927, - "step": 23892 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919187580328756, - "loss": 1.5182, - "step": 23893 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019191841980653452, - "loss": 1.478, - "step": 23894 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019191808158019346, - "loss": 1.5007, - "step": 23895 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919177433538524, - "loss": 1.4733, - "step": 23896 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019191740512751132, - "loss": 1.4106, - "step": 23897 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019191706690117026, - "loss": 1.6453, - "step": 23898 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019191672867482923, - "loss": 1.44, - "step": 23899 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019191639044848814, - "loss": 1.5332, - "step": 23900 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019191605222214708, - "loss": 1.5765, - "step": 23901 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019191571399580602, - "loss": 1.5065, - "step": 23902 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019191537576946494, - "loss": 1.6978, - "step": 23903 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019191503754312388, - "loss": 1.5597, - "step": 23904 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019191469931678282, - "loss": 1.5727, - "step": 23905 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019191436109044173, - "loss": 1.4607, - "step": 23906 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019191402286410067, - "loss": 1.7095, - "step": 23907 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019191368463775962, - "loss": 1.5883, - "step": 23908 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019191334641141853, - "loss": 1.5373, - "step": 23909 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019191300818507747, - "loss": 1.5654, - "step": 23910 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919126699587364, - "loss": 1.6207, - "step": 23911 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019191233173239533, - "loss": 1.5457, - "step": 23912 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019191199350605427, - "loss": 1.5171, - "step": 23913 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919116552797132, - "loss": 1.5565, - "step": 23914 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019191131705337212, - "loss": 1.5929, - "step": 23915 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019191097882703106, - "loss": 1.7042, - "step": 23916 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019191064060069, - "loss": 1.4291, - "step": 23917 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019191030237434892, - "loss": 1.5329, - "step": 23918 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019190996414800786, - "loss": 1.5232, - "step": 23919 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919096259216668, - "loss": 1.5061, - "step": 23920 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019190928769532571, - "loss": 1.5044, - "step": 23921 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019190894946898466, - "loss": 1.5804, - "step": 23922 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919086112426436, - "loss": 1.741, - "step": 23923 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919082730163025, - "loss": 1.5816, - "step": 23924 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019190793478996145, - "loss": 1.4859, - "step": 23925 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919075965636204, - "loss": 1.6497, - "step": 23926 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919072583372793, - "loss": 1.5588, - "step": 23927 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019190692011093825, - "loss": 1.5838, - "step": 23928 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919065818845972, - "loss": 1.4819, - "step": 23929 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919062436582561, - "loss": 1.6647, - "step": 23930 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019190590543191504, - "loss": 1.6723, - "step": 23931 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019190556720557399, - "loss": 1.6383, - "step": 23932 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919052289792329, - "loss": 1.481, - "step": 23933 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019190489075289184, - "loss": 1.5254, - "step": 23934 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019190455252655078, - "loss": 1.6625, - "step": 23935 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919042143002097, - "loss": 1.4725, - "step": 23936 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019190387607386864, - "loss": 1.4907, - "step": 23937 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019190353784752758, - "loss": 1.5451, - "step": 23938 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919031996211865, - "loss": 1.6047, - "step": 23939 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019190286139484543, - "loss": 1.62, - "step": 23940 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019190252316850437, - "loss": 1.4262, - "step": 23941 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001919021849421633, - "loss": 1.6196, - "step": 23942 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019190184671582223, - "loss": 1.3256, - "step": 23943 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019190150848948117, - "loss": 1.4961, - "step": 23944 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019190117026314009, - "loss": 1.4842, - "step": 23945 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019190083203679903, - "loss": 1.5064, - "step": 23946 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019190049381045797, - "loss": 1.4412, - "step": 23947 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019190015558411688, - "loss": 1.515, - "step": 23948 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019189981735777582, - "loss": 1.544, - "step": 23949 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019189947913143476, - "loss": 1.4211, - "step": 23950 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019189914090509368, - "loss": 1.6969, - "step": 23951 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019189880267875262, - "loss": 1.6141, - "step": 23952 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019189846445241156, - "loss": 1.5739, - "step": 23953 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019189812622607047, - "loss": 1.3435, - "step": 23954 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019189778799972944, - "loss": 1.4137, - "step": 23955 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019189744977338838, - "loss": 1.5315, - "step": 23956 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918971115470473, - "loss": 1.5926, - "step": 23957 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019189677332070624, - "loss": 1.5304, - "step": 23958 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019189643509436518, - "loss": 1.7116, - "step": 23959 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918960968680241, - "loss": 1.4707, - "step": 23960 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019189575864168303, - "loss": 1.6955, - "step": 23961 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019189542041534198, - "loss": 1.3892, - "step": 23962 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918950821890009, - "loss": 1.5057, - "step": 23963 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019189474396265983, - "loss": 1.611, - "step": 23964 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019189440573631877, - "loss": 1.5727, - "step": 23965 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019189406750997769, - "loss": 1.7166, - "step": 23966 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019189372928363663, - "loss": 1.6771, - "step": 23967 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019189339105729557, - "loss": 1.5705, - "step": 23968 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019189305283095448, - "loss": 1.5919, - "step": 23969 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019189271460461342, - "loss": 1.5011, - "step": 23970 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019189237637827236, - "loss": 1.5218, - "step": 23971 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019189203815193128, - "loss": 1.6686, - "step": 23972 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019189169992559022, - "loss": 1.4104, - "step": 23973 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019189136169924916, - "loss": 1.5035, - "step": 23974 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019189102347290808, - "loss": 1.5313, - "step": 23975 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019189068524656702, - "loss": 1.6743, - "step": 23976 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019189034702022596, - "loss": 1.6292, - "step": 23977 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019189000879388487, - "loss": 1.6112, - "step": 23978 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918896705675438, - "loss": 1.6553, - "step": 23979 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019188933234120275, - "loss": 1.8073, - "step": 23980 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019188899411486167, - "loss": 1.6381, - "step": 23981 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918886558885206, - "loss": 1.6604, - "step": 23982 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019188831766217955, - "loss": 1.4268, - "step": 23983 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019188797943583846, - "loss": 1.6383, - "step": 23984 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918876412094974, - "loss": 1.5228, - "step": 23985 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019188730298315635, - "loss": 1.4982, - "step": 23986 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019188696475681526, - "loss": 1.5859, - "step": 23987 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918866265304742, - "loss": 1.6582, - "step": 23988 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019188628830413314, - "loss": 1.5587, - "step": 23989 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019188595007779206, - "loss": 1.659, - "step": 23990 - }, - { - "epoch": 0.08, - "learning_rate": 0.000191885611851451, - "loss": 1.8343, - "step": 23991 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019188527362510994, - "loss": 1.5281, - "step": 23992 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019188493539876885, - "loss": 1.6152, - "step": 23993 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918845971724278, - "loss": 1.4111, - "step": 23994 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019188425894608674, - "loss": 1.6039, - "step": 23995 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019188392071974565, - "loss": 1.3928, - "step": 23996 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918835824934046, - "loss": 1.4628, - "step": 23997 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019188324426706353, - "loss": 1.4907, - "step": 23998 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019188290604072245, - "loss": 1.5353, - "step": 23999 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918825678143814, - "loss": 1.6055, - "step": 24000 - }, - { - "epoch": 0.08, - "eval_loss": 1.4148700420673077, - "mrc_em": 23.0, - "mrc_f1": 28.931746031746034, - "nli_acc": 0.08, - "step": 24000, - "sts_f1": 0.6976744186046511, - "ynat_f1": 0.647448750080329 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019188222958804033, - "loss": 1.5199, - "step": 24001 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019188189136169924, - "loss": 1.5443, - "step": 24002 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019188155313535818, - "loss": 1.5094, - "step": 24003 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019188121490901712, - "loss": 1.5442, - "step": 24004 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019188087668267604, - "loss": 1.4873, - "step": 24005 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019188053845633498, - "loss": 1.4055, - "step": 24006 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019188020022999392, - "loss": 1.6368, - "step": 24007 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019187986200365283, - "loss": 1.5306, - "step": 24008 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019187952377731178, - "loss": 1.6248, - "step": 24009 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019187918555097072, - "loss": 1.5293, - "step": 24010 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019187884732462963, - "loss": 1.5022, - "step": 24011 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918785090982886, - "loss": 1.5906, - "step": 24012 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918781708719475, - "loss": 1.5939, - "step": 24013 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019187783264560645, - "loss": 1.6846, - "step": 24014 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918774944192654, - "loss": 1.5888, - "step": 24015 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019187715619292434, - "loss": 1.7683, - "step": 24016 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019187681796658325, - "loss": 1.4934, - "step": 24017 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918764797402422, - "loss": 1.6762, - "step": 24018 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019187614151390113, - "loss": 1.4812, - "step": 24019 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019187580328756005, - "loss": 1.5251, - "step": 24020 - }, - { - "epoch": 0.08, - "learning_rate": 0.000191875465061219, - "loss": 1.5289, - "step": 24021 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019187512683487793, - "loss": 1.6952, - "step": 24022 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019187478860853684, - "loss": 1.5377, - "step": 24023 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019187445038219578, - "loss": 1.5142, - "step": 24024 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019187411215585473, - "loss": 1.4738, - "step": 24025 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019187377392951364, - "loss": 1.5177, - "step": 24026 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019187343570317258, - "loss": 1.4419, - "step": 24027 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019187309747683152, - "loss": 1.64, - "step": 24028 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019187275925049044, - "loss": 1.3804, - "step": 24029 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019187242102414938, - "loss": 1.4898, - "step": 24030 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019187208279780832, - "loss": 1.5828, - "step": 24031 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019187174457146723, - "loss": 1.5482, - "step": 24032 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019187140634512617, - "loss": 1.4193, - "step": 24033 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019187106811878511, - "loss": 1.6565, - "step": 24034 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019187072989244403, - "loss": 1.5687, - "step": 24035 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019187039166610297, - "loss": 1.5073, - "step": 24036 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918700534397619, - "loss": 1.5603, - "step": 24037 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019186971521342082, - "loss": 1.4298, - "step": 24038 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019186937698707977, - "loss": 1.5856, - "step": 24039 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918690387607387, - "loss": 1.5751, - "step": 24040 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019186870053439762, - "loss": 1.2902, - "step": 24041 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019186836230805656, - "loss": 1.6086, - "step": 24042 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918680240817155, - "loss": 1.7257, - "step": 24043 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019186768585537442, - "loss": 1.6522, - "step": 24044 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019186734762903336, - "loss": 1.5703, - "step": 24045 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918670094026923, - "loss": 1.6264, - "step": 24046 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019186667117635121, - "loss": 1.6222, - "step": 24047 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019186633295001016, - "loss": 1.4445, - "step": 24048 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918659947236691, - "loss": 1.6692, - "step": 24049 - }, - { - "epoch": 0.08, - "learning_rate": 0.000191865656497328, - "loss": 1.7079, - "step": 24050 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019186531827098695, - "loss": 1.4845, - "step": 24051 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918649800446459, - "loss": 1.664, - "step": 24052 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918646418183048, - "loss": 1.5079, - "step": 24053 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019186430359196375, - "loss": 1.4903, - "step": 24054 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918639653656227, - "loss": 1.7224, - "step": 24055 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918636271392816, - "loss": 1.5531, - "step": 24056 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019186328891294054, - "loss": 1.5092, - "step": 24057 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019186295068659949, - "loss": 1.5272, - "step": 24058 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918626124602584, - "loss": 1.4466, - "step": 24059 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019186227423391734, - "loss": 1.5504, - "step": 24060 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019186193600757628, - "loss": 1.4407, - "step": 24061 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918615977812352, - "loss": 1.4806, - "step": 24062 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019186125955489414, - "loss": 1.6168, - "step": 24063 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019186092132855308, - "loss": 1.472, - "step": 24064 - }, - { - "epoch": 0.08, - "learning_rate": 0.000191860583102212, - "loss": 1.4005, - "step": 24065 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019186024487587093, - "loss": 1.6593, - "step": 24066 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019185990664952987, - "loss": 1.6611, - "step": 24067 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019185956842318882, - "loss": 1.647, - "step": 24068 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019185923019684776, - "loss": 1.5639, - "step": 24069 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019185889197050667, - "loss": 1.6127, - "step": 24070 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918585537441656, - "loss": 1.5862, - "step": 24071 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019185821551782455, - "loss": 1.6772, - "step": 24072 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019185787729148347, - "loss": 1.5052, - "step": 24073 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918575390651424, - "loss": 1.5762, - "step": 24074 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019185720083880135, - "loss": 1.688, - "step": 24075 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019185686261246026, - "loss": 1.5977, - "step": 24076 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918565243861192, - "loss": 1.6172, - "step": 24077 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019185618615977815, - "loss": 1.4838, - "step": 24078 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019185584793343706, - "loss": 1.5257, - "step": 24079 - }, - { - "epoch": 0.08, - "learning_rate": 0.000191855509707096, - "loss": 1.5918, - "step": 24080 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019185517148075494, - "loss": 1.507, - "step": 24081 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019185483325441386, - "loss": 1.5887, - "step": 24082 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918544950280728, - "loss": 1.69, - "step": 24083 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019185415680173174, - "loss": 1.6664, - "step": 24084 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019185381857539065, - "loss": 1.5487, - "step": 24085 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918534803490496, - "loss": 1.5189, - "step": 24086 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019185314212270853, - "loss": 1.5385, - "step": 24087 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019185280389636745, - "loss": 1.4779, - "step": 24088 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918524656700264, - "loss": 1.7031, - "step": 24089 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019185212744368533, - "loss": 1.4875, - "step": 24090 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019185178921734424, - "loss": 1.5112, - "step": 24091 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019185145099100319, - "loss": 1.5122, - "step": 24092 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019185111276466213, - "loss": 1.4238, - "step": 24093 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019185077453832104, - "loss": 1.5817, - "step": 24094 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019185043631197998, - "loss": 1.3906, - "step": 24095 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019185009808563892, - "loss": 1.4769, - "step": 24096 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019184975985929784, - "loss": 1.7862, - "step": 24097 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019184942163295678, - "loss": 1.6186, - "step": 24098 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019184908340661572, - "loss": 1.7129, - "step": 24099 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019184874518027466, - "loss": 1.6133, - "step": 24100 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019184840695393357, - "loss": 1.7238, - "step": 24101 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019184806872759252, - "loss": 1.5388, - "step": 24102 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019184773050125146, - "loss": 1.5284, - "step": 24103 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019184739227491037, - "loss": 1.4122, - "step": 24104 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918470540485693, - "loss": 1.5814, - "step": 24105 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019184671582222825, - "loss": 1.6061, - "step": 24106 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019184637759588717, - "loss": 1.5702, - "step": 24107 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918460393695461, - "loss": 1.4806, - "step": 24108 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019184570114320505, - "loss": 1.594, - "step": 24109 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019184536291686396, - "loss": 1.7743, - "step": 24110 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918450246905229, - "loss": 1.5671, - "step": 24111 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019184468646418185, - "loss": 1.5841, - "step": 24112 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019184434823784076, - "loss": 1.565, - "step": 24113 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918440100114997, - "loss": 1.6502, - "step": 24114 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019184367178515864, - "loss": 1.6073, - "step": 24115 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019184333355881756, - "loss": 1.7159, - "step": 24116 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918429953324765, - "loss": 1.4512, - "step": 24117 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019184265710613544, - "loss": 1.6913, - "step": 24118 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019184231887979435, - "loss": 1.54, - "step": 24119 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918419806534533, - "loss": 1.3462, - "step": 24120 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019184164242711223, - "loss": 1.4146, - "step": 24121 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019184130420077115, - "loss": 1.6897, - "step": 24122 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918409659744301, - "loss": 1.5667, - "step": 24123 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019184062774808903, - "loss": 1.5565, - "step": 24124 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019184028952174797, - "loss": 1.341, - "step": 24125 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918399512954069, - "loss": 1.6576, - "step": 24126 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019183961306906583, - "loss": 1.5233, - "step": 24127 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019183927484272477, - "loss": 1.4208, - "step": 24128 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918389366163837, - "loss": 1.5949, - "step": 24129 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019183859839004262, - "loss": 1.3014, - "step": 24130 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019183826016370156, - "loss": 1.6666, - "step": 24131 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918379219373605, - "loss": 1.5562, - "step": 24132 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019183758371101942, - "loss": 1.5365, - "step": 24133 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019183724548467836, - "loss": 1.4601, - "step": 24134 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918369072583373, - "loss": 1.4299, - "step": 24135 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019183656903199622, - "loss": 1.3767, - "step": 24136 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019183623080565516, - "loss": 1.3861, - "step": 24137 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918358925793141, - "loss": 1.5788, - "step": 24138 - }, - { - "epoch": 0.08, - "learning_rate": 0.000191835554352973, - "loss": 1.7232, - "step": 24139 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019183521612663195, - "loss": 1.426, - "step": 24140 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918348779002909, - "loss": 1.4791, - "step": 24141 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918345396739498, - "loss": 1.589, - "step": 24142 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019183420144760875, - "loss": 1.5882, - "step": 24143 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918338632212677, - "loss": 1.705, - "step": 24144 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918335249949266, - "loss": 1.4127, - "step": 24145 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019183318676858555, - "loss": 1.591, - "step": 24146 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918328485422445, - "loss": 1.5844, - "step": 24147 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918325103159034, - "loss": 1.4135, - "step": 24148 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019183217208956234, - "loss": 1.6648, - "step": 24149 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019183183386322128, - "loss": 1.488, - "step": 24150 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918314956368802, - "loss": 1.4704, - "step": 24151 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019183115741053914, - "loss": 1.5167, - "step": 24152 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019183081918419808, - "loss": 1.4717, - "step": 24153 - }, - { - "epoch": 0.08, - "learning_rate": 0.000191830480957857, - "loss": 1.408, - "step": 24154 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019183014273151594, - "loss": 1.3878, - "step": 24155 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019182980450517488, - "loss": 1.526, - "step": 24156 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918294662788338, - "loss": 1.3693, - "step": 24157 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019182912805249273, - "loss": 1.6232, - "step": 24158 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019182878982615167, - "loss": 1.5065, - "step": 24159 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918284515998106, - "loss": 1.4739, - "step": 24160 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019182811337346953, - "loss": 1.6371, - "step": 24161 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019182777514712847, - "loss": 1.6213, - "step": 24162 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019182743692078738, - "loss": 1.725, - "step": 24163 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019182709869444632, - "loss": 1.4524, - "step": 24164 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019182676046810527, - "loss": 1.6107, - "step": 24165 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019182642224176418, - "loss": 1.5783, - "step": 24166 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019182608401542312, - "loss": 1.4296, - "step": 24167 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019182574578908206, - "loss": 1.5104, - "step": 24168 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019182540756274098, - "loss": 1.7025, - "step": 24169 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019182506933639992, - "loss": 1.7407, - "step": 24170 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019182473111005886, - "loss": 1.6464, - "step": 24171 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019182439288371777, - "loss": 1.5699, - "step": 24172 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918240546573767, - "loss": 1.5379, - "step": 24173 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019182371643103565, - "loss": 1.6229, - "step": 24174 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019182337820469457, - "loss": 1.493, - "step": 24175 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918230399783535, - "loss": 1.6709, - "step": 24176 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019182270175201245, - "loss": 1.6557, - "step": 24177 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019182236352567136, - "loss": 1.5199, - "step": 24178 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918220252993303, - "loss": 1.5938, - "step": 24179 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019182168707298927, - "loss": 1.4419, - "step": 24180 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918213488466482, - "loss": 1.606, - "step": 24181 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019182101062030713, - "loss": 1.4206, - "step": 24182 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019182067239396607, - "loss": 1.5577, - "step": 24183 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019182033416762498, - "loss": 1.539, - "step": 24184 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019181999594128393, - "loss": 1.4943, - "step": 24185 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019181965771494287, - "loss": 1.4564, - "step": 24186 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019181931948860178, - "loss": 1.6602, - "step": 24187 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019181898126226072, - "loss": 1.5119, - "step": 24188 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019181864303591966, - "loss": 1.6915, - "step": 24189 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019181830480957858, - "loss": 1.4281, - "step": 24190 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019181796658323752, - "loss": 1.6641, - "step": 24191 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019181762835689646, - "loss": 1.6112, - "step": 24192 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019181729013055537, - "loss": 1.582, - "step": 24193 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019181695190421431, - "loss": 1.5536, - "step": 24194 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019181661367787326, - "loss": 1.608, - "step": 24195 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019181627545153217, - "loss": 1.5446, - "step": 24196 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918159372251911, - "loss": 1.7658, - "step": 24197 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019181559899885005, - "loss": 1.4754, - "step": 24198 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019181526077250897, - "loss": 1.4926, - "step": 24199 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918149225461679, - "loss": 1.6413, - "step": 24200 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019181458431982685, - "loss": 1.4338, - "step": 24201 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019181424609348576, - "loss": 1.5761, - "step": 24202 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918139078671447, - "loss": 1.6448, - "step": 24203 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019181356964080364, - "loss": 1.5445, - "step": 24204 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019181323141446256, - "loss": 1.6517, - "step": 24205 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918128931881215, - "loss": 1.648, - "step": 24206 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019181255496178044, - "loss": 1.474, - "step": 24207 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019181221673543936, - "loss": 1.5377, - "step": 24208 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918118785090983, - "loss": 1.5268, - "step": 24209 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019181154028275724, - "loss": 1.496, - "step": 24210 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019181120205641615, - "loss": 1.5912, - "step": 24211 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918108638300751, - "loss": 1.5022, - "step": 24212 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019181052560373403, - "loss": 1.635, - "step": 24213 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019181018737739295, - "loss": 1.5065, - "step": 24214 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918098491510519, - "loss": 1.4977, - "step": 24215 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019180951092471083, - "loss": 1.6036, - "step": 24216 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019180917269836974, - "loss": 1.6747, - "step": 24217 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019180883447202869, - "loss": 1.6639, - "step": 24218 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019180849624568763, - "loss": 1.6669, - "step": 24219 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019180815801934654, - "loss": 1.6455, - "step": 24220 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019180781979300548, - "loss": 1.5745, - "step": 24221 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019180748156666442, - "loss": 1.6224, - "step": 24222 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019180714334032334, - "loss": 1.5908, - "step": 24223 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019180680511398228, - "loss": 1.698, - "step": 24224 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019180646688764122, - "loss": 1.6041, - "step": 24225 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019180612866130013, - "loss": 1.3703, - "step": 24226 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019180579043495907, - "loss": 1.5702, - "step": 24227 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019180545220861802, - "loss": 1.6219, - "step": 24228 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019180511398227693, - "loss": 1.5913, - "step": 24229 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019180477575593587, - "loss": 1.469, - "step": 24230 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918044375295948, - "loss": 1.5179, - "step": 24231 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019180409930325373, - "loss": 1.3696, - "step": 24232 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019180376107691267, - "loss": 1.4813, - "step": 24233 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918034228505716, - "loss": 1.6532, - "step": 24234 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019180308462423052, - "loss": 1.4067, - "step": 24235 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001918027463978895, - "loss": 1.5687, - "step": 24236 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019180240817154843, - "loss": 1.6292, - "step": 24237 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019180206994520735, - "loss": 1.4967, - "step": 24238 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019180173171886629, - "loss": 1.4256, - "step": 24239 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019180139349252523, - "loss": 1.5727, - "step": 24240 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019180105526618414, - "loss": 1.5159, - "step": 24241 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019180071703984308, - "loss": 1.4045, - "step": 24242 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019180037881350202, - "loss": 1.5081, - "step": 24243 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019180004058716094, - "loss": 1.4539, - "step": 24244 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019179970236081988, - "loss": 1.5934, - "step": 24245 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019179936413447882, - "loss": 1.5506, - "step": 24246 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019179902590813773, - "loss": 1.5978, - "step": 24247 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019179868768179668, - "loss": 1.4991, - "step": 24248 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019179834945545562, - "loss": 1.5428, - "step": 24249 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019179801122911453, - "loss": 1.4587, - "step": 24250 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019179767300277347, - "loss": 1.6749, - "step": 24251 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917973347764324, - "loss": 1.5504, - "step": 24252 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019179699655009133, - "loss": 1.5977, - "step": 24253 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019179665832375027, - "loss": 1.6015, - "step": 24254 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917963200974092, - "loss": 1.5799, - "step": 24255 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019179598187106812, - "loss": 1.5195, - "step": 24256 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019179564364472706, - "loss": 1.7135, - "step": 24257 - }, - { - "epoch": 0.08, - "learning_rate": 0.000191795305418386, - "loss": 1.5033, - "step": 24258 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019179496719204492, - "loss": 1.6556, - "step": 24259 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019179462896570386, - "loss": 1.5406, - "step": 24260 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917942907393628, - "loss": 1.7421, - "step": 24261 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019179395251302172, - "loss": 1.6495, - "step": 24262 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019179361428668066, - "loss": 1.5101, - "step": 24263 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917932760603396, - "loss": 1.5785, - "step": 24264 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917929378339985, - "loss": 1.5293, - "step": 24265 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019179259960765745, - "loss": 1.5301, - "step": 24266 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917922613813164, - "loss": 1.7798, - "step": 24267 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917919231549753, - "loss": 1.5821, - "step": 24268 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019179158492863425, - "loss": 1.4588, - "step": 24269 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917912467022932, - "loss": 1.6646, - "step": 24270 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917909084759521, - "loss": 1.5586, - "step": 24271 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019179057024961105, - "loss": 1.4693, - "step": 24272 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019179023202327, - "loss": 1.4898, - "step": 24273 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917898937969289, - "loss": 1.6959, - "step": 24274 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019178955557058784, - "loss": 1.6113, - "step": 24275 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019178921734424678, - "loss": 1.3199, - "step": 24276 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917888791179057, - "loss": 1.8018, - "step": 24277 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019178854089156464, - "loss": 1.4433, - "step": 24278 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019178820266522358, - "loss": 1.5176, - "step": 24279 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917878644388825, - "loss": 1.5853, - "step": 24280 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019178752621254143, - "loss": 1.603, - "step": 24281 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019178718798620038, - "loss": 1.5745, - "step": 24282 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917868497598593, - "loss": 1.6049, - "step": 24283 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019178651153351823, - "loss": 1.6371, - "step": 24284 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019178617330717717, - "loss": 1.5861, - "step": 24285 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019178583508083609, - "loss": 1.6469, - "step": 24286 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019178549685449503, - "loss": 1.6462, - "step": 24287 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019178515862815397, - "loss": 1.5229, - "step": 24288 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019178482040181288, - "loss": 1.6269, - "step": 24289 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019178448217547182, - "loss": 1.7008, - "step": 24290 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019178414394913076, - "loss": 1.5955, - "step": 24291 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917838057227897, - "loss": 1.5745, - "step": 24292 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019178346749644865, - "loss": 1.3461, - "step": 24293 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917831292701076, - "loss": 1.4891, - "step": 24294 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917827910437665, - "loss": 1.5503, - "step": 24295 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019178245281742544, - "loss": 1.487, - "step": 24296 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019178211459108438, - "loss": 1.486, - "step": 24297 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917817763647433, - "loss": 1.5433, - "step": 24298 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019178143813840224, - "loss": 1.4218, - "step": 24299 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019178109991206118, - "loss": 1.4853, - "step": 24300 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917807616857201, - "loss": 1.5542, - "step": 24301 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019178042345937904, - "loss": 1.5303, - "step": 24302 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019178008523303798, - "loss": 1.6282, - "step": 24303 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917797470066969, - "loss": 1.4682, - "step": 24304 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019177940878035583, - "loss": 1.582, - "step": 24305 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019177907055401477, - "loss": 1.5584, - "step": 24306 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917787323276737, - "loss": 1.5887, - "step": 24307 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019177839410133263, - "loss": 1.5245, - "step": 24308 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019177805587499157, - "loss": 1.4883, - "step": 24309 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019177771764865048, - "loss": 1.454, - "step": 24310 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019177737942230943, - "loss": 1.6255, - "step": 24311 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019177704119596837, - "loss": 1.4874, - "step": 24312 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019177670296962728, - "loss": 1.5648, - "step": 24313 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019177636474328622, - "loss": 1.4654, - "step": 24314 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019177602651694516, - "loss": 1.5617, - "step": 24315 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019177568829060408, - "loss": 1.6777, - "step": 24316 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019177535006426302, - "loss": 1.5546, - "step": 24317 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019177501183792196, - "loss": 1.5084, - "step": 24318 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019177467361158087, - "loss": 1.4896, - "step": 24319 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019177433538523981, - "loss": 1.5297, - "step": 24320 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019177399715889876, - "loss": 1.6295, - "step": 24321 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019177365893255767, - "loss": 1.4729, - "step": 24322 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917733207062166, - "loss": 1.525, - "step": 24323 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019177298247987555, - "loss": 1.4483, - "step": 24324 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019177264425353447, - "loss": 1.4803, - "step": 24325 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917723060271934, - "loss": 1.4312, - "step": 24326 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019177196780085235, - "loss": 1.495, - "step": 24327 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019177162957451126, - "loss": 1.6514, - "step": 24328 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917712913481702, - "loss": 1.5591, - "step": 24329 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019177095312182914, - "loss": 1.4603, - "step": 24330 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019177061489548806, - "loss": 1.605, - "step": 24331 - }, - { - "epoch": 0.08, - "learning_rate": 0.000191770276669147, - "loss": 1.8074, - "step": 24332 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019176993844280594, - "loss": 1.6279, - "step": 24333 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019176960021646485, - "loss": 1.5273, - "step": 24334 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917692619901238, - "loss": 1.593, - "step": 24335 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019176892376378274, - "loss": 1.4969, - "step": 24336 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019176858553744165, - "loss": 1.4872, - "step": 24337 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917682473111006, - "loss": 1.4732, - "step": 24338 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019176790908475953, - "loss": 1.6466, - "step": 24339 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019176757085841845, - "loss": 1.6243, - "step": 24340 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917672326320774, - "loss": 1.5118, - "step": 24341 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019176689440573633, - "loss": 1.6219, - "step": 24342 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019176655617939524, - "loss": 1.5815, - "step": 24343 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019176621795305418, - "loss": 1.5395, - "step": 24344 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019176587972671313, - "loss": 1.7788, - "step": 24345 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019176554150037204, - "loss": 1.5847, - "step": 24346 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019176520327403098, - "loss": 1.5049, - "step": 24347 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019176486504768992, - "loss": 1.5421, - "step": 24348 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019176452682134886, - "loss": 1.6162, - "step": 24349 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917641885950078, - "loss": 1.5062, - "step": 24350 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019176385036866672, - "loss": 1.4914, - "step": 24351 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019176351214232566, - "loss": 1.5202, - "step": 24352 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917631739159846, - "loss": 1.5051, - "step": 24353 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019176283568964351, - "loss": 1.4744, - "step": 24354 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019176249746330246, - "loss": 1.5177, - "step": 24355 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917621592369614, - "loss": 1.4817, - "step": 24356 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917618210106203, - "loss": 1.5013, - "step": 24357 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019176148278427925, - "loss": 1.7456, - "step": 24358 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917611445579382, - "loss": 1.7147, - "step": 24359 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917608063315971, - "loss": 1.6455, - "step": 24360 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019176046810525605, - "loss": 1.6149, - "step": 24361 - }, - { - "epoch": 0.08, - "learning_rate": 0.000191760129878915, - "loss": 1.5531, - "step": 24362 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917597916525739, - "loss": 1.6622, - "step": 24363 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019175945342623284, - "loss": 1.5445, - "step": 24364 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019175911519989179, - "loss": 1.5151, - "step": 24365 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917587769735507, - "loss": 1.529, - "step": 24366 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019175843874720964, - "loss": 1.5792, - "step": 24367 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019175810052086858, - "loss": 1.5569, - "step": 24368 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019175776229452752, - "loss": 1.5882, - "step": 24369 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019175742406818644, - "loss": 1.4717, - "step": 24370 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019175708584184538, - "loss": 1.4815, - "step": 24371 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019175674761550432, - "loss": 1.6172, - "step": 24372 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019175640938916323, - "loss": 1.5082, - "step": 24373 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019175607116282217, - "loss": 1.6817, - "step": 24374 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019175573293648112, - "loss": 1.5375, - "step": 24375 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019175539471014003, - "loss": 1.5851, - "step": 24376 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019175505648379897, - "loss": 1.4553, - "step": 24377 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917547182574579, - "loss": 1.4801, - "step": 24378 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019175438003111683, - "loss": 1.4168, - "step": 24379 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019175404180477577, - "loss": 1.4635, - "step": 24380 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917537035784347, - "loss": 1.6666, - "step": 24381 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019175336535209362, - "loss": 1.5828, - "step": 24382 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019175302712575256, - "loss": 1.6292, - "step": 24383 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917526888994115, - "loss": 1.5022, - "step": 24384 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019175235067307042, - "loss": 1.6865, - "step": 24385 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019175201244672936, - "loss": 1.513, - "step": 24386 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917516742203883, - "loss": 1.4306, - "step": 24387 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019175133599404722, - "loss": 1.5319, - "step": 24388 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019175099776770616, - "loss": 1.6282, - "step": 24389 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917506595413651, - "loss": 1.6182, - "step": 24390 - }, - { - "epoch": 0.08, - "learning_rate": 0.000191750321315024, - "loss": 1.7143, - "step": 24391 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019174998308868295, - "loss": 1.5706, - "step": 24392 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917496448623419, - "loss": 1.338, - "step": 24393 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917493066360008, - "loss": 1.5939, - "step": 24394 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019174896840965975, - "loss": 1.5832, - "step": 24395 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917486301833187, - "loss": 1.7456, - "step": 24396 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917482919569776, - "loss": 1.6912, - "step": 24397 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019174795373063655, - "loss": 1.5277, - "step": 24398 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019174761550429549, - "loss": 1.6191, - "step": 24399 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917472772779544, - "loss": 1.5705, - "step": 24400 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019174693905161334, - "loss": 1.5605, - "step": 24401 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019174660082527228, - "loss": 1.5063, - "step": 24402 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917462625989312, - "loss": 1.5555, - "step": 24403 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019174592437259014, - "loss": 1.5805, - "step": 24404 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019174558614624908, - "loss": 1.5515, - "step": 24405 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019174524791990802, - "loss": 1.8048, - "step": 24406 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019174490969356696, - "loss": 1.4886, - "step": 24407 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019174457146722588, - "loss": 1.4742, - "step": 24408 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019174423324088482, - "loss": 1.6193, - "step": 24409 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019174389501454376, - "loss": 1.5853, - "step": 24410 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019174355678820267, - "loss": 1.4625, - "step": 24411 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917432185618616, - "loss": 1.5048, - "step": 24412 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019174288033552055, - "loss": 1.6011, - "step": 24413 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019174254210917947, - "loss": 1.4733, - "step": 24414 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917422038828384, - "loss": 1.6333, - "step": 24415 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019174186565649735, - "loss": 1.6353, - "step": 24416 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019174152743015626, - "loss": 1.463, - "step": 24417 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917411892038152, - "loss": 1.5538, - "step": 24418 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019174085097747415, - "loss": 1.5777, - "step": 24419 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019174051275113306, - "loss": 1.613, - "step": 24420 - }, - { - "epoch": 0.08, - "learning_rate": 0.000191740174524792, - "loss": 1.5612, - "step": 24421 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019173983629845094, - "loss": 1.4282, - "step": 24422 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019173949807210986, - "loss": 1.6792, - "step": 24423 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917391598457688, - "loss": 1.4911, - "step": 24424 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019173882161942774, - "loss": 1.5146, - "step": 24425 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019173848339308665, - "loss": 1.5921, - "step": 24426 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917381451667456, - "loss": 1.5846, - "step": 24427 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019173780694040454, - "loss": 1.5627, - "step": 24428 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019173746871406345, - "loss": 1.5746, - "step": 24429 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917371304877224, - "loss": 1.6462, - "step": 24430 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019173679226138133, - "loss": 1.6386, - "step": 24431 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019173645403504025, - "loss": 1.4858, - "step": 24432 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917361158086992, - "loss": 1.5729, - "step": 24433 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019173577758235813, - "loss": 1.5372, - "step": 24434 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019173543935601704, - "loss": 1.561, - "step": 24435 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019173510112967598, - "loss": 1.5821, - "step": 24436 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019173476290333492, - "loss": 1.5715, - "step": 24437 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019173442467699384, - "loss": 1.3851, - "step": 24438 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019173408645065278, - "loss": 1.6493, - "step": 24439 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019173374822431172, - "loss": 1.5305, - "step": 24440 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019173340999797063, - "loss": 1.4059, - "step": 24441 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019173307177162958, - "loss": 1.4634, - "step": 24442 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019173273354528852, - "loss": 1.5747, - "step": 24443 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019173239531894743, - "loss": 1.563, - "step": 24444 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019173205709260637, - "loss": 1.6316, - "step": 24445 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917317188662653, - "loss": 1.4982, - "step": 24446 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019173138063992423, - "loss": 1.7053, - "step": 24447 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019173104241358317, - "loss": 1.6957, - "step": 24448 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917307041872421, - "loss": 1.6348, - "step": 24449 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019173036596090102, - "loss": 1.4865, - "step": 24450 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019173002773455996, - "loss": 1.5588, - "step": 24451 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917296895082189, - "loss": 1.4834, - "step": 24452 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019172935128187785, - "loss": 1.5941, - "step": 24453 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019172901305553676, - "loss": 1.5792, - "step": 24454 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917286748291957, - "loss": 1.5809, - "step": 24455 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019172833660285464, - "loss": 1.6215, - "step": 24456 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019172799837651356, - "loss": 1.4315, - "step": 24457 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917276601501725, - "loss": 1.5241, - "step": 24458 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019172732192383144, - "loss": 1.5907, - "step": 24459 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019172698369749035, - "loss": 1.6064, - "step": 24460 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019172664547114932, - "loss": 1.5992, - "step": 24461 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019172630724480824, - "loss": 1.4675, - "step": 24462 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019172596901846718, - "loss": 1.4942, - "step": 24463 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019172563079212612, - "loss": 1.559, - "step": 24464 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019172529256578503, - "loss": 1.6872, - "step": 24465 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019172495433944397, - "loss": 1.4896, - "step": 24466 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019172461611310291, - "loss": 1.772, - "step": 24467 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019172427788676183, - "loss": 1.6684, - "step": 24468 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019172393966042077, - "loss": 1.6269, - "step": 24469 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917236014340797, - "loss": 1.521, - "step": 24470 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019172326320773863, - "loss": 1.722, - "step": 24471 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019172292498139757, - "loss": 1.5888, - "step": 24472 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917225867550565, - "loss": 1.5547, - "step": 24473 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019172224852871542, - "loss": 1.5994, - "step": 24474 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019172191030237436, - "loss": 1.546, - "step": 24475 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917215720760333, - "loss": 1.5905, - "step": 24476 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019172123384969222, - "loss": 1.643, - "step": 24477 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019172089562335116, - "loss": 1.6115, - "step": 24478 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917205573970101, - "loss": 1.6334, - "step": 24479 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019172021917066901, - "loss": 1.5093, - "step": 24480 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019171988094432796, - "loss": 1.4993, - "step": 24481 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917195427179869, - "loss": 1.4305, - "step": 24482 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917192044916458, - "loss": 1.5478, - "step": 24483 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019171886626530475, - "loss": 1.4962, - "step": 24484 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917185280389637, - "loss": 1.7085, - "step": 24485 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917181898126226, - "loss": 1.5668, - "step": 24486 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019171785158628155, - "loss": 1.4435, - "step": 24487 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917175133599405, - "loss": 1.5603, - "step": 24488 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917171751335994, - "loss": 1.5766, - "step": 24489 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019171683690725834, - "loss": 1.5406, - "step": 24490 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019171649868091729, - "loss": 1.5749, - "step": 24491 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917161604545762, - "loss": 1.5683, - "step": 24492 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019171582222823514, - "loss": 1.4192, - "step": 24493 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019171548400189408, - "loss": 1.4907, - "step": 24494 - }, - { - "epoch": 0.08, - "learning_rate": 0.000191715145775553, - "loss": 1.5674, - "step": 24495 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019171480754921194, - "loss": 1.6529, - "step": 24496 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019171446932287088, - "loss": 1.6535, - "step": 24497 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917141310965298, - "loss": 1.7078, - "step": 24498 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019171379287018873, - "loss": 1.6145, - "step": 24499 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019171345464384767, - "loss": 1.5816, - "step": 24500 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917131164175066, - "loss": 1.7224, - "step": 24501 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019171277819116553, - "loss": 1.6644, - "step": 24502 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019171243996482447, - "loss": 1.5223, - "step": 24503 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019171210173848338, - "loss": 1.5272, - "step": 24504 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019171176351214233, - "loss": 1.4966, - "step": 24505 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019171142528580127, - "loss": 1.5616, - "step": 24506 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019171108705946018, - "loss": 1.5709, - "step": 24507 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019171074883311912, - "loss": 1.4419, - "step": 24508 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019171041060677806, - "loss": 1.5062, - "step": 24509 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019171007238043698, - "loss": 1.5014, - "step": 24510 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019170973415409592, - "loss": 1.5173, - "step": 24511 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019170939592775486, - "loss": 1.6124, - "step": 24512 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019170905770141377, - "loss": 1.541, - "step": 24513 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019170871947507271, - "loss": 1.5738, - "step": 24514 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019170838124873166, - "loss": 1.4709, - "step": 24515 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019170804302239057, - "loss": 1.6526, - "step": 24516 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019170770479604954, - "loss": 1.6545, - "step": 24517 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019170736656970848, - "loss": 1.5199, - "step": 24518 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917070283433674, - "loss": 1.4096, - "step": 24519 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019170669011702633, - "loss": 1.628, - "step": 24520 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019170635189068528, - "loss": 1.4889, - "step": 24521 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917060136643442, - "loss": 1.5908, - "step": 24522 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019170567543800313, - "loss": 1.6627, - "step": 24523 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019170533721166207, - "loss": 1.4306, - "step": 24524 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019170499898532099, - "loss": 1.7452, - "step": 24525 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019170466075897993, - "loss": 1.5563, - "step": 24526 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019170432253263887, - "loss": 1.47, - "step": 24527 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019170398430629778, - "loss": 1.6897, - "step": 24528 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019170364607995672, - "loss": 1.5581, - "step": 24529 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019170330785361566, - "loss": 1.569, - "step": 24530 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019170296962727458, - "loss": 1.5069, - "step": 24531 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019170263140093352, - "loss": 1.6304, - "step": 24532 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019170229317459246, - "loss": 1.4813, - "step": 24533 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019170195494825137, - "loss": 1.6869, - "step": 24534 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019170161672191032, - "loss": 1.3969, - "step": 24535 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019170127849556926, - "loss": 1.4948, - "step": 24536 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019170094026922817, - "loss": 1.726, - "step": 24537 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001917006020428871, - "loss": 1.5313, - "step": 24538 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019170026381654605, - "loss": 1.4448, - "step": 24539 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019169992559020497, - "loss": 1.526, - "step": 24540 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916995873638639, - "loss": 1.3634, - "step": 24541 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019169924913752285, - "loss": 1.534, - "step": 24542 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019169891091118176, - "loss": 1.6194, - "step": 24543 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916985726848407, - "loss": 1.4375, - "step": 24544 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019169823445849965, - "loss": 1.5565, - "step": 24545 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019169789623215856, - "loss": 1.4132, - "step": 24546 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916975580058175, - "loss": 1.6097, - "step": 24547 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019169721977947644, - "loss": 1.5979, - "step": 24548 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019169688155313536, - "loss": 1.7285, - "step": 24549 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916965433267943, - "loss": 1.503, - "step": 24550 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019169620510045324, - "loss": 1.4132, - "step": 24551 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019169586687411215, - "loss": 1.6398, - "step": 24552 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916955286477711, - "loss": 1.4942, - "step": 24553 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019169519042143003, - "loss": 1.7001, - "step": 24554 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019169485219508895, - "loss": 1.5812, - "step": 24555 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916945139687479, - "loss": 1.6138, - "step": 24556 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019169417574240683, - "loss": 1.5883, - "step": 24557 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019169383751606575, - "loss": 1.7722, - "step": 24558 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019169349928972469, - "loss": 1.5016, - "step": 24559 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019169316106338363, - "loss": 1.5115, - "step": 24560 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019169282283704254, - "loss": 1.7578, - "step": 24561 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019169248461070148, - "loss": 1.4225, - "step": 24562 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019169214638436042, - "loss": 1.5686, - "step": 24563 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019169180815801934, - "loss": 1.5901, - "step": 24564 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019169146993167828, - "loss": 1.6082, - "step": 24565 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019169113170533722, - "loss": 1.5848, - "step": 24566 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019169079347899613, - "loss": 1.6549, - "step": 24567 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019169045525265508, - "loss": 1.3793, - "step": 24568 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019169011702631402, - "loss": 1.5638, - "step": 24569 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019168977879997293, - "loss": 1.5733, - "step": 24570 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019168944057363187, - "loss": 1.576, - "step": 24571 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916891023472908, - "loss": 1.602, - "step": 24572 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019168876412094975, - "loss": 1.5711, - "step": 24573 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916884258946087, - "loss": 1.487, - "step": 24574 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019168808766826764, - "loss": 1.5408, - "step": 24575 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019168774944192655, - "loss": 1.6017, - "step": 24576 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916874112155855, - "loss": 1.6303, - "step": 24577 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019168707298924443, - "loss": 1.5287, - "step": 24578 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019168673476290335, - "loss": 1.569, - "step": 24579 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916863965365623, - "loss": 1.4697, - "step": 24580 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019168605831022123, - "loss": 1.6431, - "step": 24581 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019168572008388014, - "loss": 1.6355, - "step": 24582 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019168538185753908, - "loss": 1.7083, - "step": 24583 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019168504363119803, - "loss": 1.5384, - "step": 24584 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019168470540485694, - "loss": 1.8569, - "step": 24585 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019168436717851588, - "loss": 1.7078, - "step": 24586 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019168402895217482, - "loss": 1.6278, - "step": 24587 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019168369072583374, - "loss": 1.7852, - "step": 24588 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019168335249949268, - "loss": 1.7802, - "step": 24589 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019168301427315162, - "loss": 1.4461, - "step": 24590 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019168267604681053, - "loss": 1.3707, - "step": 24591 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019168233782046947, - "loss": 1.6667, - "step": 24592 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019168199959412841, - "loss": 1.5386, - "step": 24593 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019168166136778733, - "loss": 1.4744, - "step": 24594 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019168132314144627, - "loss": 1.6426, - "step": 24595 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916809849151052, - "loss": 1.4064, - "step": 24596 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019168064668876412, - "loss": 1.6466, - "step": 24597 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019168030846242307, - "loss": 1.601, - "step": 24598 - }, - { - "epoch": 0.08, - "learning_rate": 0.000191679970236082, - "loss": 1.3868, - "step": 24599 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019167963200974092, - "loss": 1.6355, - "step": 24600 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019167929378339986, - "loss": 1.5747, - "step": 24601 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916789555570588, - "loss": 1.4588, - "step": 24602 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019167861733071772, - "loss": 1.5527, - "step": 24603 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019167827910437666, - "loss": 1.6893, - "step": 24604 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916779408780356, - "loss": 1.6204, - "step": 24605 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916776026516945, - "loss": 1.7065, - "step": 24606 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019167726442535345, - "loss": 1.6244, - "step": 24607 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916769261990124, - "loss": 1.7803, - "step": 24608 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916765879726713, - "loss": 1.5651, - "step": 24609 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019167624974633025, - "loss": 1.5149, - "step": 24610 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916759115199892, - "loss": 1.8598, - "step": 24611 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916755732936481, - "loss": 1.5199, - "step": 24612 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019167523506730705, - "loss": 1.4444, - "step": 24613 - }, - { - "epoch": 0.08, - "learning_rate": 0.000191674896840966, - "loss": 1.4865, - "step": 24614 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916745586146249, - "loss": 1.6894, - "step": 24615 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019167422038828384, - "loss": 1.6588, - "step": 24616 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019167388216194278, - "loss": 1.4695, - "step": 24617 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916735439356017, - "loss": 1.5667, - "step": 24618 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019167320570926064, - "loss": 1.717, - "step": 24619 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019167286748291958, - "loss": 1.4911, - "step": 24620 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916725292565785, - "loss": 1.4439, - "step": 24621 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019167219103023744, - "loss": 1.4992, - "step": 24622 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019167185280389638, - "loss": 1.7929, - "step": 24623 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916715145775553, - "loss": 1.6005, - "step": 24624 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019167117635121423, - "loss": 1.5436, - "step": 24625 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019167083812487317, - "loss": 1.6772, - "step": 24626 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916704998985321, - "loss": 1.5512, - "step": 24627 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019167016167219103, - "loss": 1.655, - "step": 24628 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019166982344584997, - "loss": 1.5949, - "step": 24629 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916694852195089, - "loss": 1.5758, - "step": 24630 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019166914699316785, - "loss": 1.5388, - "step": 24631 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019166880876682677, - "loss": 1.4625, - "step": 24632 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916684705404857, - "loss": 1.6958, - "step": 24633 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019166813231414465, - "loss": 1.5282, - "step": 24634 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916677940878036, - "loss": 1.5969, - "step": 24635 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916674558614625, - "loss": 1.6262, - "step": 24636 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019166711763512144, - "loss": 1.5282, - "step": 24637 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019166677940878039, - "loss": 1.6485, - "step": 24638 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916664411824393, - "loss": 1.5083, - "step": 24639 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019166610295609824, - "loss": 1.5741, - "step": 24640 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019166576472975718, - "loss": 1.5208, - "step": 24641 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916654265034161, - "loss": 1.7284, - "step": 24642 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019166508827707504, - "loss": 1.5758, - "step": 24643 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019166475005073398, - "loss": 1.5009, - "step": 24644 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916644118243929, - "loss": 1.5114, - "step": 24645 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019166407359805183, - "loss": 1.6298, - "step": 24646 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019166373537171077, - "loss": 1.5966, - "step": 24647 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916633971453697, - "loss": 1.6474, - "step": 24648 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019166305891902863, - "loss": 1.6982, - "step": 24649 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019166272069268757, - "loss": 1.5524, - "step": 24650 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019166238246634649, - "loss": 1.4919, - "step": 24651 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019166204424000543, - "loss": 1.5711, - "step": 24652 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019166170601366437, - "loss": 1.7279, - "step": 24653 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019166136778732328, - "loss": 1.5254, - "step": 24654 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019166102956098222, - "loss": 1.7273, - "step": 24655 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019166069133464116, - "loss": 1.6641, - "step": 24656 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019166035310830008, - "loss": 1.5104, - "step": 24657 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019166001488195902, - "loss": 1.5933, - "step": 24658 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019165967665561796, - "loss": 1.5583, - "step": 24659 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019165933842927687, - "loss": 1.4502, - "step": 24660 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019165900020293582, - "loss": 1.5825, - "step": 24661 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019165866197659476, - "loss": 1.6753, - "step": 24662 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019165832375025367, - "loss": 1.4402, - "step": 24663 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916579855239126, - "loss": 1.5391, - "step": 24664 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019165764729757155, - "loss": 1.4074, - "step": 24665 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019165730907123047, - "loss": 1.5381, - "step": 24666 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916569708448894, - "loss": 1.6672, - "step": 24667 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019165663261854835, - "loss": 1.5558, - "step": 24668 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019165629439220726, - "loss": 1.5005, - "step": 24669 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916559561658662, - "loss": 1.4891, - "step": 24670 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019165561793952515, - "loss": 1.5584, - "step": 24671 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019165527971318406, - "loss": 1.4975, - "step": 24672 - }, - { - "epoch": 0.08, - "learning_rate": 0.000191654941486843, - "loss": 1.5453, - "step": 24673 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019165460326050194, - "loss": 1.5235, - "step": 24674 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019165426503416086, - "loss": 1.7904, - "step": 24675 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916539268078198, - "loss": 1.577, - "step": 24676 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019165358858147874, - "loss": 1.5881, - "step": 24677 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019165325035513765, - "loss": 1.4417, - "step": 24678 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916529121287966, - "loss": 1.6115, - "step": 24679 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019165257390245553, - "loss": 1.4964, - "step": 24680 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019165223567611445, - "loss": 1.566, - "step": 24681 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916518974497734, - "loss": 1.4657, - "step": 24682 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019165155922343233, - "loss": 1.477, - "step": 24683 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019165122099709124, - "loss": 1.6234, - "step": 24684 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019165088277075019, - "loss": 1.6683, - "step": 24685 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019165054454440913, - "loss": 1.5648, - "step": 24686 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019165020631806807, - "loss": 1.7008, - "step": 24687 - }, - { - "epoch": 0.08, - "learning_rate": 0.000191649868091727, - "loss": 1.5975, - "step": 24688 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019164952986538592, - "loss": 1.617, - "step": 24689 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019164919163904486, - "loss": 1.6463, - "step": 24690 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916488534127038, - "loss": 1.5066, - "step": 24691 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019164851518636272, - "loss": 1.7585, - "step": 24692 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019164817696002166, - "loss": 1.5512, - "step": 24693 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916478387336806, - "loss": 1.5558, - "step": 24694 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019164750050733952, - "loss": 1.5936, - "step": 24695 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019164716228099846, - "loss": 1.5242, - "step": 24696 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916468240546574, - "loss": 1.4778, - "step": 24697 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916464858283163, - "loss": 1.5614, - "step": 24698 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019164614760197525, - "loss": 1.5028, - "step": 24699 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916458093756342, - "loss": 1.4651, - "step": 24700 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916454711492931, - "loss": 1.5815, - "step": 24701 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019164513292295205, - "loss": 1.569, - "step": 24702 - }, - { - "epoch": 0.08, - "learning_rate": 0.000191644794696611, - "loss": 1.5748, - "step": 24703 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916444564702699, - "loss": 1.396, - "step": 24704 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019164411824392885, - "loss": 1.5213, - "step": 24705 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916437800175878, - "loss": 1.7636, - "step": 24706 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916434417912467, - "loss": 1.5758, - "step": 24707 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019164310356490564, - "loss": 1.4452, - "step": 24708 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019164276533856458, - "loss": 1.5318, - "step": 24709 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916424271122235, - "loss": 1.4612, - "step": 24710 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019164208888588244, - "loss": 1.6932, - "step": 24711 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019164175065954138, - "loss": 1.5156, - "step": 24712 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916414124332003, - "loss": 1.5408, - "step": 24713 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019164107420685923, - "loss": 1.6893, - "step": 24714 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019164073598051818, - "loss": 1.6609, - "step": 24715 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916403977541771, - "loss": 1.4634, - "step": 24716 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019164005952783603, - "loss": 1.6887, - "step": 24717 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019163972130149497, - "loss": 1.6785, - "step": 24718 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916393830751539, - "loss": 1.6274, - "step": 24719 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019163904484881283, - "loss": 1.7702, - "step": 24720 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019163870662247177, - "loss": 1.565, - "step": 24721 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916383683961307, - "loss": 1.5291, - "step": 24722 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019163803016978962, - "loss": 1.5089, - "step": 24723 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019163769194344856, - "loss": 1.5793, - "step": 24724 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916373537171075, - "loss": 1.5918, - "step": 24725 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019163701549076642, - "loss": 1.3973, - "step": 24726 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019163667726442536, - "loss": 1.5523, - "step": 24727 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916363390380843, - "loss": 1.3938, - "step": 24728 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019163600081174322, - "loss": 1.4867, - "step": 24729 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019163566258540216, - "loss": 1.4749, - "step": 24730 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916353243590611, - "loss": 1.4918, - "step": 24731 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019163498613272, - "loss": 1.4891, - "step": 24732 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019163464790637895, - "loss": 1.4697, - "step": 24733 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916343096800379, - "loss": 1.4751, - "step": 24734 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916339714536968, - "loss": 1.6607, - "step": 24735 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019163363322735575, - "loss": 1.4919, - "step": 24736 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916332950010147, - "loss": 1.4822, - "step": 24737 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916329567746736, - "loss": 1.5735, - "step": 24738 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019163261854833255, - "loss": 1.4125, - "step": 24739 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916322803219915, - "loss": 1.5384, - "step": 24740 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916319420956504, - "loss": 1.4367, - "step": 24741 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019163160386930937, - "loss": 1.6502, - "step": 24742 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019163126564296828, - "loss": 1.5478, - "step": 24743 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019163092741662723, - "loss": 1.6173, - "step": 24744 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019163058919028617, - "loss": 1.6382, - "step": 24745 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019163025096394508, - "loss": 1.4686, - "step": 24746 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019162991273760402, - "loss": 1.6404, - "step": 24747 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019162957451126296, - "loss": 1.4013, - "step": 24748 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019162923628492188, - "loss": 1.5614, - "step": 24749 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019162889805858082, - "loss": 1.6693, - "step": 24750 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019162855983223976, - "loss": 1.5571, - "step": 24751 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019162822160589867, - "loss": 1.4996, - "step": 24752 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019162788337955761, - "loss": 1.4552, - "step": 24753 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019162754515321656, - "loss": 1.5194, - "step": 24754 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019162720692687547, - "loss": 1.5815, - "step": 24755 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916268687005344, - "loss": 1.4077, - "step": 24756 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019162653047419335, - "loss": 1.541, - "step": 24757 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019162619224785227, - "loss": 1.4777, - "step": 24758 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916258540215112, - "loss": 1.5255, - "step": 24759 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019162551579517015, - "loss": 1.6767, - "step": 24760 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019162517756882906, - "loss": 1.6767, - "step": 24761 - }, - { - "epoch": 0.08, - "learning_rate": 0.000191624839342488, - "loss": 1.5829, - "step": 24762 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019162450111614694, - "loss": 1.6018, - "step": 24763 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019162416288980586, - "loss": 1.5939, - "step": 24764 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916238246634648, - "loss": 1.5968, - "step": 24765 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019162348643712374, - "loss": 1.593, - "step": 24766 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019162314821078265, - "loss": 1.5294, - "step": 24767 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916228099844416, - "loss": 1.6921, - "step": 24768 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019162247175810054, - "loss": 1.557, - "step": 24769 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019162213353175945, - "loss": 1.497, - "step": 24770 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916217953054184, - "loss": 1.6237, - "step": 24771 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019162145707907733, - "loss": 1.5375, - "step": 24772 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019162111885273625, - "loss": 1.4446, - "step": 24773 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916207806263952, - "loss": 1.395, - "step": 24774 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019162044240005413, - "loss": 1.5896, - "step": 24775 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019162010417371304, - "loss": 1.5184, - "step": 24776 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019161976594737198, - "loss": 1.5887, - "step": 24777 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019161942772103093, - "loss": 1.5267, - "step": 24778 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019161908949468984, - "loss": 1.6724, - "step": 24779 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019161875126834878, - "loss": 1.6987, - "step": 24780 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019161841304200772, - "loss": 1.3716, - "step": 24781 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019161807481566664, - "loss": 1.5719, - "step": 24782 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019161773658932558, - "loss": 1.634, - "step": 24783 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019161739836298452, - "loss": 1.4277, - "step": 24784 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019161706013664343, - "loss": 1.4992, - "step": 24785 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019161672191030237, - "loss": 1.5536, - "step": 24786 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019161638368396131, - "loss": 1.5805, - "step": 24787 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019161604545762023, - "loss": 1.5487, - "step": 24788 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019161570723127917, - "loss": 1.5172, - "step": 24789 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916153690049381, - "loss": 1.611, - "step": 24790 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019161503077859702, - "loss": 1.4716, - "step": 24791 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019161469255225597, - "loss": 1.5278, - "step": 24792 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916143543259149, - "loss": 1.5569, - "step": 24793 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019161401609957382, - "loss": 1.4607, - "step": 24794 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019161367787323276, - "loss": 1.5074, - "step": 24795 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916133396468917, - "loss": 1.6487, - "step": 24796 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019161300142055062, - "loss": 1.4868, - "step": 24797 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019161266319420959, - "loss": 1.5669, - "step": 24798 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019161232496786853, - "loss": 1.5586, - "step": 24799 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019161198674152744, - "loss": 1.64, - "step": 24800 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019161164851518638, - "loss": 1.6385, - "step": 24801 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019161131028884532, - "loss": 1.5499, - "step": 24802 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019161097206250424, - "loss": 1.6317, - "step": 24803 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019161063383616318, - "loss": 1.5893, - "step": 24804 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019161029560982212, - "loss": 1.5679, - "step": 24805 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019160995738348103, - "loss": 1.6561, - "step": 24806 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019160961915713997, - "loss": 1.332, - "step": 24807 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019160928093079892, - "loss": 1.385, - "step": 24808 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019160894270445783, - "loss": 1.5513, - "step": 24809 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019160860447811677, - "loss": 1.5668, - "step": 24810 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916082662517757, - "loss": 1.5593, - "step": 24811 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019160792802543463, - "loss": 1.5443, - "step": 24812 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019160758979909357, - "loss": 1.4484, - "step": 24813 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916072515727525, - "loss": 1.6211, - "step": 24814 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019160691334641142, - "loss": 1.4595, - "step": 24815 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019160657512007036, - "loss": 1.5169, - "step": 24816 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916062368937293, - "loss": 1.6307, - "step": 24817 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019160589866738822, - "loss": 1.562, - "step": 24818 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019160556044104716, - "loss": 1.4952, - "step": 24819 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916052222147061, - "loss": 1.5277, - "step": 24820 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019160488398836502, - "loss": 1.5484, - "step": 24821 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019160454576202396, - "loss": 1.6496, - "step": 24822 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916042075356829, - "loss": 1.4965, - "step": 24823 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916038693093418, - "loss": 1.6172, - "step": 24824 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019160353108300075, - "loss": 1.5297, - "step": 24825 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916031928566597, - "loss": 1.7116, - "step": 24826 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916028546303186, - "loss": 1.5943, - "step": 24827 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019160251640397755, - "loss": 1.7392, - "step": 24828 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916021781776365, - "loss": 1.5477, - "step": 24829 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916018399512954, - "loss": 1.5467, - "step": 24830 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019160150172495435, - "loss": 1.5303, - "step": 24831 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019160116349861329, - "loss": 1.4301, - "step": 24832 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001916008252722722, - "loss": 1.608, - "step": 24833 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019160048704593114, - "loss": 1.3944, - "step": 24834 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019160014881959008, - "loss": 1.6382, - "step": 24835 - }, - { - "epoch": 0.08, - "learning_rate": 0.000191599810593249, - "loss": 1.4952, - "step": 24836 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019159947236690794, - "loss": 1.5921, - "step": 24837 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019159913414056688, - "loss": 1.4373, - "step": 24838 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915987959142258, - "loss": 1.3898, - "step": 24839 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019159845768788473, - "loss": 1.652, - "step": 24840 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019159811946154368, - "loss": 1.542, - "step": 24841 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915977812352026, - "loss": 1.6549, - "step": 24842 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019159744300886153, - "loss": 1.5849, - "step": 24843 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019159710478252047, - "loss": 1.4246, - "step": 24844 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019159676655617939, - "loss": 1.6418, - "step": 24845 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019159642832983833, - "loss": 1.5759, - "step": 24846 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019159609010349727, - "loss": 1.7561, - "step": 24847 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019159575187715618, - "loss": 1.5632, - "step": 24848 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019159541365081512, - "loss": 1.4528, - "step": 24849 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019159507542447406, - "loss": 1.5552, - "step": 24850 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019159473719813298, - "loss": 1.5614, - "step": 24851 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019159439897179192, - "loss": 1.6401, - "step": 24852 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019159406074545086, - "loss": 1.4546, - "step": 24853 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915937225191098, - "loss": 1.4992, - "step": 24854 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019159338429276874, - "loss": 1.5455, - "step": 24855 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019159304606642768, - "loss": 1.6179, - "step": 24856 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915927078400866, - "loss": 1.4889, - "step": 24857 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019159236961374554, - "loss": 1.6577, - "step": 24858 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019159203138740448, - "loss": 1.5227, - "step": 24859 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915916931610634, - "loss": 1.7404, - "step": 24860 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019159135493472234, - "loss": 1.4546, - "step": 24861 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019159101670838128, - "loss": 1.4428, - "step": 24862 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915906784820402, - "loss": 1.7413, - "step": 24863 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019159034025569913, - "loss": 1.4668, - "step": 24864 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019159000202935807, - "loss": 1.4716, - "step": 24865 - }, - { - "epoch": 0.08, - "learning_rate": 0.000191589663803017, - "loss": 1.5306, - "step": 24866 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019158932557667593, - "loss": 1.3957, - "step": 24867 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019158898735033487, - "loss": 1.4881, - "step": 24868 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019158864912399378, - "loss": 1.6437, - "step": 24869 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019158831089765272, - "loss": 1.5617, - "step": 24870 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019158797267131167, - "loss": 1.6218, - "step": 24871 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019158763444497058, - "loss": 1.4682, - "step": 24872 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019158729621862952, - "loss": 1.4071, - "step": 24873 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019158695799228846, - "loss": 1.5301, - "step": 24874 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019158661976594738, - "loss": 1.6319, - "step": 24875 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019158628153960632, - "loss": 1.4518, - "step": 24876 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019158594331326526, - "loss": 1.4877, - "step": 24877 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019158560508692417, - "loss": 1.6184, - "step": 24878 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915852668605831, - "loss": 1.7315, - "step": 24879 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019158492863424205, - "loss": 1.6386, - "step": 24880 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019158459040790097, - "loss": 1.6046, - "step": 24881 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915842521815599, - "loss": 1.6306, - "step": 24882 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019158391395521885, - "loss": 1.5544, - "step": 24883 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019158357572887776, - "loss": 1.7053, - "step": 24884 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915832375025367, - "loss": 1.6719, - "step": 24885 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019158289927619565, - "loss": 1.4681, - "step": 24886 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019158256104985456, - "loss": 1.6598, - "step": 24887 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915822228235135, - "loss": 1.5469, - "step": 24888 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019158188459717244, - "loss": 1.3089, - "step": 24889 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019158154637083136, - "loss": 1.5011, - "step": 24890 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915812081444903, - "loss": 1.4572, - "step": 24891 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019158086991814924, - "loss": 1.6402, - "step": 24892 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019158053169180815, - "loss": 1.5093, - "step": 24893 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915801934654671, - "loss": 1.5898, - "step": 24894 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019157985523912604, - "loss": 1.6889, - "step": 24895 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019157951701278495, - "loss": 1.6031, - "step": 24896 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915791787864439, - "loss": 1.6455, - "step": 24897 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019157884056010283, - "loss": 1.7619, - "step": 24898 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019157850233376175, - "loss": 1.4876, - "step": 24899 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915781641074207, - "loss": 1.4596, - "step": 24900 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019157782588107963, - "loss": 1.5687, - "step": 24901 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019157748765473854, - "loss": 1.6042, - "step": 24902 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019157714942839748, - "loss": 1.5842, - "step": 24903 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019157681120205643, - "loss": 1.5941, - "step": 24904 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019157647297571534, - "loss": 1.4736, - "step": 24905 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019157613474937428, - "loss": 1.5051, - "step": 24906 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019157579652303322, - "loss": 1.4581, - "step": 24907 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019157545829669214, - "loss": 1.3609, - "step": 24908 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019157512007035108, - "loss": 1.7032, - "step": 24909 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019157478184401004, - "loss": 1.6043, - "step": 24910 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019157444361766896, - "loss": 1.6745, - "step": 24911 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915741053913279, - "loss": 1.5804, - "step": 24912 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019157376716498684, - "loss": 1.7114, - "step": 24913 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019157342893864576, - "loss": 1.6307, - "step": 24914 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915730907123047, - "loss": 1.4665, - "step": 24915 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019157275248596364, - "loss": 1.628, - "step": 24916 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019157241425962255, - "loss": 1.5825, - "step": 24917 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915720760332815, - "loss": 1.5719, - "step": 24918 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019157173780694043, - "loss": 1.5665, - "step": 24919 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019157139958059935, - "loss": 1.6341, - "step": 24920 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915710613542583, - "loss": 1.5973, - "step": 24921 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019157072312791723, - "loss": 1.5371, - "step": 24922 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019157038490157614, - "loss": 1.5955, - "step": 24923 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019157004667523509, - "loss": 1.4748, - "step": 24924 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019156970844889403, - "loss": 1.3697, - "step": 24925 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019156937022255294, - "loss": 1.4659, - "step": 24926 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019156903199621188, - "loss": 1.6523, - "step": 24927 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019156869376987082, - "loss": 1.4426, - "step": 24928 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019156835554352974, - "loss": 1.7122, - "step": 24929 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019156801731718868, - "loss": 1.5561, - "step": 24930 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019156767909084762, - "loss": 1.605, - "step": 24931 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019156734086450653, - "loss": 1.5544, - "step": 24932 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019156700263816547, - "loss": 1.3226, - "step": 24933 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019156666441182442, - "loss": 1.4783, - "step": 24934 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019156632618548333, - "loss": 1.5745, - "step": 24935 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019156598795914227, - "loss": 1.5492, - "step": 24936 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915656497328012, - "loss": 1.4827, - "step": 24937 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019156531150646013, - "loss": 1.5821, - "step": 24938 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019156497328011907, - "loss": 1.4686, - "step": 24939 - }, - { - "epoch": 0.08, - "learning_rate": 0.000191564635053778, - "loss": 1.5892, - "step": 24940 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019156429682743692, - "loss": 1.6842, - "step": 24941 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019156395860109586, - "loss": 1.5062, - "step": 24942 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915636203747548, - "loss": 1.6602, - "step": 24943 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019156328214841372, - "loss": 1.5458, - "step": 24944 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019156294392207266, - "loss": 1.652, - "step": 24945 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915626056957316, - "loss": 1.4738, - "step": 24946 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019156226746939051, - "loss": 1.4962, - "step": 24947 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019156192924304946, - "loss": 1.644, - "step": 24948 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915615910167084, - "loss": 1.5998, - "step": 24949 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915612527903673, - "loss": 1.482, - "step": 24950 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019156091456402625, - "loss": 1.7107, - "step": 24951 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915605763376852, - "loss": 1.4334, - "step": 24952 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915602381113441, - "loss": 1.4598, - "step": 24953 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019155989988500305, - "loss": 1.6783, - "step": 24954 - }, - { - "epoch": 0.08, - "learning_rate": 0.000191559561658662, - "loss": 1.5129, - "step": 24955 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915592234323209, - "loss": 1.5588, - "step": 24956 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019155888520597984, - "loss": 1.414, - "step": 24957 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019155854697963879, - "loss": 1.5092, - "step": 24958 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915582087532977, - "loss": 1.5066, - "step": 24959 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019155787052695664, - "loss": 1.6339, - "step": 24960 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019155753230061558, - "loss": 1.4756, - "step": 24961 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915571940742745, - "loss": 1.6246, - "step": 24962 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019155685584793344, - "loss": 1.5456, - "step": 24963 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019155651762159238, - "loss": 1.5184, - "step": 24964 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915561793952513, - "loss": 1.5277, - "step": 24965 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019155584116891026, - "loss": 1.6003, - "step": 24966 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019155550294256917, - "loss": 1.5744, - "step": 24967 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019155516471622812, - "loss": 1.5543, - "step": 24968 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019155482648988706, - "loss": 1.4327, - "step": 24969 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019155448826354597, - "loss": 1.4587, - "step": 24970 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915541500372049, - "loss": 1.3546, - "step": 24971 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019155381181086385, - "loss": 1.524, - "step": 24972 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019155347358452277, - "loss": 1.5578, - "step": 24973 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915531353581817, - "loss": 1.534, - "step": 24974 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019155279713184065, - "loss": 1.5623, - "step": 24975 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019155245890549956, - "loss": 1.3826, - "step": 24976 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915521206791585, - "loss": 1.7055, - "step": 24977 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019155178245281745, - "loss": 1.5134, - "step": 24978 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019155144422647636, - "loss": 1.5077, - "step": 24979 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915511060001353, - "loss": 1.4138, - "step": 24980 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019155076777379424, - "loss": 1.52, - "step": 24981 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019155042954745316, - "loss": 1.5305, - "step": 24982 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915500913211121, - "loss": 1.6912, - "step": 24983 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019154975309477104, - "loss": 1.5849, - "step": 24984 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019154941486842995, - "loss": 1.5197, - "step": 24985 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915490766420889, - "loss": 1.6253, - "step": 24986 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019154873841574783, - "loss": 1.4891, - "step": 24987 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019154840018940678, - "loss": 1.7154, - "step": 24988 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915480619630657, - "loss": 1.5431, - "step": 24989 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019154772373672463, - "loss": 1.5998, - "step": 24990 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019154738551038357, - "loss": 1.5715, - "step": 24991 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019154704728404249, - "loss": 1.4049, - "step": 24992 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019154670905770143, - "loss": 1.4863, - "step": 24993 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019154637083136037, - "loss": 1.4027, - "step": 24994 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019154603260501928, - "loss": 1.409, - "step": 24995 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019154569437867822, - "loss": 1.3806, - "step": 24996 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019154535615233716, - "loss": 1.6193, - "step": 24997 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019154501792599608, - "loss": 1.5576, - "step": 24998 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019154467969965502, - "loss": 1.4457, - "step": 24999 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019154434147331396, - "loss": 1.5023, - "step": 25000 - }, - { - "epoch": 0.08, - "eval_loss": 1.4175180288461537, - "mrc_em": 14.0, - "mrc_f1": 24.35714285714287, - "nli_acc": 0.07, - "step": 25000, - "sts_f1": 0.5454545454545454, - "ynat_f1": 0.6588431898354637 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019154400324697288, - "loss": 1.526, - "step": 25001 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019154366502063182, - "loss": 1.6128, - "step": 25002 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019154332679429076, - "loss": 1.3993, - "step": 25003 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019154298856794967, - "loss": 1.6712, - "step": 25004 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915426503416086, - "loss": 1.5274, - "step": 25005 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019154231211526755, - "loss": 1.6041, - "step": 25006 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019154197388892647, - "loss": 1.5216, - "step": 25007 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915416356625854, - "loss": 1.5817, - "step": 25008 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019154129743624435, - "loss": 1.6265, - "step": 25009 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019154095920990326, - "loss": 1.6216, - "step": 25010 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915406209835622, - "loss": 1.6834, - "step": 25011 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019154028275722115, - "loss": 1.5381, - "step": 25012 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019153994453088006, - "loss": 1.6004, - "step": 25013 - }, - { - "epoch": 0.08, - "learning_rate": 0.000191539606304539, - "loss": 1.4913, - "step": 25014 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019153926807819794, - "loss": 1.413, - "step": 25015 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019153892985185686, - "loss": 1.6817, - "step": 25016 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915385916255158, - "loss": 1.4638, - "step": 25017 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019153825339917474, - "loss": 1.6017, - "step": 25018 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019153791517283365, - "loss": 1.7477, - "step": 25019 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915375769464926, - "loss": 1.5837, - "step": 25020 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019153723872015154, - "loss": 1.6953, - "step": 25021 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019153690049381045, - "loss": 1.5783, - "step": 25022 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019153656226746942, - "loss": 1.4745, - "step": 25023 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019153622404112833, - "loss": 1.2838, - "step": 25024 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019153588581478727, - "loss": 1.5029, - "step": 25025 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019153554758844621, - "loss": 1.7949, - "step": 25026 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019153520936210513, - "loss": 1.5639, - "step": 25027 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019153487113576407, - "loss": 1.6238, - "step": 25028 - }, - { - "epoch": 0.08, - "learning_rate": 0.000191534532909423, - "loss": 1.5357, - "step": 25029 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019153419468308192, - "loss": 1.4759, - "step": 25030 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019153385645674087, - "loss": 1.6215, - "step": 25031 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915335182303998, - "loss": 1.3428, - "step": 25032 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019153318000405872, - "loss": 1.5986, - "step": 25033 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019153284177771766, - "loss": 1.6402, - "step": 25034 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915325035513766, - "loss": 1.5874, - "step": 25035 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019153216532503552, - "loss": 1.5296, - "step": 25036 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019153182709869446, - "loss": 1.5679, - "step": 25037 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915314888723534, - "loss": 1.442, - "step": 25038 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915311506460123, - "loss": 1.4406, - "step": 25039 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019153081241967125, - "loss": 1.5456, - "step": 25040 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915304741933302, - "loss": 1.5101, - "step": 25041 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915301359669891, - "loss": 1.6557, - "step": 25042 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019152979774064805, - "loss": 1.4427, - "step": 25043 - }, - { - "epoch": 0.08, - "learning_rate": 0.000191529459514307, - "loss": 1.8638, - "step": 25044 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915291212879659, - "loss": 1.5244, - "step": 25045 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019152878306162485, - "loss": 1.6945, - "step": 25046 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915284448352838, - "loss": 1.7128, - "step": 25047 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915281066089427, - "loss": 1.4178, - "step": 25048 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019152776838260164, - "loss": 1.4999, - "step": 25049 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019152743015626058, - "loss": 1.4818, - "step": 25050 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915270919299195, - "loss": 1.5866, - "step": 25051 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019152675370357844, - "loss": 1.5419, - "step": 25052 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019152641547723738, - "loss": 1.6293, - "step": 25053 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915260772508963, - "loss": 1.4671, - "step": 25054 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019152573902455524, - "loss": 1.652, - "step": 25055 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019152540079821418, - "loss": 1.3168, - "step": 25056 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915250625718731, - "loss": 1.3946, - "step": 25057 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019152472434553203, - "loss": 1.5131, - "step": 25058 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019152438611919097, - "loss": 1.6294, - "step": 25059 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915240478928499, - "loss": 1.7864, - "step": 25060 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019152370966650883, - "loss": 1.4613, - "step": 25061 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019152337144016777, - "loss": 1.6261, - "step": 25062 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019152303321382668, - "loss": 1.4728, - "step": 25063 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019152269498748563, - "loss": 1.6048, - "step": 25064 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019152235676114457, - "loss": 1.7065, - "step": 25065 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019152201853480348, - "loss": 1.6175, - "step": 25066 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019152168030846242, - "loss": 1.3921, - "step": 25067 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019152134208212136, - "loss": 1.5717, - "step": 25068 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019152100385578028, - "loss": 1.5413, - "step": 25069 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019152066562943922, - "loss": 1.5772, - "step": 25070 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019152032740309816, - "loss": 1.5982, - "step": 25071 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915199891767571, - "loss": 1.4905, - "step": 25072 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019151965095041601, - "loss": 1.68, - "step": 25073 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019151931272407496, - "loss": 1.6172, - "step": 25074 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915189744977339, - "loss": 1.5328, - "step": 25075 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915186362713928, - "loss": 1.5691, - "step": 25076 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019151829804505175, - "loss": 1.5037, - "step": 25077 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915179598187107, - "loss": 1.5614, - "step": 25078 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019151762159236963, - "loss": 1.4106, - "step": 25079 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019151728336602857, - "loss": 1.5634, - "step": 25080 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915169451396875, - "loss": 1.5803, - "step": 25081 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019151660691334643, - "loss": 1.6547, - "step": 25082 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019151626868700537, - "loss": 1.4725, - "step": 25083 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019151593046066429, - "loss": 1.6126, - "step": 25084 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019151559223432323, - "loss": 1.591, - "step": 25085 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019151525400798217, - "loss": 1.5499, - "step": 25086 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019151491578164108, - "loss": 1.4993, - "step": 25087 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019151457755530002, - "loss": 1.4875, - "step": 25088 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019151423932895896, - "loss": 1.6851, - "step": 25089 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019151390110261788, - "loss": 1.5503, - "step": 25090 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019151356287627682, - "loss": 1.4932, - "step": 25091 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019151322464993576, - "loss": 1.5438, - "step": 25092 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019151288642359467, - "loss": 1.5705, - "step": 25093 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019151254819725362, - "loss": 1.6892, - "step": 25094 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019151220997091256, - "loss": 1.5091, - "step": 25095 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019151187174457147, - "loss": 1.5818, - "step": 25096 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915115335182304, - "loss": 1.4829, - "step": 25097 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019151119529188935, - "loss": 1.4827, - "step": 25098 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019151085706554827, - "loss": 1.7587, - "step": 25099 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915105188392072, - "loss": 1.6717, - "step": 25100 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019151018061286615, - "loss": 1.6981, - "step": 25101 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019150984238652506, - "loss": 1.5543, - "step": 25102 - }, - { - "epoch": 0.08, - "learning_rate": 0.000191509504160184, - "loss": 1.3859, - "step": 25103 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019150916593384295, - "loss": 1.6094, - "step": 25104 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019150882770750186, - "loss": 1.4473, - "step": 25105 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915084894811608, - "loss": 1.4757, - "step": 25106 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019150815125481974, - "loss": 1.5873, - "step": 25107 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019150781302847866, - "loss": 1.7235, - "step": 25108 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915074748021376, - "loss": 1.7153, - "step": 25109 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019150713657579654, - "loss": 1.5696, - "step": 25110 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019150679834945545, - "loss": 1.6113, - "step": 25111 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915064601231144, - "loss": 1.6846, - "step": 25112 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019150612189677333, - "loss": 1.6124, - "step": 25113 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019150578367043225, - "loss": 1.5765, - "step": 25114 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915054454440912, - "loss": 1.5308, - "step": 25115 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019150510721775013, - "loss": 1.6114, - "step": 25116 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019150476899140904, - "loss": 1.6117, - "step": 25117 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019150443076506799, - "loss": 1.5129, - "step": 25118 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019150409253872693, - "loss": 1.4561, - "step": 25119 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019150375431238584, - "loss": 1.6773, - "step": 25120 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019150341608604478, - "loss": 1.3848, - "step": 25121 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019150307785970372, - "loss": 1.5618, - "step": 25122 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019150273963336264, - "loss": 1.6673, - "step": 25123 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019150240140702158, - "loss": 1.6847, - "step": 25124 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019150206318068052, - "loss": 1.6179, - "step": 25125 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019150172495433943, - "loss": 1.6711, - "step": 25126 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019150138672799837, - "loss": 1.6527, - "step": 25127 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019150104850165732, - "loss": 1.5891, - "step": 25128 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019150071027531623, - "loss": 1.6618, - "step": 25129 - }, - { - "epoch": 0.08, - "learning_rate": 0.00019150037204897517, - "loss": 1.6123, - "step": 25130 - }, - { - "epoch": 0.08, - "learning_rate": 0.0001915000338226341, - "loss": 1.4937, - "step": 25131 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019149969559629303, - "loss": 1.5216, - "step": 25132 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019149935736995197, - "loss": 1.4618, - "step": 25133 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914990191436109, - "loss": 1.5303, - "step": 25134 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019149868091726985, - "loss": 1.4952, - "step": 25135 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914983426909288, - "loss": 1.4653, - "step": 25136 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019149800446458773, - "loss": 1.5604, - "step": 25137 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019149766623824665, - "loss": 1.4813, - "step": 25138 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914973280119056, - "loss": 1.6522, - "step": 25139 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019149698978556453, - "loss": 1.5217, - "step": 25140 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019149665155922344, - "loss": 1.4707, - "step": 25141 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019149631333288238, - "loss": 1.4531, - "step": 25142 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019149597510654132, - "loss": 1.4639, - "step": 25143 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019149563688020024, - "loss": 1.6657, - "step": 25144 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019149529865385918, - "loss": 1.4642, - "step": 25145 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019149496042751812, - "loss": 1.6187, - "step": 25146 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019149462220117703, - "loss": 1.6558, - "step": 25147 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019149428397483598, - "loss": 1.7119, - "step": 25148 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019149394574849492, - "loss": 1.573, - "step": 25149 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019149360752215383, - "loss": 1.5158, - "step": 25150 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019149326929581277, - "loss": 1.5349, - "step": 25151 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914929310694717, - "loss": 1.4366, - "step": 25152 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019149259284313063, - "loss": 1.5609, - "step": 25153 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019149225461678957, - "loss": 1.5405, - "step": 25154 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914919163904485, - "loss": 1.53, - "step": 25155 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019149157816410742, - "loss": 1.5252, - "step": 25156 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019149123993776636, - "loss": 1.7761, - "step": 25157 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914909017114253, - "loss": 1.4596, - "step": 25158 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019149056348508422, - "loss": 1.5267, - "step": 25159 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019149022525874316, - "loss": 1.5476, - "step": 25160 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914898870324021, - "loss": 1.6296, - "step": 25161 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019148954880606102, - "loss": 1.3427, - "step": 25162 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019148921057971996, - "loss": 1.6836, - "step": 25163 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914888723533789, - "loss": 1.5713, - "step": 25164 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914885341270378, - "loss": 1.4733, - "step": 25165 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019148819590069675, - "loss": 1.474, - "step": 25166 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914878576743557, - "loss": 1.5119, - "step": 25167 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914875194480146, - "loss": 1.4784, - "step": 25168 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019148718122167355, - "loss": 1.603, - "step": 25169 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914868429953325, - "loss": 1.4328, - "step": 25170 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914865047689914, - "loss": 1.5257, - "step": 25171 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019148616654265035, - "loss": 1.5221, - "step": 25172 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914858283163093, - "loss": 1.4595, - "step": 25173 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914854900899682, - "loss": 1.6476, - "step": 25174 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019148515186362714, - "loss": 1.7069, - "step": 25175 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019148481363728608, - "loss": 1.4883, - "step": 25176 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191484475410945, - "loss": 1.5176, - "step": 25177 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019148413718460394, - "loss": 1.4404, - "step": 25178 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019148379895826288, - "loss": 1.5362, - "step": 25179 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914834607319218, - "loss": 1.5155, - "step": 25180 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019148312250558074, - "loss": 1.4836, - "step": 25181 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019148278427923968, - "loss": 1.4946, - "step": 25182 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914824460528986, - "loss": 1.605, - "step": 25183 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019148210782655753, - "loss": 1.736, - "step": 25184 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019148176960021647, - "loss": 1.4823, - "step": 25185 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914814313738754, - "loss": 1.5795, - "step": 25186 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019148109314753433, - "loss": 1.6278, - "step": 25187 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019148075492119327, - "loss": 1.504, - "step": 25188 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019148041669485218, - "loss": 1.6947, - "step": 25189 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019148007846851112, - "loss": 1.6597, - "step": 25190 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914797402421701, - "loss": 1.6912, - "step": 25191 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191479402015829, - "loss": 1.4322, - "step": 25192 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019147906378948795, - "loss": 1.611, - "step": 25193 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914787255631469, - "loss": 1.6803, - "step": 25194 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914783873368058, - "loss": 1.6942, - "step": 25195 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019147804911046474, - "loss": 1.4195, - "step": 25196 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019147771088412369, - "loss": 1.7369, - "step": 25197 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914773726577826, - "loss": 1.5979, - "step": 25198 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019147703443144154, - "loss": 1.4597, - "step": 25199 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019147669620510048, - "loss": 1.4398, - "step": 25200 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914763579787594, - "loss": 1.6006, - "step": 25201 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019147601975241834, - "loss": 1.4708, - "step": 25202 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019147568152607728, - "loss": 1.559, - "step": 25203 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914753432997362, - "loss": 1.3954, - "step": 25204 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019147500507339513, - "loss": 1.7144, - "step": 25205 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019147466684705407, - "loss": 1.628, - "step": 25206 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191474328620713, - "loss": 1.6143, - "step": 25207 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019147399039437193, - "loss": 1.7276, - "step": 25208 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019147365216803087, - "loss": 1.5054, - "step": 25209 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019147331394168978, - "loss": 1.7029, - "step": 25210 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019147297571534873, - "loss": 1.5903, - "step": 25211 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019147263748900767, - "loss": 1.571, - "step": 25212 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019147229926266658, - "loss": 1.6855, - "step": 25213 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019147196103632552, - "loss": 1.4838, - "step": 25214 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019147162280998446, - "loss": 1.7126, - "step": 25215 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019147128458364338, - "loss": 1.5967, - "step": 25216 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019147094635730232, - "loss": 1.5174, - "step": 25217 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019147060813096126, - "loss": 1.6332, - "step": 25218 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019147026990462017, - "loss": 1.4212, - "step": 25219 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019146993167827911, - "loss": 1.5928, - "step": 25220 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019146959345193806, - "loss": 1.5603, - "step": 25221 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019146925522559697, - "loss": 1.5346, - "step": 25222 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914689169992559, - "loss": 1.5338, - "step": 25223 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019146857877291485, - "loss": 1.5726, - "step": 25224 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019146824054657377, - "loss": 1.6984, - "step": 25225 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914679023202327, - "loss": 1.6523, - "step": 25226 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019146756409389165, - "loss": 1.5653, - "step": 25227 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019146722586755056, - "loss": 1.4187, - "step": 25228 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914668876412095, - "loss": 1.5802, - "step": 25229 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019146654941486844, - "loss": 1.727, - "step": 25230 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019146621118852736, - "loss": 1.6654, - "step": 25231 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914658729621863, - "loss": 1.458, - "step": 25232 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019146553473584524, - "loss": 1.6017, - "step": 25233 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019146519650950416, - "loss": 1.5161, - "step": 25234 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914648582831631, - "loss": 1.4472, - "step": 25235 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019146452005682204, - "loss": 1.4831, - "step": 25236 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019146418183048095, - "loss": 1.3201, - "step": 25237 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914638436041399, - "loss": 1.5535, - "step": 25238 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019146350537779883, - "loss": 1.5462, - "step": 25239 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019146316715145775, - "loss": 1.6615, - "step": 25240 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914628289251167, - "loss": 1.5475, - "step": 25241 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019146249069877563, - "loss": 1.6175, - "step": 25242 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019146215247243454, - "loss": 1.5455, - "step": 25243 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019146181424609349, - "loss": 1.6352, - "step": 25244 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019146147601975243, - "loss": 1.4925, - "step": 25245 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019146113779341134, - "loss": 1.5946, - "step": 25246 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914607995670703, - "loss": 1.5854, - "step": 25247 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019146046134072922, - "loss": 1.5391, - "step": 25248 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019146012311438816, - "loss": 1.7052, - "step": 25249 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914597848880471, - "loss": 1.5881, - "step": 25250 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019145944666170602, - "loss": 1.5913, - "step": 25251 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019145910843536496, - "loss": 1.5174, - "step": 25252 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914587702090239, - "loss": 1.5545, - "step": 25253 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019145843198268284, - "loss": 1.5755, - "step": 25254 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019145809375634176, - "loss": 1.5209, - "step": 25255 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914577555300007, - "loss": 1.687, - "step": 25256 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019145741730365964, - "loss": 1.5626, - "step": 25257 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019145707907731855, - "loss": 1.4842, - "step": 25258 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914567408509775, - "loss": 1.5466, - "step": 25259 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019145640262463643, - "loss": 1.6157, - "step": 25260 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019145606439829535, - "loss": 1.6465, - "step": 25261 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914557261719543, - "loss": 1.501, - "step": 25262 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019145538794561323, - "loss": 1.5273, - "step": 25263 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019145504971927215, - "loss": 1.5603, - "step": 25264 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019145471149293109, - "loss": 1.6229, - "step": 25265 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019145437326659003, - "loss": 1.5867, - "step": 25266 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019145403504024894, - "loss": 1.6085, - "step": 25267 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019145369681390788, - "loss": 1.4925, - "step": 25268 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019145335858756682, - "loss": 1.5577, - "step": 25269 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019145302036122574, - "loss": 1.5221, - "step": 25270 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019145268213488468, - "loss": 1.5696, - "step": 25271 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019145234390854362, - "loss": 1.5359, - "step": 25272 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019145200568220253, - "loss": 1.5985, - "step": 25273 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019145166745586148, - "loss": 1.5417, - "step": 25274 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019145132922952042, - "loss": 1.5096, - "step": 25275 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019145099100317933, - "loss": 1.7096, - "step": 25276 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019145065277683827, - "loss": 1.4982, - "step": 25277 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914503145504972, - "loss": 1.3701, - "step": 25278 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019144997632415613, - "loss": 1.5565, - "step": 25279 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019144963809781507, - "loss": 1.4237, - "step": 25280 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191449299871474, - "loss": 1.5151, - "step": 25281 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019144896164513292, - "loss": 1.4951, - "step": 25282 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019144862341879186, - "loss": 1.3626, - "step": 25283 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914482851924508, - "loss": 1.4044, - "step": 25284 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019144794696610972, - "loss": 1.5136, - "step": 25285 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019144760873976866, - "loss": 1.6111, - "step": 25286 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914472705134276, - "loss": 1.5051, - "step": 25287 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019144693228708652, - "loss": 1.605, - "step": 25288 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019144659406074546, - "loss": 1.6046, - "step": 25289 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914462558344044, - "loss": 1.6606, - "step": 25290 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914459176080633, - "loss": 1.5746, - "step": 25291 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019144557938172225, - "loss": 1.5296, - "step": 25292 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914452411553812, - "loss": 1.4065, - "step": 25293 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914449029290401, - "loss": 1.4593, - "step": 25294 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019144456470269905, - "loss": 1.543, - "step": 25295 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191444226476358, - "loss": 1.6014, - "step": 25296 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914438882500169, - "loss": 1.3864, - "step": 25297 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019144355002367585, - "loss": 1.4576, - "step": 25298 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914432117973348, - "loss": 1.6655, - "step": 25299 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914428735709937, - "loss": 1.6359, - "step": 25300 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019144253534465264, - "loss": 1.5797, - "step": 25301 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019144219711831158, - "loss": 1.5534, - "step": 25302 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019144185889197052, - "loss": 1.5546, - "step": 25303 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019144152066562947, - "loss": 1.7584, - "step": 25304 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019144118243928838, - "loss": 1.5288, - "step": 25305 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019144084421294732, - "loss": 1.6674, - "step": 25306 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019144050598660626, - "loss": 1.5825, - "step": 25307 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019144016776026518, - "loss": 1.6685, - "step": 25308 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019143982953392412, - "loss": 1.4531, - "step": 25309 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019143949130758306, - "loss": 1.5064, - "step": 25310 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019143915308124197, - "loss": 1.5591, - "step": 25311 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914388148549009, - "loss": 1.6186, - "step": 25312 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019143847662855985, - "loss": 1.5499, - "step": 25313 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019143813840221877, - "loss": 1.6452, - "step": 25314 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914378001758777, - "loss": 1.5601, - "step": 25315 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019143746194953665, - "loss": 1.4906, - "step": 25316 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019143712372319556, - "loss": 1.4056, - "step": 25317 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914367854968545, - "loss": 1.5994, - "step": 25318 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019143644727051345, - "loss": 1.6041, - "step": 25319 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019143610904417236, - "loss": 1.5552, - "step": 25320 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914357708178313, - "loss": 1.6088, - "step": 25321 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019143543259149024, - "loss": 1.539, - "step": 25322 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019143509436514916, - "loss": 1.5325, - "step": 25323 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914347561388081, - "loss": 1.7333, - "step": 25324 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019143441791246704, - "loss": 1.5451, - "step": 25325 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019143407968612595, - "loss": 1.4818, - "step": 25326 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914337414597849, - "loss": 1.6414, - "step": 25327 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019143340323344384, - "loss": 1.4107, - "step": 25328 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019143306500710275, - "loss": 1.5801, - "step": 25329 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914327267807617, - "loss": 1.4675, - "step": 25330 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019143238855442063, - "loss": 1.5264, - "step": 25331 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019143205032807955, - "loss": 1.6139, - "step": 25332 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914317121017385, - "loss": 1.6014, - "step": 25333 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019143137387539743, - "loss": 1.625, - "step": 25334 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019143103564905634, - "loss": 1.6568, - "step": 25335 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019143069742271528, - "loss": 1.457, - "step": 25336 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019143035919637423, - "loss": 1.5412, - "step": 25337 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019143002097003314, - "loss": 1.3765, - "step": 25338 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019142968274369208, - "loss": 1.688, - "step": 25339 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019142934451735102, - "loss": 1.44, - "step": 25340 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019142900629100996, - "loss": 1.4639, - "step": 25341 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019142866806466888, - "loss": 1.5059, - "step": 25342 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019142832983832782, - "loss": 1.678, - "step": 25343 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019142799161198676, - "loss": 1.4961, - "step": 25344 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019142765338564567, - "loss": 1.7457, - "step": 25345 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019142731515930461, - "loss": 1.463, - "step": 25346 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019142697693296356, - "loss": 1.5659, - "step": 25347 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019142663870662247, - "loss": 1.42, - "step": 25348 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914263004802814, - "loss": 1.6277, - "step": 25349 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019142596225394035, - "loss": 1.4845, - "step": 25350 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019142562402759927, - "loss": 1.5435, - "step": 25351 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914252858012582, - "loss": 1.5241, - "step": 25352 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019142494757491715, - "loss": 1.576, - "step": 25353 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019142460934857606, - "loss": 1.464, - "step": 25354 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191424271122235, - "loss": 1.5551, - "step": 25355 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019142393289589394, - "loss": 1.6717, - "step": 25356 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019142359466955286, - "loss": 1.3561, - "step": 25357 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914232564432118, - "loss": 1.4424, - "step": 25358 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019142291821687074, - "loss": 1.5365, - "step": 25359 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019142257999052968, - "loss": 1.5658, - "step": 25360 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019142224176418862, - "loss": 1.3634, - "step": 25361 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019142190353784754, - "loss": 1.6308, - "step": 25362 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019142156531150648, - "loss": 1.6237, - "step": 25363 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019142122708516542, - "loss": 1.5837, - "step": 25364 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019142088885882433, - "loss": 1.5437, - "step": 25365 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019142055063248327, - "loss": 1.5855, - "step": 25366 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019142021240614222, - "loss": 1.6337, - "step": 25367 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019141987417980113, - "loss": 1.5104, - "step": 25368 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019141953595346007, - "loss": 1.4946, - "step": 25369 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191419197727119, - "loss": 1.4931, - "step": 25370 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019141885950077793, - "loss": 1.4477, - "step": 25371 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019141852127443687, - "loss": 1.3474, - "step": 25372 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914181830480958, - "loss": 1.5073, - "step": 25373 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019141784482175472, - "loss": 1.6575, - "step": 25374 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019141750659541366, - "loss": 1.5705, - "step": 25375 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914171683690726, - "loss": 1.5532, - "step": 25376 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019141683014273152, - "loss": 1.7546, - "step": 25377 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019141649191639046, - "loss": 1.5406, - "step": 25378 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914161536900494, - "loss": 1.5832, - "step": 25379 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019141581546370831, - "loss": 1.5345, - "step": 25380 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019141547723736726, - "loss": 1.6819, - "step": 25381 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914151390110262, - "loss": 1.6835, - "step": 25382 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914148007846851, - "loss": 1.487, - "step": 25383 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019141446255834405, - "loss": 1.553, - "step": 25384 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191414124332003, - "loss": 1.477, - "step": 25385 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914137861056619, - "loss": 1.3977, - "step": 25386 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019141344787932085, - "loss": 1.5397, - "step": 25387 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914131096529798, - "loss": 1.6269, - "step": 25388 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914127714266387, - "loss": 1.6326, - "step": 25389 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019141243320029764, - "loss": 1.5152, - "step": 25390 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019141209497395659, - "loss": 1.5326, - "step": 25391 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914117567476155, - "loss": 1.4101, - "step": 25392 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019141141852127444, - "loss": 1.4742, - "step": 25393 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019141108029493338, - "loss": 1.5854, - "step": 25394 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914107420685923, - "loss": 1.5769, - "step": 25395 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019141040384225124, - "loss": 1.4639, - "step": 25396 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019141006561591018, - "loss": 1.5557, - "step": 25397 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914097273895691, - "loss": 1.8325, - "step": 25398 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019140938916322803, - "loss": 1.5358, - "step": 25399 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019140905093688697, - "loss": 1.6017, - "step": 25400 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914087127105459, - "loss": 1.672, - "step": 25401 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019140837448420483, - "loss": 1.5195, - "step": 25402 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019140803625786377, - "loss": 1.5931, - "step": 25403 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019140769803152269, - "loss": 1.5905, - "step": 25404 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019140735980518163, - "loss": 1.4516, - "step": 25405 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019140702157884057, - "loss": 1.6447, - "step": 25406 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019140668335249948, - "loss": 1.4869, - "step": 25407 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019140634512615842, - "loss": 1.6566, - "step": 25408 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019140600689981736, - "loss": 1.655, - "step": 25409 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019140566867347628, - "loss": 1.4475, - "step": 25410 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019140533044713522, - "loss": 1.6313, - "step": 25411 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019140499222079416, - "loss": 1.6801, - "step": 25412 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019140465399445307, - "loss": 1.6952, - "step": 25413 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019140431576811202, - "loss": 1.5122, - "step": 25414 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019140397754177096, - "loss": 1.5639, - "step": 25415 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914036393154299, - "loss": 1.6237, - "step": 25416 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019140330108908884, - "loss": 1.5759, - "step": 25417 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019140296286274778, - "loss": 1.6647, - "step": 25418 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914026246364067, - "loss": 1.5939, - "step": 25419 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019140228641006563, - "loss": 1.3619, - "step": 25420 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019140194818372458, - "loss": 1.4498, - "step": 25421 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001914016099573835, - "loss": 1.6471, - "step": 25422 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019140127173104243, - "loss": 1.49, - "step": 25423 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019140093350470137, - "loss": 1.5116, - "step": 25424 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019140059527836029, - "loss": 1.6696, - "step": 25425 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019140025705201923, - "loss": 1.474, - "step": 25426 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019139991882567817, - "loss": 1.5385, - "step": 25427 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019139958059933708, - "loss": 1.6192, - "step": 25428 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019139924237299602, - "loss": 1.4935, - "step": 25429 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019139890414665496, - "loss": 1.4442, - "step": 25430 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019139856592031388, - "loss": 1.5266, - "step": 25431 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019139822769397282, - "loss": 1.5347, - "step": 25432 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019139788946763176, - "loss": 1.4399, - "step": 25433 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019139755124129068, - "loss": 1.4802, - "step": 25434 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019139721301494962, - "loss": 1.4995, - "step": 25435 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019139687478860856, - "loss": 1.5772, - "step": 25436 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019139653656226747, - "loss": 1.7139, - "step": 25437 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913961983359264, - "loss": 1.5761, - "step": 25438 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019139586010958535, - "loss": 1.4061, - "step": 25439 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019139552188324427, - "loss": 1.6191, - "step": 25440 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913951836569032, - "loss": 1.5221, - "step": 25441 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019139484543056215, - "loss": 1.4617, - "step": 25442 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019139450720422106, - "loss": 1.5989, - "step": 25443 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019139416897788, - "loss": 1.7763, - "step": 25444 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019139383075153895, - "loss": 1.7226, - "step": 25445 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019139349252519786, - "loss": 1.4063, - "step": 25446 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913931542988568, - "loss": 1.6346, - "step": 25447 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019139281607251574, - "loss": 1.5139, - "step": 25448 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019139247784617466, - "loss": 1.5371, - "step": 25449 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913921396198336, - "loss": 1.5857, - "step": 25450 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019139180139349254, - "loss": 1.3899, - "step": 25451 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019139146316715145, - "loss": 1.6655, - "step": 25452 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913911249408104, - "loss": 1.4209, - "step": 25453 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019139078671446934, - "loss": 1.4315, - "step": 25454 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019139044848812825, - "loss": 1.567, - "step": 25455 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913901102617872, - "loss": 1.4981, - "step": 25456 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019138977203544613, - "loss": 1.5665, - "step": 25457 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019138943380910505, - "loss": 1.4466, - "step": 25458 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191389095582764, - "loss": 1.5841, - "step": 25459 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019138875735642293, - "loss": 1.4948, - "step": 25460 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019138841913008184, - "loss": 1.4432, - "step": 25461 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019138808090374078, - "loss": 1.6063, - "step": 25462 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019138774267739972, - "loss": 1.5368, - "step": 25463 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019138740445105864, - "loss": 1.4825, - "step": 25464 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019138706622471758, - "loss": 1.6827, - "step": 25465 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019138672799837652, - "loss": 1.521, - "step": 25466 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019138638977203543, - "loss": 1.62, - "step": 25467 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019138605154569438, - "loss": 1.569, - "step": 25468 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019138571331935332, - "loss": 1.6626, - "step": 25469 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019138537509301223, - "loss": 1.592, - "step": 25470 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019138503686667117, - "loss": 1.5689, - "step": 25471 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019138469864033014, - "loss": 1.4307, - "step": 25472 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019138436041398905, - "loss": 1.5883, - "step": 25473 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191384022187648, - "loss": 1.6091, - "step": 25474 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019138368396130694, - "loss": 1.4867, - "step": 25475 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019138334573496585, - "loss": 1.6191, - "step": 25476 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913830075086248, - "loss": 1.5465, - "step": 25477 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019138266928228373, - "loss": 1.6013, - "step": 25478 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019138233105594265, - "loss": 1.7094, - "step": 25479 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913819928296016, - "loss": 1.767, - "step": 25480 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019138165460326053, - "loss": 1.5408, - "step": 25481 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019138131637691944, - "loss": 1.5988, - "step": 25482 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019138097815057838, - "loss": 1.5085, - "step": 25483 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019138063992423733, - "loss": 1.5807, - "step": 25484 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019138030169789624, - "loss": 1.6767, - "step": 25485 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019137996347155518, - "loss": 1.7032, - "step": 25486 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019137962524521412, - "loss": 1.5054, - "step": 25487 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019137928701887304, - "loss": 1.6673, - "step": 25488 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019137894879253198, - "loss": 1.559, - "step": 25489 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019137861056619092, - "loss": 1.5843, - "step": 25490 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019137827233984983, - "loss": 1.4592, - "step": 25491 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019137793411350877, - "loss": 1.6253, - "step": 25492 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019137759588716771, - "loss": 1.6168, - "step": 25493 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019137725766082663, - "loss": 1.5289, - "step": 25494 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019137691943448557, - "loss": 1.5315, - "step": 25495 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913765812081445, - "loss": 1.5865, - "step": 25496 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019137624298180343, - "loss": 1.5268, - "step": 25497 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019137590475546237, - "loss": 1.5381, - "step": 25498 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913755665291213, - "loss": 1.7481, - "step": 25499 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019137522830278022, - "loss": 1.6203, - "step": 25500 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019137489007643916, - "loss": 1.401, - "step": 25501 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913745518500981, - "loss": 1.5638, - "step": 25502 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019137421362375702, - "loss": 1.4368, - "step": 25503 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019137387539741596, - "loss": 1.6244, - "step": 25504 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913735371710749, - "loss": 1.6208, - "step": 25505 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019137319894473381, - "loss": 1.4099, - "step": 25506 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019137286071839276, - "loss": 1.421, - "step": 25507 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913725224920517, - "loss": 1.5305, - "step": 25508 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913721842657106, - "loss": 1.3474, - "step": 25509 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019137184603936955, - "loss": 1.579, - "step": 25510 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913715078130285, - "loss": 1.5596, - "step": 25511 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913711695866874, - "loss": 1.5152, - "step": 25512 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019137083136034635, - "loss": 1.5714, - "step": 25513 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913704931340053, - "loss": 1.3654, - "step": 25514 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913701549076642, - "loss": 1.421, - "step": 25515 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019136981668132314, - "loss": 1.6423, - "step": 25516 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019136947845498209, - "loss": 1.6481, - "step": 25517 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191369140228641, - "loss": 1.6136, - "step": 25518 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019136880200229994, - "loss": 1.5841, - "step": 25519 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019136846377595888, - "loss": 1.5309, - "step": 25520 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913681255496178, - "loss": 1.6743, - "step": 25521 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019136778732327674, - "loss": 1.607, - "step": 25522 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019136744909693568, - "loss": 1.5261, - "step": 25523 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913671108705946, - "loss": 1.6074, - "step": 25524 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019136677264425353, - "loss": 1.3695, - "step": 25525 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019136643441791247, - "loss": 1.5082, - "step": 25526 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913660961915714, - "loss": 1.5666, - "step": 25527 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019136575796523036, - "loss": 1.351, - "step": 25528 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913654197388893, - "loss": 1.7769, - "step": 25529 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913650815125482, - "loss": 1.8061, - "step": 25530 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019136474328620715, - "loss": 1.4855, - "step": 25531 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913644050598661, - "loss": 1.5188, - "step": 25532 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191364066833525, - "loss": 1.5378, - "step": 25533 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019136372860718395, - "loss": 1.5194, - "step": 25534 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913633903808429, - "loss": 1.6135, - "step": 25535 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913630521545018, - "loss": 1.491, - "step": 25536 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019136271392816075, - "loss": 1.4916, - "step": 25537 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913623757018197, - "loss": 1.434, - "step": 25538 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913620374754786, - "loss": 1.5607, - "step": 25539 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019136169924913754, - "loss": 1.4961, - "step": 25540 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019136136102279648, - "loss": 1.4689, - "step": 25541 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913610227964554, - "loss": 1.5407, - "step": 25542 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019136068457011434, - "loss": 1.55, - "step": 25543 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019136034634377328, - "loss": 1.5564, - "step": 25544 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913600081174322, - "loss": 1.5296, - "step": 25545 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019135966989109113, - "loss": 1.6169, - "step": 25546 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019135933166475008, - "loss": 1.7001, - "step": 25547 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191358993438409, - "loss": 1.6271, - "step": 25548 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019135865521206793, - "loss": 1.6356, - "step": 25549 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019135831698572687, - "loss": 1.4584, - "step": 25550 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019135797875938579, - "loss": 1.5752, - "step": 25551 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019135764053304473, - "loss": 1.4547, - "step": 25552 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019135730230670367, - "loss": 1.6391, - "step": 25553 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019135696408036258, - "loss": 1.475, - "step": 25554 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019135662585402152, - "loss": 1.6115, - "step": 25555 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019135628762768046, - "loss": 1.6417, - "step": 25556 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019135594940133938, - "loss": 1.5353, - "step": 25557 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019135561117499832, - "loss": 1.722, - "step": 25558 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019135527294865726, - "loss": 1.5541, - "step": 25559 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019135493472231617, - "loss": 1.6884, - "step": 25560 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019135459649597512, - "loss": 1.6452, - "step": 25561 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019135425826963406, - "loss": 1.4703, - "step": 25562 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019135392004329297, - "loss": 1.4255, - "step": 25563 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913535818169519, - "loss": 1.5524, - "step": 25564 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019135324359061085, - "loss": 1.3849, - "step": 25565 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019135290536426977, - "loss": 1.544, - "step": 25566 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913525671379287, - "loss": 1.5496, - "step": 25567 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019135222891158765, - "loss": 1.3376, - "step": 25568 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019135189068524656, - "loss": 1.6884, - "step": 25569 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913515524589055, - "loss": 1.5133, - "step": 25570 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019135121423256445, - "loss": 1.5958, - "step": 25571 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019135087600622336, - "loss": 1.6505, - "step": 25572 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913505377798823, - "loss": 1.532, - "step": 25573 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019135019955354124, - "loss": 1.5522, - "step": 25574 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019134986132720016, - "loss": 1.4067, - "step": 25575 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913495231008591, - "loss": 1.4518, - "step": 25576 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019134918487451804, - "loss": 1.4724, - "step": 25577 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019134884664817695, - "loss": 1.5389, - "step": 25578 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913485084218359, - "loss": 1.5194, - "step": 25579 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019134817019549483, - "loss": 1.5644, - "step": 25580 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019134783196915375, - "loss": 1.5834, - "step": 25581 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913474937428127, - "loss": 1.6202, - "step": 25582 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019134715551647163, - "loss": 1.7366, - "step": 25583 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019134681729013057, - "loss": 1.5378, - "step": 25584 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913464790637895, - "loss": 1.5941, - "step": 25585 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019134614083744843, - "loss": 1.5067, - "step": 25586 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019134580261110737, - "loss": 1.5767, - "step": 25587 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913454643847663, - "loss": 1.5857, - "step": 25588 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019134512615842522, - "loss": 1.5913, - "step": 25589 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019134478793208416, - "loss": 1.5216, - "step": 25590 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913444497057431, - "loss": 1.5885, - "step": 25591 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019134411147940202, - "loss": 1.5471, - "step": 25592 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019134377325306096, - "loss": 1.5992, - "step": 25593 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913434350267199, - "loss": 1.4681, - "step": 25594 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019134309680037882, - "loss": 1.349, - "step": 25595 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019134275857403776, - "loss": 1.3918, - "step": 25596 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913424203476967, - "loss": 1.6848, - "step": 25597 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913420821213556, - "loss": 1.5218, - "step": 25598 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019134174389501455, - "loss": 1.6497, - "step": 25599 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913414056686735, - "loss": 1.6449, - "step": 25600 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913410674423324, - "loss": 1.5406, - "step": 25601 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019134072921599135, - "loss": 1.6744, - "step": 25602 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913403909896503, - "loss": 1.6157, - "step": 25603 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913400527633092, - "loss": 1.6507, - "step": 25604 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019133971453696815, - "loss": 1.6118, - "step": 25605 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913393763106271, - "loss": 1.5506, - "step": 25606 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019133903808428603, - "loss": 1.5981, - "step": 25607 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019133869985794494, - "loss": 1.6668, - "step": 25608 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019133836163160388, - "loss": 1.466, - "step": 25609 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019133802340526283, - "loss": 1.4672, - "step": 25610 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019133768517892174, - "loss": 1.645, - "step": 25611 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019133734695258068, - "loss": 1.4641, - "step": 25612 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019133700872623962, - "loss": 1.5443, - "step": 25613 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019133667049989854, - "loss": 1.601, - "step": 25614 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019133633227355748, - "loss": 1.4509, - "step": 25615 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019133599404721642, - "loss": 1.4836, - "step": 25616 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019133565582087533, - "loss": 1.5501, - "step": 25617 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019133531759453427, - "loss": 1.4198, - "step": 25618 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019133497936819321, - "loss": 1.385, - "step": 25619 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019133464114185213, - "loss": 1.5436, - "step": 25620 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019133430291551107, - "loss": 1.4443, - "step": 25621 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019133396468917, - "loss": 1.5956, - "step": 25622 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019133362646282892, - "loss": 1.5176, - "step": 25623 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019133328823648787, - "loss": 1.5354, - "step": 25624 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913329500101468, - "loss": 1.6289, - "step": 25625 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019133261178380572, - "loss": 1.3854, - "step": 25626 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019133227355746466, - "loss": 1.4907, - "step": 25627 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913319353311236, - "loss": 1.6576, - "step": 25628 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019133159710478252, - "loss": 1.5396, - "step": 25629 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019133125887844146, - "loss": 1.6129, - "step": 25630 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913309206521004, - "loss": 1.6266, - "step": 25631 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913305824257593, - "loss": 1.6565, - "step": 25632 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019133024419941825, - "loss": 1.6744, - "step": 25633 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913299059730772, - "loss": 1.6219, - "step": 25634 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913295677467361, - "loss": 1.4596, - "step": 25635 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019132922952039505, - "loss": 1.6471, - "step": 25636 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191328891294054, - "loss": 1.6353, - "step": 25637 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913285530677129, - "loss": 1.6574, - "step": 25638 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019132821484137185, - "loss": 1.5093, - "step": 25639 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913278766150308, - "loss": 1.3944, - "step": 25640 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019132753838868973, - "loss": 1.598, - "step": 25641 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019132720016234867, - "loss": 1.6299, - "step": 25642 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019132686193600758, - "loss": 1.505, - "step": 25643 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019132652370966653, - "loss": 1.5571, - "step": 25644 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019132618548332547, - "loss": 1.5302, - "step": 25645 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019132584725698438, - "loss": 1.5967, - "step": 25646 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019132550903064332, - "loss": 1.639, - "step": 25647 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019132517080430226, - "loss": 1.5549, - "step": 25648 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019132483257796118, - "loss": 1.503, - "step": 25649 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019132449435162012, - "loss": 1.6225, - "step": 25650 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019132415612527906, - "loss": 1.4521, - "step": 25651 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019132381789893797, - "loss": 1.4048, - "step": 25652 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019132347967259691, - "loss": 1.3951, - "step": 25653 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019132314144625586, - "loss": 1.5342, - "step": 25654 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019132280321991477, - "loss": 1.6328, - "step": 25655 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913224649935737, - "loss": 1.3609, - "step": 25656 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019132212676723265, - "loss": 1.6223, - "step": 25657 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019132178854089157, - "loss": 1.7809, - "step": 25658 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913214503145505, - "loss": 1.549, - "step": 25659 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019132111208820945, - "loss": 1.5745, - "step": 25660 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019132077386186836, - "loss": 1.5318, - "step": 25661 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913204356355273, - "loss": 1.7054, - "step": 25662 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019132009740918624, - "loss": 1.7463, - "step": 25663 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019131975918284516, - "loss": 1.4199, - "step": 25664 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913194209565041, - "loss": 1.5088, - "step": 25665 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019131908273016304, - "loss": 1.4489, - "step": 25666 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019131874450382196, - "loss": 1.6176, - "step": 25667 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913184062774809, - "loss": 1.6077, - "step": 25668 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019131806805113984, - "loss": 1.5322, - "step": 25669 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019131772982479875, - "loss": 1.5598, - "step": 25670 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913173915984577, - "loss": 1.4269, - "step": 25671 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019131705337211663, - "loss": 1.4054, - "step": 25672 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019131671514577555, - "loss": 1.5215, - "step": 25673 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913163769194345, - "loss": 1.6458, - "step": 25674 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019131603869309343, - "loss": 1.6033, - "step": 25675 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019131570046675234, - "loss": 1.6595, - "step": 25676 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019131536224041129, - "loss": 1.6903, - "step": 25677 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019131502401407023, - "loss": 1.5584, - "step": 25678 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019131468578772914, - "loss": 1.5257, - "step": 25679 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019131434756138808, - "loss": 1.6386, - "step": 25680 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019131400933504702, - "loss": 1.5384, - "step": 25681 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019131367110870594, - "loss": 1.4511, - "step": 25682 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019131333288236488, - "loss": 1.7349, - "step": 25683 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019131299465602382, - "loss": 1.6966, - "step": 25684 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019131265642968273, - "loss": 1.5389, - "step": 25685 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019131231820334167, - "loss": 1.5834, - "step": 25686 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019131197997700062, - "loss": 1.6271, - "step": 25687 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019131164175065953, - "loss": 1.7496, - "step": 25688 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019131130352431847, - "loss": 1.5919, - "step": 25689 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913109652979774, - "loss": 1.4942, - "step": 25690 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019131062707163633, - "loss": 1.4279, - "step": 25691 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019131028884529527, - "loss": 1.6174, - "step": 25692 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913099506189542, - "loss": 1.7016, - "step": 25693 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019130961239261315, - "loss": 1.6935, - "step": 25694 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019130927416627206, - "loss": 1.5777, - "step": 25695 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191308935939931, - "loss": 1.6022, - "step": 25696 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019130859771358995, - "loss": 1.5524, - "step": 25697 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913082594872489, - "loss": 1.6157, - "step": 25698 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019130792126090783, - "loss": 1.5864, - "step": 25699 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019130758303456674, - "loss": 1.527, - "step": 25700 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019130724480822568, - "loss": 1.4659, - "step": 25701 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019130690658188462, - "loss": 1.5822, - "step": 25702 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019130656835554354, - "loss": 1.5534, - "step": 25703 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019130623012920248, - "loss": 1.5858, - "step": 25704 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019130589190286142, - "loss": 1.3683, - "step": 25705 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019130555367652033, - "loss": 1.7816, - "step": 25706 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019130521545017928, - "loss": 1.4551, - "step": 25707 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019130487722383822, - "loss": 1.4818, - "step": 25708 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019130453899749713, - "loss": 1.5451, - "step": 25709 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019130420077115607, - "loss": 1.5293, - "step": 25710 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191303862544815, - "loss": 1.723, - "step": 25711 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019130352431847393, - "loss": 1.6704, - "step": 25712 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019130318609213287, - "loss": 1.6307, - "step": 25713 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913028478657918, - "loss": 1.7303, - "step": 25714 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019130250963945072, - "loss": 1.5343, - "step": 25715 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019130217141310966, - "loss": 1.4528, - "step": 25716 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913018331867686, - "loss": 1.4884, - "step": 25717 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019130149496042752, - "loss": 1.4887, - "step": 25718 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019130115673408646, - "loss": 1.5163, - "step": 25719 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001913008185077454, - "loss": 1.7423, - "step": 25720 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019130048028140432, - "loss": 1.6082, - "step": 25721 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019130014205506326, - "loss": 1.44, - "step": 25722 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912998038287222, - "loss": 1.5514, - "step": 25723 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912994656023811, - "loss": 1.6516, - "step": 25724 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019129912737604005, - "loss": 1.4897, - "step": 25725 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191298789149699, - "loss": 1.53, - "step": 25726 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912984509233579, - "loss": 1.5791, - "step": 25727 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019129811269701685, - "loss": 1.6672, - "step": 25728 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912977744706758, - "loss": 1.5147, - "step": 25729 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912974362443347, - "loss": 1.5541, - "step": 25730 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019129709801799365, - "loss": 1.6451, - "step": 25731 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912967597916526, - "loss": 1.5921, - "step": 25732 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912964215653115, - "loss": 1.5703, - "step": 25733 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019129608333897044, - "loss": 1.4325, - "step": 25734 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019129574511262938, - "loss": 1.519, - "step": 25735 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912954068862883, - "loss": 1.7057, - "step": 25736 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019129506865994724, - "loss": 1.5812, - "step": 25737 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019129473043360618, - "loss": 1.7005, - "step": 25738 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912943922072651, - "loss": 1.5187, - "step": 25739 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019129405398092403, - "loss": 1.5144, - "step": 25740 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019129371575458298, - "loss": 1.4913, - "step": 25741 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912933775282419, - "loss": 1.6121, - "step": 25742 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019129303930190083, - "loss": 1.6293, - "step": 25743 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019129270107555977, - "loss": 1.6564, - "step": 25744 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019129236284921869, - "loss": 1.5118, - "step": 25745 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019129202462287763, - "loss": 1.5851, - "step": 25746 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019129168639653657, - "loss": 1.6603, - "step": 25747 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019129134817019548, - "loss": 1.695, - "step": 25748 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019129100994385442, - "loss": 1.5349, - "step": 25749 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019129067171751336, - "loss": 1.6531, - "step": 25750 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019129033349117228, - "loss": 1.7051, - "step": 25751 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019128999526483122, - "loss": 1.6287, - "step": 25752 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912896570384902, - "loss": 1.6007, - "step": 25753 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912893188121491, - "loss": 1.6335, - "step": 25754 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019128898058580804, - "loss": 1.4694, - "step": 25755 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019128864235946698, - "loss": 1.6213, - "step": 25756 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912883041331259, - "loss": 1.5932, - "step": 25757 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019128796590678484, - "loss": 1.6279, - "step": 25758 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019128762768044378, - "loss": 1.5782, - "step": 25759 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912872894541027, - "loss": 1.6364, - "step": 25760 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019128695122776164, - "loss": 1.4703, - "step": 25761 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019128661300142058, - "loss": 1.6642, - "step": 25762 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912862747750795, - "loss": 1.5376, - "step": 25763 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019128593654873843, - "loss": 1.4905, - "step": 25764 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019128559832239737, - "loss": 1.5412, - "step": 25765 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912852600960563, - "loss": 1.579, - "step": 25766 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019128492186971523, - "loss": 1.4603, - "step": 25767 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019128458364337417, - "loss": 1.4784, - "step": 25768 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019128424541703308, - "loss": 1.435, - "step": 25769 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019128390719069203, - "loss": 1.4006, - "step": 25770 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019128356896435097, - "loss": 1.6197, - "step": 25771 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019128323073800988, - "loss": 1.5539, - "step": 25772 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019128289251166882, - "loss": 1.5381, - "step": 25773 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019128255428532776, - "loss": 1.5248, - "step": 25774 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019128221605898668, - "loss": 1.5576, - "step": 25775 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019128187783264562, - "loss": 1.5484, - "step": 25776 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019128153960630456, - "loss": 1.6404, - "step": 25777 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019128120137996347, - "loss": 1.5749, - "step": 25778 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019128086315362241, - "loss": 1.5576, - "step": 25779 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019128052492728136, - "loss": 1.5877, - "step": 25780 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019128018670094027, - "loss": 1.615, - "step": 25781 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912798484745992, - "loss": 1.6223, - "step": 25782 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019127951024825815, - "loss": 1.648, - "step": 25783 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019127917202191707, - "loss": 1.4344, - "step": 25784 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191278833795576, - "loss": 1.5127, - "step": 25785 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019127849556923495, - "loss": 1.5522, - "step": 25786 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019127815734289386, - "loss": 1.3989, - "step": 25787 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912778191165528, - "loss": 1.4888, - "step": 25788 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019127748089021174, - "loss": 1.6322, - "step": 25789 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019127714266387066, - "loss": 1.5367, - "step": 25790 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912768044375296, - "loss": 1.715, - "step": 25791 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019127646621118854, - "loss": 1.6118, - "step": 25792 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019127612798484745, - "loss": 1.6567, - "step": 25793 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912757897585064, - "loss": 1.4292, - "step": 25794 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019127545153216534, - "loss": 1.551, - "step": 25795 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019127511330582425, - "loss": 1.5183, - "step": 25796 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912747750794832, - "loss": 1.5625, - "step": 25797 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019127443685314213, - "loss": 1.6337, - "step": 25798 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019127409862680105, - "loss": 1.6477, - "step": 25799 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019127376040046, - "loss": 1.3706, - "step": 25800 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019127342217411893, - "loss": 1.6624, - "step": 25801 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019127308394777784, - "loss": 1.6135, - "step": 25802 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019127274572143678, - "loss": 1.7092, - "step": 25803 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019127240749509573, - "loss": 1.6274, - "step": 25804 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019127206926875464, - "loss": 1.5381, - "step": 25805 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019127173104241358, - "loss": 1.5704, - "step": 25806 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019127139281607252, - "loss": 1.8073, - "step": 25807 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019127105458973144, - "loss": 1.6274, - "step": 25808 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912707163633904, - "loss": 1.6053, - "step": 25809 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019127037813704935, - "loss": 1.5566, - "step": 25810 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019127003991070826, - "loss": 1.5582, - "step": 25811 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912697016843672, - "loss": 1.5731, - "step": 25812 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019126936345802614, - "loss": 1.5018, - "step": 25813 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019126902523168506, - "loss": 1.5554, - "step": 25814 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191268687005344, - "loss": 1.636, - "step": 25815 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019126834877900294, - "loss": 1.5749, - "step": 25816 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019126801055266185, - "loss": 1.5241, - "step": 25817 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912676723263208, - "loss": 1.5786, - "step": 25818 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019126733409997973, - "loss": 1.4254, - "step": 25819 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019126699587363865, - "loss": 1.6742, - "step": 25820 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912666576472976, - "loss": 1.6476, - "step": 25821 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019126631942095653, - "loss": 1.6253, - "step": 25822 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019126598119461544, - "loss": 1.6411, - "step": 25823 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019126564296827439, - "loss": 1.4557, - "step": 25824 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019126530474193333, - "loss": 1.5491, - "step": 25825 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019126496651559224, - "loss": 1.7658, - "step": 25826 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019126462828925118, - "loss": 1.4747, - "step": 25827 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019126429006291012, - "loss": 1.508, - "step": 25828 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019126395183656904, - "loss": 1.4755, - "step": 25829 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019126361361022798, - "loss": 1.5514, - "step": 25830 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019126327538388692, - "loss": 1.6066, - "step": 25831 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019126293715754583, - "loss": 1.4585, - "step": 25832 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019126259893120477, - "loss": 1.5589, - "step": 25833 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019126226070486372, - "loss": 1.44, - "step": 25834 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019126192247852263, - "loss": 1.6392, - "step": 25835 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019126158425218157, - "loss": 1.7183, - "step": 25836 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912612460258405, - "loss": 1.5229, - "step": 25837 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019126090779949943, - "loss": 1.4818, - "step": 25838 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019126056957315837, - "loss": 1.5165, - "step": 25839 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912602313468173, - "loss": 1.5525, - "step": 25840 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019125989312047622, - "loss": 1.5612, - "step": 25841 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019125955489413516, - "loss": 1.7374, - "step": 25842 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912592166677941, - "loss": 1.5857, - "step": 25843 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019125887844145302, - "loss": 1.6075, - "step": 25844 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019125854021511196, - "loss": 1.2945, - "step": 25845 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912582019887709, - "loss": 1.5976, - "step": 25846 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019125786376242982, - "loss": 1.4488, - "step": 25847 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019125752553608876, - "loss": 1.7046, - "step": 25848 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912571873097477, - "loss": 1.5522, - "step": 25849 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912568490834066, - "loss": 1.5087, - "step": 25850 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019125651085706555, - "loss": 1.4192, - "step": 25851 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912561726307245, - "loss": 1.6721, - "step": 25852 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912558344043834, - "loss": 1.5114, - "step": 25853 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019125549617804235, - "loss": 1.5676, - "step": 25854 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912551579517013, - "loss": 1.6044, - "step": 25855 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912548197253602, - "loss": 1.6955, - "step": 25856 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019125448149901915, - "loss": 1.3864, - "step": 25857 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019125414327267809, - "loss": 1.4829, - "step": 25858 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191253805046337, - "loss": 1.5094, - "step": 25859 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019125346681999594, - "loss": 1.6308, - "step": 25860 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019125312859365488, - "loss": 1.5209, - "step": 25861 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912527903673138, - "loss": 1.4266, - "step": 25862 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019125245214097274, - "loss": 1.5894, - "step": 25863 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019125211391463168, - "loss": 1.7725, - "step": 25864 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019125177568829062, - "loss": 1.5583, - "step": 25865 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019125143746194956, - "loss": 1.6897, - "step": 25866 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019125109923560848, - "loss": 1.6899, - "step": 25867 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019125076100926742, - "loss": 1.6165, - "step": 25868 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019125042278292636, - "loss": 1.4525, - "step": 25869 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019125008455658527, - "loss": 1.4294, - "step": 25870 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912497463302442, - "loss": 1.5304, - "step": 25871 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019124940810390315, - "loss": 1.4729, - "step": 25872 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019124906987756207, - "loss": 1.4049, - "step": 25873 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191248731651221, - "loss": 1.5574, - "step": 25874 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019124839342487995, - "loss": 1.634, - "step": 25875 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912480551985389, - "loss": 1.5427, - "step": 25876 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912477169721978, - "loss": 1.6227, - "step": 25877 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019124737874585675, - "loss": 1.4259, - "step": 25878 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912470405195157, - "loss": 1.6385, - "step": 25879 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912467022931746, - "loss": 1.6035, - "step": 25880 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019124636406683354, - "loss": 1.514, - "step": 25881 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019124602584049248, - "loss": 1.6078, - "step": 25882 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912456876141514, - "loss": 1.5059, - "step": 25883 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019124534938781034, - "loss": 1.4833, - "step": 25884 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019124501116146928, - "loss": 1.7002, - "step": 25885 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912446729351282, - "loss": 1.485, - "step": 25886 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019124433470878714, - "loss": 1.4912, - "step": 25887 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019124399648244608, - "loss": 1.5717, - "step": 25888 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191243658256105, - "loss": 1.4374, - "step": 25889 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019124332002976393, - "loss": 1.5592, - "step": 25890 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019124298180342287, - "loss": 1.5167, - "step": 25891 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912426435770818, - "loss": 1.498, - "step": 25892 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019124230535074073, - "loss": 1.5286, - "step": 25893 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019124196712439967, - "loss": 1.6036, - "step": 25894 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019124162889805858, - "loss": 1.5559, - "step": 25895 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019124129067171752, - "loss": 1.6587, - "step": 25896 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019124095244537647, - "loss": 1.7823, - "step": 25897 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019124061421903538, - "loss": 1.6473, - "step": 25898 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019124027599269432, - "loss": 1.4484, - "step": 25899 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019123993776635326, - "loss": 1.5953, - "step": 25900 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019123959954001218, - "loss": 1.6685, - "step": 25901 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019123926131367112, - "loss": 1.4896, - "step": 25902 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019123892308733006, - "loss": 1.5017, - "step": 25903 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019123858486098897, - "loss": 1.6811, - "step": 25904 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912382466346479, - "loss": 1.5322, - "step": 25905 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019123790840830685, - "loss": 1.5598, - "step": 25906 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019123757018196577, - "loss": 1.5168, - "step": 25907 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912372319556247, - "loss": 1.3576, - "step": 25908 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019123689372928365, - "loss": 1.4305, - "step": 25909 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019123655550294256, - "loss": 1.5589, - "step": 25910 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912362172766015, - "loss": 1.6817, - "step": 25911 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019123587905026045, - "loss": 1.6162, - "step": 25912 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019123554082391936, - "loss": 1.5254, - "step": 25913 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912352025975783, - "loss": 1.6783, - "step": 25914 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019123486437123724, - "loss": 1.6631, - "step": 25915 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019123452614489616, - "loss": 1.7845, - "step": 25916 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912341879185551, - "loss": 1.4445, - "step": 25917 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019123384969221404, - "loss": 1.4729, - "step": 25918 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019123351146587295, - "loss": 1.6014, - "step": 25919 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912331732395319, - "loss": 1.6022, - "step": 25920 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019123283501319084, - "loss": 1.5952, - "step": 25921 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019123249678684978, - "loss": 1.5017, - "step": 25922 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019123215856050872, - "loss": 1.6919, - "step": 25923 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019123182033416763, - "loss": 1.4639, - "step": 25924 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019123148210782657, - "loss": 1.5757, - "step": 25925 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019123114388148551, - "loss": 1.6428, - "step": 25926 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019123080565514443, - "loss": 1.4696, - "step": 25927 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019123046742880337, - "loss": 1.5489, - "step": 25928 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912301292024623, - "loss": 1.5728, - "step": 25929 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019122979097612123, - "loss": 1.4965, - "step": 25930 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019122945274978017, - "loss": 1.8254, - "step": 25931 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912291145234391, - "loss": 1.6549, - "step": 25932 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019122877629709802, - "loss": 1.6312, - "step": 25933 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019122843807075696, - "loss": 1.5607, - "step": 25934 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912280998444159, - "loss": 1.5032, - "step": 25935 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019122776161807482, - "loss": 1.5882, - "step": 25936 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019122742339173376, - "loss": 1.6678, - "step": 25937 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912270851653927, - "loss": 1.4281, - "step": 25938 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019122674693905161, - "loss": 1.5167, - "step": 25939 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019122640871271056, - "loss": 1.6143, - "step": 25940 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912260704863695, - "loss": 1.4808, - "step": 25941 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912257322600284, - "loss": 1.5204, - "step": 25942 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019122539403368735, - "loss": 1.4908, - "step": 25943 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912250558073463, - "loss": 1.529, - "step": 25944 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912247175810052, - "loss": 1.5266, - "step": 25945 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019122437935466415, - "loss": 1.511, - "step": 25946 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912240411283231, - "loss": 1.5071, - "step": 25947 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191223702901982, - "loss": 1.4821, - "step": 25948 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019122336467564094, - "loss": 1.68, - "step": 25949 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019122302644929989, - "loss": 1.5239, - "step": 25950 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912226882229588, - "loss": 1.6345, - "step": 25951 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019122234999661774, - "loss": 1.6085, - "step": 25952 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019122201177027668, - "loss": 1.4499, - "step": 25953 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912216735439356, - "loss": 1.5389, - "step": 25954 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019122133531759454, - "loss": 1.694, - "step": 25955 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019122099709125348, - "loss": 1.5311, - "step": 25956 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912206588649124, - "loss": 1.4751, - "step": 25957 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019122032063857133, - "loss": 1.5646, - "step": 25958 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019121998241223027, - "loss": 1.5776, - "step": 25959 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019121964418588922, - "loss": 1.7597, - "step": 25960 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019121930595954813, - "loss": 1.6602, - "step": 25961 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019121896773320707, - "loss": 1.664, - "step": 25962 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191218629506866, - "loss": 1.7472, - "step": 25963 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019121829128052493, - "loss": 1.6263, - "step": 25964 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019121795305418387, - "loss": 1.5943, - "step": 25965 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912176148278428, - "loss": 1.4912, - "step": 25966 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019121727660150172, - "loss": 1.5272, - "step": 25967 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019121693837516066, - "loss": 1.5408, - "step": 25968 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912166001488196, - "loss": 1.656, - "step": 25969 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019121626192247852, - "loss": 1.5014, - "step": 25970 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019121592369613746, - "loss": 1.6103, - "step": 25971 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912155854697964, - "loss": 1.5599, - "step": 25972 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019121524724345531, - "loss": 1.5168, - "step": 25973 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019121490901711426, - "loss": 1.4172, - "step": 25974 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912145707907732, - "loss": 1.6095, - "step": 25975 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912142325644321, - "loss": 1.5425, - "step": 25976 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019121389433809108, - "loss": 1.6204, - "step": 25977 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019121355611175, - "loss": 1.4884, - "step": 25978 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019121321788540893, - "loss": 1.5895, - "step": 25979 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019121287965906788, - "loss": 1.5809, - "step": 25980 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912125414327268, - "loss": 1.4633, - "step": 25981 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019121220320638573, - "loss": 1.5777, - "step": 25982 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019121186498004467, - "loss": 1.5813, - "step": 25983 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019121152675370359, - "loss": 1.5432, - "step": 25984 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019121118852736253, - "loss": 1.4385, - "step": 25985 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019121085030102147, - "loss": 1.3937, - "step": 25986 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019121051207468038, - "loss": 1.5948, - "step": 25987 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019121017384833932, - "loss": 1.6465, - "step": 25988 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019120983562199826, - "loss": 1.4821, - "step": 25989 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019120949739565718, - "loss": 1.4195, - "step": 25990 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019120915916931612, - "loss": 1.6388, - "step": 25991 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019120882094297506, - "loss": 1.4381, - "step": 25992 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019120848271663397, - "loss": 1.5775, - "step": 25993 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019120814449029292, - "loss": 1.7201, - "step": 25994 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019120780626395186, - "loss": 1.646, - "step": 25995 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019120746803761077, - "loss": 1.5198, - "step": 25996 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912071298112697, - "loss": 1.593, - "step": 25997 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019120679158492865, - "loss": 1.5924, - "step": 25998 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019120645335858757, - "loss": 1.5582, - "step": 25999 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912061151322465, - "loss": 1.5458, - "step": 26000 - }, - { - "epoch": 0.09, - "eval_loss": 1.3441256009615385, - "mrc_em": 25.0, - "mrc_f1": 34.23333333333334, - "nli_acc": 0.16, - "step": 26000, - "sts_f1": 0.6821705426356589, - "ynat_f1": 0.6445167685585804 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019120577690590545, - "loss": 1.4044, - "step": 26001 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019120543867956436, - "loss": 1.6398, - "step": 26002 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912051004532233, - "loss": 1.7013, - "step": 26003 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019120476222688225, - "loss": 1.5177, - "step": 26004 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019120442400054116, - "loss": 1.6429, - "step": 26005 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912040857742001, - "loss": 1.5715, - "step": 26006 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019120374754785904, - "loss": 1.4118, - "step": 26007 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019120340932151796, - "loss": 1.4388, - "step": 26008 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912030710951769, - "loss": 1.6472, - "step": 26009 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019120273286883584, - "loss": 1.4597, - "step": 26010 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019120239464249475, - "loss": 1.6411, - "step": 26011 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912020564161537, - "loss": 1.5772, - "step": 26012 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019120171818981263, - "loss": 1.6134, - "step": 26013 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019120137996347155, - "loss": 1.696, - "step": 26014 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001912010417371305, - "loss": 1.5918, - "step": 26015 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019120070351078943, - "loss": 1.5532, - "step": 26016 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019120036528444835, - "loss": 1.4881, - "step": 26017 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019120002705810729, - "loss": 1.6395, - "step": 26018 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019119968883176623, - "loss": 1.6471, - "step": 26019 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019119935060542514, - "loss": 1.5203, - "step": 26020 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019119901237908408, - "loss": 1.3304, - "step": 26021 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019119867415274302, - "loss": 1.6711, - "step": 26022 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019119833592640194, - "loss": 1.6427, - "step": 26023 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019119799770006088, - "loss": 1.4615, - "step": 26024 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019119765947371982, - "loss": 1.6451, - "step": 26025 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019119732124737873, - "loss": 1.5524, - "step": 26026 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019119698302103768, - "loss": 1.564, - "step": 26027 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019119664479469662, - "loss": 1.518, - "step": 26028 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019119630656835553, - "loss": 1.5635, - "step": 26029 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019119596834201447, - "loss": 1.4419, - "step": 26030 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911956301156734, - "loss": 1.6263, - "step": 26031 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019119529188933233, - "loss": 1.4995, - "step": 26032 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019119495366299127, - "loss": 1.5576, - "step": 26033 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019119461543665024, - "loss": 1.5849, - "step": 26034 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019119427721030915, - "loss": 1.6465, - "step": 26035 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911939389839681, - "loss": 1.5165, - "step": 26036 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019119360075762703, - "loss": 1.3757, - "step": 26037 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019119326253128595, - "loss": 1.3794, - "step": 26038 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911929243049449, - "loss": 1.3155, - "step": 26039 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019119258607860383, - "loss": 1.5945, - "step": 26040 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019119224785226274, - "loss": 1.4549, - "step": 26041 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019119190962592168, - "loss": 1.579, - "step": 26042 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019119157139958063, - "loss": 1.6799, - "step": 26043 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019119123317323954, - "loss": 1.4241, - "step": 26044 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019119089494689848, - "loss": 1.4262, - "step": 26045 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019119055672055742, - "loss": 1.5037, - "step": 26046 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019119021849421634, - "loss": 1.6945, - "step": 26047 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019118988026787528, - "loss": 1.481, - "step": 26048 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019118954204153422, - "loss": 1.4312, - "step": 26049 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019118920381519313, - "loss": 1.7218, - "step": 26050 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019118886558885207, - "loss": 1.5955, - "step": 26051 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019118852736251101, - "loss": 1.5363, - "step": 26052 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019118818913616993, - "loss": 1.5255, - "step": 26053 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019118785090982887, - "loss": 1.7459, - "step": 26054 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911875126834878, - "loss": 1.6713, - "step": 26055 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019118717445714672, - "loss": 1.6182, - "step": 26056 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019118683623080567, - "loss": 1.6404, - "step": 26057 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911864980044646, - "loss": 1.5561, - "step": 26058 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019118615977812352, - "loss": 1.7478, - "step": 26059 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019118582155178246, - "loss": 1.4796, - "step": 26060 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911854833254414, - "loss": 1.379, - "step": 26061 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019118514509910032, - "loss": 1.4835, - "step": 26062 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019118480687275926, - "loss": 1.6119, - "step": 26063 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911844686464182, - "loss": 1.5078, - "step": 26064 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911841304200771, - "loss": 1.6564, - "step": 26065 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019118379219373605, - "loss": 1.4286, - "step": 26066 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191183453967395, - "loss": 1.6131, - "step": 26067 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911831157410539, - "loss": 1.606, - "step": 26068 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019118277751471285, - "loss": 1.7582, - "step": 26069 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911824392883718, - "loss": 1.5403, - "step": 26070 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911821010620307, - "loss": 1.7165, - "step": 26071 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019118176283568965, - "loss": 1.6439, - "step": 26072 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911814246093486, - "loss": 1.5741, - "step": 26073 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911810863830075, - "loss": 1.4637, - "step": 26074 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019118074815666644, - "loss": 1.4151, - "step": 26075 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019118040993032538, - "loss": 1.6529, - "step": 26076 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911800717039843, - "loss": 1.4726, - "step": 26077 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019117973347764324, - "loss": 1.4748, - "step": 26078 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019117939525130218, - "loss": 1.4199, - "step": 26079 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911790570249611, - "loss": 1.6107, - "step": 26080 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019117871879862004, - "loss": 1.3803, - "step": 26081 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019117838057227898, - "loss": 1.5823, - "step": 26082 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911780423459379, - "loss": 1.5832, - "step": 26083 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019117770411959683, - "loss": 1.5561, - "step": 26084 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019117736589325577, - "loss": 1.4634, - "step": 26085 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911770276669147, - "loss": 1.486, - "step": 26086 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019117668944057363, - "loss": 1.5906, - "step": 26087 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019117635121423257, - "loss": 1.5961, - "step": 26088 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019117601298789148, - "loss": 1.5563, - "step": 26089 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019117567476155045, - "loss": 1.4418, - "step": 26090 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911753365352094, - "loss": 1.6016, - "step": 26091 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911749983088683, - "loss": 1.5607, - "step": 26092 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019117466008252725, - "loss": 1.5192, - "step": 26093 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911743218561862, - "loss": 1.671, - "step": 26094 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911739836298451, - "loss": 1.609, - "step": 26095 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019117364540350404, - "loss": 1.5109, - "step": 26096 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019117330717716299, - "loss": 1.5175, - "step": 26097 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911729689508219, - "loss": 1.5679, - "step": 26098 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019117263072448084, - "loss": 1.6445, - "step": 26099 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019117229249813978, - "loss": 1.5148, - "step": 26100 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911719542717987, - "loss": 1.6395, - "step": 26101 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019117161604545764, - "loss": 1.5155, - "step": 26102 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019117127781911658, - "loss": 1.5916, - "step": 26103 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911709395927755, - "loss": 1.6041, - "step": 26104 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019117060136643443, - "loss": 1.317, - "step": 26105 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019117026314009337, - "loss": 1.4199, - "step": 26106 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911699249137523, - "loss": 1.7503, - "step": 26107 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019116958668741123, - "loss": 1.5564, - "step": 26108 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019116924846107017, - "loss": 1.5529, - "step": 26109 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019116891023472909, - "loss": 1.5228, - "step": 26110 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019116857200838803, - "loss": 1.5799, - "step": 26111 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019116823378204697, - "loss": 1.5009, - "step": 26112 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019116789555570588, - "loss": 1.6114, - "step": 26113 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019116755732936482, - "loss": 1.6537, - "step": 26114 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019116721910302376, - "loss": 1.4617, - "step": 26115 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019116688087668268, - "loss": 1.6463, - "step": 26116 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019116654265034162, - "loss": 1.5927, - "step": 26117 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019116620442400056, - "loss": 1.5334, - "step": 26118 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019116586619765947, - "loss": 1.5957, - "step": 26119 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019116552797131842, - "loss": 1.5021, - "step": 26120 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019116518974497736, - "loss": 1.4372, - "step": 26121 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019116485151863627, - "loss": 1.4906, - "step": 26122 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911645132922952, - "loss": 1.4712, - "step": 26123 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019116417506595415, - "loss": 1.5576, - "step": 26124 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019116383683961307, - "loss": 1.475, - "step": 26125 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191163498613272, - "loss": 1.5678, - "step": 26126 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019116316038693095, - "loss": 1.4795, - "step": 26127 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019116282216058986, - "loss": 1.6094, - "step": 26128 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911624839342488, - "loss": 1.6624, - "step": 26129 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019116214570790775, - "loss": 1.694, - "step": 26130 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019116180748156666, - "loss": 1.6525, - "step": 26131 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911614692552256, - "loss": 1.4833, - "step": 26132 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019116113102888454, - "loss": 1.6439, - "step": 26133 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019116079280254346, - "loss": 1.4091, - "step": 26134 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911604545762024, - "loss": 1.5487, - "step": 26135 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019116011634986134, - "loss": 1.6034, - "step": 26136 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019115977812352025, - "loss": 1.7424, - "step": 26137 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911594398971792, - "loss": 1.5839, - "step": 26138 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019115910167083813, - "loss": 1.5677, - "step": 26139 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019115876344449705, - "loss": 1.5763, - "step": 26140 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191158425218156, - "loss": 1.4453, - "step": 26141 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019115808699181493, - "loss": 1.6834, - "step": 26142 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019115774876547384, - "loss": 1.5726, - "step": 26143 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019115741053913279, - "loss": 1.6186, - "step": 26144 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019115707231279173, - "loss": 1.5747, - "step": 26145 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019115673408645067, - "loss": 1.4891, - "step": 26146 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911563958601096, - "loss": 1.6185, - "step": 26147 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019115605763376855, - "loss": 1.4927, - "step": 26148 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019115571940742746, - "loss": 1.6888, - "step": 26149 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911553811810864, - "loss": 1.5449, - "step": 26150 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019115504295474535, - "loss": 1.5839, - "step": 26151 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019115470472840426, - "loss": 1.6767, - "step": 26152 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911543665020632, - "loss": 1.588, - "step": 26153 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019115402827572214, - "loss": 1.4858, - "step": 26154 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019115369004938106, - "loss": 1.484, - "step": 26155 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019115335182304, - "loss": 1.4455, - "step": 26156 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019115301359669894, - "loss": 1.5833, - "step": 26157 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019115267537035785, - "loss": 1.4508, - "step": 26158 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911523371440168, - "loss": 1.4863, - "step": 26159 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019115199891767574, - "loss": 1.5565, - "step": 26160 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019115166069133465, - "loss": 1.5735, - "step": 26161 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911513224649936, - "loss": 1.4251, - "step": 26162 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019115098423865253, - "loss": 1.7833, - "step": 26163 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019115064601231145, - "loss": 1.5686, - "step": 26164 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911503077859704, - "loss": 1.7554, - "step": 26165 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019114996955962933, - "loss": 1.6587, - "step": 26166 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019114963133328824, - "loss": 1.4594, - "step": 26167 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019114929310694718, - "loss": 1.5833, - "step": 26168 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019114895488060612, - "loss": 1.5634, - "step": 26169 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019114861665426504, - "loss": 1.5938, - "step": 26170 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019114827842792398, - "loss": 1.642, - "step": 26171 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019114794020158292, - "loss": 1.3795, - "step": 26172 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019114760197524183, - "loss": 1.7262, - "step": 26173 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019114726374890078, - "loss": 1.4822, - "step": 26174 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019114692552255972, - "loss": 1.53, - "step": 26175 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019114658729621863, - "loss": 1.5211, - "step": 26176 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019114624906987757, - "loss": 1.5543, - "step": 26177 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911459108435365, - "loss": 1.5004, - "step": 26178 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019114557261719543, - "loss": 1.5565, - "step": 26179 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019114523439085437, - "loss": 1.462, - "step": 26180 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911448961645133, - "loss": 1.5652, - "step": 26181 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019114455793817222, - "loss": 1.5424, - "step": 26182 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019114421971183116, - "loss": 1.4911, - "step": 26183 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911438814854901, - "loss": 1.4515, - "step": 26184 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019114354325914902, - "loss": 1.5431, - "step": 26185 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019114320503280796, - "loss": 1.6514, - "step": 26186 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911428668064669, - "loss": 1.611, - "step": 26187 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019114252858012582, - "loss": 1.427, - "step": 26188 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019114219035378476, - "loss": 1.6015, - "step": 26189 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911418521274437, - "loss": 1.3173, - "step": 26190 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911415139011026, - "loss": 1.8271, - "step": 26191 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019114117567476155, - "loss": 1.5551, - "step": 26192 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911408374484205, - "loss": 1.4896, - "step": 26193 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911404992220794, - "loss": 1.6977, - "step": 26194 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019114016099573835, - "loss": 1.4118, - "step": 26195 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911398227693973, - "loss": 1.5272, - "step": 26196 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911394845430562, - "loss": 1.3851, - "step": 26197 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019113914631671515, - "loss": 1.4551, - "step": 26198 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911388080903741, - "loss": 1.5655, - "step": 26199 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191138469864033, - "loss": 1.3609, - "step": 26200 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019113813163769194, - "loss": 1.5656, - "step": 26201 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019113779341135088, - "loss": 1.5754, - "step": 26202 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019113745518500983, - "loss": 1.5089, - "step": 26203 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019113711695866877, - "loss": 1.539, - "step": 26204 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019113677873232768, - "loss": 1.5156, - "step": 26205 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019113644050598662, - "loss": 1.5106, - "step": 26206 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019113610227964556, - "loss": 1.5623, - "step": 26207 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019113576405330448, - "loss": 1.6301, - "step": 26208 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019113542582696342, - "loss": 1.4705, - "step": 26209 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019113508760062236, - "loss": 1.525, - "step": 26210 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019113474937428127, - "loss": 1.5132, - "step": 26211 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019113441114794021, - "loss": 1.535, - "step": 26212 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019113407292159916, - "loss": 1.4681, - "step": 26213 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019113373469525807, - "loss": 1.6577, - "step": 26214 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191133396468917, - "loss": 1.4352, - "step": 26215 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019113305824257595, - "loss": 1.5428, - "step": 26216 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019113272001623487, - "loss": 1.3529, - "step": 26217 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911323817898938, - "loss": 1.5041, - "step": 26218 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019113204356355275, - "loss": 1.6072, - "step": 26219 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019113170533721166, - "loss": 1.4181, - "step": 26220 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911313671108706, - "loss": 1.5725, - "step": 26221 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019113102888452954, - "loss": 1.5626, - "step": 26222 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019113069065818846, - "loss": 1.5698, - "step": 26223 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911303524318474, - "loss": 1.5329, - "step": 26224 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019113001420550634, - "loss": 1.5538, - "step": 26225 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019112967597916525, - "loss": 1.6849, - "step": 26226 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911293377528242, - "loss": 1.5803, - "step": 26227 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019112899952648314, - "loss": 1.6552, - "step": 26228 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019112866130014208, - "loss": 1.485, - "step": 26229 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191128323073801, - "loss": 1.6276, - "step": 26230 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019112798484745993, - "loss": 1.4454, - "step": 26231 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019112764662111887, - "loss": 1.5724, - "step": 26232 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911273083947778, - "loss": 1.4412, - "step": 26233 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019112697016843673, - "loss": 1.5444, - "step": 26234 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019112663194209567, - "loss": 1.5752, - "step": 26235 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019112629371575458, - "loss": 1.401, - "step": 26236 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019112595548941353, - "loss": 1.529, - "step": 26237 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019112561726307247, - "loss": 1.4795, - "step": 26238 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019112527903673138, - "loss": 1.622, - "step": 26239 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019112494081039032, - "loss": 1.5771, - "step": 26240 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019112460258404926, - "loss": 1.7755, - "step": 26241 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019112426435770818, - "loss": 1.7196, - "step": 26242 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019112392613136712, - "loss": 1.5244, - "step": 26243 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019112358790502606, - "loss": 1.583, - "step": 26244 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019112324967868497, - "loss": 1.4131, - "step": 26245 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019112291145234391, - "loss": 1.5543, - "step": 26246 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019112257322600286, - "loss": 1.6247, - "step": 26247 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019112223499966177, - "loss": 1.5975, - "step": 26248 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911218967733207, - "loss": 1.538, - "step": 26249 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019112155854697965, - "loss": 1.6305, - "step": 26250 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019112122032063857, - "loss": 1.4316, - "step": 26251 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911208820942975, - "loss": 1.4719, - "step": 26252 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019112054386795645, - "loss": 1.6522, - "step": 26253 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019112020564161536, - "loss": 1.5949, - "step": 26254 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911198674152743, - "loss": 1.5304, - "step": 26255 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019111952918893324, - "loss": 1.565, - "step": 26256 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019111919096259216, - "loss": 1.6373, - "step": 26257 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019111885273625113, - "loss": 1.6418, - "step": 26258 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019111851450991004, - "loss": 1.378, - "step": 26259 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019111817628356898, - "loss": 1.4519, - "step": 26260 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019111783805722792, - "loss": 1.604, - "step": 26261 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019111749983088684, - "loss": 1.564, - "step": 26262 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019111716160454578, - "loss": 1.5034, - "step": 26263 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019111682337820472, - "loss": 1.4574, - "step": 26264 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019111648515186363, - "loss": 1.5785, - "step": 26265 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019111614692552257, - "loss": 1.5221, - "step": 26266 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019111580869918152, - "loss": 1.5269, - "step": 26267 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019111547047284043, - "loss": 1.7583, - "step": 26268 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019111513224649937, - "loss": 1.8176, - "step": 26269 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911147940201583, - "loss": 1.503, - "step": 26270 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019111445579381723, - "loss": 1.3961, - "step": 26271 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019111411756747617, - "loss": 1.5447, - "step": 26272 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911137793411351, - "loss": 1.4494, - "step": 26273 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019111344111479402, - "loss": 1.6741, - "step": 26274 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019111310288845296, - "loss": 1.6537, - "step": 26275 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911127646621119, - "loss": 1.5125, - "step": 26276 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019111242643577082, - "loss": 1.5138, - "step": 26277 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019111208820942976, - "loss": 1.675, - "step": 26278 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911117499830887, - "loss": 1.6613, - "step": 26279 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019111141175674762, - "loss": 1.5561, - "step": 26280 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019111107353040656, - "loss": 1.5881, - "step": 26281 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911107353040655, - "loss": 1.5493, - "step": 26282 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911103970777244, - "loss": 1.4849, - "step": 26283 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019111005885138335, - "loss": 1.5015, - "step": 26284 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911097206250423, - "loss": 1.6424, - "step": 26285 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911093823987012, - "loss": 1.6024, - "step": 26286 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019110904417236015, - "loss": 1.5227, - "step": 26287 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911087059460191, - "loss": 1.5985, - "step": 26288 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191108367719678, - "loss": 1.6015, - "step": 26289 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019110802949333695, - "loss": 1.5351, - "step": 26290 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019110769126699589, - "loss": 1.5471, - "step": 26291 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911073530406548, - "loss": 1.5535, - "step": 26292 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019110701481431374, - "loss": 1.5596, - "step": 26293 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019110667658797268, - "loss": 1.5756, - "step": 26294 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911063383616316, - "loss": 1.6751, - "step": 26295 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019110600013529054, - "loss": 1.4367, - "step": 26296 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019110566190894948, - "loss": 1.5233, - "step": 26297 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911053236826084, - "loss": 1.5705, - "step": 26298 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019110498545626733, - "loss": 1.6734, - "step": 26299 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019110464722992628, - "loss": 1.4967, - "step": 26300 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911043090035852, - "loss": 1.6353, - "step": 26301 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019110397077724413, - "loss": 1.4218, - "step": 26302 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019110363255090307, - "loss": 1.6461, - "step": 26303 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019110329432456199, - "loss": 1.5844, - "step": 26304 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019110295609822093, - "loss": 1.5023, - "step": 26305 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019110261787187987, - "loss": 1.6366, - "step": 26306 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019110227964553878, - "loss": 1.5188, - "step": 26307 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019110194141919772, - "loss": 1.498, - "step": 26308 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019110160319285666, - "loss": 1.5569, - "step": 26309 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019110126496651558, - "loss": 1.7576, - "step": 26310 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019110092674017452, - "loss": 1.5357, - "step": 26311 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019110058851383346, - "loss": 1.5096, - "step": 26312 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001911002502874924, - "loss": 1.5475, - "step": 26313 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019109991206115134, - "loss": 1.5374, - "step": 26314 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019109957383481028, - "loss": 1.4724, - "step": 26315 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910992356084692, - "loss": 1.5639, - "step": 26316 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019109889738212814, - "loss": 1.4462, - "step": 26317 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019109855915578708, - "loss": 1.4571, - "step": 26318 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191098220929446, - "loss": 1.3895, - "step": 26319 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019109788270310494, - "loss": 1.4182, - "step": 26320 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019109754447676388, - "loss": 1.5653, - "step": 26321 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910972062504228, - "loss": 1.5712, - "step": 26322 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019109686802408173, - "loss": 1.6296, - "step": 26323 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019109652979774067, - "loss": 1.5644, - "step": 26324 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910961915713996, - "loss": 1.4526, - "step": 26325 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019109585334505853, - "loss": 1.4934, - "step": 26326 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019109551511871747, - "loss": 1.7277, - "step": 26327 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019109517689237638, - "loss": 1.4913, - "step": 26328 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019109483866603532, - "loss": 1.706, - "step": 26329 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019109450043969427, - "loss": 1.5319, - "step": 26330 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019109416221335318, - "loss": 1.5853, - "step": 26331 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019109382398701212, - "loss": 1.6989, - "step": 26332 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019109348576067106, - "loss": 1.5114, - "step": 26333 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019109314753432998, - "loss": 1.5291, - "step": 26334 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019109280930798892, - "loss": 1.5897, - "step": 26335 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019109247108164786, - "loss": 1.6117, - "step": 26336 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019109213285530677, - "loss": 1.5859, - "step": 26337 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910917946289657, - "loss": 1.4566, - "step": 26338 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019109145640262465, - "loss": 1.6276, - "step": 26339 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019109111817628357, - "loss": 1.5325, - "step": 26340 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910907799499425, - "loss": 1.5688, - "step": 26341 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019109044172360145, - "loss": 1.5564, - "step": 26342 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019109010349726036, - "loss": 1.5528, - "step": 26343 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910897652709193, - "loss": 1.5932, - "step": 26344 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019108942704457825, - "loss": 1.5816, - "step": 26345 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019108908881823716, - "loss": 1.2627, - "step": 26346 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910887505918961, - "loss": 1.5873, - "step": 26347 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019108841236555504, - "loss": 1.5764, - "step": 26348 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019108807413921396, - "loss": 1.6069, - "step": 26349 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910877359128729, - "loss": 1.3348, - "step": 26350 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019108739768653184, - "loss": 1.586, - "step": 26351 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019108705946019075, - "loss": 1.5306, - "step": 26352 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910867212338497, - "loss": 1.517, - "step": 26353 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019108638300750864, - "loss": 1.5539, - "step": 26354 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019108604478116755, - "loss": 1.5997, - "step": 26355 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910857065548265, - "loss": 1.4992, - "step": 26356 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019108536832848543, - "loss": 1.4376, - "step": 26357 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019108503010214435, - "loss": 1.5407, - "step": 26358 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910846918758033, - "loss": 1.6223, - "step": 26359 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019108435364946223, - "loss": 1.5836, - "step": 26360 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019108401542312114, - "loss": 1.543, - "step": 26361 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019108367719678008, - "loss": 1.3937, - "step": 26362 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019108333897043903, - "loss": 1.5949, - "step": 26363 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019108300074409794, - "loss": 1.4215, - "step": 26364 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019108266251775688, - "loss": 1.6578, - "step": 26365 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019108232429141582, - "loss": 1.584, - "step": 26366 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019108198606507474, - "loss": 1.4245, - "step": 26367 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019108164783873368, - "loss": 1.5334, - "step": 26368 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019108130961239262, - "loss": 1.4766, - "step": 26369 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019108097138605153, - "loss": 1.6783, - "step": 26370 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910806331597105, - "loss": 1.5515, - "step": 26371 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019108029493336944, - "loss": 1.4451, - "step": 26372 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019107995670702836, - "loss": 1.6559, - "step": 26373 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910796184806873, - "loss": 1.4837, - "step": 26374 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019107928025434624, - "loss": 1.4796, - "step": 26375 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019107894202800515, - "loss": 1.6211, - "step": 26376 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910786038016641, - "loss": 1.4308, - "step": 26377 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019107826557532303, - "loss": 1.6157, - "step": 26378 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019107792734898195, - "loss": 1.4523, - "step": 26379 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910775891226409, - "loss": 1.5242, - "step": 26380 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019107725089629983, - "loss": 1.6103, - "step": 26381 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019107691266995874, - "loss": 1.5782, - "step": 26382 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019107657444361769, - "loss": 1.5696, - "step": 26383 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019107623621727663, - "loss": 1.7192, - "step": 26384 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019107589799093554, - "loss": 1.5932, - "step": 26385 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019107555976459448, - "loss": 1.6451, - "step": 26386 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019107522153825342, - "loss": 1.476, - "step": 26387 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019107488331191234, - "loss": 1.698, - "step": 26388 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019107454508557128, - "loss": 1.6805, - "step": 26389 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019107420685923022, - "loss": 1.6737, - "step": 26390 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019107386863288913, - "loss": 1.7736, - "step": 26391 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019107353040654807, - "loss": 1.7772, - "step": 26392 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019107319218020702, - "loss": 1.6146, - "step": 26393 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019107285395386593, - "loss": 1.5248, - "step": 26394 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019107251572752487, - "loss": 1.4804, - "step": 26395 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910721775011838, - "loss": 1.5405, - "step": 26396 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019107183927484273, - "loss": 1.4164, - "step": 26397 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019107150104850167, - "loss": 1.4836, - "step": 26398 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910711628221606, - "loss": 1.3093, - "step": 26399 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019107082459581952, - "loss": 1.451, - "step": 26400 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019107048636947846, - "loss": 1.5, - "step": 26401 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910701481431374, - "loss": 1.6333, - "step": 26402 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019106980991679632, - "loss": 1.747, - "step": 26403 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019106947169045526, - "loss": 1.5039, - "step": 26404 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910691334641142, - "loss": 1.6621, - "step": 26405 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019106879523777311, - "loss": 1.467, - "step": 26406 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019106845701143206, - "loss": 1.673, - "step": 26407 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191068118785091, - "loss": 1.477, - "step": 26408 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910677805587499, - "loss": 1.4469, - "step": 26409 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019106744233240885, - "loss": 1.7604, - "step": 26410 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910671041060678, - "loss": 1.6077, - "step": 26411 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910667658797267, - "loss": 1.3272, - "step": 26412 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019106642765338565, - "loss": 1.5086, - "step": 26413 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910660894270446, - "loss": 1.4483, - "step": 26414 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910657512007035, - "loss": 1.5035, - "step": 26415 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019106541297436244, - "loss": 1.5556, - "step": 26416 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019106507474802139, - "loss": 1.4848, - "step": 26417 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910647365216803, - "loss": 1.4634, - "step": 26418 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019106439829533924, - "loss": 1.5933, - "step": 26419 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019106406006899818, - "loss": 1.6457, - "step": 26420 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910637218426571, - "loss": 1.4393, - "step": 26421 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019106338361631604, - "loss": 1.608, - "step": 26422 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019106304538997498, - "loss": 1.6629, - "step": 26423 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910627071636339, - "loss": 1.5967, - "step": 26424 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019106236893729283, - "loss": 1.6358, - "step": 26425 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019106203071095177, - "loss": 1.6115, - "step": 26426 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019106169248461072, - "loss": 1.5696, - "step": 26427 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019106135425826966, - "loss": 1.5371, - "step": 26428 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910610160319286, - "loss": 1.3439, - "step": 26429 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910606778055875, - "loss": 1.5981, - "step": 26430 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019106033957924645, - "loss": 1.5871, - "step": 26431 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910600013529054, - "loss": 1.4301, - "step": 26432 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910596631265643, - "loss": 1.5509, - "step": 26433 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019105932490022325, - "loss": 1.4786, - "step": 26434 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910589866738822, - "loss": 1.5594, - "step": 26435 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910586484475411, - "loss": 1.4051, - "step": 26436 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019105831022120005, - "loss": 1.4385, - "step": 26437 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191057971994859, - "loss": 1.6415, - "step": 26438 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910576337685179, - "loss": 1.6599, - "step": 26439 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019105729554217684, - "loss": 1.5694, - "step": 26440 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019105695731583578, - "loss": 1.4626, - "step": 26441 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910566190894947, - "loss": 1.5697, - "step": 26442 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019105628086315364, - "loss": 1.5217, - "step": 26443 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019105594263681258, - "loss": 1.4724, - "step": 26444 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910556044104715, - "loss": 1.5354, - "step": 26445 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019105526618413043, - "loss": 1.6743, - "step": 26446 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019105492795778938, - "loss": 1.6948, - "step": 26447 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910545897314483, - "loss": 1.4707, - "step": 26448 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019105425150510723, - "loss": 1.524, - "step": 26449 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019105391327876617, - "loss": 1.4513, - "step": 26450 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019105357505242509, - "loss": 1.3962, - "step": 26451 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019105323682608403, - "loss": 1.5459, - "step": 26452 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019105289859974297, - "loss": 1.6533, - "step": 26453 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019105256037340188, - "loss": 1.6302, - "step": 26454 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019105222214706082, - "loss": 1.6504, - "step": 26455 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019105188392071977, - "loss": 1.5774, - "step": 26456 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019105154569437868, - "loss": 1.5438, - "step": 26457 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019105120746803762, - "loss": 1.5484, - "step": 26458 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019105086924169656, - "loss": 1.4997, - "step": 26459 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019105053101535548, - "loss": 1.7382, - "step": 26460 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019105019278901442, - "loss": 1.6396, - "step": 26461 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019104985456267336, - "loss": 1.6155, - "step": 26462 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019104951633633227, - "loss": 1.5938, - "step": 26463 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910491781099912, - "loss": 1.6008, - "step": 26464 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019104883988365015, - "loss": 1.4833, - "step": 26465 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019104850165730907, - "loss": 1.5443, - "step": 26466 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191048163430968, - "loss": 1.6571, - "step": 26467 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019104782520462695, - "loss": 1.5129, - "step": 26468 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019104748697828586, - "loss": 1.5938, - "step": 26469 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910471487519448, - "loss": 1.4807, - "step": 26470 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019104681052560375, - "loss": 1.4122, - "step": 26471 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019104647229926266, - "loss": 1.4543, - "step": 26472 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910461340729216, - "loss": 1.5374, - "step": 26473 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019104579584658054, - "loss": 1.6514, - "step": 26474 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019104545762023946, - "loss": 1.3861, - "step": 26475 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910451193938984, - "loss": 1.5596, - "step": 26476 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019104478116755734, - "loss": 1.6142, - "step": 26477 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019104444294121625, - "loss": 1.5434, - "step": 26478 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910441047148752, - "loss": 1.6756, - "step": 26479 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019104376648853414, - "loss": 1.6547, - "step": 26480 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019104342826219305, - "loss": 1.3852, - "step": 26481 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191043090035852, - "loss": 1.4299, - "step": 26482 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019104275180951093, - "loss": 1.4292, - "step": 26483 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019104241358316987, - "loss": 1.5256, - "step": 26484 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019104207535682881, - "loss": 1.5596, - "step": 26485 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019104173713048773, - "loss": 1.6175, - "step": 26486 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019104139890414667, - "loss": 1.4999, - "step": 26487 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910410606778056, - "loss": 1.4443, - "step": 26488 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019104072245146452, - "loss": 1.6267, - "step": 26489 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019104038422512347, - "loss": 1.4884, - "step": 26490 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910400459987824, - "loss": 1.4712, - "step": 26491 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019103970777244132, - "loss": 1.494, - "step": 26492 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019103936954610026, - "loss": 1.6473, - "step": 26493 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910390313197592, - "loss": 1.614, - "step": 26494 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019103869309341814, - "loss": 1.523, - "step": 26495 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019103835486707706, - "loss": 1.6348, - "step": 26496 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191038016640736, - "loss": 1.572, - "step": 26497 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019103767841439494, - "loss": 1.6165, - "step": 26498 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019103734018805385, - "loss": 1.4666, - "step": 26499 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910370019617128, - "loss": 1.5207, - "step": 26500 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019103666373537174, - "loss": 1.391, - "step": 26501 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019103632550903065, - "loss": 1.7092, - "step": 26502 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910359872826896, - "loss": 1.5893, - "step": 26503 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019103564905634853, - "loss": 1.5304, - "step": 26504 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019103531083000745, - "loss": 1.4103, - "step": 26505 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910349726036664, - "loss": 1.5205, - "step": 26506 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019103463437732533, - "loss": 1.4631, - "step": 26507 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019103429615098424, - "loss": 1.5528, - "step": 26508 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019103395792464318, - "loss": 1.6307, - "step": 26509 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019103361969830213, - "loss": 1.6402, - "step": 26510 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019103328147196104, - "loss": 1.516, - "step": 26511 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019103294324561998, - "loss": 1.4549, - "step": 26512 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019103260501927892, - "loss": 1.5434, - "step": 26513 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019103226679293784, - "loss": 1.6912, - "step": 26514 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019103192856659678, - "loss": 1.5868, - "step": 26515 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019103159034025572, - "loss": 1.6783, - "step": 26516 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019103125211391463, - "loss": 1.4644, - "step": 26517 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019103091388757357, - "loss": 1.5504, - "step": 26518 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019103057566123251, - "loss": 1.5323, - "step": 26519 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019103023743489143, - "loss": 1.6675, - "step": 26520 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019102989920855037, - "loss": 1.4949, - "step": 26521 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910295609822093, - "loss": 1.5548, - "step": 26522 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019102922275586823, - "loss": 1.4948, - "step": 26523 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019102888452952717, - "loss": 1.5391, - "step": 26524 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910285463031861, - "loss": 1.5793, - "step": 26525 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019102820807684502, - "loss": 1.5665, - "step": 26526 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019102786985050396, - "loss": 1.4838, - "step": 26527 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910275316241629, - "loss": 1.5711, - "step": 26528 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019102719339782182, - "loss": 1.5362, - "step": 26529 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019102685517148076, - "loss": 1.366, - "step": 26530 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910265169451397, - "loss": 1.678, - "step": 26531 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019102617871879861, - "loss": 1.6967, - "step": 26532 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019102584049245756, - "loss": 1.3738, - "step": 26533 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910255022661165, - "loss": 1.5429, - "step": 26534 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910251640397754, - "loss": 1.3446, - "step": 26535 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019102482581343435, - "loss": 1.6377, - "step": 26536 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910244875870933, - "loss": 1.5292, - "step": 26537 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910241493607522, - "loss": 1.4405, - "step": 26538 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019102381113441117, - "loss": 1.6328, - "step": 26539 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910234729080701, - "loss": 1.4538, - "step": 26540 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019102313468172903, - "loss": 1.4615, - "step": 26541 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019102279645538797, - "loss": 1.6293, - "step": 26542 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019102245822904689, - "loss": 1.5124, - "step": 26543 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019102212000270583, - "loss": 1.6047, - "step": 26544 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019102178177636477, - "loss": 1.6669, - "step": 26545 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019102144355002368, - "loss": 1.5698, - "step": 26546 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019102110532368262, - "loss": 1.4511, - "step": 26547 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019102076709734156, - "loss": 1.5573, - "step": 26548 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019102042887100048, - "loss": 1.6211, - "step": 26549 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019102009064465942, - "loss": 1.5274, - "step": 26550 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019101975241831836, - "loss": 1.581, - "step": 26551 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019101941419197727, - "loss": 1.8135, - "step": 26552 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019101907596563622, - "loss": 1.5916, - "step": 26553 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019101873773929516, - "loss": 1.3652, - "step": 26554 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019101839951295407, - "loss": 1.8661, - "step": 26555 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191018061286613, - "loss": 1.5149, - "step": 26556 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019101772306027195, - "loss": 1.5357, - "step": 26557 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019101738483393087, - "loss": 1.6739, - "step": 26558 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910170466075898, - "loss": 1.4619, - "step": 26559 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019101670838124875, - "loss": 1.4328, - "step": 26560 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019101637015490766, - "loss": 1.6316, - "step": 26561 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910160319285666, - "loss": 1.5718, - "step": 26562 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019101569370222555, - "loss": 1.6701, - "step": 26563 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019101535547588446, - "loss": 1.6216, - "step": 26564 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910150172495434, - "loss": 1.5542, - "step": 26565 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019101467902320234, - "loss": 1.4077, - "step": 26566 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019101434079686126, - "loss": 1.4836, - "step": 26567 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910140025705202, - "loss": 1.5639, - "step": 26568 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019101366434417914, - "loss": 1.5357, - "step": 26569 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019101332611783805, - "loss": 1.5929, - "step": 26570 - }, - { - "epoch": 0.09, - "learning_rate": 0.000191012987891497, - "loss": 1.547, - "step": 26571 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019101264966515593, - "loss": 1.6558, - "step": 26572 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019101231143881485, - "loss": 1.4202, - "step": 26573 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910119732124738, - "loss": 1.3628, - "step": 26574 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019101163498613273, - "loss": 1.7138, - "step": 26575 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019101129675979164, - "loss": 1.5423, - "step": 26576 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019101095853345059, - "loss": 1.5338, - "step": 26577 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019101062030710953, - "loss": 1.5104, - "step": 26578 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019101028208076847, - "loss": 1.6475, - "step": 26579 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019100994385442738, - "loss": 1.693, - "step": 26580 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019100960562808632, - "loss": 1.615, - "step": 26581 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019100926740174526, - "loss": 1.5232, - "step": 26582 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019100892917540418, - "loss": 1.561, - "step": 26583 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019100859094906312, - "loss": 1.6495, - "step": 26584 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019100825272272206, - "loss": 1.6672, - "step": 26585 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019100791449638097, - "loss": 1.4484, - "step": 26586 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019100757627003992, - "loss": 1.4647, - "step": 26587 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019100723804369886, - "loss": 1.6883, - "step": 26588 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019100689981735777, - "loss": 1.5426, - "step": 26589 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910065615910167, - "loss": 1.6121, - "step": 26590 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019100622336467565, - "loss": 1.441, - "step": 26591 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019100588513833457, - "loss": 1.5515, - "step": 26592 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910055469119935, - "loss": 1.5881, - "step": 26593 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019100520868565245, - "loss": 1.5342, - "step": 26594 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910048704593114, - "loss": 1.6322, - "step": 26595 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019100453223297033, - "loss": 1.4574, - "step": 26596 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019100419400662925, - "loss": 1.5248, - "step": 26597 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001910038557802882, - "loss": 1.5537, - "step": 26598 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019100351755394713, - "loss": 1.7539, - "step": 26599 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019100317932760604, - "loss": 1.5906, - "step": 26600 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019100284110126498, - "loss": 1.6095, - "step": 26601 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019100250287492392, - "loss": 1.4956, - "step": 26602 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019100216464858284, - "loss": 1.5214, - "step": 26603 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019100182642224178, - "loss": 1.5755, - "step": 26604 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019100148819590072, - "loss": 1.5127, - "step": 26605 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019100114996955963, - "loss": 1.5953, - "step": 26606 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019100081174321858, - "loss": 1.5447, - "step": 26607 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019100047351687752, - "loss": 1.4325, - "step": 26608 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019100013529053643, - "loss": 1.4596, - "step": 26609 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019099979706419537, - "loss": 1.5056, - "step": 26610 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909994588378543, - "loss": 1.3939, - "step": 26611 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019099912061151323, - "loss": 1.5015, - "step": 26612 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019099878238517217, - "loss": 1.5758, - "step": 26613 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909984441588311, - "loss": 1.5036, - "step": 26614 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019099810593249002, - "loss": 1.554, - "step": 26615 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019099776770614896, - "loss": 1.5061, - "step": 26616 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909974294798079, - "loss": 1.5987, - "step": 26617 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019099709125346682, - "loss": 1.5451, - "step": 26618 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019099675302712576, - "loss": 1.4549, - "step": 26619 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909964148007847, - "loss": 1.6736, - "step": 26620 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019099607657444362, - "loss": 1.5512, - "step": 26621 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019099573834810256, - "loss": 1.6344, - "step": 26622 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909954001217615, - "loss": 1.4822, - "step": 26623 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909950618954204, - "loss": 1.5583, - "step": 26624 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019099472366907935, - "loss": 1.4944, - "step": 26625 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909943854427383, - "loss": 1.5409, - "step": 26626 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909940472163972, - "loss": 1.665, - "step": 26627 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019099370899005615, - "loss": 1.4529, - "step": 26628 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909933707637151, - "loss": 1.542, - "step": 26629 - }, - { - "epoch": 0.09, - "learning_rate": 0.000190993032537374, - "loss": 1.5815, - "step": 26630 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019099269431103295, - "loss": 1.5638, - "step": 26631 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909923560846919, - "loss": 1.5069, - "step": 26632 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909920178583508, - "loss": 1.5008, - "step": 26633 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019099167963200974, - "loss": 1.8027, - "step": 26634 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019099134140566868, - "loss": 1.5853, - "step": 26635 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909910031793276, - "loss": 1.6409, - "step": 26636 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019099066495298654, - "loss": 1.7082, - "step": 26637 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019099032672664548, - "loss": 1.6648, - "step": 26638 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909899885003044, - "loss": 1.616, - "step": 26639 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019098965027396334, - "loss": 1.4474, - "step": 26640 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019098931204762228, - "loss": 1.6096, - "step": 26641 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909889738212812, - "loss": 1.4774, - "step": 26642 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019098863559494013, - "loss": 1.6023, - "step": 26643 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019098829736859907, - "loss": 1.4629, - "step": 26644 - }, - { - "epoch": 0.09, - "learning_rate": 0.000190987959142258, - "loss": 1.4554, - "step": 26645 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019098762091591693, - "loss": 1.4933, - "step": 26646 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019098728268957587, - "loss": 1.7132, - "step": 26647 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019098694446323478, - "loss": 1.5981, - "step": 26648 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019098660623689372, - "loss": 1.6014, - "step": 26649 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019098626801055267, - "loss": 1.5331, - "step": 26650 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909859297842116, - "loss": 1.6306, - "step": 26651 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019098559155787055, - "loss": 1.5043, - "step": 26652 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909852533315295, - "loss": 1.3715, - "step": 26653 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909849151051884, - "loss": 1.5669, - "step": 26654 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019098457687884734, - "loss": 1.7702, - "step": 26655 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019098423865250629, - "loss": 1.5183, - "step": 26656 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909839004261652, - "loss": 1.6043, - "step": 26657 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019098356219982414, - "loss": 1.6375, - "step": 26658 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019098322397348308, - "loss": 1.5209, - "step": 26659 - }, - { - "epoch": 0.09, - "learning_rate": 0.000190982885747142, - "loss": 1.5675, - "step": 26660 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019098254752080094, - "loss": 1.5753, - "step": 26661 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019098220929445988, - "loss": 1.3753, - "step": 26662 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909818710681188, - "loss": 1.4683, - "step": 26663 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019098153284177773, - "loss": 1.663, - "step": 26664 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019098119461543667, - "loss": 1.6853, - "step": 26665 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909808563890956, - "loss": 1.6592, - "step": 26666 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019098051816275453, - "loss": 1.4654, - "step": 26667 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019098017993641347, - "loss": 1.5855, - "step": 26668 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019097984171007238, - "loss": 1.3754, - "step": 26669 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019097950348373133, - "loss": 1.5692, - "step": 26670 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019097916525739027, - "loss": 1.4882, - "step": 26671 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019097882703104918, - "loss": 1.5571, - "step": 26672 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019097848880470812, - "loss": 1.5388, - "step": 26673 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019097815057836706, - "loss": 1.5869, - "step": 26674 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019097781235202598, - "loss": 1.5255, - "step": 26675 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019097747412568492, - "loss": 1.5273, - "step": 26676 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019097713589934386, - "loss": 1.6753, - "step": 26677 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019097679767300277, - "loss": 1.405, - "step": 26678 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019097645944666171, - "loss": 1.5226, - "step": 26679 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019097612122032066, - "loss": 1.5112, - "step": 26680 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019097578299397957, - "loss": 1.5187, - "step": 26681 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909754447676385, - "loss": 1.5965, - "step": 26682 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019097510654129745, - "loss": 1.7171, - "step": 26683 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019097476831495637, - "loss": 1.5094, - "step": 26684 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909744300886153, - "loss": 1.4335, - "step": 26685 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019097409186227425, - "loss": 1.5237, - "step": 26686 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019097375363593316, - "loss": 1.6634, - "step": 26687 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909734154095921, - "loss": 1.6323, - "step": 26688 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019097307718325104, - "loss": 1.4539, - "step": 26689 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019097273895690996, - "loss": 1.6152, - "step": 26690 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909724007305689, - "loss": 1.6191, - "step": 26691 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019097206250422784, - "loss": 1.5142, - "step": 26692 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019097172427788676, - "loss": 1.6489, - "step": 26693 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909713860515457, - "loss": 1.5427, - "step": 26694 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019097104782520464, - "loss": 1.5483, - "step": 26695 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019097070959886355, - "loss": 1.3291, - "step": 26696 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909703713725225, - "loss": 1.5896, - "step": 26697 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019097003314618143, - "loss": 1.4284, - "step": 26698 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019096969491984035, - "loss": 1.6366, - "step": 26699 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909693566934993, - "loss": 1.4581, - "step": 26700 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019096901846715823, - "loss": 1.5247, - "step": 26701 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019096868024081714, - "loss": 1.5593, - "step": 26702 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019096834201447609, - "loss": 1.5834, - "step": 26703 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019096800378813503, - "loss": 1.4913, - "step": 26704 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019096766556179394, - "loss": 1.5579, - "step": 26705 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019096732733545288, - "loss": 1.4441, - "step": 26706 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019096698910911182, - "loss": 1.6126, - "step": 26707 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019096665088277076, - "loss": 1.4016, - "step": 26708 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909663126564297, - "loss": 1.7146, - "step": 26709 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019096597443008865, - "loss": 1.6737, - "step": 26710 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019096563620374756, - "loss": 1.549, - "step": 26711 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909652979774065, - "loss": 1.5785, - "step": 26712 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019096495975106544, - "loss": 1.5672, - "step": 26713 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019096462152472436, - "loss": 1.5072, - "step": 26714 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909642832983833, - "loss": 1.6091, - "step": 26715 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019096394507204224, - "loss": 1.4462, - "step": 26716 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019096360684570115, - "loss": 1.5054, - "step": 26717 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909632686193601, - "loss": 1.493, - "step": 26718 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019096293039301903, - "loss": 1.5191, - "step": 26719 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019096259216667795, - "loss": 1.5189, - "step": 26720 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909622539403369, - "loss": 1.6572, - "step": 26721 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019096191571399583, - "loss": 1.438, - "step": 26722 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019096157748765475, - "loss": 1.5837, - "step": 26723 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909612392613137, - "loss": 1.5769, - "step": 26724 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019096090103497263, - "loss": 1.5402, - "step": 26725 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019096056280863154, - "loss": 1.5774, - "step": 26726 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019096022458229048, - "loss": 1.6115, - "step": 26727 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019095988635594942, - "loss": 1.6529, - "step": 26728 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019095954812960834, - "loss": 1.5852, - "step": 26729 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019095920990326728, - "loss": 1.5425, - "step": 26730 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019095887167692622, - "loss": 1.4315, - "step": 26731 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019095853345058513, - "loss": 1.6281, - "step": 26732 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019095819522424408, - "loss": 1.7035, - "step": 26733 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019095785699790302, - "loss": 1.5793, - "step": 26734 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019095751877156193, - "loss": 1.4973, - "step": 26735 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019095718054522087, - "loss": 1.5125, - "step": 26736 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909568423188798, - "loss": 1.6741, - "step": 26737 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019095650409253873, - "loss": 1.5675, - "step": 26738 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019095616586619767, - "loss": 1.5658, - "step": 26739 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909558276398566, - "loss": 1.4689, - "step": 26740 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019095548941351552, - "loss": 1.533, - "step": 26741 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019095515118717446, - "loss": 1.6472, - "step": 26742 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909548129608334, - "loss": 1.5595, - "step": 26743 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019095447473449232, - "loss": 1.4635, - "step": 26744 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019095413650815126, - "loss": 1.5188, - "step": 26745 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909537982818102, - "loss": 1.4705, - "step": 26746 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019095346005546912, - "loss": 1.3876, - "step": 26747 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019095312182912806, - "loss": 1.5533, - "step": 26748 - }, - { - "epoch": 0.09, - "learning_rate": 0.000190952783602787, - "loss": 1.571, - "step": 26749 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909524453764459, - "loss": 1.4475, - "step": 26750 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019095210715010485, - "loss": 1.3409, - "step": 26751 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909517689237638, - "loss": 1.6437, - "step": 26752 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909514306974227, - "loss": 1.6021, - "step": 26753 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019095109247108165, - "loss": 1.5737, - "step": 26754 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909507542447406, - "loss": 1.6905, - "step": 26755 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909504160183995, - "loss": 1.8169, - "step": 26756 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019095007779205845, - "loss": 1.4202, - "step": 26757 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909497395657174, - "loss": 1.5968, - "step": 26758 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909494013393763, - "loss": 1.5618, - "step": 26759 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019094906311303524, - "loss": 1.5502, - "step": 26760 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019094872488669418, - "loss": 1.6622, - "step": 26761 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909483866603531, - "loss": 1.4608, - "step": 26762 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019094804843401204, - "loss": 1.4866, - "step": 26763 - }, - { - "epoch": 0.09, - "learning_rate": 0.000190947710207671, - "loss": 1.6326, - "step": 26764 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019094737198132992, - "loss": 1.4306, - "step": 26765 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019094703375498886, - "loss": 1.5776, - "step": 26766 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909466955286478, - "loss": 1.5578, - "step": 26767 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019094635730230672, - "loss": 1.6766, - "step": 26768 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019094601907596566, - "loss": 1.4556, - "step": 26769 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909456808496246, - "loss": 1.6159, - "step": 26770 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909453426232835, - "loss": 1.7677, - "step": 26771 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019094500439694245, - "loss": 1.4333, - "step": 26772 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909446661706014, - "loss": 1.7158, - "step": 26773 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909443279442603, - "loss": 1.4774, - "step": 26774 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019094398971791925, - "loss": 1.5274, - "step": 26775 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909436514915782, - "loss": 1.4918, - "step": 26776 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909433132652371, - "loss": 1.537, - "step": 26777 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019094297503889605, - "loss": 1.7233, - "step": 26778 - }, - { - "epoch": 0.09, - "learning_rate": 0.000190942636812555, - "loss": 1.6541, - "step": 26779 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909422985862139, - "loss": 1.5206, - "step": 26780 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019094196035987284, - "loss": 1.4104, - "step": 26781 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019094162213353178, - "loss": 1.6801, - "step": 26782 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909412839071907, - "loss": 1.5157, - "step": 26783 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019094094568084964, - "loss": 1.5124, - "step": 26784 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019094060745450858, - "loss": 1.6361, - "step": 26785 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909402692281675, - "loss": 1.6242, - "step": 26786 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019093993100182644, - "loss": 1.5707, - "step": 26787 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019093959277548538, - "loss": 1.6596, - "step": 26788 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909392545491443, - "loss": 1.4829, - "step": 26789 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019093891632280323, - "loss": 1.7821, - "step": 26790 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019093857809646217, - "loss": 1.495, - "step": 26791 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909382398701211, - "loss": 1.5383, - "step": 26792 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019093790164378003, - "loss": 1.5407, - "step": 26793 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019093756341743897, - "loss": 1.4574, - "step": 26794 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019093722519109788, - "loss": 1.5634, - "step": 26795 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019093688696475683, - "loss": 1.4556, - "step": 26796 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019093654873841577, - "loss": 1.542, - "step": 26797 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019093621051207468, - "loss": 1.5442, - "step": 26798 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019093587228573362, - "loss": 1.6833, - "step": 26799 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019093553405939256, - "loss": 1.3436, - "step": 26800 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019093519583305148, - "loss": 1.7446, - "step": 26801 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019093485760671042, - "loss": 1.4219, - "step": 26802 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019093451938036936, - "loss": 1.653, - "step": 26803 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019093418115402827, - "loss": 1.5952, - "step": 26804 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019093384292768721, - "loss": 1.6255, - "step": 26805 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019093350470134616, - "loss": 1.5303, - "step": 26806 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019093316647500507, - "loss": 1.5507, - "step": 26807 - }, - { - "epoch": 0.09, - "learning_rate": 0.000190932828248664, - "loss": 1.6214, - "step": 26808 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019093249002232295, - "loss": 1.7558, - "step": 26809 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019093215179598187, - "loss": 1.5304, - "step": 26810 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909318135696408, - "loss": 1.6325, - "step": 26811 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019093147534329975, - "loss": 1.5238, - "step": 26812 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019093113711695866, - "loss": 1.4895, - "step": 26813 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909307988906176, - "loss": 1.5208, - "step": 26814 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019093046066427654, - "loss": 1.6276, - "step": 26815 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019093012243793546, - "loss": 1.5756, - "step": 26816 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909297842115944, - "loss": 1.5687, - "step": 26817 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019092944598525334, - "loss": 1.6235, - "step": 26818 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019092910775891225, - "loss": 1.5273, - "step": 26819 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019092876953257122, - "loss": 1.5695, - "step": 26820 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019092843130623014, - "loss": 1.857, - "step": 26821 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019092809307988908, - "loss": 1.6548, - "step": 26822 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019092775485354802, - "loss": 1.4209, - "step": 26823 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019092741662720693, - "loss": 1.7427, - "step": 26824 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019092707840086587, - "loss": 1.5368, - "step": 26825 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019092674017452482, - "loss": 1.5683, - "step": 26826 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019092640194818373, - "loss": 1.4642, - "step": 26827 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019092606372184267, - "loss": 1.3855, - "step": 26828 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909257254955016, - "loss": 1.4813, - "step": 26829 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019092538726916053, - "loss": 1.5442, - "step": 26830 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019092504904281947, - "loss": 1.4013, - "step": 26831 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909247108164784, - "loss": 1.4967, - "step": 26832 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019092437259013732, - "loss": 1.415, - "step": 26833 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019092403436379626, - "loss": 1.5423, - "step": 26834 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909236961374552, - "loss": 1.5118, - "step": 26835 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019092335791111412, - "loss": 1.5091, - "step": 26836 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019092301968477306, - "loss": 1.461, - "step": 26837 - }, - { - "epoch": 0.09, - "learning_rate": 0.000190922681458432, - "loss": 1.5506, - "step": 26838 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019092234323209091, - "loss": 1.6337, - "step": 26839 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019092200500574986, - "loss": 1.4658, - "step": 26840 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909216667794088, - "loss": 1.5431, - "step": 26841 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909213285530677, - "loss": 1.5486, - "step": 26842 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019092099032672665, - "loss": 1.466, - "step": 26843 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909206521003856, - "loss": 1.4296, - "step": 26844 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909203138740445, - "loss": 1.6502, - "step": 26845 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019091997564770345, - "loss": 1.4725, - "step": 26846 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909196374213624, - "loss": 1.4639, - "step": 26847 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019091929919502133, - "loss": 1.6145, - "step": 26848 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019091896096868024, - "loss": 1.6559, - "step": 26849 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019091862274233919, - "loss": 1.4771, - "step": 26850 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019091828451599813, - "loss": 1.5102, - "step": 26851 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019091794628965704, - "loss": 1.516, - "step": 26852 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019091760806331598, - "loss": 1.6016, - "step": 26853 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019091726983697492, - "loss": 1.6761, - "step": 26854 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019091693161063384, - "loss": 1.5584, - "step": 26855 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019091659338429278, - "loss": 1.5535, - "step": 26856 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019091625515795172, - "loss": 1.5192, - "step": 26857 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019091591693161063, - "loss": 1.5734, - "step": 26858 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019091557870526957, - "loss": 1.3863, - "step": 26859 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019091524047892852, - "loss": 1.6906, - "step": 26860 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019091490225258743, - "loss": 1.5163, - "step": 26861 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019091456402624637, - "loss": 1.5238, - "step": 26862 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909142257999053, - "loss": 1.643, - "step": 26863 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019091388757356423, - "loss": 1.4681, - "step": 26864 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019091354934722317, - "loss": 1.6161, - "step": 26865 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909132111208821, - "loss": 1.5058, - "step": 26866 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019091287289454102, - "loss": 1.6347, - "step": 26867 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019091253466819996, - "loss": 1.5659, - "step": 26868 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909121964418589, - "loss": 1.4071, - "step": 26869 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019091185821551782, - "loss": 1.5566, - "step": 26870 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019091151998917676, - "loss": 1.4978, - "step": 26871 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909111817628357, - "loss": 1.4745, - "step": 26872 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019091084353649462, - "loss": 1.5068, - "step": 26873 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019091050531015356, - "loss": 1.5013, - "step": 26874 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909101670838125, - "loss": 1.5722, - "step": 26875 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019090982885747144, - "loss": 1.6014, - "step": 26876 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019090949063113038, - "loss": 1.6873, - "step": 26877 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909091524047893, - "loss": 1.5778, - "step": 26878 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019090881417844823, - "loss": 1.4966, - "step": 26879 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019090847595210718, - "loss": 1.4888, - "step": 26880 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909081377257661, - "loss": 1.469, - "step": 26881 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019090779949942503, - "loss": 1.4833, - "step": 26882 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019090746127308397, - "loss": 1.65, - "step": 26883 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019090712304674289, - "loss": 1.7231, - "step": 26884 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019090678482040183, - "loss": 1.5115, - "step": 26885 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019090644659406077, - "loss": 1.4205, - "step": 26886 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019090610836771968, - "loss": 1.5538, - "step": 26887 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019090577014137862, - "loss": 1.5931, - "step": 26888 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019090543191503757, - "loss": 1.6, - "step": 26889 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019090509368869648, - "loss": 1.6933, - "step": 26890 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019090475546235542, - "loss": 1.6051, - "step": 26891 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019090441723601436, - "loss": 1.5637, - "step": 26892 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019090407900967328, - "loss": 1.6124, - "step": 26893 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019090374078333222, - "loss": 1.5058, - "step": 26894 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019090340255699116, - "loss": 1.6571, - "step": 26895 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019090306433065007, - "loss": 1.4438, - "step": 26896 - }, - { - "epoch": 0.09, - "learning_rate": 0.000190902726104309, - "loss": 1.3766, - "step": 26897 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019090238787796795, - "loss": 1.3981, - "step": 26898 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019090204965162687, - "loss": 1.4961, - "step": 26899 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909017114252858, - "loss": 1.5731, - "step": 26900 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019090137319894475, - "loss": 1.7159, - "step": 26901 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019090103497260366, - "loss": 1.6762, - "step": 26902 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001909006967462626, - "loss": 1.6335, - "step": 26903 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019090035851992155, - "loss": 1.58, - "step": 26904 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019090002029358046, - "loss": 1.4474, - "step": 26905 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908996820672394, - "loss": 1.541, - "step": 26906 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019089934384089834, - "loss": 1.6091, - "step": 26907 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019089900561455726, - "loss": 1.5277, - "step": 26908 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908986673882162, - "loss": 1.4896, - "step": 26909 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019089832916187514, - "loss": 1.6039, - "step": 26910 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019089799093553405, - "loss": 1.5126, - "step": 26911 - }, - { - "epoch": 0.09, - "learning_rate": 0.000190897652709193, - "loss": 1.432, - "step": 26912 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019089731448285194, - "loss": 1.4188, - "step": 26913 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019089697625651085, - "loss": 1.5896, - "step": 26914 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908966380301698, - "loss": 1.5593, - "step": 26915 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019089629980382873, - "loss": 1.4479, - "step": 26916 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019089596157748765, - "loss": 1.5875, - "step": 26917 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908956233511466, - "loss": 1.5692, - "step": 26918 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019089528512480553, - "loss": 1.4962, - "step": 26919 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019089494689846444, - "loss": 1.5028, - "step": 26920 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019089460867212338, - "loss": 1.4095, - "step": 26921 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019089427044578232, - "loss": 1.5312, - "step": 26922 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019089393221944124, - "loss": 1.8167, - "step": 26923 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019089359399310018, - "loss": 1.5456, - "step": 26924 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019089325576675912, - "loss": 1.4411, - "step": 26925 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019089291754041803, - "loss": 1.7014, - "step": 26926 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019089257931407698, - "loss": 1.5479, - "step": 26927 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019089224108773592, - "loss": 1.5078, - "step": 26928 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019089190286139483, - "loss": 1.5629, - "step": 26929 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019089156463505377, - "loss": 1.5901, - "step": 26930 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908912264087127, - "loss": 1.7272, - "step": 26931 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019089088818237165, - "loss": 1.7115, - "step": 26932 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908905499560306, - "loss": 1.5699, - "step": 26933 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019089021172968954, - "loss": 1.6527, - "step": 26934 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019088987350334845, - "loss": 1.4261, - "step": 26935 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908895352770074, - "loss": 1.5722, - "step": 26936 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019088919705066633, - "loss": 1.6419, - "step": 26937 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019088885882432525, - "loss": 1.6707, - "step": 26938 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908885205979842, - "loss": 1.5745, - "step": 26939 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019088818237164313, - "loss": 1.5966, - "step": 26940 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019088784414530204, - "loss": 1.4566, - "step": 26941 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019088750591896098, - "loss": 1.4563, - "step": 26942 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019088716769261993, - "loss": 1.7212, - "step": 26943 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019088682946627884, - "loss": 1.6094, - "step": 26944 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019088649123993778, - "loss": 1.3867, - "step": 26945 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019088615301359672, - "loss": 1.3544, - "step": 26946 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019088581478725564, - "loss": 1.5924, - "step": 26947 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019088547656091458, - "loss": 1.5777, - "step": 26948 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019088513833457352, - "loss": 1.5767, - "step": 26949 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019088480010823243, - "loss": 1.6624, - "step": 26950 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019088446188189137, - "loss": 1.4433, - "step": 26951 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019088412365555031, - "loss": 1.4993, - "step": 26952 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019088378542920923, - "loss": 1.6284, - "step": 26953 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019088344720286817, - "loss": 1.672, - "step": 26954 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908831089765271, - "loss": 1.6605, - "step": 26955 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019088277075018603, - "loss": 1.6059, - "step": 26956 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019088243252384497, - "loss": 1.4189, - "step": 26957 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908820942975039, - "loss": 1.5214, - "step": 26958 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019088175607116282, - "loss": 1.5754, - "step": 26959 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019088141784482176, - "loss": 1.656, - "step": 26960 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908810796184807, - "loss": 1.5078, - "step": 26961 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019088074139213962, - "loss": 1.6897, - "step": 26962 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019088040316579856, - "loss": 1.4412, - "step": 26963 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908800649394575, - "loss": 1.4354, - "step": 26964 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019087972671311641, - "loss": 1.6124, - "step": 26965 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019087938848677536, - "loss": 1.7987, - "step": 26966 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908790502604343, - "loss": 1.4058, - "step": 26967 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908787120340932, - "loss": 1.4717, - "step": 26968 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019087837380775215, - "loss": 1.5978, - "step": 26969 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908780355814111, - "loss": 1.5187, - "step": 26970 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019087769735507, - "loss": 1.5043, - "step": 26971 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019087735912872895, - "loss": 1.5447, - "step": 26972 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908770209023879, - "loss": 1.4707, - "step": 26973 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908766826760468, - "loss": 1.5006, - "step": 26974 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019087634444970574, - "loss": 1.3608, - "step": 26975 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019087600622336469, - "loss": 1.6466, - "step": 26976 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908756679970236, - "loss": 1.6649, - "step": 26977 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019087532977068254, - "loss": 1.5595, - "step": 26978 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019087499154434148, - "loss": 1.4043, - "step": 26979 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908746533180004, - "loss": 1.6512, - "step": 26980 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019087431509165934, - "loss": 1.4348, - "step": 26981 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019087397686531828, - "loss": 1.5297, - "step": 26982 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908736386389772, - "loss": 1.4578, - "step": 26983 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019087330041263613, - "loss": 1.5924, - "step": 26984 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019087296218629507, - "loss": 1.7665, - "step": 26985 - }, - { - "epoch": 0.09, - "learning_rate": 0.000190872623959954, - "loss": 1.5535, - "step": 26986 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019087228573361293, - "loss": 1.6234, - "step": 26987 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908719475072719, - "loss": 1.5332, - "step": 26988 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908716092809308, - "loss": 1.4394, - "step": 26989 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019087127105458975, - "loss": 1.6037, - "step": 26990 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908709328282487, - "loss": 1.5664, - "step": 26991 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908705946019076, - "loss": 1.6366, - "step": 26992 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019087025637556655, - "loss": 1.4857, - "step": 26993 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908699181492255, - "loss": 1.6301, - "step": 26994 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908695799228844, - "loss": 1.6167, - "step": 26995 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019086924169654335, - "loss": 1.6612, - "step": 26996 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908689034702023, - "loss": 1.5907, - "step": 26997 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908685652438612, - "loss": 1.502, - "step": 26998 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019086822701752014, - "loss": 1.5625, - "step": 26999 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019086788879117908, - "loss": 1.6647, - "step": 27000 - }, - { - "epoch": 0.09, - "eval_loss": 1.2134728064903846, - "mrc_em": 19.0, - "mrc_f1": 25.21904761904762, - "nli_acc": 0.34, - "step": 27000, - "sts_f1": 0.5652173913043478, - "ynat_f1": 0.520532745030118 - }, - { - "epoch": 0.09, - "learning_rate": 0.000190867550564838, - "loss": 1.6473, - "step": 27001 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019086721233849694, - "loss": 1.4653, - "step": 27002 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019086687411215588, - "loss": 1.6077, - "step": 27003 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908665358858148, - "loss": 1.4654, - "step": 27004 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019086619765947373, - "loss": 1.5826, - "step": 27005 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019086585943313268, - "loss": 1.5545, - "step": 27006 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908655212067916, - "loss": 1.52, - "step": 27007 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019086518298045053, - "loss": 1.5975, - "step": 27008 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019086484475410947, - "loss": 1.6476, - "step": 27009 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019086450652776839, - "loss": 1.5538, - "step": 27010 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019086416830142733, - "loss": 1.7395, - "step": 27011 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019086383007508627, - "loss": 1.6866, - "step": 27012 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019086349184874518, - "loss": 1.3394, - "step": 27013 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019086315362240412, - "loss": 1.4721, - "step": 27014 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019086281539606306, - "loss": 1.4913, - "step": 27015 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019086247716972198, - "loss": 1.4858, - "step": 27016 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019086213894338092, - "loss": 1.5263, - "step": 27017 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019086180071703986, - "loss": 1.6743, - "step": 27018 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019086146249069877, - "loss": 1.5037, - "step": 27019 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019086112426435772, - "loss": 1.6568, - "step": 27020 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019086078603801666, - "loss": 1.6715, - "step": 27021 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019086044781167557, - "loss": 1.4793, - "step": 27022 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908601095853345, - "loss": 1.5771, - "step": 27023 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019085977135899345, - "loss": 1.5886, - "step": 27024 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019085943313265237, - "loss": 1.5952, - "step": 27025 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908590949063113, - "loss": 1.5466, - "step": 27026 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019085875667997025, - "loss": 1.463, - "step": 27027 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019085841845362916, - "loss": 1.7429, - "step": 27028 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908580802272881, - "loss": 1.5097, - "step": 27029 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019085774200094705, - "loss": 1.465, - "step": 27030 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019085740377460596, - "loss": 1.5287, - "step": 27031 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908570655482649, - "loss": 1.5351, - "step": 27032 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019085672732192384, - "loss": 1.6161, - "step": 27033 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019085638909558276, - "loss": 1.6187, - "step": 27034 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908560508692417, - "loss": 1.5196, - "step": 27035 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019085571264290064, - "loss": 1.5066, - "step": 27036 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019085537441655955, - "loss": 1.551, - "step": 27037 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908550361902185, - "loss": 1.4997, - "step": 27038 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019085469796387743, - "loss": 1.5821, - "step": 27039 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019085435973753635, - "loss": 1.5946, - "step": 27040 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908540215111953, - "loss": 1.4874, - "step": 27041 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019085368328485423, - "loss": 1.3247, - "step": 27042 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019085334505851315, - "loss": 1.638, - "step": 27043 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019085300683217209, - "loss": 1.6327, - "step": 27044 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019085266860583105, - "loss": 1.554, - "step": 27045 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019085233037948997, - "loss": 1.4528, - "step": 27046 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908519921531489, - "loss": 1.5652, - "step": 27047 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019085165392680785, - "loss": 1.5774, - "step": 27048 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019085131570046676, - "loss": 1.5515, - "step": 27049 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908509774741257, - "loss": 1.532, - "step": 27050 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019085063924778465, - "loss": 1.5325, - "step": 27051 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019085030102144356, - "loss": 1.4982, - "step": 27052 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908499627951025, - "loss": 1.7828, - "step": 27053 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019084962456876144, - "loss": 1.4745, - "step": 27054 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019084928634242036, - "loss": 1.5377, - "step": 27055 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908489481160793, - "loss": 1.5509, - "step": 27056 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019084860988973824, - "loss": 1.5029, - "step": 27057 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019084827166339715, - "loss": 1.4944, - "step": 27058 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908479334370561, - "loss": 1.5563, - "step": 27059 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019084759521071504, - "loss": 1.6111, - "step": 27060 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019084725698437395, - "loss": 1.4486, - "step": 27061 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908469187580329, - "loss": 1.5305, - "step": 27062 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019084658053169183, - "loss": 1.4196, - "step": 27063 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019084624230535075, - "loss": 1.5628, - "step": 27064 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908459040790097, - "loss": 1.4863, - "step": 27065 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019084556585266863, - "loss": 1.498, - "step": 27066 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019084522762632754, - "loss": 1.4978, - "step": 27067 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019084488939998648, - "loss": 1.5663, - "step": 27068 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019084455117364543, - "loss": 1.4667, - "step": 27069 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019084421294730434, - "loss": 1.4924, - "step": 27070 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019084387472096328, - "loss": 1.3841, - "step": 27071 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019084353649462222, - "loss": 1.6355, - "step": 27072 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019084319826828114, - "loss": 1.5148, - "step": 27073 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019084286004194008, - "loss": 1.5666, - "step": 27074 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019084252181559902, - "loss": 1.7285, - "step": 27075 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019084218358925793, - "loss": 1.6399, - "step": 27076 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019084184536291687, - "loss": 1.5613, - "step": 27077 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019084150713657581, - "loss": 1.5211, - "step": 27078 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019084116891023473, - "loss": 1.623, - "step": 27079 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019084083068389367, - "loss": 1.3843, - "step": 27080 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908404924575526, - "loss": 1.4031, - "step": 27081 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019084015423121152, - "loss": 1.6889, - "step": 27082 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019083981600487047, - "loss": 1.5395, - "step": 27083 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908394777785294, - "loss": 1.1993, - "step": 27084 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019083913955218832, - "loss": 1.6878, - "step": 27085 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019083880132584726, - "loss": 1.855, - "step": 27086 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908384630995062, - "loss": 1.5418, - "step": 27087 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019083812487316512, - "loss": 1.5427, - "step": 27088 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019083778664682406, - "loss": 1.4737, - "step": 27089 - }, - { - "epoch": 0.09, - "learning_rate": 0.000190837448420483, - "loss": 1.5852, - "step": 27090 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908371101941419, - "loss": 1.6047, - "step": 27091 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019083677196780085, - "loss": 1.5227, - "step": 27092 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908364337414598, - "loss": 1.6485, - "step": 27093 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908360955151187, - "loss": 1.4763, - "step": 27094 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019083575728877765, - "loss": 1.5431, - "step": 27095 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908354190624366, - "loss": 1.772, - "step": 27096 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908350808360955, - "loss": 1.5954, - "step": 27097 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019083474260975445, - "loss": 1.4336, - "step": 27098 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908344043834134, - "loss": 1.4518, - "step": 27099 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908340661570723, - "loss": 1.5866, - "step": 27100 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019083372793073127, - "loss": 1.5185, - "step": 27101 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019083338970439018, - "loss": 1.5154, - "step": 27102 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019083305147804913, - "loss": 1.568, - "step": 27103 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019083271325170807, - "loss": 1.5911, - "step": 27104 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019083237502536698, - "loss": 1.5699, - "step": 27105 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019083203679902592, - "loss": 1.4735, - "step": 27106 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019083169857268486, - "loss": 1.3895, - "step": 27107 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019083136034634378, - "loss": 1.7128, - "step": 27108 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019083102212000272, - "loss": 1.4863, - "step": 27109 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019083068389366166, - "loss": 1.4317, - "step": 27110 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019083034566732057, - "loss": 1.5208, - "step": 27111 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019083000744097951, - "loss": 1.4666, - "step": 27112 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019082966921463846, - "loss": 1.5588, - "step": 27113 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908293309882974, - "loss": 1.5856, - "step": 27114 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908289927619563, - "loss": 1.5857, - "step": 27115 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019082865453561525, - "loss": 1.454, - "step": 27116 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908283163092742, - "loss": 1.4642, - "step": 27117 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908279780829331, - "loss": 1.5443, - "step": 27118 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019082763985659205, - "loss": 1.5608, - "step": 27119 - }, - { - "epoch": 0.09, - "learning_rate": 0.000190827301630251, - "loss": 1.6526, - "step": 27120 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908269634039099, - "loss": 1.4315, - "step": 27121 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019082662517756884, - "loss": 1.6251, - "step": 27122 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019082628695122779, - "loss": 1.7337, - "step": 27123 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908259487248867, - "loss": 1.4566, - "step": 27124 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019082561049854564, - "loss": 1.5029, - "step": 27125 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019082527227220458, - "loss": 1.7152, - "step": 27126 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908249340458635, - "loss": 1.5286, - "step": 27127 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019082459581952244, - "loss": 1.5324, - "step": 27128 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019082425759318138, - "loss": 1.7003, - "step": 27129 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908239193668403, - "loss": 1.6846, - "step": 27130 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019082358114049923, - "loss": 1.5876, - "step": 27131 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019082324291415817, - "loss": 1.6875, - "step": 27132 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908229046878171, - "loss": 1.5642, - "step": 27133 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019082256646147603, - "loss": 1.7524, - "step": 27134 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019082222823513497, - "loss": 1.5892, - "step": 27135 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019082189000879389, - "loss": 1.2501, - "step": 27136 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019082155178245283, - "loss": 1.5869, - "step": 27137 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019082121355611177, - "loss": 1.5104, - "step": 27138 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019082087532977068, - "loss": 1.6602, - "step": 27139 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019082053710342962, - "loss": 1.602, - "step": 27140 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019082019887708856, - "loss": 1.5, - "step": 27141 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019081986065074748, - "loss": 1.5829, - "step": 27142 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019081952242440642, - "loss": 1.5803, - "step": 27143 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019081918419806536, - "loss": 1.4859, - "step": 27144 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019081884597172427, - "loss": 1.4782, - "step": 27145 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019081850774538322, - "loss": 1.6388, - "step": 27146 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019081816951904216, - "loss": 1.514, - "step": 27147 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019081783129270107, - "loss": 1.5465, - "step": 27148 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019081749306636, - "loss": 1.4691, - "step": 27149 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019081715484001895, - "loss": 1.523, - "step": 27150 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019081681661367787, - "loss": 1.5042, - "step": 27151 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908164783873368, - "loss": 1.4208, - "step": 27152 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019081614016099575, - "loss": 1.595, - "step": 27153 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019081580193465466, - "loss": 1.3849, - "step": 27154 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908154637083136, - "loss": 1.5061, - "step": 27155 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019081512548197255, - "loss": 1.4818, - "step": 27156 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908147872556315, - "loss": 1.4943, - "step": 27157 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019081444902929043, - "loss": 1.6355, - "step": 27158 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019081411080294934, - "loss": 1.5338, - "step": 27159 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019081377257660828, - "loss": 1.6256, - "step": 27160 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019081343435026722, - "loss": 1.3994, - "step": 27161 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019081309612392614, - "loss": 1.656, - "step": 27162 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019081275789758508, - "loss": 1.5311, - "step": 27163 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019081241967124402, - "loss": 1.5693, - "step": 27164 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019081208144490293, - "loss": 1.5691, - "step": 27165 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019081174321856188, - "loss": 1.5338, - "step": 27166 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019081140499222082, - "loss": 1.5442, - "step": 27167 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019081106676587973, - "loss": 1.4782, - "step": 27168 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019081072853953867, - "loss": 1.6561, - "step": 27169 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908103903131976, - "loss": 1.6711, - "step": 27170 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019081005208685653, - "loss": 1.6486, - "step": 27171 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019080971386051547, - "loss": 1.4409, - "step": 27172 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908093756341744, - "loss": 1.6725, - "step": 27173 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019080903740783332, - "loss": 1.4902, - "step": 27174 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019080869918149226, - "loss": 1.5288, - "step": 27175 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908083609551512, - "loss": 1.4262, - "step": 27176 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019080802272881012, - "loss": 1.4187, - "step": 27177 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019080768450246906, - "loss": 1.6691, - "step": 27178 - }, - { - "epoch": 0.09, - "learning_rate": 0.000190807346276128, - "loss": 1.6778, - "step": 27179 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019080700804978692, - "loss": 1.3699, - "step": 27180 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019080666982344586, - "loss": 1.6314, - "step": 27181 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908063315971048, - "loss": 1.5458, - "step": 27182 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908059933707637, - "loss": 1.5914, - "step": 27183 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019080565514442265, - "loss": 1.4016, - "step": 27184 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908053169180816, - "loss": 1.6362, - "step": 27185 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908049786917405, - "loss": 1.5972, - "step": 27186 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019080464046539945, - "loss": 1.5701, - "step": 27187 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908043022390584, - "loss": 1.5898, - "step": 27188 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908039640127173, - "loss": 1.4362, - "step": 27189 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019080362578637625, - "loss": 1.5156, - "step": 27190 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908032875600352, - "loss": 1.61, - "step": 27191 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908029493336941, - "loss": 1.5672, - "step": 27192 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019080261110735304, - "loss": 1.578, - "step": 27193 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019080227288101198, - "loss": 1.5102, - "step": 27194 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908019346546709, - "loss": 1.5514, - "step": 27195 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019080159642832984, - "loss": 1.6543, - "step": 27196 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019080125820198878, - "loss": 1.5035, - "step": 27197 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001908009199756477, - "loss": 1.3986, - "step": 27198 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019080058174930663, - "loss": 1.5638, - "step": 27199 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019080024352296558, - "loss": 1.4168, - "step": 27200 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019079990529662452, - "loss": 1.5788, - "step": 27201 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019079956707028343, - "loss": 1.3859, - "step": 27202 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019079922884394237, - "loss": 1.5974, - "step": 27203 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907988906176013, - "loss": 1.6422, - "step": 27204 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019079855239126023, - "loss": 1.7038, - "step": 27205 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019079821416491917, - "loss": 1.5477, - "step": 27206 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907978759385781, - "loss": 1.6815, - "step": 27207 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019079753771223702, - "loss": 1.5901, - "step": 27208 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019079719948589596, - "loss": 1.5409, - "step": 27209 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907968612595549, - "loss": 1.6009, - "step": 27210 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019079652303321382, - "loss": 1.4857, - "step": 27211 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019079618480687276, - "loss": 1.7364, - "step": 27212 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907958465805317, - "loss": 1.4501, - "step": 27213 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019079550835419064, - "loss": 1.6097, - "step": 27214 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019079517012784958, - "loss": 1.4689, - "step": 27215 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907948319015085, - "loss": 1.4505, - "step": 27216 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019079449367516744, - "loss": 1.7247, - "step": 27217 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019079415544882638, - "loss": 1.5777, - "step": 27218 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907938172224853, - "loss": 1.4814, - "step": 27219 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019079347899614424, - "loss": 1.5167, - "step": 27220 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019079314076980318, - "loss": 1.5685, - "step": 27221 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907928025434621, - "loss": 1.6157, - "step": 27222 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019079246431712103, - "loss": 1.4862, - "step": 27223 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019079212609077997, - "loss": 1.5796, - "step": 27224 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907917878644389, - "loss": 1.6092, - "step": 27225 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019079144963809783, - "loss": 1.4106, - "step": 27226 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019079111141175677, - "loss": 1.704, - "step": 27227 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019079077318541568, - "loss": 1.6152, - "step": 27228 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019079043495907463, - "loss": 1.5423, - "step": 27229 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019079009673273357, - "loss": 1.7728, - "step": 27230 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019078975850639248, - "loss": 1.6568, - "step": 27231 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019078942028005142, - "loss": 1.7686, - "step": 27232 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019078908205371036, - "loss": 1.5909, - "step": 27233 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019078874382736928, - "loss": 1.5314, - "step": 27234 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019078840560102822, - "loss": 1.6833, - "step": 27235 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019078806737468716, - "loss": 1.5001, - "step": 27236 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019078772914834607, - "loss": 1.6012, - "step": 27237 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019078739092200501, - "loss": 1.3541, - "step": 27238 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019078705269566396, - "loss": 1.3845, - "step": 27239 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019078671446932287, - "loss": 1.4806, - "step": 27240 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907863762429818, - "loss": 1.6281, - "step": 27241 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019078603801664075, - "loss": 1.5657, - "step": 27242 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019078569979029967, - "loss": 1.5404, - "step": 27243 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907853615639586, - "loss": 1.5814, - "step": 27244 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019078502333761755, - "loss": 1.3854, - "step": 27245 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019078468511127646, - "loss": 1.4848, - "step": 27246 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907843468849354, - "loss": 1.5905, - "step": 27247 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019078400865859434, - "loss": 1.6648, - "step": 27248 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019078367043225326, - "loss": 1.5168, - "step": 27249 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907833322059122, - "loss": 1.4633, - "step": 27250 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019078299397957114, - "loss": 1.7137, - "step": 27251 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019078265575323005, - "loss": 1.5621, - "step": 27252 - }, - { - "epoch": 0.09, - "learning_rate": 0.000190782317526889, - "loss": 1.3966, - "step": 27253 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019078197930054794, - "loss": 1.6387, - "step": 27254 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019078164107420685, - "loss": 1.4215, - "step": 27255 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907813028478658, - "loss": 1.572, - "step": 27256 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019078096462152473, - "loss": 1.5812, - "step": 27257 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019078062639518365, - "loss": 1.5668, - "step": 27258 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907802881688426, - "loss": 1.6124, - "step": 27259 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019077994994250153, - "loss": 1.5843, - "step": 27260 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019077961171616044, - "loss": 1.526, - "step": 27261 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019077927348981938, - "loss": 1.4721, - "step": 27262 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019077893526347833, - "loss": 1.5874, - "step": 27263 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019077859703713724, - "loss": 1.5784, - "step": 27264 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019077825881079618, - "loss": 1.6256, - "step": 27265 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019077792058445512, - "loss": 1.7096, - "step": 27266 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019077758235811404, - "loss": 1.4708, - "step": 27267 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019077724413177298, - "loss": 1.4725, - "step": 27268 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019077690590543195, - "loss": 1.4469, - "step": 27269 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019077656767909086, - "loss": 1.572, - "step": 27270 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907762294527498, - "loss": 1.5541, - "step": 27271 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019077589122640874, - "loss": 1.4607, - "step": 27272 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019077555300006766, - "loss": 1.5547, - "step": 27273 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907752147737266, - "loss": 1.6056, - "step": 27274 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019077487654738554, - "loss": 1.38, - "step": 27275 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019077453832104445, - "loss": 1.7512, - "step": 27276 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907742000947034, - "loss": 1.5836, - "step": 27277 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019077386186836233, - "loss": 1.508, - "step": 27278 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019077352364202125, - "loss": 1.5886, - "step": 27279 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907731854156802, - "loss": 1.4495, - "step": 27280 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019077284718933913, - "loss": 1.5539, - "step": 27281 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019077250896299804, - "loss": 1.6249, - "step": 27282 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019077217073665699, - "loss": 1.5246, - "step": 27283 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019077183251031593, - "loss": 1.5699, - "step": 27284 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019077149428397484, - "loss": 1.5364, - "step": 27285 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019077115605763378, - "loss": 1.5571, - "step": 27286 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019077081783129272, - "loss": 1.5915, - "step": 27287 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019077047960495164, - "loss": 1.6723, - "step": 27288 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019077014137861058, - "loss": 1.6137, - "step": 27289 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019076980315226952, - "loss": 1.5907, - "step": 27290 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019076946492592843, - "loss": 1.4243, - "step": 27291 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019076912669958737, - "loss": 1.361, - "step": 27292 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019076878847324632, - "loss": 1.8099, - "step": 27293 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019076845024690523, - "loss": 1.4459, - "step": 27294 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019076811202056417, - "loss": 1.6024, - "step": 27295 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907677737942231, - "loss": 1.528, - "step": 27296 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019076743556788203, - "loss": 1.5348, - "step": 27297 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019076709734154097, - "loss": 1.7688, - "step": 27298 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907667591151999, - "loss": 1.5706, - "step": 27299 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019076642088885882, - "loss": 1.6699, - "step": 27300 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019076608266251776, - "loss": 1.5118, - "step": 27301 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907657444361767, - "loss": 1.4884, - "step": 27302 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019076540620983562, - "loss": 1.5483, - "step": 27303 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019076506798349456, - "loss": 1.4875, - "step": 27304 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907647297571535, - "loss": 1.4606, - "step": 27305 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019076439153081242, - "loss": 1.5944, - "step": 27306 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019076405330447136, - "loss": 1.6984, - "step": 27307 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907637150781303, - "loss": 1.6659, - "step": 27308 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907633768517892, - "loss": 1.4523, - "step": 27309 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019076303862544815, - "loss": 1.4975, - "step": 27310 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907627003991071, - "loss": 1.6323, - "step": 27311 - }, - { - "epoch": 0.09, - "learning_rate": 0.000190762362172766, - "loss": 1.5379, - "step": 27312 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019076202394642495, - "loss": 1.6517, - "step": 27313 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907616857200839, - "loss": 1.5084, - "step": 27314 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907613474937428, - "loss": 1.5918, - "step": 27315 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019076100926740175, - "loss": 1.4861, - "step": 27316 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907606710410607, - "loss": 1.6609, - "step": 27317 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907603328147196, - "loss": 1.499, - "step": 27318 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019075999458837854, - "loss": 1.552, - "step": 27319 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019075965636203748, - "loss": 1.5014, - "step": 27320 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907593181356964, - "loss": 1.5838, - "step": 27321 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019075897990935534, - "loss": 1.5885, - "step": 27322 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019075864168301428, - "loss": 1.5544, - "step": 27323 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907583034566732, - "loss": 1.5383, - "step": 27324 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019075796523033216, - "loss": 1.6383, - "step": 27325 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907576270039911, - "loss": 1.5281, - "step": 27326 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019075728877765002, - "loss": 1.4419, - "step": 27327 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019075695055130896, - "loss": 1.5299, - "step": 27328 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907566123249679, - "loss": 1.6368, - "step": 27329 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907562740986268, - "loss": 1.5653, - "step": 27330 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019075593587228575, - "loss": 1.616, - "step": 27331 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907555976459447, - "loss": 1.4456, - "step": 27332 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907552594196036, - "loss": 1.5753, - "step": 27333 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019075492119326255, - "loss": 1.7326, - "step": 27334 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907545829669215, - "loss": 1.5575, - "step": 27335 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907542447405804, - "loss": 1.6106, - "step": 27336 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019075390651423935, - "loss": 1.6108, - "step": 27337 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907535682878983, - "loss": 1.5215, - "step": 27338 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907532300615572, - "loss": 1.4876, - "step": 27339 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019075289183521614, - "loss": 1.6986, - "step": 27340 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019075255360887508, - "loss": 1.5837, - "step": 27341 - }, - { - "epoch": 0.09, - "learning_rate": 0.000190752215382534, - "loss": 1.763, - "step": 27342 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019075187715619294, - "loss": 1.4339, - "step": 27343 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019075153892985188, - "loss": 1.4115, - "step": 27344 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907512007035108, - "loss": 1.619, - "step": 27345 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019075086247716974, - "loss": 1.5638, - "step": 27346 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019075052425082868, - "loss": 1.6202, - "step": 27347 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907501860244876, - "loss": 1.5081, - "step": 27348 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019074984779814653, - "loss": 1.5229, - "step": 27349 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019074950957180547, - "loss": 1.5597, - "step": 27350 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907491713454644, - "loss": 1.5849, - "step": 27351 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019074883311912333, - "loss": 1.4106, - "step": 27352 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019074849489278227, - "loss": 1.5583, - "step": 27353 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019074815666644118, - "loss": 1.4726, - "step": 27354 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019074781844010012, - "loss": 1.509, - "step": 27355 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019074748021375907, - "loss": 1.7642, - "step": 27356 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019074714198741798, - "loss": 1.5259, - "step": 27357 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019074680376107692, - "loss": 1.6907, - "step": 27358 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019074646553473586, - "loss": 1.4582, - "step": 27359 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019074612730839478, - "loss": 1.3455, - "step": 27360 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019074578908205372, - "loss": 1.6589, - "step": 27361 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019074545085571266, - "loss": 1.5044, - "step": 27362 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019074511262937157, - "loss": 1.6201, - "step": 27363 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907447744030305, - "loss": 1.561, - "step": 27364 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019074443617668945, - "loss": 1.4453, - "step": 27365 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019074409795034837, - "loss": 1.283, - "step": 27366 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907437597240073, - "loss": 1.5851, - "step": 27367 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019074342149766625, - "loss": 1.6155, - "step": 27368 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019074308327132516, - "loss": 1.4211, - "step": 27369 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907427450449841, - "loss": 1.5739, - "step": 27370 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019074240681864305, - "loss": 1.6261, - "step": 27371 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019074206859230196, - "loss": 1.6421, - "step": 27372 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907417303659609, - "loss": 1.6231, - "step": 27373 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019074139213961984, - "loss": 1.5766, - "step": 27374 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019074105391327876, - "loss": 1.6466, - "step": 27375 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907407156869377, - "loss": 1.451, - "step": 27376 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019074037746059664, - "loss": 1.5934, - "step": 27377 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019074003923425555, - "loss": 1.5111, - "step": 27378 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907397010079145, - "loss": 1.5417, - "step": 27379 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019073936278157344, - "loss": 1.5242, - "step": 27380 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019073902455523235, - "loss": 1.5393, - "step": 27381 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019073868632889132, - "loss": 1.5184, - "step": 27382 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019073834810255026, - "loss": 1.5609, - "step": 27383 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019073800987620917, - "loss": 1.7299, - "step": 27384 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019073767164986811, - "loss": 1.6688, - "step": 27385 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019073733342352706, - "loss": 1.4729, - "step": 27386 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019073699519718597, - "loss": 1.6229, - "step": 27387 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907366569708449, - "loss": 1.6859, - "step": 27388 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019073631874450385, - "loss": 1.6826, - "step": 27389 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019073598051816277, - "loss": 1.6054, - "step": 27390 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907356422918217, - "loss": 1.634, - "step": 27391 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019073530406548065, - "loss": 1.6129, - "step": 27392 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019073496583913956, - "loss": 1.5422, - "step": 27393 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907346276127985, - "loss": 1.7054, - "step": 27394 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019073428938645744, - "loss": 1.6647, - "step": 27395 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019073395116011636, - "loss": 1.5356, - "step": 27396 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907336129337753, - "loss": 1.42, - "step": 27397 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019073327470743424, - "loss": 1.4814, - "step": 27398 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019073293648109316, - "loss": 1.5704, - "step": 27399 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907325982547521, - "loss": 1.7004, - "step": 27400 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019073226002841104, - "loss": 1.571, - "step": 27401 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019073192180206995, - "loss": 1.4563, - "step": 27402 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907315835757289, - "loss": 1.5357, - "step": 27403 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019073124534938783, - "loss": 1.4421, - "step": 27404 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019073090712304675, - "loss": 1.4633, - "step": 27405 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907305688967057, - "loss": 1.5637, - "step": 27406 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019073023067036463, - "loss": 1.5637, - "step": 27407 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019072989244402354, - "loss": 1.5482, - "step": 27408 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019072955421768249, - "loss": 1.5238, - "step": 27409 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019072921599134143, - "loss": 1.4814, - "step": 27410 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019072887776500034, - "loss": 1.4608, - "step": 27411 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019072853953865928, - "loss": 1.6615, - "step": 27412 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019072820131231822, - "loss": 1.5901, - "step": 27413 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019072786308597714, - "loss": 1.4749, - "step": 27414 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019072752485963608, - "loss": 1.6435, - "step": 27415 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019072718663329502, - "loss": 1.543, - "step": 27416 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019072684840695393, - "loss": 1.6289, - "step": 27417 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019072651018061287, - "loss": 1.4944, - "step": 27418 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019072617195427182, - "loss": 1.5963, - "step": 27419 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019072583372793073, - "loss": 1.5544, - "step": 27420 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019072549550158967, - "loss": 1.6269, - "step": 27421 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907251572752486, - "loss": 1.4955, - "step": 27422 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019072481904890753, - "loss": 1.4378, - "step": 27423 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019072448082256647, - "loss": 1.5748, - "step": 27424 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907241425962254, - "loss": 1.5328, - "step": 27425 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019072380436988432, - "loss": 1.5422, - "step": 27426 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019072346614354326, - "loss": 1.4075, - "step": 27427 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907231279172022, - "loss": 1.5667, - "step": 27428 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019072278969086112, - "loss": 1.5672, - "step": 27429 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019072245146452006, - "loss": 1.5551, - "step": 27430 - }, - { - "epoch": 0.09, - "learning_rate": 0.000190722113238179, - "loss": 1.5942, - "step": 27431 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019072177501183791, - "loss": 1.5835, - "step": 27432 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019072143678549686, - "loss": 1.4975, - "step": 27433 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907210985591558, - "loss": 1.5432, - "step": 27434 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907207603328147, - "loss": 1.5042, - "step": 27435 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019072042210647365, - "loss": 1.5122, - "step": 27436 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907200838801326, - "loss": 1.525, - "step": 27437 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019071974565379153, - "loss": 1.7324, - "step": 27438 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019071940742745048, - "loss": 1.5012, - "step": 27439 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907190692011094, - "loss": 1.5938, - "step": 27440 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019071873097476833, - "loss": 1.6215, - "step": 27441 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019071839274842727, - "loss": 1.5824, - "step": 27442 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019071805452208619, - "loss": 1.4235, - "step": 27443 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019071771629574513, - "loss": 1.635, - "step": 27444 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019071737806940407, - "loss": 1.5831, - "step": 27445 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019071703984306298, - "loss": 1.6615, - "step": 27446 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019071670161672192, - "loss": 1.4624, - "step": 27447 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019071636339038086, - "loss": 1.6164, - "step": 27448 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019071602516403978, - "loss": 1.6548, - "step": 27449 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019071568693769872, - "loss": 1.6318, - "step": 27450 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019071534871135766, - "loss": 1.5936, - "step": 27451 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019071501048501657, - "loss": 1.5815, - "step": 27452 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019071467225867552, - "loss": 1.5546, - "step": 27453 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019071433403233446, - "loss": 1.3397, - "step": 27454 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019071399580599337, - "loss": 1.7112, - "step": 27455 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907136575796523, - "loss": 1.5361, - "step": 27456 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019071331935331125, - "loss": 1.5191, - "step": 27457 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019071298112697017, - "loss": 1.6383, - "step": 27458 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907126429006291, - "loss": 1.6003, - "step": 27459 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019071230467428805, - "loss": 1.3537, - "step": 27460 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019071196644794696, - "loss": 1.4654, - "step": 27461 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907116282216059, - "loss": 1.6448, - "step": 27462 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019071128999526485, - "loss": 1.6008, - "step": 27463 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019071095176892376, - "loss": 1.7096, - "step": 27464 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907106135425827, - "loss": 1.7335, - "step": 27465 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019071027531624164, - "loss": 1.6923, - "step": 27466 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019070993708990058, - "loss": 1.5704, - "step": 27467 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907095988635595, - "loss": 1.5219, - "step": 27468 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019070926063721844, - "loss": 1.5567, - "step": 27469 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019070892241087738, - "loss": 1.541, - "step": 27470 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907085841845363, - "loss": 1.6095, - "step": 27471 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019070824595819523, - "loss": 1.4284, - "step": 27472 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019070790773185418, - "loss": 1.7012, - "step": 27473 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907075695055131, - "loss": 1.4823, - "step": 27474 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019070723127917203, - "loss": 1.7256, - "step": 27475 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019070689305283097, - "loss": 1.4301, - "step": 27476 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019070655482648989, - "loss": 1.6036, - "step": 27477 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019070621660014883, - "loss": 1.5776, - "step": 27478 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019070587837380777, - "loss": 1.5695, - "step": 27479 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019070554014746668, - "loss": 1.376, - "step": 27480 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019070520192112562, - "loss": 1.5666, - "step": 27481 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019070486369478457, - "loss": 1.4057, - "step": 27482 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019070452546844348, - "loss": 1.3921, - "step": 27483 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019070418724210242, - "loss": 1.4653, - "step": 27484 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019070384901576136, - "loss": 1.342, - "step": 27485 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019070351078942028, - "loss": 1.3538, - "step": 27486 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019070317256307922, - "loss": 1.4957, - "step": 27487 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019070283433673816, - "loss": 1.5666, - "step": 27488 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019070249611039707, - "loss": 1.5374, - "step": 27489 - }, - { - "epoch": 0.09, - "learning_rate": 0.000190702157884056, - "loss": 1.4003, - "step": 27490 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019070181965771495, - "loss": 1.4085, - "step": 27491 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019070148143137387, - "loss": 1.5991, - "step": 27492 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907011432050328, - "loss": 1.5047, - "step": 27493 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019070080497869175, - "loss": 1.6696, - "step": 27494 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001907004667523507, - "loss": 1.567, - "step": 27495 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019070012852600963, - "loss": 1.6003, - "step": 27496 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019069979029966855, - "loss": 1.4824, - "step": 27497 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906994520733275, - "loss": 1.4714, - "step": 27498 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019069911384698643, - "loss": 1.4929, - "step": 27499 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019069877562064534, - "loss": 1.5551, - "step": 27500 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019069843739430428, - "loss": 1.5239, - "step": 27501 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019069809916796323, - "loss": 1.5529, - "step": 27502 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019069776094162214, - "loss": 1.6303, - "step": 27503 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019069742271528108, - "loss": 1.6689, - "step": 27504 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019069708448894002, - "loss": 1.5546, - "step": 27505 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019069674626259894, - "loss": 1.5001, - "step": 27506 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019069640803625788, - "loss": 1.6083, - "step": 27507 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019069606980991682, - "loss": 1.5413, - "step": 27508 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019069573158357573, - "loss": 1.4478, - "step": 27509 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019069539335723467, - "loss": 1.527, - "step": 27510 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019069505513089361, - "loss": 1.6044, - "step": 27511 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019069471690455253, - "loss": 1.6667, - "step": 27512 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019069437867821147, - "loss": 1.4127, - "step": 27513 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906940404518704, - "loss": 1.4997, - "step": 27514 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019069370222552932, - "loss": 1.4895, - "step": 27515 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019069336399918827, - "loss": 1.438, - "step": 27516 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906930257728472, - "loss": 1.5924, - "step": 27517 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019069268754650612, - "loss": 1.5535, - "step": 27518 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019069234932016506, - "loss": 1.675, - "step": 27519 - }, - { - "epoch": 0.09, - "learning_rate": 0.000190692011093824, - "loss": 1.5495, - "step": 27520 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019069167286748292, - "loss": 1.459, - "step": 27521 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019069133464114186, - "loss": 1.5625, - "step": 27522 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906909964148008, - "loss": 1.5081, - "step": 27523 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906906581884597, - "loss": 1.4637, - "step": 27524 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019069031996211865, - "loss": 1.4695, - "step": 27525 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906899817357776, - "loss": 1.3206, - "step": 27526 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906896435094365, - "loss": 1.4515, - "step": 27527 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019068930528309545, - "loss": 1.6476, - "step": 27528 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906889670567544, - "loss": 1.6596, - "step": 27529 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906886288304133, - "loss": 1.6118, - "step": 27530 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019068829060407225, - "loss": 1.6276, - "step": 27531 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906879523777312, - "loss": 1.6743, - "step": 27532 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906876141513901, - "loss": 1.4938, - "step": 27533 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019068727592504904, - "loss": 1.5506, - "step": 27534 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019068693769870798, - "loss": 1.545, - "step": 27535 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906865994723669, - "loss": 1.5793, - "step": 27536 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019068626124602584, - "loss": 1.4506, - "step": 27537 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019068592301968478, - "loss": 1.4099, - "step": 27538 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906855847933437, - "loss": 1.567, - "step": 27539 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019068524656700264, - "loss": 1.6131, - "step": 27540 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019068490834066158, - "loss": 1.5778, - "step": 27541 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906845701143205, - "loss": 1.4886, - "step": 27542 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019068423188797943, - "loss": 1.607, - "step": 27543 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019068389366163837, - "loss": 1.4497, - "step": 27544 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906835554352973, - "loss": 1.5887, - "step": 27545 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019068321720895623, - "loss": 1.5775, - "step": 27546 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019068287898261517, - "loss": 1.4648, - "step": 27547 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019068254075627408, - "loss": 1.4715, - "step": 27548 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019068220252993303, - "loss": 1.4469, - "step": 27549 - }, - { - "epoch": 0.09, - "learning_rate": 0.000190681864303592, - "loss": 1.4749, - "step": 27550 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906815260772509, - "loss": 1.4276, - "step": 27551 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019068118785090985, - "loss": 1.6669, - "step": 27552 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906808496245688, - "loss": 1.6081, - "step": 27553 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906805113982277, - "loss": 1.7144, - "step": 27554 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019068017317188664, - "loss": 1.6666, - "step": 27555 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019067983494554559, - "loss": 1.5415, - "step": 27556 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906794967192045, - "loss": 1.4396, - "step": 27557 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019067915849286344, - "loss": 1.6288, - "step": 27558 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019067882026652238, - "loss": 1.6557, - "step": 27559 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906784820401813, - "loss": 1.4675, - "step": 27560 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019067814381384024, - "loss": 1.4328, - "step": 27561 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019067780558749918, - "loss": 1.533, - "step": 27562 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906774673611581, - "loss": 1.5185, - "step": 27563 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019067712913481703, - "loss": 1.561, - "step": 27564 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019067679090847597, - "loss": 1.5477, - "step": 27565 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906764526821349, - "loss": 1.6454, - "step": 27566 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019067611445579383, - "loss": 1.703, - "step": 27567 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019067577622945277, - "loss": 1.5929, - "step": 27568 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019067543800311169, - "loss": 1.5544, - "step": 27569 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019067509977677063, - "loss": 1.5833, - "step": 27570 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019067476155042957, - "loss": 1.5698, - "step": 27571 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019067442332408848, - "loss": 1.4418, - "step": 27572 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019067408509774742, - "loss": 1.7014, - "step": 27573 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019067374687140636, - "loss": 1.5479, - "step": 27574 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019067340864506528, - "loss": 1.4632, - "step": 27575 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019067307041872422, - "loss": 1.616, - "step": 27576 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019067273219238316, - "loss": 1.5754, - "step": 27577 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019067239396604207, - "loss": 1.4718, - "step": 27578 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019067205573970102, - "loss": 1.6266, - "step": 27579 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019067171751335996, - "loss": 1.4411, - "step": 27580 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019067137928701887, - "loss": 1.3991, - "step": 27581 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906710410606778, - "loss": 1.544, - "step": 27582 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019067070283433675, - "loss": 1.4566, - "step": 27583 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019067036460799567, - "loss": 1.5938, - "step": 27584 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906700263816546, - "loss": 1.6847, - "step": 27585 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019066968815531355, - "loss": 1.4164, - "step": 27586 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019066934992897246, - "loss": 1.8205, - "step": 27587 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906690117026314, - "loss": 1.5257, - "step": 27588 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019066867347629035, - "loss": 1.509, - "step": 27589 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019066833524994926, - "loss": 1.5262, - "step": 27590 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906679970236082, - "loss": 1.3712, - "step": 27591 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019066765879726714, - "loss": 1.4805, - "step": 27592 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019066732057092606, - "loss": 1.5117, - "step": 27593 - }, - { - "epoch": 0.09, - "learning_rate": 0.000190666982344585, - "loss": 1.5149, - "step": 27594 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019066664411824394, - "loss": 1.5636, - "step": 27595 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019066630589190285, - "loss": 1.6495, - "step": 27596 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906659676655618, - "loss": 1.6171, - "step": 27597 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019066562943922073, - "loss": 1.6858, - "step": 27598 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019066529121287965, - "loss": 1.6232, - "step": 27599 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906649529865386, - "loss": 1.6445, - "step": 27600 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019066461476019753, - "loss": 1.5194, - "step": 27601 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019066427653385644, - "loss": 1.4707, - "step": 27602 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019066393830751539, - "loss": 1.5833, - "step": 27603 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019066360008117433, - "loss": 1.39, - "step": 27604 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019066326185483324, - "loss": 1.6421, - "step": 27605 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906629236284922, - "loss": 1.625, - "step": 27606 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019066258540215115, - "loss": 1.5828, - "step": 27607 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019066224717581006, - "loss": 1.4689, - "step": 27608 - }, - { - "epoch": 0.09, - "learning_rate": 0.000190661908949469, - "loss": 1.4801, - "step": 27609 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019066157072312795, - "loss": 1.6533, - "step": 27610 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019066123249678686, - "loss": 1.5349, - "step": 27611 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906608942704458, - "loss": 1.6572, - "step": 27612 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019066055604410474, - "loss": 1.51, - "step": 27613 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019066021781776366, - "loss": 1.5665, - "step": 27614 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906598795914226, - "loss": 1.7902, - "step": 27615 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019065954136508154, - "loss": 1.5483, - "step": 27616 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019065920313874045, - "loss": 1.5679, - "step": 27617 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906588649123994, - "loss": 1.5588, - "step": 27618 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019065852668605834, - "loss": 1.5173, - "step": 27619 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019065818845971725, - "loss": 1.4543, - "step": 27620 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906578502333762, - "loss": 1.608, - "step": 27621 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019065751200703513, - "loss": 1.7067, - "step": 27622 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019065717378069405, - "loss": 1.475, - "step": 27623 - }, - { - "epoch": 0.09, - "learning_rate": 0.000190656835554353, - "loss": 1.6752, - "step": 27624 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019065649732801193, - "loss": 1.6923, - "step": 27625 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019065615910167084, - "loss": 1.4308, - "step": 27626 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019065582087532978, - "loss": 1.5745, - "step": 27627 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019065548264898872, - "loss": 1.6428, - "step": 27628 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019065514442264764, - "loss": 1.5798, - "step": 27629 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019065480619630658, - "loss": 1.611, - "step": 27630 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019065446796996552, - "loss": 1.614, - "step": 27631 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019065412974362443, - "loss": 1.5783, - "step": 27632 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019065379151728338, - "loss": 1.48, - "step": 27633 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019065345329094232, - "loss": 1.5361, - "step": 27634 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019065311506460123, - "loss": 1.547, - "step": 27635 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019065277683826017, - "loss": 1.4402, - "step": 27636 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906524386119191, - "loss": 1.6368, - "step": 27637 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019065210038557803, - "loss": 1.6737, - "step": 27638 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019065176215923697, - "loss": 1.6365, - "step": 27639 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906514239328959, - "loss": 1.595, - "step": 27640 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019065108570655482, - "loss": 1.7996, - "step": 27641 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019065074748021376, - "loss": 1.4698, - "step": 27642 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906504092538727, - "loss": 1.4012, - "step": 27643 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019065007102753162, - "loss": 1.4675, - "step": 27644 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019064973280119056, - "loss": 1.5818, - "step": 27645 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906493945748495, - "loss": 1.5571, - "step": 27646 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019064905634850842, - "loss": 1.561, - "step": 27647 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019064871812216736, - "loss": 1.4634, - "step": 27648 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906483798958263, - "loss": 1.5354, - "step": 27649 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906480416694852, - "loss": 1.4536, - "step": 27650 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019064770344314415, - "loss": 1.6319, - "step": 27651 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906473652168031, - "loss": 1.7586, - "step": 27652 - }, - { - "epoch": 0.09, - "learning_rate": 0.000190647026990462, - "loss": 1.6257, - "step": 27653 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019064668876412095, - "loss": 1.648, - "step": 27654 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906463505377799, - "loss": 1.4867, - "step": 27655 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906460123114388, - "loss": 1.5971, - "step": 27656 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019064567408509775, - "loss": 1.5676, - "step": 27657 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906453358587567, - "loss": 1.6051, - "step": 27658 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906449976324156, - "loss": 1.5497, - "step": 27659 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019064465940607454, - "loss": 1.5402, - "step": 27660 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019064432117973348, - "loss": 1.5126, - "step": 27661 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019064398295339243, - "loss": 1.508, - "step": 27662 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019064364472705137, - "loss": 1.5362, - "step": 27663 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906433065007103, - "loss": 1.3837, - "step": 27664 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019064296827436922, - "loss": 1.7831, - "step": 27665 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019064263004802816, - "loss": 1.4524, - "step": 27666 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906422918216871, - "loss": 1.5306, - "step": 27667 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019064195359534602, - "loss": 1.5608, - "step": 27668 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019064161536900496, - "loss": 1.4622, - "step": 27669 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906412771426639, - "loss": 1.5536, - "step": 27670 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019064093891632281, - "loss": 1.5626, - "step": 27671 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019064060068998176, - "loss": 1.3717, - "step": 27672 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906402624636407, - "loss": 1.5551, - "step": 27673 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906399242372996, - "loss": 1.4766, - "step": 27674 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019063958601095855, - "loss": 1.5316, - "step": 27675 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906392477846175, - "loss": 1.5317, - "step": 27676 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906389095582764, - "loss": 1.5517, - "step": 27677 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019063857133193535, - "loss": 1.3928, - "step": 27678 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906382331055943, - "loss": 1.6346, - "step": 27679 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906378948792532, - "loss": 1.6239, - "step": 27680 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019063755665291214, - "loss": 1.4725, - "step": 27681 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019063721842657109, - "loss": 1.5571, - "step": 27682 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019063688020023, - "loss": 1.6709, - "step": 27683 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019063654197388894, - "loss": 1.5898, - "step": 27684 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019063620374754788, - "loss": 1.529, - "step": 27685 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906358655212068, - "loss": 1.4146, - "step": 27686 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019063552729486574, - "loss": 1.6959, - "step": 27687 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019063518906852468, - "loss": 1.6028, - "step": 27688 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906348508421836, - "loss": 1.557, - "step": 27689 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019063451261584253, - "loss": 1.4851, - "step": 27690 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019063417438950147, - "loss": 1.6267, - "step": 27691 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906338361631604, - "loss": 1.5692, - "step": 27692 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019063349793681933, - "loss": 1.4959, - "step": 27693 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019063315971047827, - "loss": 1.5042, - "step": 27694 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019063282148413718, - "loss": 1.4823, - "step": 27695 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019063248325779613, - "loss": 1.532, - "step": 27696 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019063214503145507, - "loss": 1.4754, - "step": 27697 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019063180680511398, - "loss": 1.628, - "step": 27698 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019063146857877292, - "loss": 1.5906, - "step": 27699 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019063113035243186, - "loss": 1.5419, - "step": 27700 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019063079212609078, - "loss": 1.6454, - "step": 27701 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019063045389974972, - "loss": 1.6339, - "step": 27702 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019063011567340866, - "loss": 1.5009, - "step": 27703 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019062977744706757, - "loss": 1.6976, - "step": 27704 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019062943922072651, - "loss": 1.4478, - "step": 27705 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019062910099438546, - "loss": 1.6452, - "step": 27706 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019062876276804437, - "loss": 1.5302, - "step": 27707 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906284245417033, - "loss": 1.5963, - "step": 27708 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019062808631536225, - "loss": 1.5046, - "step": 27709 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019062774808902117, - "loss": 1.7079, - "step": 27710 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906274098626801, - "loss": 1.4966, - "step": 27711 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019062707163633905, - "loss": 1.5996, - "step": 27712 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019062673340999796, - "loss": 1.6788, - "step": 27713 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906263951836569, - "loss": 1.5535, - "step": 27714 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019062605695731584, - "loss": 1.4532, - "step": 27715 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019062571873097476, - "loss": 1.571, - "step": 27716 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906253805046337, - "loss": 1.6944, - "step": 27717 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019062504227829264, - "loss": 1.5192, - "step": 27718 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019062470405195158, - "loss": 1.7868, - "step": 27719 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019062436582561052, - "loss": 1.4759, - "step": 27720 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019062402759926944, - "loss": 1.4918, - "step": 27721 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019062368937292838, - "loss": 1.4853, - "step": 27722 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019062335114658732, - "loss": 1.572, - "step": 27723 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019062301292024623, - "loss": 1.6447, - "step": 27724 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019062267469390517, - "loss": 1.481, - "step": 27725 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019062233646756412, - "loss": 1.4457, - "step": 27726 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019062199824122303, - "loss": 1.6631, - "step": 27727 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019062166001488197, - "loss": 1.5822, - "step": 27728 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906213217885409, - "loss": 1.543, - "step": 27729 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019062098356219983, - "loss": 1.4594, - "step": 27730 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019062064533585877, - "loss": 1.4919, - "step": 27731 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906203071095177, - "loss": 1.5928, - "step": 27732 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019061996888317662, - "loss": 1.6683, - "step": 27733 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019061963065683556, - "loss": 1.5431, - "step": 27734 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906192924304945, - "loss": 1.5702, - "step": 27735 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019061895420415345, - "loss": 1.7213, - "step": 27736 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019061861597781236, - "loss": 1.6836, - "step": 27737 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906182777514713, - "loss": 1.5226, - "step": 27738 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019061793952513024, - "loss": 1.5566, - "step": 27739 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019061760129878916, - "loss": 1.7422, - "step": 27740 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906172630724481, - "loss": 1.5681, - "step": 27741 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019061692484610704, - "loss": 1.4234, - "step": 27742 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019061658661976595, - "loss": 1.6374, - "step": 27743 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906162483934249, - "loss": 1.5428, - "step": 27744 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019061591016708383, - "loss": 1.6754, - "step": 27745 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019061557194074275, - "loss": 1.5907, - "step": 27746 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906152337144017, - "loss": 1.5888, - "step": 27747 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019061489548806063, - "loss": 1.5241, - "step": 27748 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019061455726171955, - "loss": 1.634, - "step": 27749 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906142190353785, - "loss": 1.6162, - "step": 27750 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019061388080903743, - "loss": 1.6018, - "step": 27751 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019061354258269634, - "loss": 1.6379, - "step": 27752 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019061320435635528, - "loss": 1.6671, - "step": 27753 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019061286613001422, - "loss": 1.6617, - "step": 27754 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019061252790367314, - "loss": 1.511, - "step": 27755 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019061218967733208, - "loss": 1.4505, - "step": 27756 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019061185145099102, - "loss": 1.3702, - "step": 27757 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019061151322464993, - "loss": 1.679, - "step": 27758 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019061117499830888, - "loss": 1.4126, - "step": 27759 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019061083677196782, - "loss": 1.5454, - "step": 27760 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019061049854562673, - "loss": 1.5457, - "step": 27761 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019061016031928567, - "loss": 1.6427, - "step": 27762 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906098220929446, - "loss": 1.3835, - "step": 27763 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019060948386660353, - "loss": 1.3567, - "step": 27764 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019060914564026247, - "loss": 1.519, - "step": 27765 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906088074139214, - "loss": 1.5476, - "step": 27766 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019060846918758032, - "loss": 1.7247, - "step": 27767 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019060813096123926, - "loss": 1.633, - "step": 27768 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906077927348982, - "loss": 1.4225, - "step": 27769 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019060745450855712, - "loss": 1.476, - "step": 27770 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019060711628221606, - "loss": 1.4876, - "step": 27771 - }, - { - "epoch": 0.09, - "learning_rate": 0.000190606778055875, - "loss": 1.3753, - "step": 27772 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019060643982953392, - "loss": 1.4575, - "step": 27773 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019060610160319286, - "loss": 1.5309, - "step": 27774 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906057633768518, - "loss": 1.3644, - "step": 27775 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019060542515051074, - "loss": 1.4503, - "step": 27776 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019060508692416968, - "loss": 1.6427, - "step": 27777 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906047486978286, - "loss": 1.63, - "step": 27778 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019060441047148754, - "loss": 1.4085, - "step": 27779 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019060407224514648, - "loss": 1.487, - "step": 27780 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906037340188054, - "loss": 1.4742, - "step": 27781 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019060339579246433, - "loss": 1.5649, - "step": 27782 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019060305756612327, - "loss": 1.6112, - "step": 27783 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001906027193397822, - "loss": 1.5243, - "step": 27784 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019060238111344113, - "loss": 1.4495, - "step": 27785 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019060204288710007, - "loss": 1.4591, - "step": 27786 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019060170466075898, - "loss": 1.5931, - "step": 27787 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019060136643441792, - "loss": 1.4873, - "step": 27788 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019060102820807687, - "loss": 1.6247, - "step": 27789 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019060068998173578, - "loss": 1.4647, - "step": 27790 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019060035175539472, - "loss": 1.7472, - "step": 27791 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019060001352905366, - "loss": 1.6531, - "step": 27792 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019059967530271258, - "loss": 1.5197, - "step": 27793 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019059933707637152, - "loss": 1.5364, - "step": 27794 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019059899885003046, - "loss": 1.5613, - "step": 27795 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019059866062368937, - "loss": 1.5782, - "step": 27796 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905983223973483, - "loss": 1.4693, - "step": 27797 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019059798417100725, - "loss": 1.5318, - "step": 27798 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019059764594466617, - "loss": 1.4776, - "step": 27799 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905973077183251, - "loss": 1.6309, - "step": 27800 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019059696949198405, - "loss": 1.5108, - "step": 27801 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019059663126564296, - "loss": 1.4575, - "step": 27802 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905962930393019, - "loss": 1.4823, - "step": 27803 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019059595481296085, - "loss": 1.4639, - "step": 27804 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019059561658661976, - "loss": 1.5965, - "step": 27805 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905952783602787, - "loss": 1.4732, - "step": 27806 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019059494013393764, - "loss": 1.7083, - "step": 27807 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019059460190759656, - "loss": 1.4759, - "step": 27808 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905942636812555, - "loss": 1.4922, - "step": 27809 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019059392545491444, - "loss": 1.5315, - "step": 27810 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019059358722857335, - "loss": 1.6039, - "step": 27811 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905932490022323, - "loss": 1.7435, - "step": 27812 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019059291077589124, - "loss": 1.5827, - "step": 27813 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019059257254955015, - "loss": 1.6088, - "step": 27814 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905922343232091, - "loss": 1.5675, - "step": 27815 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019059189609686803, - "loss": 1.59, - "step": 27816 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019059155787052695, - "loss": 1.585, - "step": 27817 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905912196441859, - "loss": 1.4218, - "step": 27818 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019059088141784483, - "loss": 1.6586, - "step": 27819 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019059054319150377, - "loss": 1.4907, - "step": 27820 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019059020496516268, - "loss": 1.7525, - "step": 27821 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019058986673882163, - "loss": 1.693, - "step": 27822 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019058952851248057, - "loss": 1.5398, - "step": 27823 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019058919028613948, - "loss": 1.5326, - "step": 27824 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019058885205979842, - "loss": 1.4814, - "step": 27825 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019058851383345736, - "loss": 1.5473, - "step": 27826 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019058817560711628, - "loss": 1.4617, - "step": 27827 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019058783738077522, - "loss": 1.5702, - "step": 27828 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019058749915443416, - "loss": 1.4786, - "step": 27829 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019058716092809307, - "loss": 1.479, - "step": 27830 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019058682270175204, - "loss": 1.6153, - "step": 27831 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019058648447541096, - "loss": 1.6569, - "step": 27832 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905861462490699, - "loss": 1.6952, - "step": 27833 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019058580802272884, - "loss": 1.4684, - "step": 27834 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019058546979638775, - "loss": 1.4954, - "step": 27835 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905851315700467, - "loss": 1.6191, - "step": 27836 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019058479334370563, - "loss": 1.6662, - "step": 27837 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019058445511736455, - "loss": 1.7453, - "step": 27838 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905841168910235, - "loss": 1.6491, - "step": 27839 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019058377866468243, - "loss": 1.4803, - "step": 27840 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019058344043834134, - "loss": 1.6246, - "step": 27841 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019058310221200029, - "loss": 1.6499, - "step": 27842 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019058276398565923, - "loss": 1.5056, - "step": 27843 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019058242575931814, - "loss": 1.5756, - "step": 27844 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019058208753297708, - "loss": 1.583, - "step": 27845 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019058174930663602, - "loss": 1.736, - "step": 27846 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019058141108029494, - "loss": 1.5488, - "step": 27847 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019058107285395388, - "loss": 1.4914, - "step": 27848 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019058073462761282, - "loss": 1.4932, - "step": 27849 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019058039640127173, - "loss": 1.6015, - "step": 27850 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019058005817493067, - "loss": 1.6535, - "step": 27851 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019057971994858962, - "loss": 1.5821, - "step": 27852 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019057938172224853, - "loss": 1.5121, - "step": 27853 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019057904349590747, - "loss": 1.4487, - "step": 27854 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905787052695664, - "loss": 1.5694, - "step": 27855 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019057836704322533, - "loss": 1.4832, - "step": 27856 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019057802881688427, - "loss": 1.5993, - "step": 27857 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905776905905432, - "loss": 1.4527, - "step": 27858 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019057735236420212, - "loss": 1.5011, - "step": 27859 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019057701413786106, - "loss": 1.5102, - "step": 27860 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019057667591152, - "loss": 1.5265, - "step": 27861 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019057633768517892, - "loss": 1.554, - "step": 27862 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019057599945883786, - "loss": 1.4605, - "step": 27863 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905756612324968, - "loss": 1.5811, - "step": 27864 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019057532300615571, - "loss": 1.5698, - "step": 27865 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019057498477981466, - "loss": 1.5534, - "step": 27866 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905746465534736, - "loss": 1.5533, - "step": 27867 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905743083271325, - "loss": 1.6094, - "step": 27868 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019057397010079145, - "loss": 1.6395, - "step": 27869 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905736318744504, - "loss": 1.4622, - "step": 27870 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905732936481093, - "loss": 1.4548, - "step": 27871 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019057295542176825, - "loss": 1.5882, - "step": 27872 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905726171954272, - "loss": 1.6155, - "step": 27873 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905722789690861, - "loss": 1.6261, - "step": 27874 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019057194074274504, - "loss": 1.4888, - "step": 27875 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019057160251640399, - "loss": 1.5737, - "step": 27876 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905712642900629, - "loss": 1.5942, - "step": 27877 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019057092606372184, - "loss": 1.6484, - "step": 27878 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019057058783738078, - "loss": 1.7089, - "step": 27879 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905702496110397, - "loss": 1.5622, - "step": 27880 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019056991138469864, - "loss": 1.547, - "step": 27881 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019056957315835758, - "loss": 1.519, - "step": 27882 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905692349320165, - "loss": 1.6379, - "step": 27883 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019056889670567543, - "loss": 1.5453, - "step": 27884 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019056855847933437, - "loss": 1.5694, - "step": 27885 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905682202529933, - "loss": 1.521, - "step": 27886 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019056788202665226, - "loss": 1.5855, - "step": 27887 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905675438003112, - "loss": 1.5556, - "step": 27888 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905672055739701, - "loss": 1.646, - "step": 27889 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019056686734762905, - "loss": 1.7967, - "step": 27890 - }, - { - "epoch": 0.09, - "learning_rate": 0.000190566529121288, - "loss": 1.5581, - "step": 27891 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905661908949469, - "loss": 1.5433, - "step": 27892 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019056585266860585, - "loss": 1.416, - "step": 27893 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905655144422648, - "loss": 1.4553, - "step": 27894 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905651762159237, - "loss": 1.6126, - "step": 27895 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019056483798958265, - "loss": 1.5414, - "step": 27896 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905644997632416, - "loss": 1.7068, - "step": 27897 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905641615369005, - "loss": 1.6179, - "step": 27898 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019056382331055944, - "loss": 1.4715, - "step": 27899 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019056348508421838, - "loss": 1.4613, - "step": 27900 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905631468578773, - "loss": 1.613, - "step": 27901 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019056280863153624, - "loss": 1.6694, - "step": 27902 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019056247040519518, - "loss": 1.5265, - "step": 27903 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905621321788541, - "loss": 1.4937, - "step": 27904 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019056179395251303, - "loss": 1.5402, - "step": 27905 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019056145572617198, - "loss": 1.4836, - "step": 27906 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905611174998309, - "loss": 1.5645, - "step": 27907 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019056077927348983, - "loss": 1.5181, - "step": 27908 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019056044104714877, - "loss": 1.6793, - "step": 27909 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905601028208077, - "loss": 1.586, - "step": 27910 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019055976459446663, - "loss": 1.5288, - "step": 27911 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019055942636812557, - "loss": 1.7497, - "step": 27912 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019055908814178448, - "loss": 1.5788, - "step": 27913 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019055874991544342, - "loss": 1.602, - "step": 27914 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019055841168910237, - "loss": 1.5427, - "step": 27915 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019055807346276128, - "loss": 1.5925, - "step": 27916 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019055773523642022, - "loss": 1.5474, - "step": 27917 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019055739701007916, - "loss": 1.6244, - "step": 27918 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019055705878373808, - "loss": 1.4626, - "step": 27919 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019055672055739702, - "loss": 1.6538, - "step": 27920 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019055638233105596, - "loss": 1.5531, - "step": 27921 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019055604410471487, - "loss": 1.4929, - "step": 27922 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905557058783738, - "loss": 1.6206, - "step": 27923 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019055536765203275, - "loss": 1.5983, - "step": 27924 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019055502942569167, - "loss": 1.4288, - "step": 27925 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905546911993506, - "loss": 1.4797, - "step": 27926 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019055435297300955, - "loss": 1.4951, - "step": 27927 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019055401474666846, - "loss": 1.4764, - "step": 27928 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905536765203274, - "loss": 1.4834, - "step": 27929 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019055333829398635, - "loss": 1.5344, - "step": 27930 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019055300006764526, - "loss": 1.7389, - "step": 27931 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905526618413042, - "loss": 1.553, - "step": 27932 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019055232361496314, - "loss": 1.5156, - "step": 27933 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019055198538862206, - "loss": 1.4578, - "step": 27934 - }, - { - "epoch": 0.09, - "learning_rate": 0.000190551647162281, - "loss": 1.5973, - "step": 27935 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019055130893593994, - "loss": 1.6571, - "step": 27936 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019055097070959885, - "loss": 1.5605, - "step": 27937 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905506324832578, - "loss": 1.6294, - "step": 27938 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019055029425691674, - "loss": 1.5363, - "step": 27939 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019054995603057565, - "loss": 1.5527, - "step": 27940 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905496178042346, - "loss": 1.4139, - "step": 27941 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019054927957789353, - "loss": 1.5022, - "step": 27942 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019054894135155247, - "loss": 1.4505, - "step": 27943 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019054860312521141, - "loss": 1.6194, - "step": 27944 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019054826489887036, - "loss": 1.5103, - "step": 27945 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019054792667252927, - "loss": 1.6399, - "step": 27946 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905475884461882, - "loss": 1.6145, - "step": 27947 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019054725021984715, - "loss": 1.5924, - "step": 27948 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019054691199350607, - "loss": 1.5177, - "step": 27949 - }, - { - "epoch": 0.09, - "learning_rate": 0.000190546573767165, - "loss": 1.5258, - "step": 27950 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019054623554082395, - "loss": 1.361, - "step": 27951 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019054589731448286, - "loss": 1.479, - "step": 27952 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905455590881418, - "loss": 1.5254, - "step": 27953 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019054522086180074, - "loss": 1.4463, - "step": 27954 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019054488263545966, - "loss": 1.5007, - "step": 27955 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905445444091186, - "loss": 1.8391, - "step": 27956 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019054420618277754, - "loss": 1.4088, - "step": 27957 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019054386795643645, - "loss": 1.5121, - "step": 27958 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905435297300954, - "loss": 1.5, - "step": 27959 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019054319150375434, - "loss": 1.6634, - "step": 27960 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019054285327741325, - "loss": 1.6565, - "step": 27961 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905425150510722, - "loss": 1.4591, - "step": 27962 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019054217682473113, - "loss": 1.665, - "step": 27963 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019054183859839005, - "loss": 1.5949, - "step": 27964 - }, - { - "epoch": 0.09, - "learning_rate": 0.000190541500372049, - "loss": 1.5596, - "step": 27965 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019054116214570793, - "loss": 1.5618, - "step": 27966 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019054082391936684, - "loss": 1.5944, - "step": 27967 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019054048569302578, - "loss": 1.5782, - "step": 27968 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019054014746668473, - "loss": 1.6019, - "step": 27969 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019053980924034364, - "loss": 1.5289, - "step": 27970 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019053947101400258, - "loss": 1.6339, - "step": 27971 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019053913278766152, - "loss": 1.5618, - "step": 27972 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019053879456132044, - "loss": 1.6611, - "step": 27973 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019053845633497938, - "loss": 1.4709, - "step": 27974 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019053811810863832, - "loss": 1.5434, - "step": 27975 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019053777988229723, - "loss": 1.5679, - "step": 27976 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019053744165595617, - "loss": 1.4041, - "step": 27977 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019053710342961511, - "loss": 1.6212, - "step": 27978 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019053676520327403, - "loss": 1.5246, - "step": 27979 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019053642697693297, - "loss": 1.4372, - "step": 27980 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905360887505919, - "loss": 1.4716, - "step": 27981 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019053575052425083, - "loss": 1.5133, - "step": 27982 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019053541229790977, - "loss": 1.5838, - "step": 27983 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905350740715687, - "loss": 1.5639, - "step": 27984 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019053473584522762, - "loss": 1.4317, - "step": 27985 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019053439761888656, - "loss": 1.5259, - "step": 27986 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905340593925455, - "loss": 1.5697, - "step": 27987 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019053372116620442, - "loss": 1.5208, - "step": 27988 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019053338293986336, - "loss": 1.5437, - "step": 27989 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905330447135223, - "loss": 1.4958, - "step": 27990 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019053270648718121, - "loss": 1.4542, - "step": 27991 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019053236826084016, - "loss": 1.5342, - "step": 27992 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905320300344991, - "loss": 1.6513, - "step": 27993 - }, - { - "epoch": 0.09, - "learning_rate": 0.000190531691808158, - "loss": 1.592, - "step": 27994 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019053135358181695, - "loss": 1.7253, - "step": 27995 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905310153554759, - "loss": 1.599, - "step": 27996 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905306771291348, - "loss": 1.5744, - "step": 27997 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019053033890279375, - "loss": 1.5657, - "step": 27998 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905300006764527, - "loss": 1.3763, - "step": 27999 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019052966245011163, - "loss": 1.5203, - "step": 28000 - }, - { - "epoch": 0.09, - "eval_loss": 1.3677509014423077, - "mrc_em": 27.0, - "mrc_f1": 36.114285714285714, - "nli_acc": 0.1, - "step": 28000, - "sts_f1": 0.6391752577319587, - "ynat_f1": 0.7008936368069495 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019052932422377057, - "loss": 1.5353, - "step": 28001 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905289859974295, - "loss": 1.7316, - "step": 28002 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019052864777108843, - "loss": 1.675, - "step": 28003 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019052830954474737, - "loss": 1.7842, - "step": 28004 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905279713184063, - "loss": 1.5671, - "step": 28005 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019052763309206522, - "loss": 1.3323, - "step": 28006 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019052729486572416, - "loss": 1.6291, - "step": 28007 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905269566393831, - "loss": 1.4182, - "step": 28008 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019052661841304202, - "loss": 1.5286, - "step": 28009 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019052628018670096, - "loss": 1.5842, - "step": 28010 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905259419603599, - "loss": 1.6137, - "step": 28011 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019052560373401882, - "loss": 1.7629, - "step": 28012 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019052526550767776, - "loss": 1.5822, - "step": 28013 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905249272813367, - "loss": 1.4164, - "step": 28014 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905245890549956, - "loss": 1.3006, - "step": 28015 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019052425082865455, - "loss": 1.6731, - "step": 28016 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905239126023135, - "loss": 1.6088, - "step": 28017 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905235743759724, - "loss": 1.5211, - "step": 28018 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019052323614963135, - "loss": 1.5802, - "step": 28019 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905228979232903, - "loss": 1.6804, - "step": 28020 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905225596969492, - "loss": 1.4611, - "step": 28021 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019052222147060815, - "loss": 1.6954, - "step": 28022 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905218832442671, - "loss": 1.541, - "step": 28023 - }, - { - "epoch": 0.09, - "learning_rate": 0.000190521545017926, - "loss": 1.6654, - "step": 28024 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019052120679158494, - "loss": 1.508, - "step": 28025 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019052086856524388, - "loss": 1.731, - "step": 28026 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905205303389028, - "loss": 1.4866, - "step": 28027 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019052019211256174, - "loss": 1.4641, - "step": 28028 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019051985388622068, - "loss": 1.4349, - "step": 28029 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905195156598796, - "loss": 1.5937, - "step": 28030 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019051917743353853, - "loss": 1.6214, - "step": 28031 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019051883920719748, - "loss": 1.6504, - "step": 28032 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905185009808564, - "loss": 1.498, - "step": 28033 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019051816275451533, - "loss": 1.8089, - "step": 28034 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019051782452817427, - "loss": 1.5242, - "step": 28035 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019051748630183319, - "loss": 1.6273, - "step": 28036 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019051714807549213, - "loss": 1.6482, - "step": 28037 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019051680984915107, - "loss": 1.6528, - "step": 28038 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019051647162280998, - "loss": 1.3828, - "step": 28039 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019051613339646892, - "loss": 1.5898, - "step": 28040 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019051579517012786, - "loss": 1.7246, - "step": 28041 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019051545694378678, - "loss": 1.7351, - "step": 28042 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019051511871744572, - "loss": 1.4951, - "step": 28043 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019051478049110466, - "loss": 1.5772, - "step": 28044 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019051444226476357, - "loss": 1.4585, - "step": 28045 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019051410403842252, - "loss": 1.6748, - "step": 28046 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019051376581208146, - "loss": 1.4755, - "step": 28047 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019051342758574037, - "loss": 1.5436, - "step": 28048 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905130893593993, - "loss": 1.7535, - "step": 28049 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019051275113305825, - "loss": 1.5112, - "step": 28050 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019051241290671717, - "loss": 1.6287, - "step": 28051 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905120746803761, - "loss": 1.5429, - "step": 28052 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019051173645403505, - "loss": 1.6077, - "step": 28053 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019051139822769396, - "loss": 1.4308, - "step": 28054 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905110600013529, - "loss": 1.4865, - "step": 28055 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019051072177501185, - "loss": 1.658, - "step": 28056 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905103835486708, - "loss": 1.628, - "step": 28057 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019051004532232973, - "loss": 1.4169, - "step": 28058 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019050970709598864, - "loss": 1.7181, - "step": 28059 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019050936886964758, - "loss": 1.4892, - "step": 28060 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019050903064330652, - "loss": 1.7142, - "step": 28061 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019050869241696544, - "loss": 1.4224, - "step": 28062 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019050835419062438, - "loss": 1.4707, - "step": 28063 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019050801596428332, - "loss": 1.5596, - "step": 28064 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019050767773794223, - "loss": 1.5633, - "step": 28065 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019050733951160118, - "loss": 1.7022, - "step": 28066 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019050700128526012, - "loss": 1.57, - "step": 28067 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019050666305891903, - "loss": 1.336, - "step": 28068 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019050632483257797, - "loss": 1.702, - "step": 28069 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905059866062369, - "loss": 1.5642, - "step": 28070 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019050564837989583, - "loss": 1.6301, - "step": 28071 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019050531015355477, - "loss": 1.6459, - "step": 28072 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905049719272137, - "loss": 1.3994, - "step": 28073 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019050463370087262, - "loss": 1.4886, - "step": 28074 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019050429547453157, - "loss": 1.643, - "step": 28075 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905039572481905, - "loss": 1.5917, - "step": 28076 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019050361902184942, - "loss": 1.4537, - "step": 28077 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019050328079550836, - "loss": 1.6033, - "step": 28078 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905029425691673, - "loss": 1.4902, - "step": 28079 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019050260434282622, - "loss": 1.601, - "step": 28080 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019050226611648516, - "loss": 1.4509, - "step": 28081 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905019278901441, - "loss": 1.4917, - "step": 28082 - }, - { - "epoch": 0.09, - "learning_rate": 0.000190501589663803, - "loss": 1.5342, - "step": 28083 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019050125143746195, - "loss": 1.4995, - "step": 28084 - }, - { - "epoch": 0.09, - "learning_rate": 0.0001905009132111209, - "loss": 1.6355, - "step": 28085 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019050057498477984, - "loss": 1.4268, - "step": 28086 - }, - { - "epoch": 0.09, - "learning_rate": 0.00019050023675843875, - "loss": 1.4665, - "step": 28087 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904998985320977, - "loss": 1.62, - "step": 28088 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019049956030575663, - "loss": 1.5748, - "step": 28089 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019049922207941555, - "loss": 1.5768, - "step": 28090 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904988838530745, - "loss": 1.6858, - "step": 28091 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019049854562673343, - "loss": 1.566, - "step": 28092 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019049820740039234, - "loss": 1.5603, - "step": 28093 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019049786917405128, - "loss": 1.6666, - "step": 28094 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019049753094771023, - "loss": 1.4856, - "step": 28095 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019049719272136914, - "loss": 1.6317, - "step": 28096 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019049685449502808, - "loss": 1.4427, - "step": 28097 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019049651626868702, - "loss": 1.3455, - "step": 28098 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019049617804234594, - "loss": 1.548, - "step": 28099 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019049583981600488, - "loss": 1.6431, - "step": 28100 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019049550158966382, - "loss": 1.5934, - "step": 28101 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019049516336332273, - "loss": 1.7022, - "step": 28102 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019049482513698167, - "loss": 1.5907, - "step": 28103 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019049448691064061, - "loss": 1.4548, - "step": 28104 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019049414868429953, - "loss": 1.3627, - "step": 28105 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019049381045795847, - "loss": 1.4764, - "step": 28106 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904934722316174, - "loss": 1.6354, - "step": 28107 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019049313400527632, - "loss": 1.4743, - "step": 28108 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019049279577893527, - "loss": 1.4046, - "step": 28109 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904924575525942, - "loss": 1.5395, - "step": 28110 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019049211932625312, - "loss": 1.3821, - "step": 28111 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904917810999121, - "loss": 1.603, - "step": 28112 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190491442873571, - "loss": 1.6484, - "step": 28113 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019049110464722994, - "loss": 1.6612, - "step": 28114 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019049076642088889, - "loss": 1.4735, - "step": 28115 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904904281945478, - "loss": 1.5575, - "step": 28116 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019049008996820674, - "loss": 1.5473, - "step": 28117 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019048975174186568, - "loss": 1.5077, - "step": 28118 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904894135155246, - "loss": 1.5689, - "step": 28119 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019048907528918354, - "loss": 1.6164, - "step": 28120 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019048873706284248, - "loss": 1.3737, - "step": 28121 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904883988365014, - "loss": 1.6011, - "step": 28122 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019048806061016033, - "loss": 1.497, - "step": 28123 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019048772238381927, - "loss": 1.725, - "step": 28124 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904873841574782, - "loss": 1.7119, - "step": 28125 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019048704593113713, - "loss": 1.5638, - "step": 28126 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019048670770479607, - "loss": 1.6592, - "step": 28127 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019048636947845498, - "loss": 1.6186, - "step": 28128 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019048603125211393, - "loss": 1.4783, - "step": 28129 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019048569302577287, - "loss": 1.5319, - "step": 28130 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019048535479943178, - "loss": 1.5505, - "step": 28131 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019048501657309072, - "loss": 1.6669, - "step": 28132 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019048467834674966, - "loss": 1.6006, - "step": 28133 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019048434012040858, - "loss": 1.6706, - "step": 28134 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019048400189406752, - "loss": 1.5057, - "step": 28135 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019048366366772646, - "loss": 1.6458, - "step": 28136 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019048332544138537, - "loss": 1.5792, - "step": 28137 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019048298721504431, - "loss": 1.5431, - "step": 28138 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019048264898870326, - "loss": 1.5369, - "step": 28139 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019048231076236217, - "loss": 1.5737, - "step": 28140 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904819725360211, - "loss": 1.6257, - "step": 28141 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019048163430968005, - "loss": 1.6403, - "step": 28142 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019048129608333897, - "loss": 1.4201, - "step": 28143 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904809578569979, - "loss": 1.5159, - "step": 28144 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019048061963065685, - "loss": 1.5787, - "step": 28145 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019048028140431576, - "loss": 1.5966, - "step": 28146 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904799431779747, - "loss": 1.5484, - "step": 28147 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019047960495163364, - "loss": 1.4654, - "step": 28148 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019047926672529256, - "loss": 1.4283, - "step": 28149 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904789284989515, - "loss": 1.4699, - "step": 28150 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019047859027261044, - "loss": 1.4065, - "step": 28151 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019047825204626936, - "loss": 1.563, - "step": 28152 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904779138199283, - "loss": 1.4983, - "step": 28153 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019047757559358724, - "loss": 1.5894, - "step": 28154 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019047723736724615, - "loss": 1.5495, - "step": 28155 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904768991409051, - "loss": 1.4776, - "step": 28156 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019047656091456403, - "loss": 1.5368, - "step": 28157 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019047622268822295, - "loss": 1.5466, - "step": 28158 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904758844618819, - "loss": 1.6187, - "step": 28159 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019047554623554083, - "loss": 1.5729, - "step": 28160 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019047520800919974, - "loss": 1.5846, - "step": 28161 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019047486978285869, - "loss": 1.5551, - "step": 28162 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019047453155651763, - "loss": 1.5063, - "step": 28163 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019047419333017654, - "loss": 1.4215, - "step": 28164 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019047385510383548, - "loss": 1.4087, - "step": 28165 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019047351687749442, - "loss": 1.5868, - "step": 28166 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019047317865115334, - "loss": 1.5417, - "step": 28167 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904728404248123, - "loss": 1.764, - "step": 28168 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019047250219847125, - "loss": 1.4267, - "step": 28169 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019047216397213016, - "loss": 1.6861, - "step": 28170 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904718257457891, - "loss": 1.703, - "step": 28171 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019047148751944804, - "loss": 1.5619, - "step": 28172 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019047114929310696, - "loss": 1.5404, - "step": 28173 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904708110667659, - "loss": 1.4176, - "step": 28174 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019047047284042484, - "loss": 1.6503, - "step": 28175 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019047013461408375, - "loss": 1.459, - "step": 28176 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904697963877427, - "loss": 1.7343, - "step": 28177 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019046945816140164, - "loss": 1.414, - "step": 28178 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019046911993506055, - "loss": 1.4218, - "step": 28179 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904687817087195, - "loss": 1.443, - "step": 28180 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019046844348237843, - "loss": 1.701, - "step": 28181 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019046810525603735, - "loss": 1.4312, - "step": 28182 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904677670296963, - "loss": 1.7216, - "step": 28183 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019046742880335523, - "loss": 1.6129, - "step": 28184 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019046709057701414, - "loss": 1.5484, - "step": 28185 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019046675235067308, - "loss": 1.5248, - "step": 28186 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019046641412433202, - "loss": 1.5385, - "step": 28187 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019046607589799094, - "loss": 1.7302, - "step": 28188 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019046573767164988, - "loss": 1.5384, - "step": 28189 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019046539944530882, - "loss": 1.5161, - "step": 28190 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019046506121896773, - "loss": 1.7525, - "step": 28191 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019046472299262668, - "loss": 1.7107, - "step": 28192 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019046438476628562, - "loss": 1.583, - "step": 28193 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019046404653994453, - "loss": 1.5694, - "step": 28194 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019046370831360347, - "loss": 1.6678, - "step": 28195 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904633700872624, - "loss": 1.4058, - "step": 28196 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019046303186092133, - "loss": 1.4418, - "step": 28197 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019046269363458027, - "loss": 1.4171, - "step": 28198 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904623554082392, - "loss": 1.6476, - "step": 28199 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019046201718189812, - "loss": 1.6911, - "step": 28200 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019046167895555706, - "loss": 1.5472, - "step": 28201 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190461340729216, - "loss": 1.8506, - "step": 28202 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019046100250287492, - "loss": 1.626, - "step": 28203 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019046066427653386, - "loss": 1.7288, - "step": 28204 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904603260501928, - "loss": 1.5014, - "step": 28205 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019045998782385172, - "loss": 1.5108, - "step": 28206 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019045964959751066, - "loss": 1.6131, - "step": 28207 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904593113711696, - "loss": 1.6369, - "step": 28208 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904589731448285, - "loss": 1.4135, - "step": 28209 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019045863491848745, - "loss": 1.5184, - "step": 28210 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904582966921464, - "loss": 1.6307, - "step": 28211 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904579584658053, - "loss": 1.7448, - "step": 28212 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019045762023946425, - "loss": 1.4915, - "step": 28213 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904572820131232, - "loss": 1.6323, - "step": 28214 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904569437867821, - "loss": 1.3999, - "step": 28215 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019045660556044105, - "loss": 1.4892, - "step": 28216 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904562673341, - "loss": 1.5883, - "step": 28217 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904559291077589, - "loss": 1.4721, - "step": 28218 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019045559088141784, - "loss": 1.5953, - "step": 28219 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019045525265507678, - "loss": 1.3772, - "step": 28220 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904549144287357, - "loss": 1.4139, - "step": 28221 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019045457620239464, - "loss": 1.5889, - "step": 28222 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019045423797605358, - "loss": 1.5149, - "step": 28223 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019045389974971252, - "loss": 1.4814, - "step": 28224 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019045356152337146, - "loss": 1.4933, - "step": 28225 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904532232970304, - "loss": 1.525, - "step": 28226 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019045288507068932, - "loss": 1.404, - "step": 28227 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019045254684434826, - "loss": 1.5248, - "step": 28228 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904522086180072, - "loss": 1.596, - "step": 28229 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904518703916661, - "loss": 1.4916, - "step": 28230 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019045153216532505, - "loss": 1.6586, - "step": 28231 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190451193938984, - "loss": 1.7476, - "step": 28232 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904508557126429, - "loss": 1.4922, - "step": 28233 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019045051748630185, - "loss": 1.2622, - "step": 28234 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904501792599608, - "loss": 1.6057, - "step": 28235 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904498410336197, - "loss": 1.5715, - "step": 28236 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019044950280727865, - "loss": 1.6249, - "step": 28237 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904491645809376, - "loss": 1.451, - "step": 28238 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904488263545965, - "loss": 1.4857, - "step": 28239 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019044848812825544, - "loss": 1.5617, - "step": 28240 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019044814990191438, - "loss": 1.3558, - "step": 28241 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904478116755733, - "loss": 1.5432, - "step": 28242 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019044747344923224, - "loss": 1.5854, - "step": 28243 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019044713522289118, - "loss": 1.4223, - "step": 28244 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904467969965501, - "loss": 1.4844, - "step": 28245 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019044645877020904, - "loss": 1.6611, - "step": 28246 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019044612054386798, - "loss": 1.5605, - "step": 28247 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904457823175269, - "loss": 1.2989, - "step": 28248 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019044544409118583, - "loss": 1.5564, - "step": 28249 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019044510586484477, - "loss": 1.5153, - "step": 28250 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904447676385037, - "loss": 1.5491, - "step": 28251 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019044442941216263, - "loss": 1.4136, - "step": 28252 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019044409118582157, - "loss": 1.5245, - "step": 28253 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019044375295948048, - "loss": 1.5189, - "step": 28254 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019044341473313943, - "loss": 1.5515, - "step": 28255 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019044307650679837, - "loss": 1.4934, - "step": 28256 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019044273828045728, - "loss": 1.6375, - "step": 28257 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019044240005411622, - "loss": 1.414, - "step": 28258 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019044206182777516, - "loss": 1.4411, - "step": 28259 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019044172360143408, - "loss": 1.3584, - "step": 28260 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019044138537509302, - "loss": 1.4528, - "step": 28261 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019044104714875196, - "loss": 1.7324, - "step": 28262 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019044070892241087, - "loss": 1.5333, - "step": 28263 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019044037069606981, - "loss": 1.7354, - "step": 28264 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019044003246972876, - "loss": 1.5685, - "step": 28265 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019043969424338767, - "loss": 1.4558, - "step": 28266 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904393560170466, - "loss": 1.6203, - "step": 28267 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019043901779070555, - "loss": 1.6426, - "step": 28268 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019043867956436447, - "loss": 1.5698, - "step": 28269 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904383413380234, - "loss": 1.5945, - "step": 28270 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019043800311168235, - "loss": 1.5982, - "step": 28271 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019043766488534126, - "loss": 1.5358, - "step": 28272 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904373266590002, - "loss": 1.4711, - "step": 28273 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019043698843265914, - "loss": 1.4869, - "step": 28274 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019043665020631806, - "loss": 1.5448, - "step": 28275 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190436311979977, - "loss": 1.6635, - "step": 28276 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019043597375363594, - "loss": 1.4757, - "step": 28277 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019043563552729485, - "loss": 1.6544, - "step": 28278 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904352973009538, - "loss": 1.5707, - "step": 28279 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019043495907461276, - "loss": 1.6365, - "step": 28280 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019043462084827168, - "loss": 1.4165, - "step": 28281 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019043428262193062, - "loss": 1.5516, - "step": 28282 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019043394439558956, - "loss": 1.411, - "step": 28283 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019043360616924847, - "loss": 1.4949, - "step": 28284 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019043326794290742, - "loss": 1.5525, - "step": 28285 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019043292971656636, - "loss": 1.6141, - "step": 28286 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019043259149022527, - "loss": 1.7929, - "step": 28287 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904322532638842, - "loss": 1.4718, - "step": 28288 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019043191503754315, - "loss": 1.5626, - "step": 28289 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019043157681120207, - "loss": 1.4371, - "step": 28290 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190431238584861, - "loss": 1.6342, - "step": 28291 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019043090035851995, - "loss": 1.5776, - "step": 28292 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019043056213217886, - "loss": 1.4399, - "step": 28293 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904302239058378, - "loss": 1.6327, - "step": 28294 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019042988567949675, - "loss": 1.5672, - "step": 28295 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019042954745315566, - "loss": 1.6457, - "step": 28296 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904292092268146, - "loss": 1.5959, - "step": 28297 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019042887100047354, - "loss": 1.6905, - "step": 28298 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019042853277413246, - "loss": 1.5092, - "step": 28299 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904281945477914, - "loss": 1.6006, - "step": 28300 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019042785632145034, - "loss": 1.522, - "step": 28301 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019042751809510925, - "loss": 1.4361, - "step": 28302 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904271798687682, - "loss": 1.5402, - "step": 28303 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019042684164242713, - "loss": 1.4033, - "step": 28304 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019042650341608605, - "loss": 1.6978, - "step": 28305 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190426165189745, - "loss": 1.5251, - "step": 28306 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019042582696340393, - "loss": 1.3524, - "step": 28307 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019042548873706284, - "loss": 1.5241, - "step": 28308 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019042515051072179, - "loss": 1.4311, - "step": 28309 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019042481228438073, - "loss": 1.4962, - "step": 28310 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019042447405803964, - "loss": 1.6542, - "step": 28311 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019042413583169858, - "loss": 1.513, - "step": 28312 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019042379760535752, - "loss": 1.5523, - "step": 28313 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019042345937901644, - "loss": 1.4114, - "step": 28314 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019042312115267538, - "loss": 1.6571, - "step": 28315 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019042278292633432, - "loss": 1.5096, - "step": 28316 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019042244469999323, - "loss": 1.4567, - "step": 28317 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019042210647365217, - "loss": 1.6006, - "step": 28318 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019042176824731112, - "loss": 1.5336, - "step": 28319 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019042143002097003, - "loss": 1.4578, - "step": 28320 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019042109179462897, - "loss": 1.5131, - "step": 28321 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904207535682879, - "loss": 1.4593, - "step": 28322 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019042041534194683, - "loss": 1.524, - "step": 28323 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019042007711560577, - "loss": 1.5221, - "step": 28324 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904197388892647, - "loss": 1.5558, - "step": 28325 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019041940066292362, - "loss": 1.7102, - "step": 28326 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019041906243658256, - "loss": 1.5469, - "step": 28327 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904187242102415, - "loss": 1.6321, - "step": 28328 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019041838598390042, - "loss": 1.4268, - "step": 28329 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019041804775755936, - "loss": 1.5519, - "step": 28330 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904177095312183, - "loss": 1.7267, - "step": 28331 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019041737130487722, - "loss": 1.4469, - "step": 28332 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019041703307853616, - "loss": 1.7498, - "step": 28333 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904166948521951, - "loss": 1.4754, - "step": 28334 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190416356625854, - "loss": 1.488, - "step": 28335 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019041601839951298, - "loss": 1.6574, - "step": 28336 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904156801731719, - "loss": 1.7369, - "step": 28337 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019041534194683083, - "loss": 1.5962, - "step": 28338 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019041500372048978, - "loss": 1.5699, - "step": 28339 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904146654941487, - "loss": 1.4688, - "step": 28340 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019041432726780763, - "loss": 1.5804, - "step": 28341 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019041398904146657, - "loss": 1.5734, - "step": 28342 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904136508151255, - "loss": 1.4319, - "step": 28343 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019041331258878443, - "loss": 1.4144, - "step": 28344 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019041297436244337, - "loss": 1.5087, - "step": 28345 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019041263613610228, - "loss": 1.3911, - "step": 28346 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019041229790976122, - "loss": 1.4663, - "step": 28347 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019041195968342017, - "loss": 1.6023, - "step": 28348 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019041162145707908, - "loss": 1.4975, - "step": 28349 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019041128323073802, - "loss": 1.4935, - "step": 28350 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019041094500439696, - "loss": 1.6617, - "step": 28351 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019041060677805588, - "loss": 1.3889, - "step": 28352 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019041026855171482, - "loss": 1.4994, - "step": 28353 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019040993032537376, - "loss": 1.5182, - "step": 28354 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904095920990327, - "loss": 1.6243, - "step": 28355 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904092538726916, - "loss": 1.4872, - "step": 28356 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019040891564635055, - "loss": 1.4834, - "step": 28357 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904085774200095, - "loss": 1.5124, - "step": 28358 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904082391936684, - "loss": 1.6986, - "step": 28359 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019040790096732735, - "loss": 1.7093, - "step": 28360 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904075627409863, - "loss": 1.5796, - "step": 28361 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904072245146452, - "loss": 1.355, - "step": 28362 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019040688628830415, - "loss": 1.5471, - "step": 28363 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904065480619631, - "loss": 1.5413, - "step": 28364 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190406209835622, - "loss": 1.6294, - "step": 28365 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019040587160928094, - "loss": 1.607, - "step": 28366 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019040553338293988, - "loss": 1.6216, - "step": 28367 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904051951565988, - "loss": 1.5114, - "step": 28368 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019040485693025774, - "loss": 1.5474, - "step": 28369 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019040451870391668, - "loss": 1.5303, - "step": 28370 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904041804775756, - "loss": 1.5377, - "step": 28371 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019040384225123454, - "loss": 1.6137, - "step": 28372 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019040350402489348, - "loss": 1.5793, - "step": 28373 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904031657985524, - "loss": 1.6048, - "step": 28374 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019040282757221133, - "loss": 1.4749, - "step": 28375 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019040248934587027, - "loss": 1.6096, - "step": 28376 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001904021511195292, - "loss": 1.6042, - "step": 28377 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019040181289318813, - "loss": 1.696, - "step": 28378 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019040147466684707, - "loss": 1.6297, - "step": 28379 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019040113644050598, - "loss": 1.5021, - "step": 28380 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019040079821416492, - "loss": 1.6255, - "step": 28381 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019040045998782387, - "loss": 1.6604, - "step": 28382 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019040012176148278, - "loss": 1.5045, - "step": 28383 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019039978353514172, - "loss": 1.4798, - "step": 28384 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019039944530880066, - "loss": 1.7785, - "step": 28385 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019039910708245958, - "loss": 1.549, - "step": 28386 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019039876885611852, - "loss": 1.4029, - "step": 28387 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019039843062977746, - "loss": 1.6007, - "step": 28388 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019039809240343637, - "loss": 1.7747, - "step": 28389 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903977541770953, - "loss": 1.4685, - "step": 28390 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019039741595075425, - "loss": 1.6427, - "step": 28391 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019039707772441317, - "loss": 1.6104, - "step": 28392 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019039673949807214, - "loss": 1.4823, - "step": 28393 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019039640127173105, - "loss": 1.6338, - "step": 28394 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019039606304539, - "loss": 1.5617, - "step": 28395 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019039572481904893, - "loss": 1.5839, - "step": 28396 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019039538659270785, - "loss": 1.5988, - "step": 28397 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903950483663668, - "loss": 1.5589, - "step": 28398 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019039471014002573, - "loss": 1.3639, - "step": 28399 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019039437191368464, - "loss": 1.4473, - "step": 28400 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019039403368734358, - "loss": 1.4215, - "step": 28401 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019039369546100253, - "loss": 1.6419, - "step": 28402 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019039335723466144, - "loss": 1.5526, - "step": 28403 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019039301900832038, - "loss": 1.5384, - "step": 28404 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019039268078197932, - "loss": 1.4426, - "step": 28405 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019039234255563824, - "loss": 1.5746, - "step": 28406 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019039200432929718, - "loss": 1.4189, - "step": 28407 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019039166610295612, - "loss": 1.4894, - "step": 28408 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019039132787661503, - "loss": 1.4658, - "step": 28409 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019039098965027397, - "loss": 1.4056, - "step": 28410 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019039065142393291, - "loss": 1.6449, - "step": 28411 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019039031319759183, - "loss": 1.3918, - "step": 28412 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019038997497125077, - "loss": 1.493, - "step": 28413 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903896367449097, - "loss": 1.4452, - "step": 28414 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019038929851856863, - "loss": 1.4217, - "step": 28415 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019038896029222757, - "loss": 1.4673, - "step": 28416 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903886220658865, - "loss": 1.7267, - "step": 28417 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019038828383954542, - "loss": 1.7595, - "step": 28418 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019038794561320436, - "loss": 1.4717, - "step": 28419 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903876073868633, - "loss": 1.5899, - "step": 28420 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019038726916052222, - "loss": 1.515, - "step": 28421 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019038693093418116, - "loss": 1.5681, - "step": 28422 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903865927078401, - "loss": 1.5589, - "step": 28423 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019038625448149901, - "loss": 1.443, - "step": 28424 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019038591625515796, - "loss": 1.5534, - "step": 28425 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903855780288169, - "loss": 1.5548, - "step": 28426 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903852398024758, - "loss": 1.4898, - "step": 28427 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019038490157613475, - "loss": 1.8133, - "step": 28428 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903845633497937, - "loss": 1.6895, - "step": 28429 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903842251234526, - "loss": 1.5659, - "step": 28430 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019038388689711155, - "loss": 1.6508, - "step": 28431 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903835486707705, - "loss": 1.5004, - "step": 28432 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903832104444294, - "loss": 1.5454, - "step": 28433 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019038287221808834, - "loss": 1.5003, - "step": 28434 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019038253399174729, - "loss": 1.4528, - "step": 28435 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903821957654062, - "loss": 1.614, - "step": 28436 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019038185753906514, - "loss": 1.3885, - "step": 28437 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019038151931272408, - "loss": 1.531, - "step": 28438 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019038118108638302, - "loss": 1.5396, - "step": 28439 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019038084286004194, - "loss": 1.5927, - "step": 28440 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019038050463370088, - "loss": 1.5848, - "step": 28441 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019038016640735982, - "loss": 1.4746, - "step": 28442 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019037982818101873, - "loss": 1.5191, - "step": 28443 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019037948995467767, - "loss": 1.6343, - "step": 28444 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019037915172833662, - "loss": 1.3445, - "step": 28445 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019037881350199553, - "loss": 1.5251, - "step": 28446 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019037847527565447, - "loss": 1.6128, - "step": 28447 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903781370493134, - "loss": 1.4971, - "step": 28448 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019037779882297235, - "loss": 1.5308, - "step": 28449 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903774605966313, - "loss": 1.4923, - "step": 28450 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903771223702902, - "loss": 1.3738, - "step": 28451 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019037678414394915, - "loss": 1.7059, - "step": 28452 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903764459176081, - "loss": 1.4082, - "step": 28453 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190376107691267, - "loss": 1.4949, - "step": 28454 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019037576946492595, - "loss": 1.6081, - "step": 28455 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903754312385849, - "loss": 1.6691, - "step": 28456 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903750930122438, - "loss": 1.6459, - "step": 28457 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019037475478590274, - "loss": 1.5914, - "step": 28458 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019037441655956168, - "loss": 1.4371, - "step": 28459 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903740783332206, - "loss": 1.3434, - "step": 28460 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019037374010687954, - "loss": 1.5617, - "step": 28461 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019037340188053848, - "loss": 1.4166, - "step": 28462 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903730636541974, - "loss": 1.5293, - "step": 28463 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019037272542785633, - "loss": 1.4714, - "step": 28464 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019037238720151528, - "loss": 1.5968, - "step": 28465 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903720489751742, - "loss": 1.5781, - "step": 28466 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019037171074883313, - "loss": 1.3785, - "step": 28467 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019037137252249207, - "loss": 1.5308, - "step": 28468 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019037103429615099, - "loss": 1.5651, - "step": 28469 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019037069606980993, - "loss": 1.6197, - "step": 28470 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019037035784346887, - "loss": 1.3898, - "step": 28471 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019037001961712778, - "loss": 1.6179, - "step": 28472 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019036968139078672, - "loss": 1.5988, - "step": 28473 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019036934316444566, - "loss": 1.6074, - "step": 28474 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019036900493810458, - "loss": 1.5349, - "step": 28475 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019036866671176352, - "loss": 1.5224, - "step": 28476 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019036832848542246, - "loss": 1.4209, - "step": 28477 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019036799025908137, - "loss": 1.6952, - "step": 28478 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019036765203274032, - "loss": 1.8001, - "step": 28479 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019036731380639926, - "loss": 1.4385, - "step": 28480 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019036697558005817, - "loss": 1.4694, - "step": 28481 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903666373537171, - "loss": 1.4951, - "step": 28482 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019036629912737605, - "loss": 1.6343, - "step": 28483 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019036596090103497, - "loss": 1.5506, - "step": 28484 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903656226746939, - "loss": 1.5516, - "step": 28485 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019036528444835285, - "loss": 1.55, - "step": 28486 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019036494622201176, - "loss": 1.5833, - "step": 28487 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903646079956707, - "loss": 1.4734, - "step": 28488 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019036426976932965, - "loss": 1.5114, - "step": 28489 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019036393154298856, - "loss": 1.4618, - "step": 28490 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903635933166475, - "loss": 1.3626, - "step": 28491 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019036325509030644, - "loss": 1.6808, - "step": 28492 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019036291686396536, - "loss": 1.7466, - "step": 28493 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903625786376243, - "loss": 1.4704, - "step": 28494 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019036224041128324, - "loss": 1.3867, - "step": 28495 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019036190218494215, - "loss": 1.4573, - "step": 28496 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903615639586011, - "loss": 1.4789, - "step": 28497 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019036122573226003, - "loss": 1.8125, - "step": 28498 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019036088750591895, - "loss": 1.5907, - "step": 28499 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903605492795779, - "loss": 1.5799, - "step": 28500 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019036021105323683, - "loss": 1.6569, - "step": 28501 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019035987282689575, - "loss": 1.7097, - "step": 28502 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019035953460055469, - "loss": 1.6574, - "step": 28503 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019035919637421363, - "loss": 1.5479, - "step": 28504 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019035885814787257, - "loss": 1.488, - "step": 28505 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903585199215315, - "loss": 1.3238, - "step": 28506 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019035818169519045, - "loss": 1.5208, - "step": 28507 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019035784346884937, - "loss": 1.5024, - "step": 28508 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903575052425083, - "loss": 1.5657, - "step": 28509 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019035716701616725, - "loss": 1.5865, - "step": 28510 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019035682878982616, - "loss": 1.4471, - "step": 28511 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903564905634851, - "loss": 1.3861, - "step": 28512 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019035615233714404, - "loss": 1.5371, - "step": 28513 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019035581411080296, - "loss": 1.654, - "step": 28514 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903554758844619, - "loss": 1.4242, - "step": 28515 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019035513765812084, - "loss": 1.6524, - "step": 28516 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019035479943177975, - "loss": 1.4933, - "step": 28517 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903544612054387, - "loss": 1.5514, - "step": 28518 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019035412297909764, - "loss": 1.7897, - "step": 28519 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019035378475275655, - "loss": 1.6502, - "step": 28520 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903534465264155, - "loss": 1.6444, - "step": 28521 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019035310830007443, - "loss": 1.4071, - "step": 28522 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019035277007373335, - "loss": 1.5695, - "step": 28523 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903524318473923, - "loss": 1.659, - "step": 28524 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019035209362105123, - "loss": 1.6053, - "step": 28525 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019035175539471014, - "loss": 1.4668, - "step": 28526 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019035141716836908, - "loss": 1.6851, - "step": 28527 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019035107894202803, - "loss": 1.6438, - "step": 28528 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019035074071568694, - "loss": 1.4533, - "step": 28529 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019035040248934588, - "loss": 1.5976, - "step": 28530 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019035006426300482, - "loss": 1.595, - "step": 28531 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019034972603666374, - "loss": 1.4619, - "step": 28532 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019034938781032268, - "loss": 1.5899, - "step": 28533 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019034904958398162, - "loss": 1.3854, - "step": 28534 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019034871135764053, - "loss": 1.5126, - "step": 28535 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019034837313129947, - "loss": 1.6634, - "step": 28536 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019034803490495841, - "loss": 1.6151, - "step": 28537 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019034769667861733, - "loss": 1.4956, - "step": 28538 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019034735845227627, - "loss": 1.3854, - "step": 28539 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903470202259352, - "loss": 1.5717, - "step": 28540 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019034668199959412, - "loss": 1.5952, - "step": 28541 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019034634377325307, - "loss": 1.6728, - "step": 28542 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190346005546912, - "loss": 1.5015, - "step": 28543 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019034566732057092, - "loss": 1.596, - "step": 28544 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019034532909422986, - "loss": 1.5408, - "step": 28545 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903449908678888, - "loss": 1.7355, - "step": 28546 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019034465264154772, - "loss": 1.7414, - "step": 28547 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019034431441520666, - "loss": 1.562, - "step": 28548 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903439761888656, - "loss": 1.5099, - "step": 28549 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903436379625245, - "loss": 1.5934, - "step": 28550 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019034329973618345, - "loss": 1.3419, - "step": 28551 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903429615098424, - "loss": 1.4392, - "step": 28552 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903426232835013, - "loss": 1.507, - "step": 28553 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019034228505716025, - "loss": 1.4871, - "step": 28554 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903419468308192, - "loss": 1.5019, - "step": 28555 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903416086044781, - "loss": 1.4452, - "step": 28556 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019034127037813705, - "loss": 1.497, - "step": 28557 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190340932151796, - "loss": 1.5634, - "step": 28558 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903405939254549, - "loss": 1.5366, - "step": 28559 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019034025569911384, - "loss": 1.6059, - "step": 28560 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903399174727728, - "loss": 1.5185, - "step": 28561 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019033957924643173, - "loss": 1.4642, - "step": 28562 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019033924102009067, - "loss": 1.5641, - "step": 28563 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903389027937496, - "loss": 1.5362, - "step": 28564 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019033856456740852, - "loss": 1.516, - "step": 28565 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019033822634106746, - "loss": 1.4741, - "step": 28566 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903378881147264, - "loss": 1.4455, - "step": 28567 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019033754988838532, - "loss": 1.4834, - "step": 28568 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019033721166204426, - "loss": 1.4292, - "step": 28569 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903368734357032, - "loss": 1.6037, - "step": 28570 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019033653520936211, - "loss": 1.4148, - "step": 28571 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019033619698302106, - "loss": 1.5965, - "step": 28572 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019033585875668, - "loss": 1.5971, - "step": 28573 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903355205303389, - "loss": 1.5498, - "step": 28574 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019033518230399785, - "loss": 1.4537, - "step": 28575 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903348440776568, - "loss": 1.5696, - "step": 28576 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903345058513157, - "loss": 1.5303, - "step": 28577 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019033416762497465, - "loss": 1.4923, - "step": 28578 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903338293986336, - "loss": 1.3504, - "step": 28579 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903334911722925, - "loss": 1.4524, - "step": 28580 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019033315294595144, - "loss": 1.6194, - "step": 28581 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019033281471961039, - "loss": 1.5339, - "step": 28582 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903324764932693, - "loss": 1.5014, - "step": 28583 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019033213826692824, - "loss": 1.4903, - "step": 28584 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019033180004058718, - "loss": 1.4837, - "step": 28585 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903314618142461, - "loss": 1.4145, - "step": 28586 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019033112358790504, - "loss": 1.2121, - "step": 28587 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019033078536156398, - "loss": 1.6299, - "step": 28588 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903304471352229, - "loss": 1.4299, - "step": 28589 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019033010890888183, - "loss": 1.5568, - "step": 28590 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019032977068254077, - "loss": 1.5967, - "step": 28591 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903294324561997, - "loss": 1.5222, - "step": 28592 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019032909422985863, - "loss": 1.521, - "step": 28593 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019032875600351757, - "loss": 1.6915, - "step": 28594 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019032841777717649, - "loss": 1.4524, - "step": 28595 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019032807955083543, - "loss": 1.5483, - "step": 28596 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019032774132449437, - "loss": 1.5521, - "step": 28597 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019032740309815328, - "loss": 1.7477, - "step": 28598 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019032706487181222, - "loss": 1.4545, - "step": 28599 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019032672664547116, - "loss": 1.7319, - "step": 28600 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019032638841913008, - "loss": 1.5034, - "step": 28601 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019032605019278902, - "loss": 1.5957, - "step": 28602 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019032571196644796, - "loss": 1.531, - "step": 28603 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019032537374010687, - "loss": 1.4549, - "step": 28604 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019032503551376582, - "loss": 1.4543, - "step": 28605 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019032469728742476, - "loss": 1.5713, - "step": 28606 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019032435906108367, - "loss": 1.6265, - "step": 28607 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903240208347426, - "loss": 1.6073, - "step": 28608 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019032368260840155, - "loss": 1.4499, - "step": 28609 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019032334438206047, - "loss": 1.4995, - "step": 28610 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903230061557194, - "loss": 1.6229, - "step": 28611 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019032266792937835, - "loss": 1.6196, - "step": 28612 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019032232970303726, - "loss": 1.5853, - "step": 28613 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903219914766962, - "loss": 1.6157, - "step": 28614 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019032165325035515, - "loss": 1.3254, - "step": 28615 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019032131502401406, - "loss": 1.4814, - "step": 28616 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019032097679767303, - "loss": 1.4581, - "step": 28617 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019032063857133194, - "loss": 1.6414, - "step": 28618 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019032030034499088, - "loss": 1.6533, - "step": 28619 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019031996211864982, - "loss": 1.6102, - "step": 28620 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019031962389230874, - "loss": 1.6253, - "step": 28621 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019031928566596768, - "loss": 1.6154, - "step": 28622 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019031894743962662, - "loss": 1.6364, - "step": 28623 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019031860921328556, - "loss": 1.6058, - "step": 28624 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019031827098694448, - "loss": 1.6126, - "step": 28625 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019031793276060342, - "loss": 1.6194, - "step": 28626 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019031759453426236, - "loss": 1.6062, - "step": 28627 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019031725630792127, - "loss": 1.5152, - "step": 28628 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903169180815802, - "loss": 1.5827, - "step": 28629 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019031657985523915, - "loss": 1.5737, - "step": 28630 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019031624162889807, - "loss": 1.4271, - "step": 28631 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190315903402557, - "loss": 1.6618, - "step": 28632 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019031556517621595, - "loss": 1.6075, - "step": 28633 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019031522694987486, - "loss": 1.4532, - "step": 28634 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903148887235338, - "loss": 1.5001, - "step": 28635 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019031455049719275, - "loss": 1.6428, - "step": 28636 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019031421227085166, - "loss": 1.4753, - "step": 28637 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903138740445106, - "loss": 1.3207, - "step": 28638 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019031353581816954, - "loss": 1.5978, - "step": 28639 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019031319759182846, - "loss": 1.5773, - "step": 28640 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903128593654874, - "loss": 1.5997, - "step": 28641 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019031252113914634, - "loss": 1.7107, - "step": 28642 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019031218291280525, - "loss": 1.5775, - "step": 28643 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903118446864642, - "loss": 1.5014, - "step": 28644 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019031150646012314, - "loss": 1.6706, - "step": 28645 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019031116823378205, - "loss": 1.4618, - "step": 28646 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190310830007441, - "loss": 1.4982, - "step": 28647 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019031049178109993, - "loss": 1.5642, - "step": 28648 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019031015355475885, - "loss": 1.6726, - "step": 28649 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903098153284178, - "loss": 1.608, - "step": 28650 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019030947710207673, - "loss": 1.5159, - "step": 28651 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019030913887573564, - "loss": 1.5292, - "step": 28652 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019030880064939458, - "loss": 1.4571, - "step": 28653 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019030846242305352, - "loss": 1.5143, - "step": 28654 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019030812419671244, - "loss": 1.4966, - "step": 28655 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019030778597037138, - "loss": 1.5773, - "step": 28656 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019030744774403032, - "loss": 1.533, - "step": 28657 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019030710951768923, - "loss": 1.5733, - "step": 28658 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019030677129134818, - "loss": 1.5183, - "step": 28659 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019030643306500712, - "loss": 1.6258, - "step": 28660 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019030609483866603, - "loss": 1.4658, - "step": 28661 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019030575661232497, - "loss": 1.3916, - "step": 28662 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903054183859839, - "loss": 1.4248, - "step": 28663 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019030508015964283, - "loss": 1.5452, - "step": 28664 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019030474193330177, - "loss": 1.4843, - "step": 28665 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903044037069607, - "loss": 1.6107, - "step": 28666 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019030406548061962, - "loss": 1.6991, - "step": 28667 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019030372725427856, - "loss": 1.5913, - "step": 28668 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903033890279375, - "loss": 1.7476, - "step": 28669 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019030305080159642, - "loss": 1.5972, - "step": 28670 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019030271257525536, - "loss": 1.3753, - "step": 28671 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903023743489143, - "loss": 1.483, - "step": 28672 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019030203612257324, - "loss": 1.4808, - "step": 28673 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019030169789623218, - "loss": 1.5003, - "step": 28674 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903013596698911, - "loss": 1.4784, - "step": 28675 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019030102144355004, - "loss": 1.4488, - "step": 28676 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019030068321720898, - "loss": 1.6087, - "step": 28677 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001903003449908679, - "loss": 1.554, - "step": 28678 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019030000676452684, - "loss": 1.7691, - "step": 28679 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019029966853818578, - "loss": 1.5112, - "step": 28680 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902993303118447, - "loss": 1.4229, - "step": 28681 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019029899208550363, - "loss": 1.5259, - "step": 28682 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019029865385916257, - "loss": 1.5924, - "step": 28683 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902983156328215, - "loss": 1.5684, - "step": 28684 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019029797740648043, - "loss": 1.4666, - "step": 28685 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019029763918013937, - "loss": 1.5332, - "step": 28686 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019029730095379828, - "loss": 1.5864, - "step": 28687 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019029696272745723, - "loss": 1.4904, - "step": 28688 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019029662450111617, - "loss": 1.4545, - "step": 28689 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019029628627477508, - "loss": 1.676, - "step": 28690 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019029594804843402, - "loss": 1.5068, - "step": 28691 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019029560982209296, - "loss": 1.5725, - "step": 28692 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019029527159575188, - "loss": 1.6236, - "step": 28693 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019029493336941082, - "loss": 1.4116, - "step": 28694 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019029459514306976, - "loss": 1.4843, - "step": 28695 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019029425691672867, - "loss": 1.371, - "step": 28696 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019029391869038761, - "loss": 1.4237, - "step": 28697 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019029358046404656, - "loss": 1.4741, - "step": 28698 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019029324223770547, - "loss": 1.5921, - "step": 28699 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902929040113644, - "loss": 1.5195, - "step": 28700 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019029256578502335, - "loss": 1.5674, - "step": 28701 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019029222755868227, - "loss": 1.4832, - "step": 28702 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902918893323412, - "loss": 1.5595, - "step": 28703 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019029155110600015, - "loss": 1.5118, - "step": 28704 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019029121287965906, - "loss": 1.5826, - "step": 28705 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190290874653318, - "loss": 1.659, - "step": 28706 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019029053642697694, - "loss": 1.3856, - "step": 28707 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019029019820063589, - "loss": 1.6275, - "step": 28708 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902898599742948, - "loss": 1.6598, - "step": 28709 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019028952174795374, - "loss": 1.5874, - "step": 28710 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019028918352161268, - "loss": 1.5086, - "step": 28711 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902888452952716, - "loss": 1.4387, - "step": 28712 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019028850706893054, - "loss": 1.449, - "step": 28713 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019028816884258948, - "loss": 1.6085, - "step": 28714 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902878306162484, - "loss": 1.4684, - "step": 28715 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019028749238990733, - "loss": 1.511, - "step": 28716 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019028715416356627, - "loss": 1.572, - "step": 28717 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902868159372252, - "loss": 1.6374, - "step": 28718 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019028647771088413, - "loss": 1.6117, - "step": 28719 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019028613948454307, - "loss": 1.7134, - "step": 28720 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019028580125820198, - "loss": 1.569, - "step": 28721 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019028546303186093, - "loss": 1.6905, - "step": 28722 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019028512480551987, - "loss": 1.6107, - "step": 28723 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019028478657917878, - "loss": 1.5285, - "step": 28724 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019028444835283772, - "loss": 1.4954, - "step": 28725 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019028411012649666, - "loss": 1.4924, - "step": 28726 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019028377190015558, - "loss": 1.6074, - "step": 28727 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019028343367381452, - "loss": 1.4795, - "step": 28728 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019028309544747346, - "loss": 1.4872, - "step": 28729 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902827572211324, - "loss": 1.5126, - "step": 28730 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019028241899479134, - "loss": 1.4901, - "step": 28731 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019028208076845026, - "loss": 1.4656, - "step": 28732 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902817425421092, - "loss": 1.5501, - "step": 28733 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019028140431576814, - "loss": 1.4868, - "step": 28734 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019028106608942705, - "loss": 1.6058, - "step": 28735 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190280727863086, - "loss": 1.5489, - "step": 28736 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019028038963674493, - "loss": 1.6061, - "step": 28737 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019028005141040385, - "loss": 1.444, - "step": 28738 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902797131840628, - "loss": 1.4421, - "step": 28739 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019027937495772173, - "loss": 1.6401, - "step": 28740 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019027903673138064, - "loss": 1.6729, - "step": 28741 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019027869850503959, - "loss": 1.6597, - "step": 28742 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019027836027869853, - "loss": 1.5208, - "step": 28743 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019027802205235744, - "loss": 1.7494, - "step": 28744 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019027768382601638, - "loss": 1.526, - "step": 28745 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019027734559967532, - "loss": 1.5762, - "step": 28746 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019027700737333424, - "loss": 1.4506, - "step": 28747 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019027666914699318, - "loss": 1.6099, - "step": 28748 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019027633092065212, - "loss": 1.5757, - "step": 28749 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019027599269431103, - "loss": 1.4672, - "step": 28750 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019027565446796997, - "loss": 1.5549, - "step": 28751 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019027531624162892, - "loss": 1.4693, - "step": 28752 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019027497801528783, - "loss": 1.6157, - "step": 28753 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019027463978894677, - "loss": 1.6953, - "step": 28754 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902743015626057, - "loss": 1.6178, - "step": 28755 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019027396333626463, - "loss": 1.6158, - "step": 28756 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019027362510992357, - "loss": 1.4746, - "step": 28757 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902732868835825, - "loss": 1.6076, - "step": 28758 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019027294865724142, - "loss": 1.383, - "step": 28759 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019027261043090036, - "loss": 1.3946, - "step": 28760 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902722722045593, - "loss": 1.6445, - "step": 28761 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019027193397821822, - "loss": 1.5772, - "step": 28762 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019027159575187716, - "loss": 1.4644, - "step": 28763 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902712575255361, - "loss": 1.4798, - "step": 28764 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019027091929919502, - "loss": 1.3945, - "step": 28765 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019027058107285396, - "loss": 1.4849, - "step": 28766 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902702428465129, - "loss": 1.554, - "step": 28767 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902699046201718, - "loss": 1.7106, - "step": 28768 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019026956639383075, - "loss": 1.6027, - "step": 28769 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902692281674897, - "loss": 1.4239, - "step": 28770 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902688899411486, - "loss": 1.5482, - "step": 28771 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019026855171480755, - "loss": 1.5684, - "step": 28772 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902682134884665, - "loss": 1.5602, - "step": 28773 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902678752621254, - "loss": 1.5285, - "step": 28774 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019026753703578435, - "loss": 1.3663, - "step": 28775 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902671988094433, - "loss": 1.5554, - "step": 28776 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902668605831022, - "loss": 1.5374, - "step": 28777 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019026652235676114, - "loss": 1.4303, - "step": 28778 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019026618413042008, - "loss": 1.7394, - "step": 28779 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190265845904079, - "loss": 1.6324, - "step": 28780 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019026550767773794, - "loss": 1.474, - "step": 28781 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019026516945139688, - "loss": 1.373, - "step": 28782 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902648312250558, - "loss": 1.5788, - "step": 28783 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019026449299871473, - "loss": 1.541, - "step": 28784 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019026415477237368, - "loss": 1.5311, - "step": 28785 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019026381654603262, - "loss": 1.5948, - "step": 28786 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019026347831969156, - "loss": 1.4991, - "step": 28787 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902631400933505, - "loss": 1.5854, - "step": 28788 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902628018670094, - "loss": 1.4939, - "step": 28789 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019026246364066835, - "loss": 1.7119, - "step": 28790 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902621254143273, - "loss": 1.529, - "step": 28791 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902617871879862, - "loss": 1.4552, - "step": 28792 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019026144896164515, - "loss": 1.6358, - "step": 28793 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902611107353041, - "loss": 1.4266, - "step": 28794 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190260772508963, - "loss": 1.6465, - "step": 28795 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019026043428262195, - "loss": 1.6235, - "step": 28796 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902600960562809, - "loss": 1.6375, - "step": 28797 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902597578299398, - "loss": 1.5239, - "step": 28798 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019025941960359874, - "loss": 1.773, - "step": 28799 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019025908137725768, - "loss": 1.6322, - "step": 28800 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902587431509166, - "loss": 1.446, - "step": 28801 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019025840492457554, - "loss": 1.4532, - "step": 28802 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019025806669823448, - "loss": 1.7422, - "step": 28803 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902577284718934, - "loss": 1.5914, - "step": 28804 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019025739024555234, - "loss": 1.5793, - "step": 28805 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019025705201921128, - "loss": 1.4907, - "step": 28806 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902567137928702, - "loss": 1.6784, - "step": 28807 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019025637556652913, - "loss": 1.4851, - "step": 28808 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019025603734018807, - "loss": 1.5911, - "step": 28809 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190255699113847, - "loss": 1.6108, - "step": 28810 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019025536088750593, - "loss": 1.6062, - "step": 28811 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019025502266116487, - "loss": 1.5861, - "step": 28812 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019025468443482378, - "loss": 1.5261, - "step": 28813 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019025434620848272, - "loss": 1.735, - "step": 28814 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019025400798214167, - "loss": 1.5117, - "step": 28815 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019025366975580058, - "loss": 1.4839, - "step": 28816 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019025333152945952, - "loss": 1.6283, - "step": 28817 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019025299330311846, - "loss": 1.5285, - "step": 28818 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019025265507677738, - "loss": 1.6222, - "step": 28819 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019025231685043632, - "loss": 1.5681, - "step": 28820 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019025197862409526, - "loss": 1.464, - "step": 28821 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019025164039775417, - "loss": 1.4738, - "step": 28822 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902513021714131, - "loss": 1.5566, - "step": 28823 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019025096394507205, - "loss": 1.522, - "step": 28824 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019025062571873097, - "loss": 1.5049, - "step": 28825 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902502874923899, - "loss": 1.5004, - "step": 28826 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019024994926604885, - "loss": 1.4757, - "step": 28827 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019024961103970776, - "loss": 1.5162, - "step": 28828 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902492728133667, - "loss": 1.6375, - "step": 28829 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019024893458702565, - "loss": 1.5271, - "step": 28830 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019024859636068456, - "loss": 1.5338, - "step": 28831 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902482581343435, - "loss": 1.5891, - "step": 28832 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019024791990800244, - "loss": 1.6339, - "step": 28833 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019024758168166136, - "loss": 1.5598, - "step": 28834 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902472434553203, - "loss": 1.4756, - "step": 28835 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019024690522897924, - "loss": 1.5896, - "step": 28836 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019024656700263815, - "loss": 1.5356, - "step": 28837 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902462287762971, - "loss": 1.5171, - "step": 28838 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019024589054995604, - "loss": 1.6391, - "step": 28839 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019024555232361495, - "loss": 1.4679, - "step": 28840 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902452140972739, - "loss": 1.4344, - "step": 28841 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019024487587093286, - "loss": 1.4918, - "step": 28842 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019024453764459177, - "loss": 1.5172, - "step": 28843 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019024419941825071, - "loss": 1.5282, - "step": 28844 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019024386119190966, - "loss": 1.6118, - "step": 28845 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019024352296556857, - "loss": 1.3545, - "step": 28846 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902431847392275, - "loss": 1.4891, - "step": 28847 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019024284651288645, - "loss": 1.471, - "step": 28848 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019024250828654537, - "loss": 1.4351, - "step": 28849 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902421700602043, - "loss": 1.6746, - "step": 28850 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019024183183386325, - "loss": 1.5396, - "step": 28851 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019024149360752216, - "loss": 1.5262, - "step": 28852 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902411553811811, - "loss": 1.546, - "step": 28853 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019024081715484004, - "loss": 1.6165, - "step": 28854 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019024047892849896, - "loss": 1.606, - "step": 28855 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902401407021579, - "loss": 1.584, - "step": 28856 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019023980247581684, - "loss": 1.6437, - "step": 28857 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019023946424947576, - "loss": 1.5536, - "step": 28858 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902391260231347, - "loss": 1.5718, - "step": 28859 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019023878779679364, - "loss": 1.4367, - "step": 28860 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019023844957045255, - "loss": 1.4908, - "step": 28861 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902381113441115, - "loss": 1.4346, - "step": 28862 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019023777311777043, - "loss": 1.5519, - "step": 28863 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019023743489142935, - "loss": 1.4992, - "step": 28864 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902370966650883, - "loss": 1.6085, - "step": 28865 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019023675843874723, - "loss": 1.4806, - "step": 28866 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019023642021240614, - "loss": 1.4569, - "step": 28867 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019023608198606509, - "loss": 1.6551, - "step": 28868 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019023574375972403, - "loss": 1.5371, - "step": 28869 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019023540553338294, - "loss": 1.5755, - "step": 28870 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019023506730704188, - "loss": 1.4164, - "step": 28871 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019023472908070082, - "loss": 1.5075, - "step": 28872 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019023439085435974, - "loss": 1.6433, - "step": 28873 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019023405262801868, - "loss": 1.6336, - "step": 28874 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019023371440167762, - "loss": 1.5803, - "step": 28875 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019023337617533653, - "loss": 1.67, - "step": 28876 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019023303794899547, - "loss": 1.533, - "step": 28877 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019023269972265442, - "loss": 1.7164, - "step": 28878 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019023236149631333, - "loss": 1.521, - "step": 28879 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019023202326997227, - "loss": 1.3867, - "step": 28880 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902316850436312, - "loss": 1.6, - "step": 28881 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019023134681729013, - "loss": 1.4559, - "step": 28882 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019023100859094907, - "loss": 1.6741, - "step": 28883 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190230670364608, - "loss": 1.6618, - "step": 28884 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019023033213826692, - "loss": 1.576, - "step": 28885 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019022999391192586, - "loss": 1.4111, - "step": 28886 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902296556855848, - "loss": 1.52, - "step": 28887 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019022931745924372, - "loss": 1.7353, - "step": 28888 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019022897923290266, - "loss": 1.587, - "step": 28889 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902286410065616, - "loss": 1.4337, - "step": 28890 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019022830278022051, - "loss": 1.5091, - "step": 28891 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019022796455387946, - "loss": 1.685, - "step": 28892 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902276263275384, - "loss": 1.5005, - "step": 28893 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902272881011973, - "loss": 1.602, - "step": 28894 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019022694987485625, - "loss": 1.4694, - "step": 28895 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902266116485152, - "loss": 1.5561, - "step": 28896 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902262734221741, - "loss": 1.5955, - "step": 28897 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019022593519583308, - "loss": 1.5676, - "step": 28898 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019022559696949202, - "loss": 1.6568, - "step": 28899 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019022525874315093, - "loss": 1.4672, - "step": 28900 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019022492051680987, - "loss": 1.4126, - "step": 28901 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902245822904688, - "loss": 1.5612, - "step": 28902 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019022424406412773, - "loss": 1.6098, - "step": 28903 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019022390583778667, - "loss": 1.4328, - "step": 28904 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902235676114456, - "loss": 1.5271, - "step": 28905 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019022322938510452, - "loss": 1.4901, - "step": 28906 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019022289115876346, - "loss": 1.6566, - "step": 28907 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902225529324224, - "loss": 1.5352, - "step": 28908 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019022221470608132, - "loss": 1.3877, - "step": 28909 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019022187647974026, - "loss": 1.6566, - "step": 28910 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902215382533992, - "loss": 1.4723, - "step": 28911 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019022120002705812, - "loss": 1.6283, - "step": 28912 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019022086180071706, - "loss": 1.6116, - "step": 28913 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190220523574376, - "loss": 1.5001, - "step": 28914 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902201853480349, - "loss": 1.4349, - "step": 28915 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019021984712169385, - "loss": 1.4567, - "step": 28916 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902195088953528, - "loss": 1.6681, - "step": 28917 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902191706690117, - "loss": 1.5529, - "step": 28918 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019021883244267065, - "loss": 1.454, - "step": 28919 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902184942163296, - "loss": 1.5791, - "step": 28920 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902181559899885, - "loss": 1.5182, - "step": 28921 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019021781776364745, - "loss": 1.6428, - "step": 28922 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902174795373064, - "loss": 1.4307, - "step": 28923 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902171413109653, - "loss": 1.5375, - "step": 28924 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019021680308462424, - "loss": 1.6683, - "step": 28925 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019021646485828318, - "loss": 1.5755, - "step": 28926 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902161266319421, - "loss": 1.598, - "step": 28927 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019021578840560104, - "loss": 1.6648, - "step": 28928 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019021545017925998, - "loss": 1.5576, - "step": 28929 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902151119529189, - "loss": 1.575, - "step": 28930 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019021477372657783, - "loss": 1.6702, - "step": 28931 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019021443550023678, - "loss": 1.5144, - "step": 28932 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902140972738957, - "loss": 1.5687, - "step": 28933 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019021375904755463, - "loss": 1.5849, - "step": 28934 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019021342082121357, - "loss": 1.5318, - "step": 28935 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902130825948725, - "loss": 1.5572, - "step": 28936 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019021274436853143, - "loss": 1.7548, - "step": 28937 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019021240614219037, - "loss": 1.5389, - "step": 28938 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019021206791584928, - "loss": 1.5706, - "step": 28939 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019021172968950822, - "loss": 1.6709, - "step": 28940 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019021139146316717, - "loss": 1.3446, - "step": 28941 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019021105323682608, - "loss": 1.4095, - "step": 28942 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019021071501048502, - "loss": 1.7145, - "step": 28943 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019021037678414396, - "loss": 1.6116, - "step": 28944 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019021003855780288, - "loss": 1.6501, - "step": 28945 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019020970033146182, - "loss": 1.5435, - "step": 28946 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019020936210512076, - "loss": 1.4624, - "step": 28947 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019020902387877967, - "loss": 1.5522, - "step": 28948 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902086856524386, - "loss": 1.583, - "step": 28949 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019020834742609755, - "loss": 1.5056, - "step": 28950 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019020800919975647, - "loss": 1.6943, - "step": 28951 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902076709734154, - "loss": 1.5701, - "step": 28952 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019020733274707435, - "loss": 1.4922, - "step": 28953 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902069945207333, - "loss": 1.6315, - "step": 28954 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019020665629439223, - "loss": 1.3268, - "step": 28955 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019020631806805115, - "loss": 1.4724, - "step": 28956 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001902059798417101, - "loss": 1.6047, - "step": 28957 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019020564161536903, - "loss": 1.4726, - "step": 28958 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019020530338902794, - "loss": 1.5908, - "step": 28959 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019020496516268688, - "loss": 1.4899, - "step": 28960 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019020462693634583, - "loss": 1.5703, - "step": 28961 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019020428871000474, - "loss": 1.6148, - "step": 28962 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019020395048366368, - "loss": 1.5169, - "step": 28963 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019020361225732262, - "loss": 1.6415, - "step": 28964 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019020327403098154, - "loss": 1.5374, - "step": 28965 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019020293580464048, - "loss": 1.4794, - "step": 28966 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019020259757829942, - "loss": 1.6572, - "step": 28967 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019020225935195833, - "loss": 1.5769, - "step": 28968 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019020192112561727, - "loss": 1.534, - "step": 28969 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019020158289927621, - "loss": 1.7781, - "step": 28970 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019020124467293513, - "loss": 1.5581, - "step": 28971 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019020090644659407, - "loss": 1.5571, - "step": 28972 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190200568220253, - "loss": 1.7785, - "step": 28973 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019020022999391195, - "loss": 1.772, - "step": 28974 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019019989176757087, - "loss": 1.6759, - "step": 28975 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901995535412298, - "loss": 1.5366, - "step": 28976 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019019921531488875, - "loss": 1.4876, - "step": 28977 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019019887708854766, - "loss": 1.5634, - "step": 28978 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901985388622066, - "loss": 1.6539, - "step": 28979 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019019820063586554, - "loss": 1.307, - "step": 28980 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019019786240952446, - "loss": 1.3447, - "step": 28981 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901975241831834, - "loss": 1.4932, - "step": 28982 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019019718595684234, - "loss": 1.6604, - "step": 28983 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019019684773050125, - "loss": 1.4274, - "step": 28984 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901965095041602, - "loss": 1.6039, - "step": 28985 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019019617127781914, - "loss": 1.4524, - "step": 28986 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019019583305147805, - "loss": 1.6203, - "step": 28987 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190195494825137, - "loss": 1.5462, - "step": 28988 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019019515659879593, - "loss": 1.6438, - "step": 28989 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019019481837245485, - "loss": 1.5289, - "step": 28990 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901944801461138, - "loss": 1.4116, - "step": 28991 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019019414191977273, - "loss": 1.5624, - "step": 28992 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019019380369343164, - "loss": 1.6056, - "step": 28993 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019019346546709058, - "loss": 1.7111, - "step": 28994 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019019312724074953, - "loss": 1.5035, - "step": 28995 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019019278901440844, - "loss": 1.5224, - "step": 28996 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019019245078806738, - "loss": 1.7179, - "step": 28997 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019019211256172632, - "loss": 1.6364, - "step": 28998 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019019177433538524, - "loss": 1.6006, - "step": 28999 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019019143610904418, - "loss": 1.544, - "step": 29000 - }, - { - "epoch": 0.1, - "eval_loss": 1.4643179086538463, - "mrc_em": 24.0, - "mrc_f1": 29.78571428571429, - "nli_acc": 0.05, - "step": 29000, - "sts_f1": 0.6024096385542168, - "ynat_f1": 0.7430898551866294 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019019109788270312, - "loss": 1.5523, - "step": 29001 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019019075965636203, - "loss": 1.49, - "step": 29002 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019019042143002097, - "loss": 1.4172, - "step": 29003 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019019008320367991, - "loss": 1.6299, - "step": 29004 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019018974497733883, - "loss": 1.6427, - "step": 29005 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019018940675099777, - "loss": 1.4928, - "step": 29006 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901890685246567, - "loss": 1.7883, - "step": 29007 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019018873029831563, - "loss": 1.5848, - "step": 29008 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019018839207197457, - "loss": 1.5361, - "step": 29009 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901880538456335, - "loss": 1.4408, - "step": 29010 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019018771561929245, - "loss": 1.6302, - "step": 29011 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901873773929514, - "loss": 1.5758, - "step": 29012 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901870391666103, - "loss": 1.5796, - "step": 29013 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019018670094026924, - "loss": 1.5916, - "step": 29014 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019018636271392819, - "loss": 1.5876, - "step": 29015 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901860244875871, - "loss": 1.5556, - "step": 29016 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019018568626124604, - "loss": 1.4945, - "step": 29017 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019018534803490498, - "loss": 1.6006, - "step": 29018 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901850098085639, - "loss": 1.5107, - "step": 29019 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019018467158222284, - "loss": 1.6788, - "step": 29020 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019018433335588178, - "loss": 1.5855, - "step": 29021 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901839951295407, - "loss": 1.5187, - "step": 29022 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019018365690319963, - "loss": 1.6747, - "step": 29023 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019018331867685857, - "loss": 1.4921, - "step": 29024 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901829804505175, - "loss": 1.6405, - "step": 29025 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019018264222417643, - "loss": 1.4463, - "step": 29026 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019018230399783537, - "loss": 1.5461, - "step": 29027 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019018196577149429, - "loss": 1.4138, - "step": 29028 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019018162754515323, - "loss": 1.468, - "step": 29029 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019018128931881217, - "loss": 1.6437, - "step": 29030 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019018095109247108, - "loss": 1.4616, - "step": 29031 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019018061286613002, - "loss": 1.5857, - "step": 29032 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019018027463978896, - "loss": 1.6361, - "step": 29033 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019017993641344788, - "loss": 1.4738, - "step": 29034 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019017959818710682, - "loss": 1.588, - "step": 29035 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019017925996076576, - "loss": 1.7473, - "step": 29036 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019017892173442467, - "loss": 1.5674, - "step": 29037 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019017858350808362, - "loss": 1.5245, - "step": 29038 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019017824528174256, - "loss": 1.5279, - "step": 29039 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019017790705540147, - "loss": 1.6357, - "step": 29040 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901775688290604, - "loss": 1.4576, - "step": 29041 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019017723060271935, - "loss": 1.6528, - "step": 29042 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019017689237637827, - "loss": 1.632, - "step": 29043 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901765541500372, - "loss": 1.5177, - "step": 29044 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019017621592369615, - "loss": 1.6147, - "step": 29045 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019017587769735506, - "loss": 1.528, - "step": 29046 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190175539471014, - "loss": 1.7063, - "step": 29047 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019017520124467295, - "loss": 1.5744, - "step": 29048 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019017486301833186, - "loss": 1.6406, - "step": 29049 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901745247919908, - "loss": 1.4883, - "step": 29050 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019017418656564974, - "loss": 1.6297, - "step": 29051 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019017384833930866, - "loss": 1.5628, - "step": 29052 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901735101129676, - "loss": 1.7522, - "step": 29053 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019017317188662654, - "loss": 1.474, - "step": 29054 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019017283366028545, - "loss": 1.4898, - "step": 29055 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901724954339444, - "loss": 1.4557, - "step": 29056 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019017215720760333, - "loss": 1.4784, - "step": 29057 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019017181898126225, - "loss": 1.4389, - "step": 29058 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901714807549212, - "loss": 1.413, - "step": 29059 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019017114252858013, - "loss": 1.3818, - "step": 29060 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019017080430223907, - "loss": 1.6344, - "step": 29061 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019017046607589799, - "loss": 1.6267, - "step": 29062 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019017012784955693, - "loss": 1.4609, - "step": 29063 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019016978962321587, - "loss": 1.5809, - "step": 29064 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019016945139687478, - "loss": 1.4717, - "step": 29065 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019016911317053372, - "loss": 1.5979, - "step": 29066 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019016877494419266, - "loss": 1.5669, - "step": 29067 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901684367178516, - "loss": 1.643, - "step": 29068 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019016809849151055, - "loss": 1.5994, - "step": 29069 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019016776026516946, - "loss": 1.7017, - "step": 29070 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901674220388284, - "loss": 1.5774, - "step": 29071 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019016708381248734, - "loss": 1.5963, - "step": 29072 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019016674558614626, - "loss": 1.5757, - "step": 29073 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901664073598052, - "loss": 1.6748, - "step": 29074 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019016606913346414, - "loss": 1.585, - "step": 29075 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019016573090712305, - "loss": 1.5235, - "step": 29076 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190165392680782, - "loss": 1.621, - "step": 29077 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019016505445444094, - "loss": 1.4672, - "step": 29078 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019016471622809985, - "loss": 1.5854, - "step": 29079 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901643780017588, - "loss": 1.4815, - "step": 29080 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019016403977541773, - "loss": 1.6011, - "step": 29081 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019016370154907665, - "loss": 1.5408, - "step": 29082 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901633633227356, - "loss": 1.63, - "step": 29083 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019016302509639453, - "loss": 1.5383, - "step": 29084 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019016268687005344, - "loss": 1.6342, - "step": 29085 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019016234864371238, - "loss": 1.4451, - "step": 29086 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019016201041737132, - "loss": 1.382, - "step": 29087 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019016167219103024, - "loss": 1.4604, - "step": 29088 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019016133396468918, - "loss": 1.4488, - "step": 29089 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019016099573834812, - "loss": 1.5227, - "step": 29090 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019016065751200703, - "loss": 1.4558, - "step": 29091 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019016031928566598, - "loss": 1.6125, - "step": 29092 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019015998105932492, - "loss": 1.5488, - "step": 29093 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019015964283298383, - "loss": 1.5843, - "step": 29094 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019015930460664277, - "loss": 1.6084, - "step": 29095 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901589663803017, - "loss": 1.5609, - "step": 29096 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019015862815396063, - "loss": 1.533, - "step": 29097 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019015828992761957, - "loss": 1.56, - "step": 29098 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901579517012785, - "loss": 1.5933, - "step": 29099 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019015761347493742, - "loss": 1.5892, - "step": 29100 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019015727524859637, - "loss": 1.5971, - "step": 29101 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901569370222553, - "loss": 1.504, - "step": 29102 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019015659879591422, - "loss": 1.5524, - "step": 29103 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019015626056957316, - "loss": 1.6049, - "step": 29104 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901559223432321, - "loss": 1.7104, - "step": 29105 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019015558411689102, - "loss": 1.4555, - "step": 29106 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019015524589054996, - "loss": 1.4947, - "step": 29107 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901549076642089, - "loss": 1.6919, - "step": 29108 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901545694378678, - "loss": 1.435, - "step": 29109 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019015423121152675, - "loss": 1.4387, - "step": 29110 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901538929851857, - "loss": 1.5427, - "step": 29111 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901535547588446, - "loss": 1.5617, - "step": 29112 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019015321653250355, - "loss": 1.6536, - "step": 29113 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901528783061625, - "loss": 1.4941, - "step": 29114 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901525400798214, - "loss": 1.4576, - "step": 29115 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019015220185348035, - "loss": 1.7813, - "step": 29116 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901518636271393, - "loss": 1.5367, - "step": 29117 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901515254007982, - "loss": 1.6181, - "step": 29118 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019015118717445714, - "loss": 1.6637, - "step": 29119 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019015084894811608, - "loss": 1.4067, - "step": 29120 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190150510721775, - "loss": 1.6494, - "step": 29121 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019015017249543394, - "loss": 1.6843, - "step": 29122 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901498342690929, - "loss": 1.4875, - "step": 29123 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019014949604275182, - "loss": 1.442, - "step": 29124 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019014915781641076, - "loss": 1.4089, - "step": 29125 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901488195900697, - "loss": 1.4264, - "step": 29126 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019014848136372862, - "loss": 1.5066, - "step": 29127 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019014814313738756, - "loss": 1.6136, - "step": 29128 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901478049110465, - "loss": 1.5703, - "step": 29129 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019014746668470541, - "loss": 1.5084, - "step": 29130 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019014712845836436, - "loss": 1.5587, - "step": 29131 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901467902320233, - "loss": 1.5219, - "step": 29132 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901464520056822, - "loss": 1.5479, - "step": 29133 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019014611377934115, - "loss": 1.6386, - "step": 29134 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901457755530001, - "loss": 1.4141, - "step": 29135 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190145437326659, - "loss": 1.6101, - "step": 29136 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019014509910031795, - "loss": 1.5552, - "step": 29137 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901447608739769, - "loss": 1.477, - "step": 29138 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901444226476358, - "loss": 1.5115, - "step": 29139 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019014408442129474, - "loss": 1.4676, - "step": 29140 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019014374619495369, - "loss": 1.6509, - "step": 29141 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901434079686126, - "loss": 1.6246, - "step": 29142 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019014306974227154, - "loss": 1.5114, - "step": 29143 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019014273151593048, - "loss": 1.514, - "step": 29144 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901423932895894, - "loss": 1.6823, - "step": 29145 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019014205506324834, - "loss": 1.628, - "step": 29146 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019014171683690728, - "loss": 1.4029, - "step": 29147 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901413786105662, - "loss": 1.5817, - "step": 29148 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019014104038422513, - "loss": 1.5258, - "step": 29149 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019014070215788407, - "loss": 1.5875, - "step": 29150 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190140363931543, - "loss": 1.586, - "step": 29151 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019014002570520193, - "loss": 1.4986, - "step": 29152 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019013968747886087, - "loss": 1.566, - "step": 29153 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019013934925251978, - "loss": 1.681, - "step": 29154 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019013901102617873, - "loss": 1.5529, - "step": 29155 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019013867279983767, - "loss": 1.5111, - "step": 29156 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019013833457349658, - "loss": 1.4475, - "step": 29157 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019013799634715552, - "loss": 1.4926, - "step": 29158 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019013765812081446, - "loss": 1.5291, - "step": 29159 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019013731989447338, - "loss": 1.6492, - "step": 29160 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019013698166813232, - "loss": 1.5278, - "step": 29161 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019013664344179126, - "loss": 1.5973, - "step": 29162 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019013630521545017, - "loss": 1.5603, - "step": 29163 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019013596698910911, - "loss": 1.4137, - "step": 29164 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019013562876276806, - "loss": 1.5609, - "step": 29165 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019013529053642697, - "loss": 1.4155, - "step": 29166 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901349523100859, - "loss": 1.46, - "step": 29167 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019013461408374485, - "loss": 1.5371, - "step": 29168 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019013427585740377, - "loss": 1.5031, - "step": 29169 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901339376310627, - "loss": 1.4723, - "step": 29170 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019013359940472165, - "loss": 1.5883, - "step": 29171 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019013326117838056, - "loss": 1.381, - "step": 29172 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901329229520395, - "loss": 1.7336, - "step": 29173 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019013258472569844, - "loss": 1.5352, - "step": 29174 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019013224649935736, - "loss": 1.5118, - "step": 29175 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901319082730163, - "loss": 1.5005, - "step": 29176 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019013157004667524, - "loss": 1.512, - "step": 29177 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019013123182033416, - "loss": 1.4512, - "step": 29178 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019013089359399312, - "loss": 1.6516, - "step": 29179 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019013055536765206, - "loss": 1.5912, - "step": 29180 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019013021714131098, - "loss": 1.5074, - "step": 29181 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019012987891496992, - "loss": 1.554, - "step": 29182 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019012954068862886, - "loss": 1.5605, - "step": 29183 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019012920246228777, - "loss": 1.4196, - "step": 29184 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019012886423594672, - "loss": 1.5955, - "step": 29185 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019012852600960566, - "loss": 1.6218, - "step": 29186 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019012818778326457, - "loss": 1.6457, - "step": 29187 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901278495569235, - "loss": 1.6902, - "step": 29188 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019012751133058245, - "loss": 1.619, - "step": 29189 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019012717310424137, - "loss": 1.5566, - "step": 29190 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901268348779003, - "loss": 1.5615, - "step": 29191 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019012649665155925, - "loss": 1.5977, - "step": 29192 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019012615842521816, - "loss": 1.6295, - "step": 29193 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901258201988771, - "loss": 1.5155, - "step": 29194 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019012548197253605, - "loss": 1.5227, - "step": 29195 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019012514374619496, - "loss": 1.7675, - "step": 29196 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901248055198539, - "loss": 1.6514, - "step": 29197 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019012446729351284, - "loss": 1.4527, - "step": 29198 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019012412906717176, - "loss": 1.4775, - "step": 29199 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901237908408307, - "loss": 1.5663, - "step": 29200 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019012345261448964, - "loss": 1.3412, - "step": 29201 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019012311438814855, - "loss": 1.4888, - "step": 29202 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901227761618075, - "loss": 1.4782, - "step": 29203 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019012243793546644, - "loss": 1.5002, - "step": 29204 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019012209970912535, - "loss": 1.5044, - "step": 29205 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901217614827843, - "loss": 1.3788, - "step": 29206 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019012142325644323, - "loss": 1.6357, - "step": 29207 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019012108503010215, - "loss": 1.4499, - "step": 29208 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901207468037611, - "loss": 1.5438, - "step": 29209 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019012040857742003, - "loss": 1.5798, - "step": 29210 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019012007035107894, - "loss": 1.4697, - "step": 29211 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019011973212473788, - "loss": 1.6345, - "step": 29212 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019011939389839682, - "loss": 1.6133, - "step": 29213 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019011905567205574, - "loss": 1.7452, - "step": 29214 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019011871744571468, - "loss": 1.681, - "step": 29215 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019011837921937362, - "loss": 1.5222, - "step": 29216 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019011804099303253, - "loss": 1.7293, - "step": 29217 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019011770276669148, - "loss": 1.6063, - "step": 29218 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019011736454035042, - "loss": 1.4901, - "step": 29219 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019011702631400933, - "loss": 1.526, - "step": 29220 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019011668808766827, - "loss": 1.5502, - "step": 29221 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901163498613272, - "loss": 1.5425, - "step": 29222 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019011601163498613, - "loss": 1.6318, - "step": 29223 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019011567340864507, - "loss": 1.6736, - "step": 29224 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190115335182304, - "loss": 1.5842, - "step": 29225 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019011499695596292, - "loss": 1.6853, - "step": 29226 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019011465872962186, - "loss": 1.5283, - "step": 29227 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901143205032808, - "loss": 1.5795, - "step": 29228 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019011398227693972, - "loss": 1.4708, - "step": 29229 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019011364405059866, - "loss": 1.5642, - "step": 29230 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901133058242576, - "loss": 1.4956, - "step": 29231 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019011296759791652, - "loss": 1.4802, - "step": 29232 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019011262937157546, - "loss": 1.5525, - "step": 29233 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901122911452344, - "loss": 1.4829, - "step": 29234 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019011195291889334, - "loss": 1.5469, - "step": 29235 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019011161469255228, - "loss": 1.6244, - "step": 29236 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901112764662112, - "loss": 1.5103, - "step": 29237 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019011093823987014, - "loss": 1.6541, - "step": 29238 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019011060001352908, - "loss": 1.6357, - "step": 29239 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190110261787188, - "loss": 1.6638, - "step": 29240 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019010992356084693, - "loss": 1.5142, - "step": 29241 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019010958533450587, - "loss": 1.6118, - "step": 29242 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019010924710816481, - "loss": 1.6353, - "step": 29243 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019010890888182373, - "loss": 1.3978, - "step": 29244 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019010857065548267, - "loss": 1.5419, - "step": 29245 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901082324291416, - "loss": 1.5469, - "step": 29246 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019010789420280052, - "loss": 1.5681, - "step": 29247 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019010755597645947, - "loss": 1.4352, - "step": 29248 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901072177501184, - "loss": 1.5201, - "step": 29249 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019010687952377732, - "loss": 1.5425, - "step": 29250 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019010654129743626, - "loss": 1.48, - "step": 29251 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901062030710952, - "loss": 1.5001, - "step": 29252 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019010586484475412, - "loss": 1.5091, - "step": 29253 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019010552661841306, - "loss": 1.4791, - "step": 29254 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190105188392072, - "loss": 1.6809, - "step": 29255 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901048501657309, - "loss": 1.5112, - "step": 29256 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019010451193938985, - "loss": 1.4161, - "step": 29257 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901041737130488, - "loss": 1.69, - "step": 29258 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901038354867077, - "loss": 1.5929, - "step": 29259 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019010349726036665, - "loss": 1.5559, - "step": 29260 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901031590340256, - "loss": 1.6765, - "step": 29261 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901028208076845, - "loss": 1.648, - "step": 29262 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019010248258134345, - "loss": 1.6472, - "step": 29263 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901021443550024, - "loss": 1.7296, - "step": 29264 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901018061286613, - "loss": 1.4679, - "step": 29265 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019010146790232024, - "loss": 1.6466, - "step": 29266 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019010112967597918, - "loss": 1.5202, - "step": 29267 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001901007914496381, - "loss": 1.5601, - "step": 29268 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019010045322329704, - "loss": 1.6092, - "step": 29269 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019010011499695598, - "loss": 1.3932, - "step": 29270 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900997767706149, - "loss": 1.6007, - "step": 29271 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019009943854427384, - "loss": 1.6179, - "step": 29272 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019009910031793278, - "loss": 1.5422, - "step": 29273 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900987620915917, - "loss": 1.4191, - "step": 29274 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019009842386525063, - "loss": 1.5739, - "step": 29275 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019009808563890957, - "loss": 1.4949, - "step": 29276 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900977474125685, - "loss": 1.4098, - "step": 29277 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019009740918622743, - "loss": 1.7173, - "step": 29278 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019009707095988637, - "loss": 1.7445, - "step": 29279 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019009673273354528, - "loss": 1.4306, - "step": 29280 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019009639450720423, - "loss": 1.5251, - "step": 29281 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019009605628086317, - "loss": 1.5902, - "step": 29282 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019009571805452208, - "loss": 1.4769, - "step": 29283 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019009537982818102, - "loss": 1.5113, - "step": 29284 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019009504160183996, - "loss": 1.52, - "step": 29285 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019009470337549888, - "loss": 1.396, - "step": 29286 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019009436514915782, - "loss": 1.6267, - "step": 29287 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019009402692281676, - "loss": 1.5472, - "step": 29288 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019009368869647567, - "loss": 1.6147, - "step": 29289 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019009335047013461, - "loss": 1.5656, - "step": 29290 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019009301224379356, - "loss": 1.4749, - "step": 29291 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900926740174525, - "loss": 1.4706, - "step": 29292 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019009233579111144, - "loss": 1.6559, - "step": 29293 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019009199756477035, - "loss": 1.4673, - "step": 29294 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900916593384293, - "loss": 1.6412, - "step": 29295 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019009132111208823, - "loss": 1.5877, - "step": 29296 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019009098288574715, - "loss": 1.493, - "step": 29297 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900906446594061, - "loss": 1.5442, - "step": 29298 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019009030643306503, - "loss": 1.5195, - "step": 29299 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019008996820672394, - "loss": 1.5738, - "step": 29300 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019008962998038289, - "loss": 1.5813, - "step": 29301 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019008929175404183, - "loss": 1.5494, - "step": 29302 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019008895352770074, - "loss": 1.48, - "step": 29303 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019008861530135968, - "loss": 1.3832, - "step": 29304 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019008827707501862, - "loss": 1.6733, - "step": 29305 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019008793884867754, - "loss": 1.5332, - "step": 29306 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019008760062233648, - "loss": 1.6594, - "step": 29307 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019008726239599542, - "loss": 1.4749, - "step": 29308 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019008692416965433, - "loss": 1.5101, - "step": 29309 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019008658594331327, - "loss": 1.5797, - "step": 29310 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019008624771697222, - "loss": 1.6376, - "step": 29311 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019008590949063113, - "loss": 1.4648, - "step": 29312 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019008557126429007, - "loss": 1.5297, - "step": 29313 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190085233037949, - "loss": 1.5524, - "step": 29314 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019008489481160793, - "loss": 1.5524, - "step": 29315 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019008455658526687, - "loss": 1.5614, - "step": 29316 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900842183589258, - "loss": 1.5088, - "step": 29317 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019008388013258472, - "loss": 1.628, - "step": 29318 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019008354190624366, - "loss": 1.5651, - "step": 29319 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900832036799026, - "loss": 1.5212, - "step": 29320 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019008286545356152, - "loss": 1.4473, - "step": 29321 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019008252722722046, - "loss": 1.7096, - "step": 29322 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900821890008794, - "loss": 1.3636, - "step": 29323 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019008185077453831, - "loss": 1.4272, - "step": 29324 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019008151254819726, - "loss": 1.6807, - "step": 29325 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900811743218562, - "loss": 1.5042, - "step": 29326 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019008083609551514, - "loss": 1.419, - "step": 29327 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019008049786917405, - "loss": 1.668, - "step": 29328 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190080159642833, - "loss": 1.4966, - "step": 29329 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019007982141649193, - "loss": 1.5983, - "step": 29330 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019007948319015085, - "loss": 1.564, - "step": 29331 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900791449638098, - "loss": 1.6013, - "step": 29332 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019007880673746873, - "loss": 1.5654, - "step": 29333 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019007846851112764, - "loss": 1.5203, - "step": 29334 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019007813028478659, - "loss": 1.5878, - "step": 29335 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019007779205844553, - "loss": 1.5316, - "step": 29336 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019007745383210444, - "loss": 1.3886, - "step": 29337 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019007711560576338, - "loss": 1.6695, - "step": 29338 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019007677737942232, - "loss": 1.4951, - "step": 29339 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019007643915308124, - "loss": 1.4268, - "step": 29340 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019007610092674018, - "loss": 1.5455, - "step": 29341 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019007576270039912, - "loss": 1.6679, - "step": 29342 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019007542447405803, - "loss": 1.6603, - "step": 29343 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019007508624771697, - "loss": 1.501, - "step": 29344 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019007474802137592, - "loss": 1.5265, - "step": 29345 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019007440979503483, - "loss": 1.5549, - "step": 29346 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900740715686938, - "loss": 1.5319, - "step": 29347 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900737333423527, - "loss": 1.7716, - "step": 29348 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019007339511601165, - "loss": 1.4115, - "step": 29349 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900730568896706, - "loss": 1.6132, - "step": 29350 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900727186633295, - "loss": 1.4875, - "step": 29351 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019007238043698845, - "loss": 1.4591, - "step": 29352 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900720422106474, - "loss": 1.4139, - "step": 29353 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900717039843063, - "loss": 1.5317, - "step": 29354 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019007136575796525, - "loss": 1.414, - "step": 29355 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900710275316242, - "loss": 1.6231, - "step": 29356 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900706893052831, - "loss": 1.4395, - "step": 29357 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019007035107894204, - "loss": 1.596, - "step": 29358 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019007001285260098, - "loss": 1.5732, - "step": 29359 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900696746262599, - "loss": 1.5401, - "step": 29360 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019006933639991884, - "loss": 1.6107, - "step": 29361 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019006899817357778, - "loss": 1.2943, - "step": 29362 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900686599472367, - "loss": 1.626, - "step": 29363 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019006832172089563, - "loss": 1.5928, - "step": 29364 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019006798349455458, - "loss": 1.7231, - "step": 29365 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900676452682135, - "loss": 1.6266, - "step": 29366 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019006730704187243, - "loss": 1.6777, - "step": 29367 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019006696881553137, - "loss": 1.4845, - "step": 29368 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900666305891903, - "loss": 1.4417, - "step": 29369 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019006629236284923, - "loss": 1.4479, - "step": 29370 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019006595413650817, - "loss": 1.4075, - "step": 29371 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019006561591016708, - "loss": 1.7274, - "step": 29372 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019006527768382602, - "loss": 1.6677, - "step": 29373 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019006493945748497, - "loss": 1.5482, - "step": 29374 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019006460123114388, - "loss": 1.5674, - "step": 29375 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019006426300480282, - "loss": 1.5471, - "step": 29376 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019006392477846176, - "loss": 1.3887, - "step": 29377 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019006358655212068, - "loss": 1.6665, - "step": 29378 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019006324832577962, - "loss": 1.4201, - "step": 29379 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019006291009943856, - "loss": 1.4655, - "step": 29380 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019006257187309747, - "loss": 1.5297, - "step": 29381 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900622336467564, - "loss": 1.4269, - "step": 29382 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019006189542041535, - "loss": 1.6404, - "step": 29383 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019006155719407427, - "loss": 1.5977, - "step": 29384 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900612189677332, - "loss": 1.5051, - "step": 29385 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019006088074139215, - "loss": 1.5036, - "step": 29386 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019006054251505106, - "loss": 1.4719, - "step": 29387 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019006020428871, - "loss": 1.5495, - "step": 29388 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019005986606236895, - "loss": 1.5868, - "step": 29389 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019005952783602786, - "loss": 1.4848, - "step": 29390 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900591896096868, - "loss": 1.486, - "step": 29391 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019005885138334574, - "loss": 1.4148, - "step": 29392 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019005851315700466, - "loss": 1.5875, - "step": 29393 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900581749306636, - "loss": 1.508, - "step": 29394 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019005783670432254, - "loss": 1.5568, - "step": 29395 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019005749847798145, - "loss": 1.7066, - "step": 29396 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900571602516404, - "loss": 1.5824, - "step": 29397 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019005682202529934, - "loss": 1.6715, - "step": 29398 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019005648379895825, - "loss": 1.5822, - "step": 29399 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900561455726172, - "loss": 1.675, - "step": 29400 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019005580734627613, - "loss": 1.5373, - "step": 29401 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019005546911993505, - "loss": 1.4869, - "step": 29402 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190055130893594, - "loss": 1.8107, - "step": 29403 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019005479266725296, - "loss": 1.4895, - "step": 29404 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019005445444091187, - "loss": 1.6458, - "step": 29405 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900541162145708, - "loss": 1.5352, - "step": 29406 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019005377798822975, - "loss": 1.5239, - "step": 29407 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019005343976188867, - "loss": 1.4806, - "step": 29408 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900531015355476, - "loss": 1.5249, - "step": 29409 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019005276330920655, - "loss": 1.5661, - "step": 29410 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019005242508286546, - "loss": 1.6177, - "step": 29411 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900520868565244, - "loss": 1.694, - "step": 29412 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019005174863018334, - "loss": 1.55, - "step": 29413 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019005141040384226, - "loss": 1.6521, - "step": 29414 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900510721775012, - "loss": 1.4423, - "step": 29415 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019005073395116014, - "loss": 1.5997, - "step": 29416 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019005039572481905, - "loss": 1.5401, - "step": 29417 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190050057498478, - "loss": 1.7449, - "step": 29418 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019004971927213694, - "loss": 1.5039, - "step": 29419 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019004938104579585, - "loss": 1.6617, - "step": 29420 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900490428194548, - "loss": 1.3858, - "step": 29421 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019004870459311373, - "loss": 1.4141, - "step": 29422 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019004836636677265, - "loss": 1.4224, - "step": 29423 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900480281404316, - "loss": 1.6376, - "step": 29424 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019004768991409053, - "loss": 1.5517, - "step": 29425 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019004735168774944, - "loss": 1.515, - "step": 29426 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019004701346140838, - "loss": 1.6996, - "step": 29427 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019004667523506733, - "loss": 1.5305, - "step": 29428 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019004633700872624, - "loss": 1.4828, - "step": 29429 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019004599878238518, - "loss": 1.4596, - "step": 29430 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019004566055604412, - "loss": 1.5256, - "step": 29431 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019004532232970304, - "loss": 1.3894, - "step": 29432 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019004498410336198, - "loss": 1.5227, - "step": 29433 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019004464587702092, - "loss": 1.4899, - "step": 29434 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019004430765067983, - "loss": 1.4828, - "step": 29435 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019004396942433877, - "loss": 1.589, - "step": 29436 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019004363119799771, - "loss": 1.4537, - "step": 29437 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019004329297165663, - "loss": 1.66, - "step": 29438 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019004295474531557, - "loss": 1.4706, - "step": 29439 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900426165189745, - "loss": 1.6846, - "step": 29440 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019004227829263343, - "loss": 1.2645, - "step": 29441 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019004194006629237, - "loss": 1.5573, - "step": 29442 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900416018399513, - "loss": 1.6867, - "step": 29443 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019004126361361022, - "loss": 1.6512, - "step": 29444 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019004092538726916, - "loss": 1.4984, - "step": 29445 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900405871609281, - "loss": 1.6275, - "step": 29446 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019004024893458702, - "loss": 1.4334, - "step": 29447 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019003991070824596, - "loss": 1.6174, - "step": 29448 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900395724819049, - "loss": 1.6872, - "step": 29449 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019003923425556381, - "loss": 1.5348, - "step": 29450 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019003889602922276, - "loss": 1.6122, - "step": 29451 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900385578028817, - "loss": 1.4745, - "step": 29452 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900382195765406, - "loss": 1.5468, - "step": 29453 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019003788135019955, - "loss": 1.4782, - "step": 29454 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900375431238585, - "loss": 1.6504, - "step": 29455 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900372048975174, - "loss": 1.3308, - "step": 29456 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019003686667117635, - "loss": 1.6105, - "step": 29457 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900365284448353, - "loss": 1.5405, - "step": 29458 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900361902184942, - "loss": 1.6033, - "step": 29459 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019003585199215317, - "loss": 1.3889, - "step": 29460 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900355137658121, - "loss": 1.5232, - "step": 29461 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019003517553947103, - "loss": 1.4192, - "step": 29462 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019003483731312997, - "loss": 1.5063, - "step": 29463 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900344990867889, - "loss": 1.5402, - "step": 29464 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019003416086044782, - "loss": 1.6461, - "step": 29465 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019003382263410676, - "loss": 1.5567, - "step": 29466 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900334844077657, - "loss": 1.5409, - "step": 29467 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019003314618142462, - "loss": 1.5801, - "step": 29468 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019003280795508356, - "loss": 1.7512, - "step": 29469 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900324697287425, - "loss": 1.48, - "step": 29470 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019003213150240142, - "loss": 1.3498, - "step": 29471 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019003179327606036, - "loss": 1.4589, - "step": 29472 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900314550497193, - "loss": 1.687, - "step": 29473 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900311168233782, - "loss": 1.4875, - "step": 29474 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019003077859703715, - "loss": 1.568, - "step": 29475 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900304403706961, - "loss": 1.6646, - "step": 29476 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190030102144355, - "loss": 1.4674, - "step": 29477 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019002976391801395, - "loss": 1.5997, - "step": 29478 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900294256916729, - "loss": 1.5688, - "step": 29479 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900290874653318, - "loss": 1.5556, - "step": 29480 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019002874923899075, - "loss": 1.5677, - "step": 29481 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900284110126497, - "loss": 1.5407, - "step": 29482 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900280727863086, - "loss": 1.5251, - "step": 29483 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019002773455996754, - "loss": 1.5311, - "step": 29484 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019002739633362648, - "loss": 1.4549, - "step": 29485 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900270581072854, - "loss": 1.5968, - "step": 29486 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019002671988094434, - "loss": 1.6027, - "step": 29487 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019002638165460328, - "loss": 1.5917, - "step": 29488 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900260434282622, - "loss": 1.6128, - "step": 29489 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019002570520192113, - "loss": 1.4788, - "step": 29490 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019002536697558008, - "loss": 1.4642, - "step": 29491 - }, - { - "epoch": 0.1, - "learning_rate": 0.000190025028749239, - "loss": 1.5175, - "step": 29492 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019002469052289793, - "loss": 1.6048, - "step": 29493 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019002435229655687, - "loss": 1.4574, - "step": 29494 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019002401407021579, - "loss": 1.4216, - "step": 29495 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019002367584387473, - "loss": 1.5313, - "step": 29496 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019002333761753367, - "loss": 1.5456, - "step": 29497 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019002299939119258, - "loss": 1.6891, - "step": 29498 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019002266116485152, - "loss": 1.6078, - "step": 29499 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019002232293851046, - "loss": 1.5145, - "step": 29500 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019002198471216938, - "loss": 1.4452, - "step": 29501 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019002164648582832, - "loss": 1.5873, - "step": 29502 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019002130825948726, - "loss": 1.5035, - "step": 29503 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019002097003314617, - "loss": 1.5469, - "step": 29504 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019002063180680512, - "loss": 1.4758, - "step": 29505 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019002029358046406, - "loss": 1.5024, - "step": 29506 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019001995535412297, - "loss": 1.757, - "step": 29507 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900196171277819, - "loss": 1.4553, - "step": 29508 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019001927890144085, - "loss": 1.6118, - "step": 29509 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019001894067509977, - "loss": 1.5575, - "step": 29510 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900186024487587, - "loss": 1.5068, - "step": 29511 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019001826422241765, - "loss": 1.5516, - "step": 29512 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019001792599607656, - "loss": 1.682, - "step": 29513 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900175877697355, - "loss": 1.4848, - "step": 29514 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019001724954339445, - "loss": 1.6157, - "step": 29515 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900169113170534, - "loss": 1.6404, - "step": 29516 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019001657309071233, - "loss": 1.4318, - "step": 29517 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019001623486437127, - "loss": 1.5064, - "step": 29518 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019001589663803018, - "loss": 1.6486, - "step": 29519 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019001555841168912, - "loss": 1.6373, - "step": 29520 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019001522018534807, - "loss": 1.6529, - "step": 29521 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019001488195900698, - "loss": 1.6715, - "step": 29522 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019001454373266592, - "loss": 1.7157, - "step": 29523 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019001420550632486, - "loss": 1.5349, - "step": 29524 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019001386727998378, - "loss": 1.5242, - "step": 29525 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019001352905364272, - "loss": 1.7659, - "step": 29526 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019001319082730166, - "loss": 1.6189, - "step": 29527 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019001285260096057, - "loss": 1.4456, - "step": 29528 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019001251437461951, - "loss": 1.7197, - "step": 29529 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019001217614827845, - "loss": 1.6462, - "step": 29530 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019001183792193737, - "loss": 1.4423, - "step": 29531 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900114996955963, - "loss": 1.537, - "step": 29532 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019001116146925525, - "loss": 1.5032, - "step": 29533 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019001082324291417, - "loss": 1.6358, - "step": 29534 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900104850165731, - "loss": 1.4892, - "step": 29535 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019001014679023205, - "loss": 1.6234, - "step": 29536 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019000980856389096, - "loss": 1.5329, - "step": 29537 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900094703375499, - "loss": 1.6234, - "step": 29538 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019000913211120884, - "loss": 1.6649, - "step": 29539 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019000879388486776, - "loss": 1.5617, - "step": 29540 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900084556585267, - "loss": 1.4896, - "step": 29541 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019000811743218564, - "loss": 1.646, - "step": 29542 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019000777920584455, - "loss": 1.5669, - "step": 29543 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900074409795035, - "loss": 1.5738, - "step": 29544 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019000710275316244, - "loss": 1.5791, - "step": 29545 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019000676452682135, - "loss": 1.6059, - "step": 29546 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900064263004803, - "loss": 1.5658, - "step": 29547 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019000608807413923, - "loss": 1.6832, - "step": 29548 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019000574984779815, - "loss": 1.5599, - "step": 29549 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001900054116214571, - "loss": 1.5963, - "step": 29550 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019000507339511603, - "loss": 1.6161, - "step": 29551 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019000473516877494, - "loss": 1.528, - "step": 29552 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019000439694243388, - "loss": 1.7275, - "step": 29553 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019000405871609283, - "loss": 1.5496, - "step": 29554 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019000372048975174, - "loss": 1.4264, - "step": 29555 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019000338226341068, - "loss": 1.6351, - "step": 29556 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019000304403706962, - "loss": 1.5003, - "step": 29557 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019000270581072854, - "loss": 1.483, - "step": 29558 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019000236758438748, - "loss": 1.7159, - "step": 29559 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019000202935804642, - "loss": 1.6121, - "step": 29560 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019000169113170533, - "loss": 1.5603, - "step": 29561 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019000135290536427, - "loss": 1.6419, - "step": 29562 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019000101467902321, - "loss": 1.3237, - "step": 29563 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019000067645268213, - "loss": 1.4135, - "step": 29564 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019000033822634107, - "loss": 1.3822, - "step": 29565 - }, - { - "epoch": 0.1, - "learning_rate": 0.00019, - "loss": 1.4286, - "step": 29566 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018999966177365892, - "loss": 1.5626, - "step": 29567 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018999932354731787, - "loss": 1.536, - "step": 29568 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899989853209768, - "loss": 1.5693, - "step": 29569 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018999864709463572, - "loss": 1.6351, - "step": 29570 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018999830886829466, - "loss": 1.6422, - "step": 29571 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899979706419536, - "loss": 1.6885, - "step": 29572 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018999763241561254, - "loss": 1.656, - "step": 29573 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018999729418927149, - "loss": 1.5895, - "step": 29574 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899969559629304, - "loss": 1.7627, - "step": 29575 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018999661773658934, - "loss": 1.5527, - "step": 29576 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018999627951024828, - "loss": 1.6675, - "step": 29577 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899959412839072, - "loss": 1.6178, - "step": 29578 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018999560305756614, - "loss": 1.3207, - "step": 29579 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018999526483122508, - "loss": 1.5743, - "step": 29580 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189994926604884, - "loss": 1.6073, - "step": 29581 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018999458837854293, - "loss": 1.6306, - "step": 29582 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018999425015220187, - "loss": 1.556, - "step": 29583 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899939119258608, - "loss": 1.6747, - "step": 29584 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018999357369951973, - "loss": 1.6808, - "step": 29585 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018999323547317867, - "loss": 1.5785, - "step": 29586 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018999289724683758, - "loss": 1.6838, - "step": 29587 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018999255902049653, - "loss": 1.6236, - "step": 29588 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018999222079415547, - "loss": 1.5298, - "step": 29589 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018999188256781438, - "loss": 1.5699, - "step": 29590 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018999154434147332, - "loss": 1.4438, - "step": 29591 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018999120611513226, - "loss": 1.5879, - "step": 29592 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018999086788879118, - "loss": 1.7189, - "step": 29593 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018999052966245012, - "loss": 1.5344, - "step": 29594 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018999019143610906, - "loss": 1.4621, - "step": 29595 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189989853209768, - "loss": 1.4512, - "step": 29596 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018998951498342691, - "loss": 1.4862, - "step": 29597 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018998917675708586, - "loss": 1.6464, - "step": 29598 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899888385307448, - "loss": 1.5145, - "step": 29599 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899885003044037, - "loss": 1.4805, - "step": 29600 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018998816207806265, - "loss": 1.5574, - "step": 29601 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899878238517216, - "loss": 1.4737, - "step": 29602 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899874856253805, - "loss": 1.6895, - "step": 29603 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018998714739903945, - "loss": 1.4678, - "step": 29604 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899868091726984, - "loss": 1.2853, - "step": 29605 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899864709463573, - "loss": 1.6488, - "step": 29606 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018998613272001624, - "loss": 1.593, - "step": 29607 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018998579449367519, - "loss": 1.4898, - "step": 29608 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899854562673341, - "loss": 1.4286, - "step": 29609 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018998511804099304, - "loss": 1.6366, - "step": 29610 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018998477981465198, - "loss": 1.5299, - "step": 29611 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899844415883109, - "loss": 1.5172, - "step": 29612 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018998410336196984, - "loss": 1.371, - "step": 29613 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018998376513562878, - "loss": 1.5829, - "step": 29614 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899834269092877, - "loss": 1.6242, - "step": 29615 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018998308868294663, - "loss": 1.6014, - "step": 29616 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018998275045660557, - "loss": 1.5702, - "step": 29617 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899824122302645, - "loss": 1.6481, - "step": 29618 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018998207400392343, - "loss": 1.3225, - "step": 29619 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018998173577758237, - "loss": 1.4206, - "step": 29620 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018998139755124129, - "loss": 1.6443, - "step": 29621 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018998105932490023, - "loss": 1.5817, - "step": 29622 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018998072109855917, - "loss": 1.4847, - "step": 29623 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018998038287221808, - "loss": 1.4828, - "step": 29624 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018998004464587702, - "loss": 1.5504, - "step": 29625 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018997970641953596, - "loss": 1.6296, - "step": 29626 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018997936819319488, - "loss": 1.4442, - "step": 29627 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018997902996685385, - "loss": 1.4656, - "step": 29628 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018997869174051276, - "loss": 1.583, - "step": 29629 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899783535141717, - "loss": 1.5931, - "step": 29630 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018997801528783064, - "loss": 1.4959, - "step": 29631 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018997767706148956, - "loss": 1.4749, - "step": 29632 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899773388351485, - "loss": 1.5486, - "step": 29633 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018997700060880744, - "loss": 1.4951, - "step": 29634 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018997666238246635, - "loss": 1.474, - "step": 29635 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899763241561253, - "loss": 1.594, - "step": 29636 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018997598592978424, - "loss": 1.5148, - "step": 29637 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018997564770344315, - "loss": 1.4848, - "step": 29638 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899753094771021, - "loss": 1.6143, - "step": 29639 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018997497125076103, - "loss": 1.4524, - "step": 29640 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018997463302441995, - "loss": 1.6032, - "step": 29641 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899742947980789, - "loss": 1.4938, - "step": 29642 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018997395657173783, - "loss": 1.7186, - "step": 29643 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018997361834539674, - "loss": 1.5354, - "step": 29644 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018997328011905568, - "loss": 1.5538, - "step": 29645 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018997294189271462, - "loss": 1.628, - "step": 29646 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018997260366637354, - "loss": 1.3608, - "step": 29647 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018997226544003248, - "loss": 1.683, - "step": 29648 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018997192721369142, - "loss": 1.48, - "step": 29649 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018997158898735033, - "loss": 1.4346, - "step": 29650 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018997125076100928, - "loss": 1.5334, - "step": 29651 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018997091253466822, - "loss": 1.4544, - "step": 29652 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018997057430832713, - "loss": 1.5545, - "step": 29653 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018997023608198607, - "loss": 1.6266, - "step": 29654 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189969897855645, - "loss": 1.552, - "step": 29655 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018996955962930393, - "loss": 1.5057, - "step": 29656 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018996922140296287, - "loss": 1.3871, - "step": 29657 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899688831766218, - "loss": 1.4557, - "step": 29658 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018996854495028072, - "loss": 1.5277, - "step": 29659 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018996820672393966, - "loss": 1.5799, - "step": 29660 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899678684975986, - "loss": 1.4717, - "step": 29661 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018996753027125752, - "loss": 1.4949, - "step": 29662 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018996719204491646, - "loss": 1.45, - "step": 29663 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899668538185754, - "loss": 1.5176, - "step": 29664 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018996651559223432, - "loss": 1.5471, - "step": 29665 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018996617736589326, - "loss": 1.7196, - "step": 29666 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899658391395522, - "loss": 1.539, - "step": 29667 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899655009132111, - "loss": 1.542, - "step": 29668 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018996516268687005, - "loss": 1.5323, - "step": 29669 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189964824460529, - "loss": 1.4523, - "step": 29670 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899644862341879, - "loss": 1.4534, - "step": 29671 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018996414800784685, - "loss": 1.5254, - "step": 29672 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899638097815058, - "loss": 1.51, - "step": 29673 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899634715551647, - "loss": 1.5665, - "step": 29674 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018996313332882365, - "loss": 1.4761, - "step": 29675 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899627951024826, - "loss": 1.7358, - "step": 29676 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899624568761415, - "loss": 1.5312, - "step": 29677 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018996211864980044, - "loss": 1.6539, - "step": 29678 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018996178042345938, - "loss": 1.7571, - "step": 29679 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018996144219711832, - "loss": 1.5974, - "step": 29680 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018996110397077724, - "loss": 1.6118, - "step": 29681 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018996076574443618, - "loss": 1.5857, - "step": 29682 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018996042751809512, - "loss": 1.5506, - "step": 29683 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018996008929175406, - "loss": 1.6173, - "step": 29684 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189959751065413, - "loss": 1.4581, - "step": 29685 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018995941283907192, - "loss": 1.6027, - "step": 29686 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018995907461273086, - "loss": 1.5805, - "step": 29687 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899587363863898, - "loss": 1.5059, - "step": 29688 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899583981600487, - "loss": 1.499, - "step": 29689 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018995805993370765, - "loss": 1.7187, - "step": 29690 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899577217073666, - "loss": 1.4903, - "step": 29691 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899573834810255, - "loss": 1.5287, - "step": 29692 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018995704525468445, - "loss": 1.5974, - "step": 29693 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899567070283434, - "loss": 1.546, - "step": 29694 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899563688020023, - "loss": 1.4917, - "step": 29695 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018995603057566125, - "loss": 1.529, - "step": 29696 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899556923493202, - "loss": 1.5054, - "step": 29697 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899553541229791, - "loss": 1.5737, - "step": 29698 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018995501589663804, - "loss": 1.5357, - "step": 29699 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018995467767029698, - "loss": 1.6571, - "step": 29700 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899543394439559, - "loss": 1.5175, - "step": 29701 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018995400121761484, - "loss": 1.4406, - "step": 29702 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018995366299127378, - "loss": 1.4334, - "step": 29703 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899533247649327, - "loss": 1.51, - "step": 29704 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018995298653859164, - "loss": 1.5671, - "step": 29705 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018995264831225058, - "loss": 1.5651, - "step": 29706 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899523100859095, - "loss": 1.4224, - "step": 29707 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018995197185956843, - "loss": 1.4904, - "step": 29708 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018995163363322737, - "loss": 1.5904, - "step": 29709 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899512954068863, - "loss": 1.618, - "step": 29710 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018995095718054523, - "loss": 1.5788, - "step": 29711 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018995061895420417, - "loss": 1.3755, - "step": 29712 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018995028072786308, - "loss": 1.4988, - "step": 29713 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018994994250152203, - "loss": 1.3645, - "step": 29714 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018994960427518097, - "loss": 1.4832, - "step": 29715 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018994926604883988, - "loss": 1.643, - "step": 29716 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018994892782249882, - "loss": 1.5017, - "step": 29717 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018994858959615776, - "loss": 1.6159, - "step": 29718 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018994825136981668, - "loss": 1.5206, - "step": 29719 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018994791314347562, - "loss": 1.4847, - "step": 29720 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018994757491713456, - "loss": 1.4038, - "step": 29721 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018994723669079347, - "loss": 1.4425, - "step": 29722 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018994689846445241, - "loss": 1.5491, - "step": 29723 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018994656023811136, - "loss": 1.4836, - "step": 29724 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018994622201177027, - "loss": 1.5257, - "step": 29725 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899458837854292, - "loss": 1.5107, - "step": 29726 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018994554555908815, - "loss": 1.4967, - "step": 29727 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018994520733274707, - "loss": 1.5771, - "step": 29728 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189944869106406, - "loss": 1.6058, - "step": 29729 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018994453088006495, - "loss": 1.6343, - "step": 29730 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018994419265372386, - "loss": 1.508, - "step": 29731 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899438544273828, - "loss": 1.5566, - "step": 29732 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018994351620104174, - "loss": 1.3529, - "step": 29733 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018994317797470066, - "loss": 1.4865, - "step": 29734 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899428397483596, - "loss": 1.6007, - "step": 29735 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018994250152201854, - "loss": 1.6586, - "step": 29736 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018994216329567745, - "loss": 1.4868, - "step": 29737 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899418250693364, - "loss": 1.5142, - "step": 29738 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018994148684299534, - "loss": 1.4905, - "step": 29739 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018994114861665425, - "loss": 1.7056, - "step": 29740 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018994081039031322, - "loss": 1.6948, - "step": 29741 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018994047216397216, - "loss": 1.5542, - "step": 29742 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018994013393763107, - "loss": 1.6116, - "step": 29743 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018993979571129002, - "loss": 1.4983, - "step": 29744 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018993945748494896, - "loss": 1.587, - "step": 29745 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018993911925860787, - "loss": 1.4754, - "step": 29746 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899387810322668, - "loss": 1.6524, - "step": 29747 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018993844280592575, - "loss": 1.4678, - "step": 29748 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018993810457958467, - "loss": 1.5616, - "step": 29749 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899377663532436, - "loss": 1.5462, - "step": 29750 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018993742812690255, - "loss": 1.5323, - "step": 29751 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018993708990056146, - "loss": 1.6314, - "step": 29752 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899367516742204, - "loss": 1.5055, - "step": 29753 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018993641344787935, - "loss": 1.5033, - "step": 29754 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018993607522153826, - "loss": 1.7024, - "step": 29755 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899357369951972, - "loss": 1.6134, - "step": 29756 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018993539876885614, - "loss": 1.5334, - "step": 29757 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018993506054251506, - "loss": 1.5779, - "step": 29758 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189934722316174, - "loss": 1.5136, - "step": 29759 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018993438408983294, - "loss": 1.6259, - "step": 29760 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018993404586349185, - "loss": 1.5717, - "step": 29761 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899337076371508, - "loss": 1.612, - "step": 29762 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018993336941080973, - "loss": 1.4705, - "step": 29763 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018993303118446865, - "loss": 1.5459, - "step": 29764 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899326929581276, - "loss": 1.6295, - "step": 29765 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018993235473178653, - "loss": 1.7957, - "step": 29766 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018993201650544544, - "loss": 1.6756, - "step": 29767 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018993167827910439, - "loss": 1.567, - "step": 29768 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018993134005276333, - "loss": 1.5202, - "step": 29769 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018993100182642224, - "loss": 1.5182, - "step": 29770 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018993066360008118, - "loss": 1.5141, - "step": 29771 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018993032537374012, - "loss": 1.6954, - "step": 29772 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018992998714739904, - "loss": 1.5467, - "step": 29773 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018992964892105798, - "loss": 1.6637, - "step": 29774 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018992931069471692, - "loss": 1.5153, - "step": 29775 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018992897246837583, - "loss": 1.5276, - "step": 29776 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018992863424203477, - "loss": 1.4842, - "step": 29777 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018992829601569372, - "loss": 1.6103, - "step": 29778 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018992795778935263, - "loss": 1.4587, - "step": 29779 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018992761956301157, - "loss": 1.3844, - "step": 29780 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899272813366705, - "loss": 1.6128, - "step": 29781 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018992694311032943, - "loss": 1.5336, - "step": 29782 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018992660488398837, - "loss": 1.6007, - "step": 29783 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899262666576473, - "loss": 1.3832, - "step": 29784 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018992592843130622, - "loss": 1.4685, - "step": 29785 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018992559020496516, - "loss": 1.7637, - "step": 29786 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899252519786241, - "loss": 1.4455, - "step": 29787 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018992491375228302, - "loss": 1.3956, - "step": 29788 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018992457552594196, - "loss": 1.4567, - "step": 29789 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899242372996009, - "loss": 1.6585, - "step": 29790 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018992389907325982, - "loss": 1.569, - "step": 29791 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018992356084691876, - "loss": 1.6275, - "step": 29792 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899232226205777, - "loss": 1.6698, - "step": 29793 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899228843942366, - "loss": 1.3657, - "step": 29794 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018992254616789555, - "loss": 1.5962, - "step": 29795 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899222079415545, - "loss": 1.5783, - "step": 29796 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018992186971521344, - "loss": 1.5976, - "step": 29797 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018992153148887238, - "loss": 1.6051, - "step": 29798 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018992119326253132, - "loss": 1.6527, - "step": 29799 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018992085503619023, - "loss": 1.5536, - "step": 29800 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018992051680984917, - "loss": 1.742, - "step": 29801 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018992017858350811, - "loss": 1.6355, - "step": 29802 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018991984035716703, - "loss": 1.6789, - "step": 29803 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018991950213082597, - "loss": 1.5669, - "step": 29804 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899191639044849, - "loss": 1.6192, - "step": 29805 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018991882567814382, - "loss": 1.6219, - "step": 29806 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018991848745180277, - "loss": 1.575, - "step": 29807 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899181492254617, - "loss": 1.586, - "step": 29808 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018991781099912062, - "loss": 1.3472, - "step": 29809 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018991747277277956, - "loss": 1.4433, - "step": 29810 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899171345464385, - "loss": 1.7337, - "step": 29811 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018991679632009742, - "loss": 1.6439, - "step": 29812 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018991645809375636, - "loss": 1.3053, - "step": 29813 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899161198674153, - "loss": 1.6145, - "step": 29814 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899157816410742, - "loss": 1.5, - "step": 29815 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018991544341473315, - "loss": 1.5024, - "step": 29816 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899151051883921, - "loss": 1.558, - "step": 29817 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189914766962051, - "loss": 1.5246, - "step": 29818 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018991442873570995, - "loss": 1.4574, - "step": 29819 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899140905093689, - "loss": 1.5139, - "step": 29820 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899137522830278, - "loss": 1.432, - "step": 29821 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018991341405668675, - "loss": 1.4577, - "step": 29822 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899130758303457, - "loss": 1.6325, - "step": 29823 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899127376040046, - "loss": 1.5093, - "step": 29824 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018991239937766354, - "loss": 1.5151, - "step": 29825 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018991206115132248, - "loss": 1.5352, - "step": 29826 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899117229249814, - "loss": 1.479, - "step": 29827 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018991138469864034, - "loss": 1.6119, - "step": 29828 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018991104647229928, - "loss": 1.563, - "step": 29829 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899107082459582, - "loss": 1.5179, - "step": 29830 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018991037001961714, - "loss": 1.5233, - "step": 29831 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018991003179327608, - "loss": 1.4727, - "step": 29832 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189909693566935, - "loss": 1.5069, - "step": 29833 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018990935534059393, - "loss": 1.4778, - "step": 29834 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018990901711425287, - "loss": 1.431, - "step": 29835 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899086788879118, - "loss": 1.5549, - "step": 29836 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018990834066157073, - "loss": 1.6008, - "step": 29837 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018990800243522967, - "loss": 1.6328, - "step": 29838 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018990766420888858, - "loss": 1.5799, - "step": 29839 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018990732598254752, - "loss": 1.5375, - "step": 29840 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018990698775620647, - "loss": 1.6048, - "step": 29841 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018990664952986538, - "loss": 1.5699, - "step": 29842 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018990631130352432, - "loss": 1.5409, - "step": 29843 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018990597307718326, - "loss": 1.5205, - "step": 29844 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018990563485084218, - "loss": 1.4605, - "step": 29845 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018990529662450112, - "loss": 1.6876, - "step": 29846 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018990495839816006, - "loss": 1.4207, - "step": 29847 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018990462017181897, - "loss": 1.5712, - "step": 29848 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899042819454779, - "loss": 1.4757, - "step": 29849 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018990394371913685, - "loss": 1.3746, - "step": 29850 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018990360549279577, - "loss": 1.5185, - "step": 29851 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899032672664547, - "loss": 1.412, - "step": 29852 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018990292904011365, - "loss": 1.5107, - "step": 29853 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899025908137726, - "loss": 1.4736, - "step": 29854 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018990225258743153, - "loss": 1.5165, - "step": 29855 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018990191436109045, - "loss": 1.4673, - "step": 29856 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001899015761347494, - "loss": 1.391, - "step": 29857 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018990123790840833, - "loss": 1.5854, - "step": 29858 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018990089968206724, - "loss": 1.5334, - "step": 29859 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018990056145572618, - "loss": 1.56, - "step": 29860 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018990022322938513, - "loss": 1.7416, - "step": 29861 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018989988500304407, - "loss": 1.5753, - "step": 29862 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018989954677670298, - "loss": 1.6422, - "step": 29863 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018989920855036192, - "loss": 1.35, - "step": 29864 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018989887032402086, - "loss": 1.6067, - "step": 29865 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018989853209767978, - "loss": 1.4232, - "step": 29866 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018989819387133872, - "loss": 1.5738, - "step": 29867 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018989785564499766, - "loss": 1.4629, - "step": 29868 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018989751741865657, - "loss": 1.3955, - "step": 29869 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018989717919231551, - "loss": 1.6646, - "step": 29870 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018989684096597446, - "loss": 1.5414, - "step": 29871 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018989650273963337, - "loss": 1.5108, - "step": 29872 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898961645132923, - "loss": 1.5953, - "step": 29873 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018989582628695125, - "loss": 1.6711, - "step": 29874 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018989548806061017, - "loss": 1.5573, - "step": 29875 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898951498342691, - "loss": 1.6329, - "step": 29876 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018989481160792805, - "loss": 1.6825, - "step": 29877 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018989447338158696, - "loss": 1.5229, - "step": 29878 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898941351552459, - "loss": 1.4092, - "step": 29879 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018989379692890484, - "loss": 1.7514, - "step": 29880 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018989345870256376, - "loss": 1.4117, - "step": 29881 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898931204762227, - "loss": 1.5038, - "step": 29882 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018989278224988164, - "loss": 1.5174, - "step": 29883 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018989244402354056, - "loss": 1.4978, - "step": 29884 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898921057971995, - "loss": 1.5379, - "step": 29885 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018989176757085844, - "loss": 1.6365, - "step": 29886 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018989142934451735, - "loss": 1.4476, - "step": 29887 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898910911181763, - "loss": 1.6952, - "step": 29888 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018989075289183523, - "loss": 1.4923, - "step": 29889 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018989041466549415, - "loss": 1.4949, - "step": 29890 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898900764391531, - "loss": 1.4954, - "step": 29891 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018988973821281203, - "loss": 1.5758, - "step": 29892 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018988939998647094, - "loss": 1.4394, - "step": 29893 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018988906176012989, - "loss": 1.6003, - "step": 29894 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018988872353378883, - "loss": 1.4185, - "step": 29895 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018988838530744774, - "loss": 1.5771, - "step": 29896 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018988804708110668, - "loss": 1.6499, - "step": 29897 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018988770885476562, - "loss": 1.5755, - "step": 29898 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018988737062842454, - "loss": 1.5682, - "step": 29899 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018988703240208348, - "loss": 1.5646, - "step": 29900 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018988669417574242, - "loss": 1.5686, - "step": 29901 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018988635594940133, - "loss": 1.4954, - "step": 29902 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018988601772306027, - "loss": 1.4331, - "step": 29903 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018988567949671922, - "loss": 1.5821, - "step": 29904 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018988534127037813, - "loss": 1.4384, - "step": 29905 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018988500304403707, - "loss": 1.421, - "step": 29906 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189884664817696, - "loss": 1.5827, - "step": 29907 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018988432659135493, - "loss": 1.5779, - "step": 29908 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898839883650139, - "loss": 1.603, - "step": 29909 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898836501386728, - "loss": 1.419, - "step": 29910 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018988331191233175, - "loss": 1.5855, - "step": 29911 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898829736859907, - "loss": 1.6842, - "step": 29912 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898826354596496, - "loss": 1.3761, - "step": 29913 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018988229723330855, - "loss": 1.5612, - "step": 29914 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898819590069675, - "loss": 1.6805, - "step": 29915 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898816207806264, - "loss": 1.5335, - "step": 29916 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018988128255428534, - "loss": 1.471, - "step": 29917 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018988094432794428, - "loss": 1.4563, - "step": 29918 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898806061016032, - "loss": 1.4149, - "step": 29919 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018988026787526214, - "loss": 1.3642, - "step": 29920 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018987992964892108, - "loss": 1.7446, - "step": 29921 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018987959142258, - "loss": 1.6278, - "step": 29922 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018987925319623893, - "loss": 1.4889, - "step": 29923 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018987891496989788, - "loss": 1.4403, - "step": 29924 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898785767435568, - "loss": 1.6077, - "step": 29925 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018987823851721573, - "loss": 1.4012, - "step": 29926 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018987790029087467, - "loss": 1.4007, - "step": 29927 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018987756206453359, - "loss": 1.6516, - "step": 29928 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018987722383819253, - "loss": 1.5758, - "step": 29929 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018987688561185147, - "loss": 1.6432, - "step": 29930 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018987654738551038, - "loss": 1.4369, - "step": 29931 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018987620915916932, - "loss": 1.5825, - "step": 29932 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018987587093282826, - "loss": 1.5799, - "step": 29933 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018987553270648718, - "loss": 1.3923, - "step": 29934 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018987519448014612, - "loss": 1.3976, - "step": 29935 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018987485625380506, - "loss": 1.8235, - "step": 29936 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018987451802746397, - "loss": 1.5638, - "step": 29937 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018987417980112292, - "loss": 1.4663, - "step": 29938 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018987384157478186, - "loss": 1.492, - "step": 29939 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018987350334844077, - "loss": 1.5017, - "step": 29940 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898731651220997, - "loss": 1.5916, - "step": 29941 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018987282689575865, - "loss": 1.5289, - "step": 29942 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018987248866941757, - "loss": 1.6711, - "step": 29943 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898721504430765, - "loss": 1.8807, - "step": 29944 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018987181221673545, - "loss": 1.7795, - "step": 29945 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898714739903944, - "loss": 1.7052, - "step": 29946 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898711357640533, - "loss": 1.3911, - "step": 29947 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018987079753771225, - "loss": 1.6683, - "step": 29948 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898704593113712, - "loss": 1.506, - "step": 29949 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898701210850301, - "loss": 1.5032, - "step": 29950 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018986978285868904, - "loss": 1.6369, - "step": 29951 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018986944463234798, - "loss": 1.4225, - "step": 29952 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898691064060069, - "loss": 1.3662, - "step": 29953 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018986876817966584, - "loss": 1.5975, - "step": 29954 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018986842995332478, - "loss": 1.5423, - "step": 29955 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898680917269837, - "loss": 1.6507, - "step": 29956 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018986775350064263, - "loss": 1.3828, - "step": 29957 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018986741527430158, - "loss": 1.401, - "step": 29958 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898670770479605, - "loss": 1.4655, - "step": 29959 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018986673882161943, - "loss": 1.4748, - "step": 29960 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018986640059527837, - "loss": 1.5775, - "step": 29961 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898660623689373, - "loss": 1.5402, - "step": 29962 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018986572414259623, - "loss": 1.5964, - "step": 29963 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018986538591625517, - "loss": 1.6315, - "step": 29964 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898650476899141, - "loss": 1.5231, - "step": 29965 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018986470946357305, - "loss": 1.6607, - "step": 29966 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018986437123723197, - "loss": 1.5011, - "step": 29967 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898640330108909, - "loss": 1.5091, - "step": 29968 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018986369478454985, - "loss": 1.5368, - "step": 29969 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018986335655820876, - "loss": 1.6425, - "step": 29970 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898630183318677, - "loss": 1.6507, - "step": 29971 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018986268010552664, - "loss": 1.4567, - "step": 29972 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018986234187918556, - "loss": 1.5919, - "step": 29973 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898620036528445, - "loss": 1.5715, - "step": 29974 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018986166542650344, - "loss": 1.4498, - "step": 29975 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018986132720016235, - "loss": 1.5768, - "step": 29976 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898609889738213, - "loss": 1.5135, - "step": 29977 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018986065074748024, - "loss": 1.7068, - "step": 29978 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018986031252113915, - "loss": 1.4864, - "step": 29979 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898599742947981, - "loss": 1.5262, - "step": 29980 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018985963606845703, - "loss": 1.627, - "step": 29981 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018985929784211595, - "loss": 1.3828, - "step": 29982 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898589596157749, - "loss": 1.6394, - "step": 29983 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018985862138943383, - "loss": 1.5909, - "step": 29984 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018985828316309274, - "loss": 1.485, - "step": 29985 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018985794493675168, - "loss": 1.4429, - "step": 29986 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018985760671041063, - "loss": 1.7101, - "step": 29987 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018985726848406954, - "loss": 1.6041, - "step": 29988 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018985693025772848, - "loss": 1.6334, - "step": 29989 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018985659203138742, - "loss": 1.8521, - "step": 29990 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018985625380504634, - "loss": 1.6725, - "step": 29991 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018985591557870528, - "loss": 1.7327, - "step": 29992 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018985557735236422, - "loss": 1.544, - "step": 29993 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018985523912602313, - "loss": 1.5557, - "step": 29994 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018985490089968207, - "loss": 1.7278, - "step": 29995 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018985456267334101, - "loss": 1.6142, - "step": 29996 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018985422444699993, - "loss": 1.5879, - "step": 29997 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018985388622065887, - "loss": 1.4178, - "step": 29998 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898535479943178, - "loss": 1.4697, - "step": 29999 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018985320976797672, - "loss": 1.6311, - "step": 30000 - }, - { - "epoch": 0.1, - "eval_loss": 1.3839956430288463, - "mrc_em": 25.0, - "mrc_f1": 32.16666666666667, - "nli_acc": 0.11, - "step": 30000, - "sts_f1": 0.6181818181818182, - "ynat_f1": 0.7443722943722945 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018985287154163567, - "loss": 1.578, - "step": 30001 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898525333152946, - "loss": 1.4313, - "step": 30002 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018985219508895352, - "loss": 1.6053, - "step": 30003 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018985185686261246, - "loss": 1.5051, - "step": 30004 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898515186362714, - "loss": 1.6153, - "step": 30005 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018985118040993032, - "loss": 1.5329, - "step": 30006 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018985084218358926, - "loss": 1.5295, - "step": 30007 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898505039572482, - "loss": 1.5272, - "step": 30008 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898501657309071, - "loss": 1.4727, - "step": 30009 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018984982750456605, - "loss": 1.555, - "step": 30010 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189849489278225, - "loss": 1.4732, - "step": 30011 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898491510518839, - "loss": 1.3831, - "step": 30012 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018984881282554285, - "loss": 1.5206, - "step": 30013 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898484745992018, - "loss": 1.5162, - "step": 30014 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898481363728607, - "loss": 1.5319, - "step": 30015 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018984779814651965, - "loss": 1.5157, - "step": 30016 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898474599201786, - "loss": 1.5074, - "step": 30017 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898471216938375, - "loss": 1.4854, - "step": 30018 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018984678346749644, - "loss": 1.6644, - "step": 30019 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018984644524115538, - "loss": 1.6541, - "step": 30020 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018984610701481433, - "loss": 1.5894, - "step": 30021 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018984576878847327, - "loss": 1.6183, - "step": 30022 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898454305621322, - "loss": 1.5525, - "step": 30023 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018984509233579112, - "loss": 1.5133, - "step": 30024 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018984475410945006, - "loss": 1.5729, - "step": 30025 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189844415883109, - "loss": 1.5708, - "step": 30026 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018984407765676792, - "loss": 1.5731, - "step": 30027 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018984373943042686, - "loss": 1.5446, - "step": 30028 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898434012040858, - "loss": 1.6566, - "step": 30029 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018984306297774471, - "loss": 1.4929, - "step": 30030 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018984272475140366, - "loss": 1.5369, - "step": 30031 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898423865250626, - "loss": 1.6136, - "step": 30032 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898420482987215, - "loss": 1.5174, - "step": 30033 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018984171007238045, - "loss": 1.5696, - "step": 30034 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898413718460394, - "loss": 1.4808, - "step": 30035 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898410336196983, - "loss": 1.4897, - "step": 30036 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018984069539335725, - "loss": 1.3856, - "step": 30037 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898403571670162, - "loss": 1.5737, - "step": 30038 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898400189406751, - "loss": 1.5861, - "step": 30039 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018983968071433404, - "loss": 1.5886, - "step": 30040 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018983934248799299, - "loss": 1.5874, - "step": 30041 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898390042616519, - "loss": 1.5373, - "step": 30042 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018983866603531084, - "loss": 1.5206, - "step": 30043 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018983832780896978, - "loss": 1.4974, - "step": 30044 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898379895826287, - "loss": 1.5397, - "step": 30045 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018983765135628764, - "loss": 1.3955, - "step": 30046 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018983731312994658, - "loss": 1.4732, - "step": 30047 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898369749036055, - "loss": 1.5198, - "step": 30048 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018983663667726443, - "loss": 1.6654, - "step": 30049 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018983629845092337, - "loss": 1.5031, - "step": 30050 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898359602245823, - "loss": 1.5434, - "step": 30051 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018983562199824123, - "loss": 1.5743, - "step": 30052 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018983528377190017, - "loss": 1.5363, - "step": 30053 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018983494554555909, - "loss": 1.6059, - "step": 30054 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018983460731921803, - "loss": 1.667, - "step": 30055 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018983426909287697, - "loss": 1.5458, - "step": 30056 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018983393086653588, - "loss": 1.5896, - "step": 30057 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018983359264019482, - "loss": 1.4574, - "step": 30058 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018983325441385376, - "loss": 1.4525, - "step": 30059 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018983291618751268, - "loss": 1.53, - "step": 30060 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018983257796117162, - "loss": 1.694, - "step": 30061 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018983223973483056, - "loss": 1.4935, - "step": 30062 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018983190150848947, - "loss": 1.6449, - "step": 30063 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018983156328214842, - "loss": 1.5795, - "step": 30064 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018983122505580736, - "loss": 1.3726, - "step": 30065 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018983088682946627, - "loss": 1.494, - "step": 30066 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898305486031252, - "loss": 1.5345, - "step": 30067 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018983021037678415, - "loss": 1.5797, - "step": 30068 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018982987215044307, - "loss": 1.4744, - "step": 30069 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189829533924102, - "loss": 1.5195, - "step": 30070 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018982919569776095, - "loss": 1.5786, - "step": 30071 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018982885747141986, - "loss": 1.6712, - "step": 30072 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898285192450788, - "loss": 1.424, - "step": 30073 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018982818101873775, - "loss": 1.5147, - "step": 30074 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018982784279239666, - "loss": 1.6073, - "step": 30075 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898275045660556, - "loss": 1.5827, - "step": 30076 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018982716633971454, - "loss": 1.5444, - "step": 30077 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018982682811337348, - "loss": 1.6824, - "step": 30078 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018982648988703242, - "loss": 1.6689, - "step": 30079 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018982615166069137, - "loss": 1.5926, - "step": 30080 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018982581343435028, - "loss": 1.5793, - "step": 30081 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018982547520800922, - "loss": 1.5816, - "step": 30082 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018982513698166816, - "loss": 1.577, - "step": 30083 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018982479875532708, - "loss": 1.4594, - "step": 30084 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018982446052898602, - "loss": 1.6068, - "step": 30085 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018982412230264496, - "loss": 1.5599, - "step": 30086 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018982378407630387, - "loss": 1.664, - "step": 30087 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898234458499628, - "loss": 1.6413, - "step": 30088 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018982310762362175, - "loss": 1.5221, - "step": 30089 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018982276939728067, - "loss": 1.5293, - "step": 30090 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898224311709396, - "loss": 1.6107, - "step": 30091 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018982209294459855, - "loss": 1.7216, - "step": 30092 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018982175471825746, - "loss": 1.6581, - "step": 30093 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898214164919164, - "loss": 1.5701, - "step": 30094 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018982107826557535, - "loss": 1.3895, - "step": 30095 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018982074003923426, - "loss": 1.4442, - "step": 30096 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898204018128932, - "loss": 1.5859, - "step": 30097 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018982006358655214, - "loss": 1.3982, - "step": 30098 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018981972536021106, - "loss": 1.5659, - "step": 30099 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018981938713387, - "loss": 1.5298, - "step": 30100 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018981904890752894, - "loss": 1.3473, - "step": 30101 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018981871068118785, - "loss": 1.6579, - "step": 30102 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898183724548468, - "loss": 1.6161, - "step": 30103 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018981803422850574, - "loss": 1.6273, - "step": 30104 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018981769600216465, - "loss": 1.5686, - "step": 30105 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898173577758236, - "loss": 1.608, - "step": 30106 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018981701954948253, - "loss": 1.5403, - "step": 30107 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018981668132314145, - "loss": 1.5224, - "step": 30108 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898163430968004, - "loss": 1.6256, - "step": 30109 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018981600487045933, - "loss": 1.4606, - "step": 30110 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018981566664411824, - "loss": 1.5546, - "step": 30111 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018981532841777718, - "loss": 1.4117, - "step": 30112 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018981499019143612, - "loss": 1.5844, - "step": 30113 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018981465196509504, - "loss": 1.4761, - "step": 30114 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018981431373875398, - "loss": 1.4696, - "step": 30115 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018981397551241292, - "loss": 1.6039, - "step": 30116 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018981363728607183, - "loss": 1.5338, - "step": 30117 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018981329905973078, - "loss": 1.6708, - "step": 30118 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018981296083338972, - "loss": 1.5318, - "step": 30119 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018981262260704863, - "loss": 1.5157, - "step": 30120 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018981228438070757, - "loss": 1.2951, - "step": 30121 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898119461543665, - "loss": 1.7301, - "step": 30122 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018981160792802543, - "loss": 1.5582, - "step": 30123 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018981126970168437, - "loss": 1.5416, - "step": 30124 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898109314753433, - "loss": 1.5155, - "step": 30125 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018981059324900222, - "loss": 1.5005, - "step": 30126 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018981025502266117, - "loss": 1.6308, - "step": 30127 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898099167963201, - "loss": 1.3894, - "step": 30128 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018980957856997902, - "loss": 1.4873, - "step": 30129 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018980924034363796, - "loss": 1.667, - "step": 30130 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898089021172969, - "loss": 1.612, - "step": 30131 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018980856389095582, - "loss": 1.7797, - "step": 30132 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018980822566461476, - "loss": 1.5416, - "step": 30133 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018980788743827373, - "loss": 1.5811, - "step": 30134 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018980754921193264, - "loss": 1.5517, - "step": 30135 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018980721098559158, - "loss": 1.6908, - "step": 30136 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018980687275925052, - "loss": 1.5689, - "step": 30137 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018980653453290944, - "loss": 1.5386, - "step": 30138 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018980619630656838, - "loss": 1.4903, - "step": 30139 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018980585808022732, - "loss": 1.6537, - "step": 30140 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018980551985388623, - "loss": 1.4778, - "step": 30141 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018980518162754517, - "loss": 1.5747, - "step": 30142 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018980484340120411, - "loss": 1.5277, - "step": 30143 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018980450517486303, - "loss": 1.5618, - "step": 30144 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018980416694852197, - "loss": 1.5041, - "step": 30145 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898038287221809, - "loss": 1.5891, - "step": 30146 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018980349049583983, - "loss": 1.6355, - "step": 30147 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018980315226949877, - "loss": 1.685, - "step": 30148 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898028140431577, - "loss": 1.5845, - "step": 30149 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018980247581681662, - "loss": 1.6834, - "step": 30150 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018980213759047556, - "loss": 1.6235, - "step": 30151 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898017993641345, - "loss": 1.7168, - "step": 30152 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018980146113779342, - "loss": 1.5464, - "step": 30153 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018980112291145236, - "loss": 1.719, - "step": 30154 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001898007846851113, - "loss": 1.5311, - "step": 30155 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018980044645877021, - "loss": 1.6749, - "step": 30156 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018980010823242916, - "loss": 1.5653, - "step": 30157 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897997700060881, - "loss": 1.592, - "step": 30158 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189799431779747, - "loss": 1.4883, - "step": 30159 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018979909355340595, - "loss": 1.5622, - "step": 30160 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897987553270649, - "loss": 1.6125, - "step": 30161 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897984171007238, - "loss": 1.5562, - "step": 30162 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018979807887438275, - "loss": 1.3672, - "step": 30163 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897977406480417, - "loss": 1.5781, - "step": 30164 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897974024217006, - "loss": 1.4699, - "step": 30165 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018979706419535954, - "loss": 1.5747, - "step": 30166 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018979672596901849, - "loss": 1.5574, - "step": 30167 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897963877426774, - "loss": 1.2988, - "step": 30168 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018979604951633634, - "loss": 1.5532, - "step": 30169 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018979571128999528, - "loss": 1.535, - "step": 30170 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897953730636542, - "loss": 1.6989, - "step": 30171 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018979503483731314, - "loss": 1.4928, - "step": 30172 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018979469661097208, - "loss": 1.3773, - "step": 30173 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189794358384631, - "loss": 1.3848, - "step": 30174 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018979402015828993, - "loss": 1.5303, - "step": 30175 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018979368193194887, - "loss": 1.4655, - "step": 30176 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897933437056078, - "loss": 1.5458, - "step": 30177 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018979300547926673, - "loss": 1.6777, - "step": 30178 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018979266725292567, - "loss": 1.7644, - "step": 30179 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018979232902658458, - "loss": 1.6349, - "step": 30180 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018979199080024353, - "loss": 1.5419, - "step": 30181 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018979165257390247, - "loss": 1.5725, - "step": 30182 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018979131434756138, - "loss": 1.356, - "step": 30183 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018979097612122032, - "loss": 1.492, - "step": 30184 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018979063789487926, - "loss": 1.6621, - "step": 30185 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018979029966853818, - "loss": 1.5943, - "step": 30186 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018978996144219712, - "loss": 1.4908, - "step": 30187 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018978962321585606, - "loss": 1.5501, - "step": 30188 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018978928498951497, - "loss": 1.6507, - "step": 30189 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018978894676317394, - "loss": 1.6984, - "step": 30190 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018978860853683286, - "loss": 1.5453, - "step": 30191 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897882703104918, - "loss": 1.6221, - "step": 30192 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018978793208415074, - "loss": 1.4453, - "step": 30193 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018978759385780965, - "loss": 1.5076, - "step": 30194 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897872556314686, - "loss": 1.5538, - "step": 30195 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018978691740512753, - "loss": 1.618, - "step": 30196 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018978657917878645, - "loss": 1.4639, - "step": 30197 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897862409524454, - "loss": 1.6694, - "step": 30198 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018978590272610433, - "loss": 1.5145, - "step": 30199 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018978556449976324, - "loss": 1.6524, - "step": 30200 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018978522627342219, - "loss": 1.708, - "step": 30201 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018978488804708113, - "loss": 1.4943, - "step": 30202 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018978454982074004, - "loss": 1.5851, - "step": 30203 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018978421159439898, - "loss": 1.6231, - "step": 30204 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018978387336805792, - "loss": 1.5115, - "step": 30205 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018978353514171684, - "loss": 1.5168, - "step": 30206 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018978319691537578, - "loss": 1.4634, - "step": 30207 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018978285868903472, - "loss": 1.3904, - "step": 30208 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018978252046269363, - "loss": 1.5306, - "step": 30209 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018978218223635257, - "loss": 1.551, - "step": 30210 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018978184401001152, - "loss": 1.6386, - "step": 30211 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018978150578367043, - "loss": 1.4547, - "step": 30212 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018978116755732937, - "loss": 1.6286, - "step": 30213 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897808293309883, - "loss": 1.421, - "step": 30214 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018978049110464725, - "loss": 1.4919, - "step": 30215 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018978015287830617, - "loss": 1.5705, - "step": 30216 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897798146519651, - "loss": 1.4848, - "step": 30217 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018977947642562405, - "loss": 1.6686, - "step": 30218 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018977913819928296, - "loss": 1.5839, - "step": 30219 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897787999729419, - "loss": 1.5074, - "step": 30220 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018977846174660085, - "loss": 1.4901, - "step": 30221 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018977812352025976, - "loss": 1.4207, - "step": 30222 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897777852939187, - "loss": 1.5452, - "step": 30223 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018977744706757764, - "loss": 1.5991, - "step": 30224 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018977710884123656, - "loss": 1.7401, - "step": 30225 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897767706148955, - "loss": 1.4129, - "step": 30226 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018977643238855444, - "loss": 1.695, - "step": 30227 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018977609416221335, - "loss": 1.6936, - "step": 30228 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897757559358723, - "loss": 1.4886, - "step": 30229 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018977541770953124, - "loss": 1.5281, - "step": 30230 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018977507948319015, - "loss": 1.425, - "step": 30231 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897747412568491, - "loss": 1.4722, - "step": 30232 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018977440303050803, - "loss": 1.606, - "step": 30233 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018977406480416695, - "loss": 1.3871, - "step": 30234 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897737265778259, - "loss": 1.6445, - "step": 30235 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018977338835148483, - "loss": 1.4402, - "step": 30236 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018977305012514374, - "loss": 1.6384, - "step": 30237 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018977271189880268, - "loss": 1.658, - "step": 30238 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018977237367246162, - "loss": 1.5703, - "step": 30239 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018977203544612054, - "loss": 1.5838, - "step": 30240 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018977169721977948, - "loss": 1.4991, - "step": 30241 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018977135899343842, - "loss": 1.716, - "step": 30242 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018977102076709733, - "loss": 1.3522, - "step": 30243 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018977068254075628, - "loss": 1.648, - "step": 30244 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018977034431441522, - "loss": 1.4957, - "step": 30245 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018977000608807416, - "loss": 1.4984, - "step": 30246 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897696678617331, - "loss": 1.4698, - "step": 30247 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189769329635392, - "loss": 1.5767, - "step": 30248 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018976899140905095, - "loss": 1.413, - "step": 30249 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897686531827099, - "loss": 1.5084, - "step": 30250 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897683149563688, - "loss": 1.5686, - "step": 30251 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018976797673002775, - "loss": 1.4971, - "step": 30252 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897676385036867, - "loss": 1.4135, - "step": 30253 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897673002773456, - "loss": 1.5725, - "step": 30254 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018976696205100455, - "loss": 1.4722, - "step": 30255 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897666238246635, - "loss": 1.6161, - "step": 30256 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897662855983224, - "loss": 1.5776, - "step": 30257 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018976594737198134, - "loss": 1.7838, - "step": 30258 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018976560914564028, - "loss": 1.6697, - "step": 30259 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897652709192992, - "loss": 1.4953, - "step": 30260 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018976493269295814, - "loss": 1.3391, - "step": 30261 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018976459446661708, - "loss": 1.5733, - "step": 30262 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189764256240276, - "loss": 1.5427, - "step": 30263 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018976391801393494, - "loss": 1.6394, - "step": 30264 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018976357978759388, - "loss": 1.6206, - "step": 30265 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897632415612528, - "loss": 1.4083, - "step": 30266 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018976290333491173, - "loss": 1.4816, - "step": 30267 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018976256510857067, - "loss": 1.5564, - "step": 30268 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897622268822296, - "loss": 1.3847, - "step": 30269 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018976188865588853, - "loss": 1.4892, - "step": 30270 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018976155042954747, - "loss": 1.4405, - "step": 30271 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018976121220320638, - "loss": 1.6008, - "step": 30272 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018976087397686532, - "loss": 1.6367, - "step": 30273 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018976053575052427, - "loss": 1.7185, - "step": 30274 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018976019752418318, - "loss": 1.7517, - "step": 30275 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018975985929784212, - "loss": 1.6092, - "step": 30276 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018975952107150106, - "loss": 1.4844, - "step": 30277 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018975918284515998, - "loss": 1.4141, - "step": 30278 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018975884461881892, - "loss": 1.6038, - "step": 30279 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018975850639247786, - "loss": 1.5997, - "step": 30280 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018975816816613677, - "loss": 1.4496, - "step": 30281 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897578299397957, - "loss": 1.4716, - "step": 30282 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018975749171345465, - "loss": 1.5928, - "step": 30283 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018975715348711357, - "loss": 1.6163, - "step": 30284 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897568152607725, - "loss": 1.559, - "step": 30285 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018975647703443145, - "loss": 1.7147, - "step": 30286 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018975613880809037, - "loss": 1.592, - "step": 30287 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897558005817493, - "loss": 1.6582, - "step": 30288 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018975546235540825, - "loss": 1.5293, - "step": 30289 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018975512412906716, - "loss": 1.437, - "step": 30290 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897547859027261, - "loss": 1.5515, - "step": 30291 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018975444767638504, - "loss": 1.5674, - "step": 30292 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018975410945004396, - "loss": 1.5408, - "step": 30293 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897537712237029, - "loss": 1.4049, - "step": 30294 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018975343299736184, - "loss": 1.493, - "step": 30295 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018975309477102075, - "loss": 1.6845, - "step": 30296 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897527565446797, - "loss": 1.5747, - "step": 30297 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018975241831833864, - "loss": 1.5793, - "step": 30298 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018975208009199758, - "loss": 1.4248, - "step": 30299 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897517418656565, - "loss": 1.5328, - "step": 30300 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018975140363931543, - "loss": 1.594, - "step": 30301 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018975106541297437, - "loss": 1.531, - "step": 30302 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018975072718663331, - "loss": 1.6442, - "step": 30303 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018975038896029226, - "loss": 1.4847, - "step": 30304 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018975005073395117, - "loss": 1.5245, - "step": 30305 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897497125076101, - "loss": 1.6364, - "step": 30306 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018974937428126905, - "loss": 1.6196, - "step": 30307 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018974903605492797, - "loss": 1.5451, - "step": 30308 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897486978285869, - "loss": 1.623, - "step": 30309 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018974835960224585, - "loss": 1.536, - "step": 30310 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018974802137590476, - "loss": 1.5179, - "step": 30311 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897476831495637, - "loss": 1.5357, - "step": 30312 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018974734492322264, - "loss": 1.541, - "step": 30313 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018974700669688156, - "loss": 1.5963, - "step": 30314 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897466684705405, - "loss": 1.7595, - "step": 30315 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018974633024419944, - "loss": 1.7538, - "step": 30316 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018974599201785836, - "loss": 1.4881, - "step": 30317 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897456537915173, - "loss": 1.5505, - "step": 30318 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018974531556517624, - "loss": 1.2814, - "step": 30319 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018974497733883515, - "loss": 1.5485, - "step": 30320 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897446391124941, - "loss": 1.639, - "step": 30321 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018974430088615303, - "loss": 1.441, - "step": 30322 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018974396265981195, - "loss": 1.5829, - "step": 30323 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897436244334709, - "loss": 1.594, - "step": 30324 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018974328620712983, - "loss": 1.7107, - "step": 30325 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018974294798078874, - "loss": 1.5935, - "step": 30326 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018974260975444769, - "loss": 1.3497, - "step": 30327 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018974227152810663, - "loss": 1.3781, - "step": 30328 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018974193330176554, - "loss": 1.4677, - "step": 30329 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018974159507542448, - "loss": 1.2945, - "step": 30330 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018974125684908342, - "loss": 1.5877, - "step": 30331 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018974091862274234, - "loss": 1.5959, - "step": 30332 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018974058039640128, - "loss": 1.5638, - "step": 30333 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018974024217006022, - "loss": 1.4871, - "step": 30334 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018973990394371913, - "loss": 1.4695, - "step": 30335 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018973956571737807, - "loss": 1.5226, - "step": 30336 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018973922749103702, - "loss": 1.6248, - "step": 30337 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018973888926469593, - "loss": 1.3852, - "step": 30338 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018973855103835487, - "loss": 1.5793, - "step": 30339 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897382128120138, - "loss": 1.3355, - "step": 30340 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018973787458567273, - "loss": 1.5461, - "step": 30341 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018973753635933167, - "loss": 1.2944, - "step": 30342 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897371981329906, - "loss": 1.4915, - "step": 30343 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018973685990664952, - "loss": 1.5662, - "step": 30344 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018973652168030846, - "loss": 1.6486, - "step": 30345 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897361834539674, - "loss": 1.2831, - "step": 30346 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018973584522762632, - "loss": 1.6312, - "step": 30347 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018973550700128526, - "loss": 1.6291, - "step": 30348 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897351687749442, - "loss": 1.5104, - "step": 30349 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018973483054860311, - "loss": 1.5454, - "step": 30350 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018973449232226206, - "loss": 1.5293, - "step": 30351 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189734154095921, - "loss": 1.5671, - "step": 30352 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897338158695799, - "loss": 1.6631, - "step": 30353 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018973347764323885, - "loss": 1.487, - "step": 30354 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897331394168978, - "loss": 1.6008, - "step": 30355 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897328011905567, - "loss": 1.5851, - "step": 30356 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018973246296421565, - "loss": 1.4973, - "step": 30357 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018973212473787462, - "loss": 1.4961, - "step": 30358 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018973178651153353, - "loss": 1.4851, - "step": 30359 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018973144828519247, - "loss": 1.6001, - "step": 30360 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897311100588514, - "loss": 1.5215, - "step": 30361 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018973077183251033, - "loss": 1.5529, - "step": 30362 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018973043360616927, - "loss": 1.6099, - "step": 30363 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897300953798282, - "loss": 1.3656, - "step": 30364 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018972975715348712, - "loss": 1.5646, - "step": 30365 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018972941892714606, - "loss": 1.5612, - "step": 30366 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189729080700805, - "loss": 1.5251, - "step": 30367 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018972874247446392, - "loss": 1.6042, - "step": 30368 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018972840424812286, - "loss": 1.6005, - "step": 30369 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897280660217818, - "loss": 1.543, - "step": 30370 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018972772779544072, - "loss": 1.7061, - "step": 30371 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018972738956909966, - "loss": 1.38, - "step": 30372 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897270513427586, - "loss": 1.4143, - "step": 30373 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897267131164175, - "loss": 1.5148, - "step": 30374 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018972637489007645, - "loss": 1.6916, - "step": 30375 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897260366637354, - "loss": 1.5245, - "step": 30376 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897256984373943, - "loss": 1.5964, - "step": 30377 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018972536021105325, - "loss": 1.6028, - "step": 30378 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897250219847122, - "loss": 1.4889, - "step": 30379 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897246837583711, - "loss": 1.7006, - "step": 30380 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018972434553203005, - "loss": 1.541, - "step": 30381 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189724007305689, - "loss": 1.5774, - "step": 30382 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897236690793479, - "loss": 1.5292, - "step": 30383 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018972333085300684, - "loss": 1.495, - "step": 30384 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018972299262666578, - "loss": 1.5502, - "step": 30385 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897226544003247, - "loss": 1.533, - "step": 30386 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018972231617398364, - "loss": 1.5897, - "step": 30387 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018972197794764258, - "loss": 1.4351, - "step": 30388 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897216397213015, - "loss": 1.4377, - "step": 30389 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018972130149496044, - "loss": 1.631, - "step": 30390 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018972096326861938, - "loss": 1.4771, - "step": 30391 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897206250422783, - "loss": 1.3856, - "step": 30392 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018972028681593723, - "loss": 1.4665, - "step": 30393 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018971994858959617, - "loss": 1.6975, - "step": 30394 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897196103632551, - "loss": 1.5741, - "step": 30395 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018971927213691403, - "loss": 1.5438, - "step": 30396 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018971893391057297, - "loss": 1.5141, - "step": 30397 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018971859568423188, - "loss": 1.6134, - "step": 30398 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018971825745789082, - "loss": 1.6018, - "step": 30399 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018971791923154977, - "loss": 1.4213, - "step": 30400 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018971758100520868, - "loss": 1.6248, - "step": 30401 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018971724277886762, - "loss": 1.3621, - "step": 30402 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018971690455252656, - "loss": 1.6074, - "step": 30403 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018971656632618548, - "loss": 1.3569, - "step": 30404 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018971622809984442, - "loss": 1.5587, - "step": 30405 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018971588987350336, - "loss": 1.5449, - "step": 30406 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018971555164716227, - "loss": 1.6545, - "step": 30407 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897152134208212, - "loss": 1.7552, - "step": 30408 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018971487519448015, - "loss": 1.4483, - "step": 30409 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018971453696813907, - "loss": 1.6662, - "step": 30410 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189714198741798, - "loss": 1.4891, - "step": 30411 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018971386051545695, - "loss": 1.488, - "step": 30412 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018971352228911586, - "loss": 1.5505, - "step": 30413 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897131840627748, - "loss": 1.4917, - "step": 30414 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018971284583643377, - "loss": 1.5702, - "step": 30415 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897125076100927, - "loss": 1.6569, - "step": 30416 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018971216938375163, - "loss": 1.5901, - "step": 30417 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018971183115741057, - "loss": 1.5573, - "step": 30418 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018971149293106948, - "loss": 1.5369, - "step": 30419 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018971115470472843, - "loss": 1.4064, - "step": 30420 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018971081647838737, - "loss": 1.5967, - "step": 30421 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018971047825204628, - "loss": 1.6392, - "step": 30422 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018971014002570522, - "loss": 1.5942, - "step": 30423 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018970980179936416, - "loss": 1.584, - "step": 30424 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018970946357302308, - "loss": 1.6348, - "step": 30425 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018970912534668202, - "loss": 1.5029, - "step": 30426 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018970878712034096, - "loss": 1.559, - "step": 30427 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018970844889399987, - "loss": 1.6296, - "step": 30428 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018970811066765881, - "loss": 1.5183, - "step": 30429 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018970777244131776, - "loss": 1.606, - "step": 30430 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018970743421497667, - "loss": 1.429, - "step": 30431 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897070959886356, - "loss": 1.5813, - "step": 30432 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018970675776229455, - "loss": 1.4951, - "step": 30433 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018970641953595347, - "loss": 1.4683, - "step": 30434 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897060813096124, - "loss": 1.6233, - "step": 30435 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018970574308327135, - "loss": 1.4497, - "step": 30436 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018970540485693026, - "loss": 1.6924, - "step": 30437 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897050666305892, - "loss": 1.5829, - "step": 30438 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018970472840424814, - "loss": 1.5214, - "step": 30439 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018970439017790706, - "loss": 1.4303, - "step": 30440 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189704051951566, - "loss": 1.5203, - "step": 30441 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018970371372522494, - "loss": 1.6031, - "step": 30442 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018970337549888385, - "loss": 1.324, - "step": 30443 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897030372725428, - "loss": 1.6805, - "step": 30444 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018970269904620174, - "loss": 1.5261, - "step": 30445 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018970236081986065, - "loss": 1.4838, - "step": 30446 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897020225935196, - "loss": 1.5728, - "step": 30447 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018970168436717853, - "loss": 1.7502, - "step": 30448 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018970134614083745, - "loss": 1.5142, - "step": 30449 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001897010079144964, - "loss": 1.6307, - "step": 30450 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018970066968815533, - "loss": 1.4884, - "step": 30451 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018970033146181424, - "loss": 1.5529, - "step": 30452 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018969999323547318, - "loss": 1.7011, - "step": 30453 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018969965500913213, - "loss": 1.5018, - "step": 30454 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018969931678279104, - "loss": 1.5641, - "step": 30455 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018969897855644998, - "loss": 1.4989, - "step": 30456 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018969864033010892, - "loss": 1.609, - "step": 30457 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018969830210376784, - "loss": 1.6308, - "step": 30458 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018969796387742678, - "loss": 1.6041, - "step": 30459 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018969762565108572, - "loss": 1.6241, - "step": 30460 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018969728742474463, - "loss": 1.7081, - "step": 30461 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018969694919840357, - "loss": 1.6043, - "step": 30462 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018969661097206251, - "loss": 1.708, - "step": 30463 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018969627274572143, - "loss": 1.4343, - "step": 30464 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018969593451938037, - "loss": 1.3876, - "step": 30465 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896955962930393, - "loss": 1.5782, - "step": 30466 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018969525806669823, - "loss": 1.4064, - "step": 30467 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018969491984035717, - "loss": 1.6391, - "step": 30468 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896945816140161, - "loss": 1.5523, - "step": 30469 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018969424338767502, - "loss": 1.6199, - "step": 30470 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189693905161334, - "loss": 1.5206, - "step": 30471 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896935669349929, - "loss": 1.4765, - "step": 30472 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018969322870865184, - "loss": 1.6249, - "step": 30473 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018969289048231079, - "loss": 1.5476, - "step": 30474 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896925522559697, - "loss": 1.655, - "step": 30475 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018969221402962864, - "loss": 1.6131, - "step": 30476 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018969187580328758, - "loss": 1.5282, - "step": 30477 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896915375769465, - "loss": 1.5476, - "step": 30478 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018969119935060544, - "loss": 1.5933, - "step": 30479 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018969086112426438, - "loss": 1.68, - "step": 30480 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018969052289792332, - "loss": 1.4905, - "step": 30481 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018969018467158223, - "loss": 1.4588, - "step": 30482 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018968984644524117, - "loss": 1.5689, - "step": 30483 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018968950821890012, - "loss": 1.5115, - "step": 30484 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018968916999255903, - "loss": 1.4683, - "step": 30485 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018968883176621797, - "loss": 1.5009, - "step": 30486 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896884935398769, - "loss": 1.3937, - "step": 30487 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018968815531353583, - "loss": 1.4945, - "step": 30488 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018968781708719477, - "loss": 1.7135, - "step": 30489 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896874788608537, - "loss": 1.4971, - "step": 30490 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018968714063451262, - "loss": 1.6449, - "step": 30491 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018968680240817156, - "loss": 1.4826, - "step": 30492 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896864641818305, - "loss": 1.5138, - "step": 30493 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018968612595548942, - "loss": 1.5357, - "step": 30494 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018968578772914836, - "loss": 1.5563, - "step": 30495 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896854495028073, - "loss": 1.5107, - "step": 30496 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018968511127646622, - "loss": 1.5312, - "step": 30497 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018968477305012516, - "loss": 1.663, - "step": 30498 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896844348237841, - "loss": 1.5442, - "step": 30499 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189684096597443, - "loss": 1.5692, - "step": 30500 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018968375837110195, - "loss": 1.5845, - "step": 30501 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896834201447609, - "loss": 1.6252, - "step": 30502 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896830819184198, - "loss": 1.5755, - "step": 30503 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018968274369207875, - "loss": 1.419, - "step": 30504 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896824054657377, - "loss": 1.5204, - "step": 30505 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896820672393966, - "loss": 1.6146, - "step": 30506 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018968172901305555, - "loss": 1.5617, - "step": 30507 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896813907867145, - "loss": 1.4224, - "step": 30508 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896810525603734, - "loss": 1.6591, - "step": 30509 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018968071433403234, - "loss": 1.4879, - "step": 30510 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018968037610769128, - "loss": 1.5246, - "step": 30511 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896800378813502, - "loss": 1.6072, - "step": 30512 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018967969965500914, - "loss": 1.6388, - "step": 30513 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018967936142866808, - "loss": 1.5721, - "step": 30514 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189679023202327, - "loss": 1.6463, - "step": 30515 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018967868497598593, - "loss": 1.5072, - "step": 30516 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018967834674964488, - "loss": 1.4636, - "step": 30517 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896780085233038, - "loss": 1.6231, - "step": 30518 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018967767029696273, - "loss": 1.6221, - "step": 30519 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018967733207062167, - "loss": 1.6683, - "step": 30520 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018967699384428059, - "loss": 1.4577, - "step": 30521 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018967665561793953, - "loss": 1.5515, - "step": 30522 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018967631739159847, - "loss": 1.7042, - "step": 30523 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018967597916525738, - "loss": 1.459, - "step": 30524 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018967564093891632, - "loss": 1.733, - "step": 30525 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018967530271257526, - "loss": 1.5775, - "step": 30526 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896749644862342, - "loss": 1.7056, - "step": 30527 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018967462625989315, - "loss": 1.71, - "step": 30528 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018967428803355206, - "loss": 1.3859, - "step": 30529 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189673949807211, - "loss": 1.5427, - "step": 30530 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018967361158086994, - "loss": 1.5533, - "step": 30531 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018967327335452886, - "loss": 1.4636, - "step": 30532 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896729351281878, - "loss": 1.4205, - "step": 30533 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018967259690184674, - "loss": 1.4393, - "step": 30534 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018967225867550565, - "loss": 1.4221, - "step": 30535 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896719204491646, - "loss": 1.4941, - "step": 30536 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018967158222282354, - "loss": 1.4867, - "step": 30537 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018967124399648245, - "loss": 1.3409, - "step": 30538 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896709057701414, - "loss": 1.4051, - "step": 30539 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018967056754380033, - "loss": 1.5727, - "step": 30540 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018967022931745925, - "loss": 1.5463, - "step": 30541 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896698910911182, - "loss": 1.6183, - "step": 30542 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018966955286477713, - "loss": 1.4569, - "step": 30543 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018966921463843604, - "loss": 1.4934, - "step": 30544 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018966887641209498, - "loss": 1.3678, - "step": 30545 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018966853818575392, - "loss": 1.5491, - "step": 30546 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018966819995941284, - "loss": 1.4683, - "step": 30547 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018966786173307178, - "loss": 1.463, - "step": 30548 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018966752350673072, - "loss": 1.6059, - "step": 30549 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018966718528038963, - "loss": 1.5622, - "step": 30550 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018966684705404858, - "loss": 1.4433, - "step": 30551 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018966650882770752, - "loss": 1.5366, - "step": 30552 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018966617060136643, - "loss": 1.5843, - "step": 30553 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018966583237502537, - "loss": 1.4573, - "step": 30554 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018966549414868431, - "loss": 1.6378, - "step": 30555 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018966515592234323, - "loss": 1.5022, - "step": 30556 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018966481769600217, - "loss": 1.4708, - "step": 30557 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896644794696611, - "loss": 1.6164, - "step": 30558 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018966414124332002, - "loss": 1.5784, - "step": 30559 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018966380301697897, - "loss": 1.3104, - "step": 30560 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896634647906379, - "loss": 1.5155, - "step": 30561 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018966312656429682, - "loss": 1.5024, - "step": 30562 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018966278833795576, - "loss": 1.5074, - "step": 30563 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896624501116147, - "loss": 1.4594, - "step": 30564 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018966211188527362, - "loss": 1.5203, - "step": 30565 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018966177365893256, - "loss": 1.6707, - "step": 30566 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896614354325915, - "loss": 1.5902, - "step": 30567 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018966109720625044, - "loss": 1.4577, - "step": 30568 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018966075897990935, - "loss": 1.5231, - "step": 30569 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896604207535683, - "loss": 1.6276, - "step": 30570 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018966008252722724, - "loss": 1.5036, - "step": 30571 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018965974430088615, - "loss": 1.3739, - "step": 30572 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896594060745451, - "loss": 1.3742, - "step": 30573 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018965906784820403, - "loss": 1.5971, - "step": 30574 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018965872962186295, - "loss": 1.4742, - "step": 30575 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896583913955219, - "loss": 1.5893, - "step": 30576 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018965805316918083, - "loss": 1.4935, - "step": 30577 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018965771494283974, - "loss": 1.4021, - "step": 30578 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018965737671649868, - "loss": 1.5257, - "step": 30579 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018965703849015763, - "loss": 1.5287, - "step": 30580 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018965670026381654, - "loss": 1.5284, - "step": 30581 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018965636203747548, - "loss": 1.5106, - "step": 30582 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018965602381113442, - "loss": 1.5767, - "step": 30583 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018965568558479336, - "loss": 1.5695, - "step": 30584 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896553473584523, - "loss": 1.6723, - "step": 30585 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018965500913211122, - "loss": 1.7094, - "step": 30586 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018965467090577016, - "loss": 1.6378, - "step": 30587 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896543326794291, - "loss": 1.4546, - "step": 30588 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018965399445308801, - "loss": 1.5229, - "step": 30589 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018965365622674696, - "loss": 1.548, - "step": 30590 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896533180004059, - "loss": 1.5289, - "step": 30591 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896529797740648, - "loss": 1.4975, - "step": 30592 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018965264154772375, - "loss": 1.7474, - "step": 30593 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896523033213827, - "loss": 1.4541, - "step": 30594 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896519650950416, - "loss": 1.5192, - "step": 30595 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018965162686870055, - "loss": 1.6234, - "step": 30596 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896512886423595, - "loss": 1.6349, - "step": 30597 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896509504160184, - "loss": 1.6402, - "step": 30598 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018965061218967734, - "loss": 1.5421, - "step": 30599 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018965027396333629, - "loss": 1.5155, - "step": 30600 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896499357369952, - "loss": 1.5051, - "step": 30601 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018964959751065414, - "loss": 1.6604, - "step": 30602 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018964925928431308, - "loss": 1.4905, - "step": 30603 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189648921057972, - "loss": 1.4604, - "step": 30604 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018964858283163094, - "loss": 1.4886, - "step": 30605 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018964824460528988, - "loss": 1.517, - "step": 30606 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896479063789488, - "loss": 1.4592, - "step": 30607 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018964756815260773, - "loss": 1.7406, - "step": 30608 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018964722992626667, - "loss": 1.6959, - "step": 30609 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896468916999256, - "loss": 1.634, - "step": 30610 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018964655347358453, - "loss": 1.4281, - "step": 30611 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018964621524724347, - "loss": 1.58, - "step": 30612 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018964587702090238, - "loss": 1.3953, - "step": 30613 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018964553879456133, - "loss": 1.5526, - "step": 30614 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018964520056822027, - "loss": 1.4809, - "step": 30615 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018964486234187918, - "loss": 1.4219, - "step": 30616 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018964452411553812, - "loss": 1.5034, - "step": 30617 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018964418588919706, - "loss": 1.4332, - "step": 30618 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018964384766285598, - "loss": 1.3535, - "step": 30619 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018964350943651492, - "loss": 1.7139, - "step": 30620 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018964317121017386, - "loss": 1.5078, - "step": 30621 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018964283298383277, - "loss": 1.5703, - "step": 30622 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018964249475749171, - "loss": 1.47, - "step": 30623 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018964215653115066, - "loss": 1.4574, - "step": 30624 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018964181830480957, - "loss": 1.4441, - "step": 30625 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896414800784685, - "loss": 1.5456, - "step": 30626 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018964114185212745, - "loss": 1.4522, - "step": 30627 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018964080362578637, - "loss": 1.6431, - "step": 30628 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896404653994453, - "loss": 1.5066, - "step": 30629 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018964012717310425, - "loss": 1.5497, - "step": 30630 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018963978894676316, - "loss": 1.6063, - "step": 30631 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896394507204221, - "loss": 1.472, - "step": 30632 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018963911249408104, - "loss": 1.4971, - "step": 30633 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018963877426773996, - "loss": 1.5454, - "step": 30634 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896384360413989, - "loss": 1.5397, - "step": 30635 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018963809781505784, - "loss": 1.5066, - "step": 30636 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018963775958871676, - "loss": 1.3777, - "step": 30637 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896374213623757, - "loss": 1.7341, - "step": 30638 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018963708313603466, - "loss": 1.3797, - "step": 30639 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018963674490969358, - "loss": 1.59, - "step": 30640 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018963640668335252, - "loss": 1.6555, - "step": 30641 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018963606845701146, - "loss": 1.5532, - "step": 30642 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018963573023067037, - "loss": 1.5744, - "step": 30643 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018963539200432932, - "loss": 1.3508, - "step": 30644 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018963505377798826, - "loss": 1.5122, - "step": 30645 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018963471555164717, - "loss": 1.4158, - "step": 30646 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896343773253061, - "loss": 1.4621, - "step": 30647 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018963403909896505, - "loss": 1.5174, - "step": 30648 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018963370087262397, - "loss": 1.4643, - "step": 30649 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896333626462829, - "loss": 1.5257, - "step": 30650 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018963302441994185, - "loss": 1.6264, - "step": 30651 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018963268619360076, - "loss": 1.642, - "step": 30652 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896323479672597, - "loss": 1.6035, - "step": 30653 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018963200974091865, - "loss": 1.5202, - "step": 30654 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018963167151457756, - "loss": 1.464, - "step": 30655 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896313332882365, - "loss": 1.554, - "step": 30656 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018963099506189544, - "loss": 1.4233, - "step": 30657 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018963065683555436, - "loss": 1.6012, - "step": 30658 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896303186092133, - "loss": 1.4995, - "step": 30659 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018962998038287224, - "loss": 1.5268, - "step": 30660 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018962964215653115, - "loss": 1.569, - "step": 30661 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896293039301901, - "loss": 1.3807, - "step": 30662 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018962896570384904, - "loss": 1.49, - "step": 30663 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018962862747750795, - "loss": 1.4968, - "step": 30664 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896282892511669, - "loss": 1.3634, - "step": 30665 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018962795102482583, - "loss": 1.5762, - "step": 30666 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018962761279848475, - "loss": 1.6042, - "step": 30667 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896272745721437, - "loss": 1.5244, - "step": 30668 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018962693634580263, - "loss": 1.5433, - "step": 30669 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018962659811946154, - "loss": 1.5588, - "step": 30670 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018962625989312048, - "loss": 1.4583, - "step": 30671 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018962592166677942, - "loss": 1.5357, - "step": 30672 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018962558344043834, - "loss": 1.665, - "step": 30673 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018962524521409728, - "loss": 1.4646, - "step": 30674 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018962490698775622, - "loss": 1.5329, - "step": 30675 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018962456876141513, - "loss": 1.4533, - "step": 30676 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018962423053507408, - "loss": 1.4768, - "step": 30677 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018962389230873302, - "loss": 1.5978, - "step": 30678 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018962355408239193, - "loss": 1.4143, - "step": 30679 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018962321585605087, - "loss": 1.4291, - "step": 30680 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896228776297098, - "loss": 1.6106, - "step": 30681 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018962253940336873, - "loss": 1.6096, - "step": 30682 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018962220117702767, - "loss": 1.4566, - "step": 30683 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896218629506866, - "loss": 1.3344, - "step": 30684 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018962152472434552, - "loss": 1.4611, - "step": 30685 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018962118649800446, - "loss": 1.4738, - "step": 30686 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896208482716634, - "loss": 1.5403, - "step": 30687 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018962051004532232, - "loss": 1.5812, - "step": 30688 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018962017181898126, - "loss": 1.6133, - "step": 30689 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896198335926402, - "loss": 1.6649, - "step": 30690 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018961949536629912, - "loss": 1.5606, - "step": 30691 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018961915713995806, - "loss": 1.4927, - "step": 30692 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189618818913617, - "loss": 1.3856, - "step": 30693 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896184806872759, - "loss": 1.602, - "step": 30694 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018961814246093488, - "loss": 1.4758, - "step": 30695 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018961780423459382, - "loss": 1.5484, - "step": 30696 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018961746600825274, - "loss": 1.4591, - "step": 30697 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018961712778191168, - "loss": 1.5572, - "step": 30698 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018961678955557062, - "loss": 1.5685, - "step": 30699 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018961645132922953, - "loss": 1.4757, - "step": 30700 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018961611310288847, - "loss": 1.4268, - "step": 30701 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018961577487654741, - "loss": 1.7165, - "step": 30702 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018961543665020633, - "loss": 1.5151, - "step": 30703 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018961509842386527, - "loss": 1.609, - "step": 30704 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896147601975242, - "loss": 1.5648, - "step": 30705 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018961442197118312, - "loss": 1.4421, - "step": 30706 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018961408374484207, - "loss": 1.4923, - "step": 30707 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189613745518501, - "loss": 1.4322, - "step": 30708 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018961340729215992, - "loss": 1.6102, - "step": 30709 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018961306906581886, - "loss": 1.3752, - "step": 30710 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896127308394778, - "loss": 1.5836, - "step": 30711 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018961239261313672, - "loss": 1.41, - "step": 30712 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018961205438679566, - "loss": 1.5397, - "step": 30713 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896117161604546, - "loss": 1.5652, - "step": 30714 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896113779341135, - "loss": 1.7019, - "step": 30715 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018961103970777245, - "loss": 1.5388, - "step": 30716 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896107014814314, - "loss": 1.6167, - "step": 30717 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896103632550903, - "loss": 1.4364, - "step": 30718 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018961002502874925, - "loss": 1.531, - "step": 30719 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896096868024082, - "loss": 1.4313, - "step": 30720 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896093485760671, - "loss": 1.748, - "step": 30721 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018960901034972605, - "loss": 1.4642, - "step": 30722 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189608672123385, - "loss": 1.6141, - "step": 30723 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896083338970439, - "loss": 1.429, - "step": 30724 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018960799567070284, - "loss": 1.694, - "step": 30725 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018960765744436178, - "loss": 1.4993, - "step": 30726 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896073192180207, - "loss": 1.7707, - "step": 30727 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018960698099167964, - "loss": 1.5866, - "step": 30728 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018960664276533858, - "loss": 1.5397, - "step": 30729 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896063045389975, - "loss": 1.6017, - "step": 30730 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018960596631265644, - "loss": 1.4933, - "step": 30731 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018960562808631538, - "loss": 1.6382, - "step": 30732 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896052898599743, - "loss": 1.5105, - "step": 30733 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018960495163363323, - "loss": 1.4441, - "step": 30734 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018960461340729217, - "loss": 1.4862, - "step": 30735 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001896042751809511, - "loss": 1.4494, - "step": 30736 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018960393695461003, - "loss": 1.6395, - "step": 30737 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018960359872826897, - "loss": 1.5849, - "step": 30738 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018960326050192788, - "loss": 1.3664, - "step": 30739 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018960292227558683, - "loss": 1.7131, - "step": 30740 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018960258404924577, - "loss": 1.4128, - "step": 30741 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018960224582290468, - "loss": 1.4018, - "step": 30742 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018960190759656362, - "loss": 1.5408, - "step": 30743 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018960156937022256, - "loss": 1.5399, - "step": 30744 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018960123114388148, - "loss": 1.5744, - "step": 30745 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018960089291754042, - "loss": 1.5218, - "step": 30746 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018960055469119936, - "loss": 1.6244, - "step": 30747 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018960021646485827, - "loss": 1.6945, - "step": 30748 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018959987823851721, - "loss": 1.5539, - "step": 30749 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018959954001217616, - "loss": 1.3765, - "step": 30750 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018959920178583507, - "loss": 1.5752, - "step": 30751 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018959886355949404, - "loss": 1.545, - "step": 30752 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018959852533315298, - "loss": 1.632, - "step": 30753 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895981871068119, - "loss": 1.5264, - "step": 30754 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018959784888047083, - "loss": 1.4691, - "step": 30755 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018959751065412977, - "loss": 1.5259, - "step": 30756 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895971724277887, - "loss": 1.4018, - "step": 30757 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018959683420144763, - "loss": 1.6353, - "step": 30758 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018959649597510657, - "loss": 1.498, - "step": 30759 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018959615774876549, - "loss": 1.4956, - "step": 30760 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018959581952242443, - "loss": 1.3583, - "step": 30761 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018959548129608337, - "loss": 1.3794, - "step": 30762 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018959514306974228, - "loss": 1.6723, - "step": 30763 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018959480484340122, - "loss": 1.5152, - "step": 30764 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018959446661706016, - "loss": 1.5275, - "step": 30765 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018959412839071908, - "loss": 1.4714, - "step": 30766 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018959379016437802, - "loss": 1.4744, - "step": 30767 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018959345193803696, - "loss": 1.5438, - "step": 30768 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018959311371169587, - "loss": 1.3817, - "step": 30769 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018959277548535482, - "loss": 1.5661, - "step": 30770 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018959243725901376, - "loss": 1.7118, - "step": 30771 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018959209903267267, - "loss": 1.532, - "step": 30772 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895917608063316, - "loss": 1.5326, - "step": 30773 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018959142257999055, - "loss": 1.6411, - "step": 30774 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018959108435364947, - "loss": 1.5585, - "step": 30775 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895907461273084, - "loss": 1.5938, - "step": 30776 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018959040790096735, - "loss": 1.5905, - "step": 30777 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018959006967462626, - "loss": 1.6793, - "step": 30778 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895897314482852, - "loss": 1.3956, - "step": 30779 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018958939322194415, - "loss": 1.5182, - "step": 30780 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018958905499560306, - "loss": 1.3584, - "step": 30781 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189588716769262, - "loss": 1.505, - "step": 30782 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018958837854292094, - "loss": 1.4976, - "step": 30783 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018958804031657986, - "loss": 1.488, - "step": 30784 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895877020902388, - "loss": 1.7282, - "step": 30785 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018958736386389774, - "loss": 1.5689, - "step": 30786 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018958702563755665, - "loss": 1.8113, - "step": 30787 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895866874112156, - "loss": 1.5765, - "step": 30788 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018958634918487453, - "loss": 1.5252, - "step": 30789 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018958601095853345, - "loss": 1.6041, - "step": 30790 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895856727321924, - "loss": 1.498, - "step": 30791 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018958533450585133, - "loss": 1.5483, - "step": 30792 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018958499627951024, - "loss": 1.5985, - "step": 30793 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018958465805316919, - "loss": 1.4818, - "step": 30794 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018958431982682813, - "loss": 1.5512, - "step": 30795 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018958398160048704, - "loss": 1.6468, - "step": 30796 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018958364337414598, - "loss": 1.5514, - "step": 30797 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018958330514780492, - "loss": 1.5391, - "step": 30798 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018958296692146384, - "loss": 1.4643, - "step": 30799 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018958262869512278, - "loss": 1.682, - "step": 30800 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018958229046878172, - "loss": 1.5927, - "step": 30801 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018958195224244063, - "loss": 1.8063, - "step": 30802 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018958161401609957, - "loss": 1.4963, - "step": 30803 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018958127578975852, - "loss": 1.4739, - "step": 30804 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018958093756341743, - "loss": 1.6015, - "step": 30805 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018958059933707637, - "loss": 1.5342, - "step": 30806 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895802611107353, - "loss": 1.5802, - "step": 30807 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018957992288439425, - "loss": 1.4287, - "step": 30808 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895795846580532, - "loss": 1.5065, - "step": 30809 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895792464317121, - "loss": 1.6121, - "step": 30810 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018957890820537105, - "loss": 1.5092, - "step": 30811 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018957856997903, - "loss": 1.5908, - "step": 30812 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895782317526889, - "loss": 1.5126, - "step": 30813 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018957789352634785, - "loss": 1.6003, - "step": 30814 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895775553000068, - "loss": 1.5324, - "step": 30815 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895772170736657, - "loss": 1.4562, - "step": 30816 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018957687884732464, - "loss": 1.5946, - "step": 30817 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018957654062098358, - "loss": 1.6112, - "step": 30818 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895762023946425, - "loss": 1.4831, - "step": 30819 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018957586416830144, - "loss": 1.4845, - "step": 30820 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018957552594196038, - "loss": 1.3189, - "step": 30821 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895751877156193, - "loss": 1.5241, - "step": 30822 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018957484948927824, - "loss": 1.5799, - "step": 30823 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018957451126293718, - "loss": 1.7577, - "step": 30824 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895741730365961, - "loss": 1.6232, - "step": 30825 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018957383481025503, - "loss": 1.3535, - "step": 30826 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018957349658391397, - "loss": 1.3824, - "step": 30827 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895731583575729, - "loss": 1.6011, - "step": 30828 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018957282013123183, - "loss": 1.4623, - "step": 30829 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018957248190489077, - "loss": 1.6104, - "step": 30830 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018957214367854968, - "loss": 1.7175, - "step": 30831 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018957180545220862, - "loss": 1.6582, - "step": 30832 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018957146722586757, - "loss": 1.6482, - "step": 30833 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895711289995265, - "loss": 1.4455, - "step": 30834 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018957079077318542, - "loss": 1.6296, - "step": 30835 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018957045254684436, - "loss": 1.5627, - "step": 30836 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895701143205033, - "loss": 1.4134, - "step": 30837 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018956977609416222, - "loss": 1.3673, - "step": 30838 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018956943786782116, - "loss": 1.5296, - "step": 30839 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895690996414801, - "loss": 1.5152, - "step": 30840 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189568761415139, - "loss": 1.6, - "step": 30841 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018956842318879795, - "loss": 1.5819, - "step": 30842 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895680849624569, - "loss": 1.4491, - "step": 30843 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895677467361158, - "loss": 1.5128, - "step": 30844 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018956740850977475, - "loss": 1.3611, - "step": 30845 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895670702834337, - "loss": 1.6149, - "step": 30846 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895667320570926, - "loss": 1.5574, - "step": 30847 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018956639383075155, - "loss": 1.5937, - "step": 30848 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895660556044105, - "loss": 1.531, - "step": 30849 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895657173780694, - "loss": 1.5783, - "step": 30850 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018956537915172834, - "loss": 1.4522, - "step": 30851 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018956504092538728, - "loss": 1.536, - "step": 30852 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895647026990462, - "loss": 1.5881, - "step": 30853 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018956436447270514, - "loss": 1.5158, - "step": 30854 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018956402624636408, - "loss": 1.6096, - "step": 30855 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189563688020023, - "loss": 1.5364, - "step": 30856 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018956334979368194, - "loss": 1.5173, - "step": 30857 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018956301156734088, - "loss": 1.3571, - "step": 30858 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895626733409998, - "loss": 1.5427, - "step": 30859 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018956233511465873, - "loss": 1.4949, - "step": 30860 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018956199688831767, - "loss": 1.5566, - "step": 30861 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895616586619766, - "loss": 1.5304, - "step": 30862 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018956132043563553, - "loss": 1.4439, - "step": 30863 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018956098220929447, - "loss": 1.4705, - "step": 30864 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895606439829534, - "loss": 1.527, - "step": 30865 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018956030575661235, - "loss": 1.4858, - "step": 30866 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018955996753027127, - "loss": 1.5243, - "step": 30867 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895596293039302, - "loss": 1.6086, - "step": 30868 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018955929107758915, - "loss": 1.6731, - "step": 30869 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018955895285124806, - "loss": 1.5675, - "step": 30870 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189558614624907, - "loss": 1.4809, - "step": 30871 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018955827639856594, - "loss": 1.5267, - "step": 30872 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018955793817222486, - "loss": 1.4163, - "step": 30873 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895575999458838, - "loss": 1.6948, - "step": 30874 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018955726171954274, - "loss": 1.522, - "step": 30875 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018955692349320165, - "loss": 1.6843, - "step": 30876 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895565852668606, - "loss": 1.4292, - "step": 30877 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018955624704051954, - "loss": 1.6303, - "step": 30878 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018955590881417845, - "loss": 1.4137, - "step": 30879 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895555705878374, - "loss": 1.5106, - "step": 30880 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018955523236149633, - "loss": 1.7345, - "step": 30881 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018955489413515525, - "loss": 1.5124, - "step": 30882 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895545559088142, - "loss": 1.4596, - "step": 30883 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018955421768247313, - "loss": 1.6246, - "step": 30884 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018955387945613204, - "loss": 1.6228, - "step": 30885 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018955354122979098, - "loss": 1.4331, - "step": 30886 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018955320300344993, - "loss": 1.5923, - "step": 30887 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018955286477710884, - "loss": 1.6543, - "step": 30888 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018955252655076778, - "loss": 1.6419, - "step": 30889 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018955218832442672, - "loss": 1.445, - "step": 30890 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018955185009808564, - "loss": 1.4877, - "step": 30891 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018955151187174458, - "loss": 1.526, - "step": 30892 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018955117364540352, - "loss": 1.494, - "step": 30893 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018955083541906243, - "loss": 1.3596, - "step": 30894 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018955049719272137, - "loss": 1.5668, - "step": 30895 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018955015896638031, - "loss": 1.7044, - "step": 30896 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018954982074003923, - "loss": 1.6834, - "step": 30897 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018954948251369817, - "loss": 1.6374, - "step": 30898 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895491442873571, - "loss": 1.7698, - "step": 30899 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018954880606101603, - "loss": 1.4224, - "step": 30900 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018954846783467497, - "loss": 1.4735, - "step": 30901 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895481296083339, - "loss": 1.4271, - "step": 30902 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018954779138199282, - "loss": 1.6152, - "step": 30903 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018954745315565176, - "loss": 1.4401, - "step": 30904 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895471149293107, - "loss": 1.5061, - "step": 30905 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018954677670296962, - "loss": 1.6089, - "step": 30906 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018954643847662856, - "loss": 1.5253, - "step": 30907 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895461002502875, - "loss": 1.5094, - "step": 30908 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018954576202394641, - "loss": 1.6535, - "step": 30909 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018954542379760536, - "loss": 1.3804, - "step": 30910 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895450855712643, - "loss": 1.4565, - "step": 30911 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895447473449232, - "loss": 1.5328, - "step": 30912 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018954440911858215, - "loss": 1.5343, - "step": 30913 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895440708922411, - "loss": 1.5127, - "step": 30914 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895437326659, - "loss": 1.6221, - "step": 30915 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018954339443955895, - "loss": 1.6732, - "step": 30916 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895430562132179, - "loss": 1.7477, - "step": 30917 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895427179868768, - "loss": 1.5232, - "step": 30918 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018954237976053574, - "loss": 1.5762, - "step": 30919 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895420415341947, - "loss": 1.5044, - "step": 30920 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018954170330785363, - "loss": 1.4302, - "step": 30921 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018954136508151257, - "loss": 1.5143, - "step": 30922 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895410268551715, - "loss": 1.4838, - "step": 30923 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018954068862883042, - "loss": 1.693, - "step": 30924 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018954035040248936, - "loss": 1.471, - "step": 30925 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895400121761483, - "loss": 1.6547, - "step": 30926 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018953967394980722, - "loss": 1.7103, - "step": 30927 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018953933572346616, - "loss": 1.6164, - "step": 30928 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895389974971251, - "loss": 1.5185, - "step": 30929 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018953865927078402, - "loss": 1.5756, - "step": 30930 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018953832104444296, - "loss": 1.4731, - "step": 30931 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895379828181019, - "loss": 1.5762, - "step": 30932 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895376445917608, - "loss": 1.6798, - "step": 30933 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018953730636541975, - "loss": 1.6313, - "step": 30934 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895369681390787, - "loss": 1.6299, - "step": 30935 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895366299127376, - "loss": 1.6126, - "step": 30936 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018953629168639655, - "loss": 1.4601, - "step": 30937 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895359534600555, - "loss": 1.5149, - "step": 30938 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895356152337144, - "loss": 1.71, - "step": 30939 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018953527700737335, - "loss": 1.6158, - "step": 30940 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895349387810323, - "loss": 1.6865, - "step": 30941 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895346005546912, - "loss": 1.4511, - "step": 30942 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018953426232835014, - "loss": 1.5685, - "step": 30943 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018953392410200908, - "loss": 1.5728, - "step": 30944 - }, - { - "epoch": 0.1, - "learning_rate": 0.000189533585875668, - "loss": 1.6138, - "step": 30945 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018953324764932694, - "loss": 1.4819, - "step": 30946 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018953290942298588, - "loss": 1.6263, - "step": 30947 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895325711966448, - "loss": 1.5386, - "step": 30948 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018953223297030373, - "loss": 1.5203, - "step": 30949 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018953189474396268, - "loss": 1.6163, - "step": 30950 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895315565176216, - "loss": 1.5785, - "step": 30951 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018953121829128053, - "loss": 1.6246, - "step": 30952 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018953088006493947, - "loss": 1.5257, - "step": 30953 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018953054183859839, - "loss": 1.4452, - "step": 30954 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018953020361225733, - "loss": 1.5862, - "step": 30955 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018952986538591627, - "loss": 1.6392, - "step": 30956 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018952952715957518, - "loss": 1.4248, - "step": 30957 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018952918893323412, - "loss": 1.6033, - "step": 30958 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018952885070689306, - "loss": 1.5906, - "step": 30959 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018952851248055198, - "loss": 1.4579, - "step": 30960 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018952817425421092, - "loss": 1.657, - "step": 30961 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018952783602786986, - "loss": 1.4224, - "step": 30962 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018952749780152877, - "loss": 1.6298, - "step": 30963 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018952715957518772, - "loss": 1.6229, - "step": 30964 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018952682134884666, - "loss": 1.5133, - "step": 30965 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018952648312250557, - "loss": 1.3794, - "step": 30966 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895261448961645, - "loss": 1.5184, - "step": 30967 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018952580666982345, - "loss": 1.4831, - "step": 30968 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018952546844348237, - "loss": 1.5364, - "step": 30969 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895251302171413, - "loss": 1.5604, - "step": 30970 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018952479199080025, - "loss": 1.4063, - "step": 30971 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018952445376445916, - "loss": 1.509, - "step": 30972 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895241155381181, - "loss": 1.4591, - "step": 30973 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018952377731177705, - "loss": 1.7653, - "step": 30974 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018952343908543596, - "loss": 1.7987, - "step": 30975 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018952310085909493, - "loss": 1.7097, - "step": 30976 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018952276263275387, - "loss": 1.4516, - "step": 30977 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018952242440641278, - "loss": 1.5767, - "step": 30978 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018952208618007172, - "loss": 1.6276, - "step": 30979 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018952174795373067, - "loss": 1.4767, - "step": 30980 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018952140972738958, - "loss": 1.5914, - "step": 30981 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018952107150104852, - "loss": 1.4771, - "step": 30982 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018952073327470746, - "loss": 1.7767, - "step": 30983 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018952039504836638, - "loss": 1.4456, - "step": 30984 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018952005682202532, - "loss": 1.4922, - "step": 30985 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018951971859568426, - "loss": 1.4909, - "step": 30986 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018951938036934317, - "loss": 1.5356, - "step": 30987 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018951904214300211, - "loss": 1.6842, - "step": 30988 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018951870391666105, - "loss": 1.4238, - "step": 30989 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018951836569031997, - "loss": 1.5562, - "step": 30990 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895180274639789, - "loss": 1.4462, - "step": 30991 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018951768923763785, - "loss": 1.6456, - "step": 30992 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018951735101129677, - "loss": 1.528, - "step": 30993 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895170127849557, - "loss": 1.4065, - "step": 30994 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018951667455861465, - "loss": 1.5086, - "step": 30995 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018951633633227356, - "loss": 1.4861, - "step": 30996 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895159981059325, - "loss": 1.5151, - "step": 30997 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018951565987959144, - "loss": 1.4605, - "step": 30998 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018951532165325036, - "loss": 1.4275, - "step": 30999 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895149834269093, - "loss": 1.5151, - "step": 31000 - }, - { - "epoch": 0.1, - "eval_loss": 1.3699857271634615, - "mrc_em": 15.0, - "mrc_f1": 27.133333333333347, - "nli_acc": 0.18, - "step": 31000, - "sts_f1": 0.5405405405405405, - "ynat_f1": 0.5828359006879273 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018951464520056824, - "loss": 1.3227, - "step": 31001 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018951430697422715, - "loss": 1.6645, - "step": 31002 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895139687478861, - "loss": 1.3116, - "step": 31003 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018951363052154504, - "loss": 1.5851, - "step": 31004 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018951329229520395, - "loss": 1.6841, - "step": 31005 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895129540688629, - "loss": 1.5587, - "step": 31006 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018951261584252183, - "loss": 1.5757, - "step": 31007 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018951227761618075, - "loss": 1.6655, - "step": 31008 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895119393898397, - "loss": 1.4152, - "step": 31009 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018951160116349863, - "loss": 1.6381, - "step": 31010 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018951126293715754, - "loss": 1.5694, - "step": 31011 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018951092471081648, - "loss": 1.5035, - "step": 31012 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018951058648447543, - "loss": 1.3404, - "step": 31013 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018951024825813434, - "loss": 1.4382, - "step": 31014 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018950991003179328, - "loss": 1.5141, - "step": 31015 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018950957180545222, - "loss": 1.493, - "step": 31016 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018950923357911114, - "loss": 1.592, - "step": 31017 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018950889535277008, - "loss": 1.5905, - "step": 31018 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018950855712642902, - "loss": 1.6759, - "step": 31019 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018950821890008793, - "loss": 1.5385, - "step": 31020 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018950788067374687, - "loss": 1.4872, - "step": 31021 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018950754244740581, - "loss": 1.4868, - "step": 31022 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018950720422106473, - "loss": 1.549, - "step": 31023 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018950686599472367, - "loss": 1.5592, - "step": 31024 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895065277683826, - "loss": 1.5414, - "step": 31025 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018950618954204152, - "loss": 1.6027, - "step": 31026 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018950585131570047, - "loss": 1.6075, - "step": 31027 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895055130893594, - "loss": 1.5215, - "step": 31028 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018950517486301832, - "loss": 1.5751, - "step": 31029 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018950483663667726, - "loss": 1.7721, - "step": 31030 - }, - { - "epoch": 0.1, - "learning_rate": 0.0001895044984103362, - "loss": 1.6283, - "step": 31031 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018950416018399514, - "loss": 1.4805, - "step": 31032 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018950382195765409, - "loss": 1.4502, - "step": 31033 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018950348373131303, - "loss": 1.6864, - "step": 31034 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018950314550497194, - "loss": 1.7522, - "step": 31035 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018950280727863088, - "loss": 1.6572, - "step": 31036 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018950246905228982, - "loss": 1.4944, - "step": 31037 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018950213082594874, - "loss": 1.5768, - "step": 31038 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018950179259960768, - "loss": 1.5152, - "step": 31039 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018950145437326662, - "loss": 1.5106, - "step": 31040 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018950111614692553, - "loss": 1.6529, - "step": 31041 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018950077792058447, - "loss": 1.3962, - "step": 31042 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018950043969424342, - "loss": 1.4989, - "step": 31043 - }, - { - "epoch": 0.1, - "learning_rate": 0.00018950010146790233, - "loss": 1.5297, - "step": 31044 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018949976324156127, - "loss": 1.5902, - "step": 31045 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894994250152202, - "loss": 1.566, - "step": 31046 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018949908678887913, - "loss": 1.4514, - "step": 31047 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018949874856253807, - "loss": 1.541, - "step": 31048 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189498410336197, - "loss": 1.4755, - "step": 31049 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018949807210985592, - "loss": 1.5973, - "step": 31050 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018949773388351486, - "loss": 1.5968, - "step": 31051 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894973956571738, - "loss": 1.6323, - "step": 31052 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018949705743083272, - "loss": 1.3635, - "step": 31053 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018949671920449166, - "loss": 1.3921, - "step": 31054 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894963809781506, - "loss": 1.6143, - "step": 31055 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018949604275180951, - "loss": 1.5385, - "step": 31056 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018949570452546846, - "loss": 1.4019, - "step": 31057 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894953662991274, - "loss": 1.5987, - "step": 31058 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894950280727863, - "loss": 1.6167, - "step": 31059 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018949468984644525, - "loss": 1.5861, - "step": 31060 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894943516201042, - "loss": 1.4726, - "step": 31061 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894940133937631, - "loss": 1.514, - "step": 31062 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018949367516742205, - "loss": 1.6165, - "step": 31063 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189493336941081, - "loss": 1.6369, - "step": 31064 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894929987147399, - "loss": 1.508, - "step": 31065 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018949266048839884, - "loss": 1.4804, - "step": 31066 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018949232226205779, - "loss": 1.4298, - "step": 31067 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894919840357167, - "loss": 1.4263, - "step": 31068 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018949164580937564, - "loss": 1.6444, - "step": 31069 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018949130758303458, - "loss": 1.4923, - "step": 31070 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894909693566935, - "loss": 1.5065, - "step": 31071 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018949063113035244, - "loss": 1.5583, - "step": 31072 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018949029290401138, - "loss": 1.5536, - "step": 31073 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894899546776703, - "loss": 1.6926, - "step": 31074 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018948961645132923, - "loss": 1.5923, - "step": 31075 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018948927822498817, - "loss": 1.6335, - "step": 31076 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894889399986471, - "loss": 1.5857, - "step": 31077 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018948860177230603, - "loss": 1.4939, - "step": 31078 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018948826354596497, - "loss": 1.5097, - "step": 31079 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018948792531962389, - "loss": 1.5794, - "step": 31080 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018948758709328283, - "loss": 1.4931, - "step": 31081 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018948724886694177, - "loss": 1.5074, - "step": 31082 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018948691064060068, - "loss": 1.5697, - "step": 31083 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018948657241425962, - "loss": 1.7055, - "step": 31084 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018948623418791856, - "loss": 1.4441, - "step": 31085 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018948589596157748, - "loss": 1.7394, - "step": 31086 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018948555773523642, - "loss": 1.467, - "step": 31087 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018948521950889536, - "loss": 1.585, - "step": 31088 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894848812825543, - "loss": 1.4495, - "step": 31089 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018948454305621324, - "loss": 1.52, - "step": 31090 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018948420482987216, - "loss": 1.5492, - "step": 31091 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894838666035311, - "loss": 1.6693, - "step": 31092 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018948352837719004, - "loss": 1.5966, - "step": 31093 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018948319015084895, - "loss": 1.5408, - "step": 31094 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894828519245079, - "loss": 1.4344, - "step": 31095 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018948251369816684, - "loss": 1.4846, - "step": 31096 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018948217547182575, - "loss": 1.4367, - "step": 31097 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894818372454847, - "loss": 1.4131, - "step": 31098 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018948149901914363, - "loss": 1.6182, - "step": 31099 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018948116079280255, - "loss": 1.5421, - "step": 31100 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894808225664615, - "loss": 1.5255, - "step": 31101 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018948048434012043, - "loss": 1.5585, - "step": 31102 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018948014611377937, - "loss": 1.5163, - "step": 31103 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018947980788743828, - "loss": 1.5717, - "step": 31104 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018947946966109722, - "loss": 1.6303, - "step": 31105 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018947913143475617, - "loss": 1.4551, - "step": 31106 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018947879320841508, - "loss": 1.5737, - "step": 31107 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018947845498207402, - "loss": 1.5829, - "step": 31108 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018947811675573296, - "loss": 1.4613, - "step": 31109 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018947777852939188, - "loss": 1.4988, - "step": 31110 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018947744030305082, - "loss": 1.6594, - "step": 31111 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018947710207670976, - "loss": 1.624, - "step": 31112 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018947676385036867, - "loss": 1.4862, - "step": 31113 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894764256240276, - "loss": 1.5179, - "step": 31114 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018947608739768655, - "loss": 1.4315, - "step": 31115 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018947574917134547, - "loss": 1.3102, - "step": 31116 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894754109450044, - "loss": 1.4008, - "step": 31117 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018947507271866335, - "loss": 1.5539, - "step": 31118 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018947473449232226, - "loss": 1.5979, - "step": 31119 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894743962659812, - "loss": 1.4842, - "step": 31120 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018947405803964015, - "loss": 1.6401, - "step": 31121 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018947371981329906, - "loss": 1.6515, - "step": 31122 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189473381586958, - "loss": 1.7997, - "step": 31123 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018947304336061694, - "loss": 1.684, - "step": 31124 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018947270513427586, - "loss": 1.6601, - "step": 31125 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894723669079348, - "loss": 1.5519, - "step": 31126 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018947202868159374, - "loss": 1.4293, - "step": 31127 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018947169045525265, - "loss": 1.2974, - "step": 31128 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894713522289116, - "loss": 1.4745, - "step": 31129 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018947101400257054, - "loss": 1.5821, - "step": 31130 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018947067577622945, - "loss": 1.6938, - "step": 31131 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894703375498884, - "loss": 1.4197, - "step": 31132 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018946999932354733, - "loss": 1.5963, - "step": 31133 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018946966109720625, - "loss": 1.6427, - "step": 31134 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894693228708652, - "loss": 1.4915, - "step": 31135 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018946898464452413, - "loss": 1.6623, - "step": 31136 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018946864641818304, - "loss": 1.2772, - "step": 31137 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018946830819184198, - "loss": 1.5566, - "step": 31138 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018946796996550092, - "loss": 1.6774, - "step": 31139 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018946763173915984, - "loss": 1.4273, - "step": 31140 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018946729351281878, - "loss": 1.6146, - "step": 31141 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018946695528647772, - "loss": 1.4746, - "step": 31142 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018946661706013663, - "loss": 1.4248, - "step": 31143 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018946627883379558, - "loss": 1.5018, - "step": 31144 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018946594060745452, - "loss": 1.6607, - "step": 31145 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018946560238111346, - "loss": 1.5769, - "step": 31146 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894652641547724, - "loss": 1.4061, - "step": 31147 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018946492592843131, - "loss": 1.6569, - "step": 31148 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018946458770209025, - "loss": 1.4617, - "step": 31149 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894642494757492, - "loss": 1.5852, - "step": 31150 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894639112494081, - "loss": 1.4348, - "step": 31151 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018946357302306705, - "loss": 1.622, - "step": 31152 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189463234796726, - "loss": 1.4811, - "step": 31153 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894628965703849, - "loss": 1.6466, - "step": 31154 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018946255834404385, - "loss": 1.4299, - "step": 31155 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894622201177028, - "loss": 1.4233, - "step": 31156 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894618818913617, - "loss": 1.5017, - "step": 31157 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018946154366502064, - "loss": 1.6657, - "step": 31158 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018946120543867958, - "loss": 1.6776, - "step": 31159 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894608672123385, - "loss": 1.4734, - "step": 31160 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018946052898599744, - "loss": 1.6588, - "step": 31161 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018946019075965638, - "loss": 1.6837, - "step": 31162 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894598525333153, - "loss": 1.5343, - "step": 31163 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018945951430697424, - "loss": 1.7042, - "step": 31164 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018945917608063318, - "loss": 1.6635, - "step": 31165 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894588378542921, - "loss": 1.5181, - "step": 31166 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018945849962795103, - "loss": 1.4756, - "step": 31167 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018945816140160997, - "loss": 1.3404, - "step": 31168 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894578231752689, - "loss": 1.5421, - "step": 31169 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018945748494892783, - "loss": 1.5869, - "step": 31170 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018945714672258677, - "loss": 1.7525, - "step": 31171 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018945680849624568, - "loss": 1.6244, - "step": 31172 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018945647026990463, - "loss": 1.466, - "step": 31173 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018945613204356357, - "loss": 1.4756, - "step": 31174 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018945579381722248, - "loss": 1.7676, - "step": 31175 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018945545559088142, - "loss": 1.6668, - "step": 31176 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018945511736454036, - "loss": 1.6797, - "step": 31177 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018945477913819928, - "loss": 1.5331, - "step": 31178 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018945444091185822, - "loss": 1.5951, - "step": 31179 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018945410268551716, - "loss": 1.6251, - "step": 31180 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018945376445917607, - "loss": 1.7031, - "step": 31181 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018945342623283501, - "loss": 1.6083, - "step": 31182 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018945308800649396, - "loss": 1.51, - "step": 31183 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018945274978015287, - "loss": 1.4349, - "step": 31184 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894524115538118, - "loss": 1.5114, - "step": 31185 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018945207332747075, - "loss": 1.5822, - "step": 31186 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894517351011297, - "loss": 1.5141, - "step": 31187 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894513968747886, - "loss": 1.5219, - "step": 31188 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018945105864844755, - "loss": 1.4362, - "step": 31189 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894507204221065, - "loss": 1.5159, - "step": 31190 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894503821957654, - "loss": 1.696, - "step": 31191 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018945004396942434, - "loss": 1.5383, - "step": 31192 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018944970574308329, - "loss": 1.5098, - "step": 31193 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894493675167422, - "loss": 1.5561, - "step": 31194 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018944902929040114, - "loss": 1.5147, - "step": 31195 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018944869106406008, - "loss": 1.5061, - "step": 31196 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189448352837719, - "loss": 1.6165, - "step": 31197 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018944801461137794, - "loss": 1.5066, - "step": 31198 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018944767638503688, - "loss": 1.4129, - "step": 31199 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894473381586958, - "loss": 1.4555, - "step": 31200 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018944699993235476, - "loss": 1.4928, - "step": 31201 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018944666170601367, - "loss": 1.5781, - "step": 31202 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018944632347967262, - "loss": 1.5966, - "step": 31203 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018944598525333156, - "loss": 1.4892, - "step": 31204 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018944564702699047, - "loss": 1.6631, - "step": 31205 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894453088006494, - "loss": 1.4068, - "step": 31206 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018944497057430835, - "loss": 1.4478, - "step": 31207 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018944463234796727, - "loss": 1.5349, - "step": 31208 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894442941216262, - "loss": 1.4032, - "step": 31209 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018944395589528515, - "loss": 1.5247, - "step": 31210 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018944361766894406, - "loss": 1.6562, - "step": 31211 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189443279442603, - "loss": 1.5905, - "step": 31212 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018944294121626195, - "loss": 1.5556, - "step": 31213 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018944260298992086, - "loss": 1.4707, - "step": 31214 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894422647635798, - "loss": 1.5414, - "step": 31215 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018944192653723874, - "loss": 1.6632, - "step": 31216 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018944158831089766, - "loss": 1.6974, - "step": 31217 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894412500845566, - "loss": 1.3354, - "step": 31218 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018944091185821554, - "loss": 1.4312, - "step": 31219 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018944057363187445, - "loss": 1.4592, - "step": 31220 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894402354055334, - "loss": 1.4783, - "step": 31221 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018943989717919233, - "loss": 1.5268, - "step": 31222 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018943955895285125, - "loss": 1.4738, - "step": 31223 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894392207265102, - "loss": 1.4866, - "step": 31224 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018943888250016913, - "loss": 1.4445, - "step": 31225 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018943854427382804, - "loss": 1.6549, - "step": 31226 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018943820604748699, - "loss": 1.3529, - "step": 31227 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018943786782114593, - "loss": 1.4844, - "step": 31228 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018943752959480484, - "loss": 1.7697, - "step": 31229 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018943719136846378, - "loss": 1.4839, - "step": 31230 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018943685314212272, - "loss": 1.6879, - "step": 31231 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018943651491578164, - "loss": 1.3907, - "step": 31232 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018943617668944058, - "loss": 1.5136, - "step": 31233 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018943583846309952, - "loss": 1.4516, - "step": 31234 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018943550023675843, - "loss": 1.5403, - "step": 31235 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018943516201041737, - "loss": 1.7504, - "step": 31236 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018943482378407632, - "loss": 1.6779, - "step": 31237 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018943448555773523, - "loss": 1.4842, - "step": 31238 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018943414733139417, - "loss": 1.5326, - "step": 31239 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894338091050531, - "loss": 1.321, - "step": 31240 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018943347087871203, - "loss": 1.4807, - "step": 31241 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018943313265237097, - "loss": 1.5648, - "step": 31242 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894327944260299, - "loss": 1.4426, - "step": 31243 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018943245619968882, - "loss": 1.5844, - "step": 31244 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018943211797334776, - "loss": 1.6221, - "step": 31245 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894317797470067, - "loss": 1.5478, - "step": 31246 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018943144152066562, - "loss": 1.4926, - "step": 31247 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018943110329432456, - "loss": 1.4344, - "step": 31248 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894307650679835, - "loss": 1.2823, - "step": 31249 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018943042684164242, - "loss": 1.5626, - "step": 31250 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018943008861530136, - "loss": 1.5995, - "step": 31251 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894297503889603, - "loss": 1.4507, - "step": 31252 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894294121626192, - "loss": 1.5365, - "step": 31253 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018942907393627815, - "loss": 1.5561, - "step": 31254 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894287357099371, - "loss": 1.5483, - "step": 31255 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189428397483596, - "loss": 1.6152, - "step": 31256 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018942805925725498, - "loss": 1.5471, - "step": 31257 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018942772103091392, - "loss": 1.5686, - "step": 31258 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018942738280457283, - "loss": 1.6612, - "step": 31259 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018942704457823177, - "loss": 1.5418, - "step": 31260 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018942670635189071, - "loss": 1.4831, - "step": 31261 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018942636812554963, - "loss": 1.5799, - "step": 31262 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018942602989920857, - "loss": 1.5044, - "step": 31263 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894256916728675, - "loss": 1.4841, - "step": 31264 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018942535344652642, - "loss": 1.4549, - "step": 31265 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018942501522018537, - "loss": 1.4182, - "step": 31266 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894246769938443, - "loss": 1.4497, - "step": 31267 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018942433876750322, - "loss": 1.4663, - "step": 31268 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018942400054116216, - "loss": 1.6059, - "step": 31269 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894236623148211, - "loss": 1.6093, - "step": 31270 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018942332408848002, - "loss": 1.5087, - "step": 31271 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018942298586213896, - "loss": 1.5466, - "step": 31272 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894226476357979, - "loss": 1.4414, - "step": 31273 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894223094094568, - "loss": 1.5596, - "step": 31274 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018942197118311575, - "loss": 1.4642, - "step": 31275 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894216329567747, - "loss": 1.6106, - "step": 31276 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894212947304336, - "loss": 1.5441, - "step": 31277 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018942095650409255, - "loss": 1.6332, - "step": 31278 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894206182777515, - "loss": 1.6235, - "step": 31279 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894202800514104, - "loss": 1.4836, - "step": 31280 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018941994182506935, - "loss": 1.4765, - "step": 31281 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894196035987283, - "loss": 1.6507, - "step": 31282 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894192653723872, - "loss": 1.6547, - "step": 31283 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018941892714604614, - "loss": 1.4831, - "step": 31284 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018941858891970508, - "loss": 1.5849, - "step": 31285 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189418250693364, - "loss": 1.5349, - "step": 31286 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018941791246702294, - "loss": 1.6253, - "step": 31287 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018941757424068188, - "loss": 1.5128, - "step": 31288 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894172360143408, - "loss": 1.7333, - "step": 31289 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018941689778799974, - "loss": 1.6047, - "step": 31290 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018941655956165868, - "loss": 1.4192, - "step": 31291 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894162213353176, - "loss": 1.5883, - "step": 31292 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018941588310897653, - "loss": 1.5949, - "step": 31293 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018941554488263547, - "loss": 1.5123, - "step": 31294 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894152066562944, - "loss": 1.6501, - "step": 31295 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018941486842995333, - "loss": 1.6047, - "step": 31296 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018941453020361227, - "loss": 1.5909, - "step": 31297 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018941419197727118, - "loss": 1.4513, - "step": 31298 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018941385375093012, - "loss": 1.409, - "step": 31299 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018941351552458907, - "loss": 1.6367, - "step": 31300 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018941317729824798, - "loss": 1.6105, - "step": 31301 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018941283907190692, - "loss": 1.6267, - "step": 31302 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018941250084556586, - "loss": 1.5247, - "step": 31303 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018941216261922478, - "loss": 1.5807, - "step": 31304 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018941182439288372, - "loss": 1.5009, - "step": 31305 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018941148616654266, - "loss": 1.5903, - "step": 31306 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018941114794020157, - "loss": 1.6341, - "step": 31307 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894108097138605, - "loss": 1.5443, - "step": 31308 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018941047148751945, - "loss": 1.539, - "step": 31309 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018941013326117837, - "loss": 1.5386, - "step": 31310 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894097950348373, - "loss": 1.6657, - "step": 31311 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018940945680849625, - "loss": 1.5709, - "step": 31312 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894091185821552, - "loss": 1.4121, - "step": 31313 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018940878035581413, - "loss": 1.5884, - "step": 31314 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018940844212947307, - "loss": 1.4532, - "step": 31315 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189408103903132, - "loss": 1.5534, - "step": 31316 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018940776567679093, - "loss": 1.6582, - "step": 31317 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018940742745044987, - "loss": 1.6669, - "step": 31318 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018940708922410878, - "loss": 1.4355, - "step": 31319 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018940675099776773, - "loss": 1.5118, - "step": 31320 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018940641277142667, - "loss": 1.5502, - "step": 31321 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018940607454508558, - "loss": 1.4946, - "step": 31322 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018940573631874452, - "loss": 1.4861, - "step": 31323 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018940539809240346, - "loss": 1.5457, - "step": 31324 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018940505986606238, - "loss": 1.5541, - "step": 31325 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018940472163972132, - "loss": 1.5725, - "step": 31326 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018940438341338026, - "loss": 1.5243, - "step": 31327 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018940404518703917, - "loss": 1.549, - "step": 31328 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018940370696069811, - "loss": 1.5684, - "step": 31329 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018940336873435706, - "loss": 1.5192, - "step": 31330 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018940303050801597, - "loss": 1.5771, - "step": 31331 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894026922816749, - "loss": 1.6174, - "step": 31332 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018940235405533385, - "loss": 1.5528, - "step": 31333 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018940201582899277, - "loss": 1.5355, - "step": 31334 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894016776026517, - "loss": 1.541, - "step": 31335 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018940133937631065, - "loss": 1.3696, - "step": 31336 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018940100114996956, - "loss": 1.4902, - "step": 31337 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001894006629236285, - "loss": 1.4105, - "step": 31338 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018940032469728744, - "loss": 1.6221, - "step": 31339 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018939998647094636, - "loss": 1.3864, - "step": 31340 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893996482446053, - "loss": 1.6214, - "step": 31341 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018939931001826424, - "loss": 1.5417, - "step": 31342 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018939897179192316, - "loss": 1.6357, - "step": 31343 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893986335655821, - "loss": 1.5791, - "step": 31344 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018939829533924104, - "loss": 1.6141, - "step": 31345 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018939795711289995, - "loss": 1.4236, - "step": 31346 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893976188865589, - "loss": 1.5466, - "step": 31347 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018939728066021783, - "loss": 1.4488, - "step": 31348 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018939694243387675, - "loss": 1.5004, - "step": 31349 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893966042075357, - "loss": 1.3921, - "step": 31350 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018939626598119463, - "loss": 1.6831, - "step": 31351 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018939592775485354, - "loss": 1.5863, - "step": 31352 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018939558952851249, - "loss": 1.5693, - "step": 31353 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018939525130217143, - "loss": 1.5439, - "step": 31354 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018939491307583034, - "loss": 1.472, - "step": 31355 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018939457484948928, - "loss": 1.438, - "step": 31356 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018939423662314822, - "loss": 1.4744, - "step": 31357 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018939389839680714, - "loss": 1.5264, - "step": 31358 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018939356017046608, - "loss": 1.4469, - "step": 31359 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018939322194412502, - "loss": 1.5136, - "step": 31360 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018939288371778393, - "loss": 1.5695, - "step": 31361 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018939254549144287, - "loss": 1.4231, - "step": 31362 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018939220726510182, - "loss": 1.6724, - "step": 31363 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018939186903876073, - "loss": 1.3467, - "step": 31364 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018939153081241967, - "loss": 1.5744, - "step": 31365 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893911925860786, - "loss": 1.357, - "step": 31366 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018939085435973753, - "loss": 1.525, - "step": 31367 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018939051613339647, - "loss": 1.5475, - "step": 31368 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018939017790705544, - "loss": 1.6312, - "step": 31369 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018938983968071435, - "loss": 1.4838, - "step": 31370 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893895014543733, - "loss": 1.506, - "step": 31371 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018938916322803223, - "loss": 1.3997, - "step": 31372 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018938882500169115, - "loss": 1.4703, - "step": 31373 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893884867753501, - "loss": 1.648, - "step": 31374 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018938814854900903, - "loss": 1.6486, - "step": 31375 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018938781032266794, - "loss": 1.5468, - "step": 31376 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018938747209632688, - "loss": 1.4493, - "step": 31377 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018938713386998582, - "loss": 1.535, - "step": 31378 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018938679564364474, - "loss": 1.6168, - "step": 31379 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018938645741730368, - "loss": 1.4595, - "step": 31380 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018938611919096262, - "loss": 1.5469, - "step": 31381 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018938578096462153, - "loss": 1.7072, - "step": 31382 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018938544273828048, - "loss": 1.5306, - "step": 31383 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018938510451193942, - "loss": 1.6038, - "step": 31384 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018938476628559833, - "loss": 1.5212, - "step": 31385 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018938442805925727, - "loss": 1.5877, - "step": 31386 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893840898329162, - "loss": 1.5767, - "step": 31387 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018938375160657513, - "loss": 1.762, - "step": 31388 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018938341338023407, - "loss": 1.6446, - "step": 31389 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189383075153893, - "loss": 1.5389, - "step": 31390 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018938273692755192, - "loss": 1.5774, - "step": 31391 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018938239870121086, - "loss": 1.5994, - "step": 31392 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893820604748698, - "loss": 1.3376, - "step": 31393 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018938172224852872, - "loss": 1.6563, - "step": 31394 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018938138402218766, - "loss": 1.5415, - "step": 31395 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893810457958466, - "loss": 1.5342, - "step": 31396 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018938070756950552, - "loss": 1.5866, - "step": 31397 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018938036934316446, - "loss": 1.5133, - "step": 31398 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893800311168234, - "loss": 1.6156, - "step": 31399 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893796928904823, - "loss": 1.6943, - "step": 31400 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018937935466414125, - "loss": 1.5425, - "step": 31401 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893790164378002, - "loss": 1.4239, - "step": 31402 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893786782114591, - "loss": 1.5114, - "step": 31403 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018937833998511805, - "loss": 1.7005, - "step": 31404 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189378001758777, - "loss": 1.4572, - "step": 31405 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893776635324359, - "loss": 1.4686, - "step": 31406 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018937732530609485, - "loss": 1.3482, - "step": 31407 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893769870797538, - "loss": 1.4609, - "step": 31408 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893766488534127, - "loss": 1.6399, - "step": 31409 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018937631062707164, - "loss": 1.4878, - "step": 31410 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018937597240073058, - "loss": 1.3784, - "step": 31411 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893756341743895, - "loss": 1.6565, - "step": 31412 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018937529594804844, - "loss": 1.6212, - "step": 31413 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018937495772170738, - "loss": 1.4016, - "step": 31414 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893746194953663, - "loss": 1.7107, - "step": 31415 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018937428126902524, - "loss": 1.4881, - "step": 31416 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018937394304268418, - "loss": 1.7024, - "step": 31417 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893736048163431, - "loss": 1.6747, - "step": 31418 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018937326659000203, - "loss": 1.4369, - "step": 31419 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018937292836366097, - "loss": 1.4898, - "step": 31420 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893725901373199, - "loss": 1.5087, - "step": 31421 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018937225191097883, - "loss": 1.5719, - "step": 31422 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018937191368463777, - "loss": 1.5083, - "step": 31423 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018937157545829668, - "loss": 1.6005, - "step": 31424 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018937123723195562, - "loss": 1.629, - "step": 31425 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018937089900561457, - "loss": 1.3597, - "step": 31426 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893705607792735, - "loss": 1.5766, - "step": 31427 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018937022255293245, - "loss": 1.4896, - "step": 31428 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018936988432659136, - "loss": 1.5575, - "step": 31429 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893695461002503, - "loss": 1.6764, - "step": 31430 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018936920787390924, - "loss": 1.471, - "step": 31431 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018936886964756816, - "loss": 1.553, - "step": 31432 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893685314212271, - "loss": 1.5209, - "step": 31433 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018936819319488604, - "loss": 1.5479, - "step": 31434 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018936785496854495, - "loss": 1.6233, - "step": 31435 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893675167422039, - "loss": 1.5708, - "step": 31436 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018936717851586284, - "loss": 1.4246, - "step": 31437 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018936684028952175, - "loss": 1.5831, - "step": 31438 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893665020631807, - "loss": 1.5843, - "step": 31439 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018936616383683963, - "loss": 1.3957, - "step": 31440 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018936582561049855, - "loss": 1.5486, - "step": 31441 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893654873841575, - "loss": 1.6186, - "step": 31442 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018936514915781643, - "loss": 1.6495, - "step": 31443 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018936481093147534, - "loss": 1.3966, - "step": 31444 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018936447270513428, - "loss": 1.6617, - "step": 31445 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018936413447879323, - "loss": 1.5214, - "step": 31446 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018936379625245214, - "loss": 1.457, - "step": 31447 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018936345802611108, - "loss": 1.5809, - "step": 31448 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018936311979977002, - "loss": 1.5182, - "step": 31449 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018936278157342894, - "loss": 1.4722, - "step": 31450 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018936244334708788, - "loss": 1.5529, - "step": 31451 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018936210512074682, - "loss": 1.6193, - "step": 31452 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018936176689440573, - "loss": 1.5681, - "step": 31453 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018936142866806467, - "loss": 1.4982, - "step": 31454 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018936109044172361, - "loss": 1.4807, - "step": 31455 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018936075221538256, - "loss": 1.312, - "step": 31456 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018936041398904147, - "loss": 1.6004, - "step": 31457 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893600757627004, - "loss": 1.5359, - "step": 31458 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018935973753635935, - "loss": 1.5806, - "step": 31459 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018935939931001827, - "loss": 1.4512, - "step": 31460 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893590610836772, - "loss": 1.3014, - "step": 31461 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018935872285733615, - "loss": 1.6025, - "step": 31462 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018935838463099506, - "loss": 1.5704, - "step": 31463 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189358046404654, - "loss": 1.5191, - "step": 31464 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018935770817831294, - "loss": 1.4657, - "step": 31465 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018935736995197186, - "loss": 1.5936, - "step": 31466 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893570317256308, - "loss": 1.6525, - "step": 31467 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018935669349928974, - "loss": 1.4728, - "step": 31468 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018935635527294865, - "loss": 1.4636, - "step": 31469 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893560170466076, - "loss": 1.5619, - "step": 31470 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018935567882026654, - "loss": 1.4819, - "step": 31471 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018935534059392545, - "loss": 1.5266, - "step": 31472 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893550023675844, - "loss": 1.5928, - "step": 31473 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018935466414124333, - "loss": 1.6455, - "step": 31474 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018935432591490225, - "loss": 1.5844, - "step": 31475 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893539876885612, - "loss": 1.5776, - "step": 31476 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018935364946222013, - "loss": 1.7622, - "step": 31477 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018935331123587904, - "loss": 1.5795, - "step": 31478 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018935297300953798, - "loss": 1.6512, - "step": 31479 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018935263478319693, - "loss": 1.4852, - "step": 31480 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018935229655685584, - "loss": 1.5975, - "step": 31481 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893519583305148, - "loss": 1.4651, - "step": 31482 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018935162010417372, - "loss": 1.321, - "step": 31483 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018935128187783266, - "loss": 1.4907, - "step": 31484 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893509436514916, - "loss": 1.461, - "step": 31485 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018935060542515052, - "loss": 1.6406, - "step": 31486 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018935026719880946, - "loss": 1.6002, - "step": 31487 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893499289724684, - "loss": 1.4951, - "step": 31488 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018934959074612731, - "loss": 1.5806, - "step": 31489 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018934925251978626, - "loss": 1.5388, - "step": 31490 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893489142934452, - "loss": 1.384, - "step": 31491 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893485760671041, - "loss": 1.4842, - "step": 31492 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018934823784076305, - "loss": 1.5871, - "step": 31493 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189347899614422, - "loss": 1.4385, - "step": 31494 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893475613880809, - "loss": 1.5173, - "step": 31495 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018934722316173985, - "loss": 1.5416, - "step": 31496 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893468849353988, - "loss": 1.5378, - "step": 31497 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893465467090577, - "loss": 1.481, - "step": 31498 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018934620848271664, - "loss": 1.5427, - "step": 31499 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018934587025637559, - "loss": 1.4385, - "step": 31500 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893455320300345, - "loss": 1.5354, - "step": 31501 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018934519380369344, - "loss": 1.4945, - "step": 31502 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018934485557735238, - "loss": 1.387, - "step": 31503 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893445173510113, - "loss": 1.6772, - "step": 31504 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018934417912467024, - "loss": 1.6268, - "step": 31505 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018934384089832918, - "loss": 1.5026, - "step": 31506 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893435026719881, - "loss": 1.5244, - "step": 31507 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018934316444564703, - "loss": 1.4204, - "step": 31508 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018934282621930597, - "loss": 1.6492, - "step": 31509 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893424879929649, - "loss": 1.5337, - "step": 31510 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018934214976662383, - "loss": 1.4883, - "step": 31511 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018934181154028277, - "loss": 1.6059, - "step": 31512 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018934147331394169, - "loss": 1.5072, - "step": 31513 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018934113508760063, - "loss": 1.6099, - "step": 31514 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018934079686125957, - "loss": 1.5235, - "step": 31515 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018934045863491848, - "loss": 1.4956, - "step": 31516 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018934012040857742, - "loss": 1.6134, - "step": 31517 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018933978218223636, - "loss": 1.5378, - "step": 31518 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018933944395589528, - "loss": 1.4523, - "step": 31519 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018933910572955422, - "loss": 1.5872, - "step": 31520 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018933876750321316, - "loss": 1.4572, - "step": 31521 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018933842927687207, - "loss": 1.7253, - "step": 31522 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018933809105053102, - "loss": 1.408, - "step": 31523 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018933775282418996, - "loss": 1.536, - "step": 31524 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018933741459784887, - "loss": 1.611, - "step": 31525 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893370763715078, - "loss": 1.521, - "step": 31526 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018933673814516675, - "loss": 1.3925, - "step": 31527 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018933639991882567, - "loss": 1.4999, - "step": 31528 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893360616924846, - "loss": 1.6229, - "step": 31529 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018933572346614355, - "loss": 1.52, - "step": 31530 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018933538523980246, - "loss": 1.5398, - "step": 31531 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893350470134614, - "loss": 1.5649, - "step": 31532 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018933470878712035, - "loss": 1.7779, - "step": 31533 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018933437056077926, - "loss": 1.3937, - "step": 31534 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893340323344382, - "loss": 1.5686, - "step": 31535 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018933369410809714, - "loss": 1.5625, - "step": 31536 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018933335588175606, - "loss": 1.552, - "step": 31537 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018933301765541502, - "loss": 1.3401, - "step": 31538 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018933267942907397, - "loss": 1.453, - "step": 31539 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018933234120273288, - "loss": 1.5669, - "step": 31540 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018933200297639182, - "loss": 1.5235, - "step": 31541 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018933166475005076, - "loss": 1.5595, - "step": 31542 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018933132652370968, - "loss": 1.5411, - "step": 31543 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018933098829736862, - "loss": 1.6365, - "step": 31544 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018933065007102756, - "loss": 1.581, - "step": 31545 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018933031184468647, - "loss": 1.6032, - "step": 31546 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893299736183454, - "loss": 1.579, - "step": 31547 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018932963539200435, - "loss": 1.6051, - "step": 31548 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018932929716566327, - "loss": 1.5275, - "step": 31549 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893289589393222, - "loss": 1.5274, - "step": 31550 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018932862071298115, - "loss": 1.5789, - "step": 31551 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018932828248664006, - "loss": 1.5058, - "step": 31552 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189327944260299, - "loss": 1.5974, - "step": 31553 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018932760603395795, - "loss": 1.494, - "step": 31554 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018932726780761686, - "loss": 1.5046, - "step": 31555 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893269295812758, - "loss": 1.7562, - "step": 31556 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018932659135493474, - "loss": 1.4966, - "step": 31557 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018932625312859366, - "loss": 1.5526, - "step": 31558 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893259149022526, - "loss": 1.5145, - "step": 31559 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018932557667591154, - "loss": 1.6482, - "step": 31560 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018932523844957045, - "loss": 1.6081, - "step": 31561 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893249002232294, - "loss": 1.5466, - "step": 31562 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018932456199688834, - "loss": 1.5583, - "step": 31563 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018932422377054725, - "loss": 1.6503, - "step": 31564 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893238855442062, - "loss": 1.4587, - "step": 31565 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018932354731786513, - "loss": 1.5657, - "step": 31566 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018932320909152405, - "loss": 1.6236, - "step": 31567 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189322870865183, - "loss": 1.4161, - "step": 31568 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018932253263884193, - "loss": 1.6018, - "step": 31569 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018932219441250084, - "loss": 1.5017, - "step": 31570 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018932185618615978, - "loss": 1.4837, - "step": 31571 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018932151795981872, - "loss": 1.351, - "step": 31572 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018932117973347764, - "loss": 1.6249, - "step": 31573 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018932084150713658, - "loss": 1.5076, - "step": 31574 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018932050328079552, - "loss": 1.6174, - "step": 31575 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018932016505445443, - "loss": 1.476, - "step": 31576 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018931982682811338, - "loss": 1.5757, - "step": 31577 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018931948860177232, - "loss": 1.5191, - "step": 31578 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018931915037543123, - "loss": 1.3886, - "step": 31579 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018931881214909017, - "loss": 1.6575, - "step": 31580 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018931847392274911, - "loss": 1.7353, - "step": 31581 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018931813569640803, - "loss": 1.4783, - "step": 31582 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018931779747006697, - "loss": 1.3744, - "step": 31583 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893174592437259, - "loss": 1.488, - "step": 31584 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018931712101738482, - "loss": 1.5577, - "step": 31585 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018931678279104377, - "loss": 1.6091, - "step": 31586 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893164445647027, - "loss": 1.4787, - "step": 31587 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018931610633836162, - "loss": 1.5507, - "step": 31588 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018931576811202056, - "loss": 1.4433, - "step": 31589 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893154298856795, - "loss": 1.4492, - "step": 31590 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018931509165933842, - "loss": 1.5523, - "step": 31591 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018931475343299736, - "loss": 1.6818, - "step": 31592 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893144152066563, - "loss": 1.5113, - "step": 31593 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018931407698031524, - "loss": 1.2375, - "step": 31594 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018931373875397418, - "loss": 1.5023, - "step": 31595 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018931340052763312, - "loss": 1.5614, - "step": 31596 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018931306230129204, - "loss": 1.4819, - "step": 31597 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018931272407495098, - "loss": 1.4478, - "step": 31598 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018931238584860992, - "loss": 1.6323, - "step": 31599 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018931204762226883, - "loss": 1.5265, - "step": 31600 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018931170939592777, - "loss": 1.4909, - "step": 31601 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018931137116958671, - "loss": 1.5288, - "step": 31602 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018931103294324563, - "loss": 1.5745, - "step": 31603 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018931069471690457, - "loss": 1.4862, - "step": 31604 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893103564905635, - "loss": 1.5537, - "step": 31605 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018931001826422243, - "loss": 1.5637, - "step": 31606 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018930968003788137, - "loss": 1.5644, - "step": 31607 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893093418115403, - "loss": 1.5251, - "step": 31608 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018930900358519922, - "loss": 1.463, - "step": 31609 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018930866535885816, - "loss": 1.6307, - "step": 31610 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893083271325171, - "loss": 1.5267, - "step": 31611 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018930798890617602, - "loss": 1.5138, - "step": 31612 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018930765067983496, - "loss": 1.5159, - "step": 31613 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893073124534939, - "loss": 1.5801, - "step": 31614 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018930697422715281, - "loss": 1.5924, - "step": 31615 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018930663600081176, - "loss": 1.4631, - "step": 31616 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893062977744707, - "loss": 1.6328, - "step": 31617 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893059595481296, - "loss": 1.5992, - "step": 31618 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018930562132178855, - "loss": 1.702, - "step": 31619 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893052830954475, - "loss": 1.4714, - "step": 31620 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893049448691064, - "loss": 1.3613, - "step": 31621 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018930460664276535, - "loss": 1.6248, - "step": 31622 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893042684164243, - "loss": 1.502, - "step": 31623 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893039301900832, - "loss": 1.3896, - "step": 31624 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018930359196374214, - "loss": 1.3501, - "step": 31625 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018930325373740109, - "loss": 1.4705, - "step": 31626 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018930291551106, - "loss": 1.5021, - "step": 31627 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018930257728471894, - "loss": 1.5251, - "step": 31628 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018930223905837788, - "loss": 1.5189, - "step": 31629 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893019008320368, - "loss": 1.568, - "step": 31630 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018930156260569574, - "loss": 1.5647, - "step": 31631 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018930122437935468, - "loss": 1.6422, - "step": 31632 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001893008861530136, - "loss": 1.6536, - "step": 31633 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018930054792667253, - "loss": 1.5454, - "step": 31634 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018930020970033147, - "loss": 1.5491, - "step": 31635 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892998714739904, - "loss": 1.5631, - "step": 31636 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018929953324764933, - "loss": 1.6033, - "step": 31637 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018929919502130827, - "loss": 1.629, - "step": 31638 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018929885679496718, - "loss": 1.4902, - "step": 31639 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018929851856862613, - "loss": 1.3914, - "step": 31640 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018929818034228507, - "loss": 1.7076, - "step": 31641 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018929784211594398, - "loss": 1.5988, - "step": 31642 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018929750388960292, - "loss": 1.581, - "step": 31643 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018929716566326186, - "loss": 1.4633, - "step": 31644 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018929682743692078, - "loss": 1.7238, - "step": 31645 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018929648921057972, - "loss": 1.5571, - "step": 31646 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018929615098423866, - "loss": 1.7185, - "step": 31647 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018929581275789757, - "loss": 1.3684, - "step": 31648 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018929547453155651, - "loss": 1.4684, - "step": 31649 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018929513630521548, - "loss": 1.5153, - "step": 31650 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892947980788744, - "loss": 1.3987, - "step": 31651 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018929445985253334, - "loss": 1.5782, - "step": 31652 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018929412162619228, - "loss": 1.5632, - "step": 31653 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892937833998512, - "loss": 1.4916, - "step": 31654 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018929344517351013, - "loss": 1.7805, - "step": 31655 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018929310694716908, - "loss": 1.4308, - "step": 31656 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189292768720828, - "loss": 1.3593, - "step": 31657 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018929243049448693, - "loss": 1.7869, - "step": 31658 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018929209226814587, - "loss": 1.4527, - "step": 31659 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018929175404180479, - "loss": 1.4876, - "step": 31660 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018929141581546373, - "loss": 1.5981, - "step": 31661 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018929107758912267, - "loss": 1.4495, - "step": 31662 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018929073936278158, - "loss": 1.3618, - "step": 31663 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018929040113644052, - "loss": 1.4932, - "step": 31664 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018929006291009946, - "loss": 1.4799, - "step": 31665 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018928972468375838, - "loss": 1.4319, - "step": 31666 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018928938645741732, - "loss": 1.5961, - "step": 31667 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018928904823107626, - "loss": 1.6545, - "step": 31668 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018928871000473517, - "loss": 1.4775, - "step": 31669 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018928837177839412, - "loss": 1.4758, - "step": 31670 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018928803355205306, - "loss": 1.6292, - "step": 31671 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018928769532571197, - "loss": 1.5719, - "step": 31672 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892873570993709, - "loss": 1.4549, - "step": 31673 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018928701887302985, - "loss": 1.3241, - "step": 31674 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018928668064668877, - "loss": 1.5291, - "step": 31675 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892863424203477, - "loss": 1.5134, - "step": 31676 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018928600419400665, - "loss": 1.4578, - "step": 31677 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018928566596766556, - "loss": 1.5203, - "step": 31678 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892853277413245, - "loss": 1.5008, - "step": 31679 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018928498951498345, - "loss": 1.406, - "step": 31680 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018928465128864236, - "loss": 1.5789, - "step": 31681 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892843130623013, - "loss": 1.5444, - "step": 31682 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018928397483596024, - "loss": 1.7159, - "step": 31683 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018928363660961916, - "loss": 1.4975, - "step": 31684 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892832983832781, - "loss": 1.4925, - "step": 31685 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018928296015693704, - "loss": 1.5751, - "step": 31686 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018928262193059595, - "loss": 1.6429, - "step": 31687 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892822837042549, - "loss": 1.5829, - "step": 31688 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018928194547791384, - "loss": 1.5547, - "step": 31689 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018928160725157275, - "loss": 1.5372, - "step": 31690 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892812690252317, - "loss": 1.5053, - "step": 31691 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018928093079889063, - "loss": 1.6812, - "step": 31692 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018928059257254955, - "loss": 1.5133, - "step": 31693 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892802543462085, - "loss": 1.537, - "step": 31694 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018927991611986743, - "loss": 1.552, - "step": 31695 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018927957789352634, - "loss": 1.5655, - "step": 31696 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018927923966718528, - "loss": 1.4326, - "step": 31697 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018927890144084422, - "loss": 1.567, - "step": 31698 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018927856321450314, - "loss": 1.5375, - "step": 31699 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018927822498816208, - "loss": 1.5169, - "step": 31700 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018927788676182102, - "loss": 1.5545, - "step": 31701 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018927754853547993, - "loss": 1.653, - "step": 31702 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018927721030913888, - "loss": 1.4542, - "step": 31703 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018927687208279782, - "loss": 1.5861, - "step": 31704 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018927653385645673, - "loss": 1.5929, - "step": 31705 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892761956301157, - "loss": 1.5438, - "step": 31706 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892758574037746, - "loss": 1.4619, - "step": 31707 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018927551917743355, - "loss": 1.5298, - "step": 31708 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892751809510925, - "loss": 1.6189, - "step": 31709 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892748427247514, - "loss": 1.532, - "step": 31710 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018927450449841035, - "loss": 1.6514, - "step": 31711 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892741662720693, - "loss": 1.3757, - "step": 31712 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892738280457282, - "loss": 1.6969, - "step": 31713 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018927348981938715, - "loss": 1.4338, - "step": 31714 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892731515930461, - "loss": 1.5704, - "step": 31715 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189272813366705, - "loss": 1.5454, - "step": 31716 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018927247514036394, - "loss": 1.3598, - "step": 31717 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018927213691402288, - "loss": 1.4803, - "step": 31718 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892717986876818, - "loss": 1.6574, - "step": 31719 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018927146046134074, - "loss": 1.5264, - "step": 31720 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018927112223499968, - "loss": 1.5731, - "step": 31721 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018927078400865862, - "loss": 1.4826, - "step": 31722 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018927044578231754, - "loss": 1.6502, - "step": 31723 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018927010755597648, - "loss": 1.5425, - "step": 31724 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018926976932963542, - "loss": 1.4307, - "step": 31725 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018926943110329433, - "loss": 1.5331, - "step": 31726 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018926909287695327, - "loss": 1.5358, - "step": 31727 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018926875465061221, - "loss": 1.5139, - "step": 31728 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018926841642427113, - "loss": 1.6796, - "step": 31729 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018926807819793007, - "loss": 1.4902, - "step": 31730 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189267739971589, - "loss": 1.3722, - "step": 31731 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018926740174524792, - "loss": 1.7022, - "step": 31732 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018926706351890687, - "loss": 1.7448, - "step": 31733 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892667252925658, - "loss": 1.4733, - "step": 31734 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018926638706622472, - "loss": 1.7082, - "step": 31735 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018926604883988366, - "loss": 1.4874, - "step": 31736 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892657106135426, - "loss": 1.739, - "step": 31737 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018926537238720152, - "loss": 1.4627, - "step": 31738 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018926503416086046, - "loss": 1.5332, - "step": 31739 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892646959345194, - "loss": 1.5238, - "step": 31740 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892643577081783, - "loss": 1.5564, - "step": 31741 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018926401948183725, - "loss": 1.6237, - "step": 31742 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892636812554962, - "loss": 1.4598, - "step": 31743 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892633430291551, - "loss": 1.5262, - "step": 31744 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018926300480281405, - "loss": 1.5843, - "step": 31745 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189262666576473, - "loss": 1.3954, - "step": 31746 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892623283501319, - "loss": 1.622, - "step": 31747 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018926199012379085, - "loss": 1.5665, - "step": 31748 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892616518974498, - "loss": 1.5585, - "step": 31749 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892613136711087, - "loss": 1.4473, - "step": 31750 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018926097544476764, - "loss": 1.4917, - "step": 31751 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018926063721842658, - "loss": 1.5859, - "step": 31752 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892602989920855, - "loss": 1.5585, - "step": 31753 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018925996076574444, - "loss": 1.6967, - "step": 31754 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018925962253940338, - "loss": 1.5448, - "step": 31755 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892592843130623, - "loss": 1.9165, - "step": 31756 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018925894608672124, - "loss": 1.4736, - "step": 31757 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018925860786038018, - "loss": 1.5823, - "step": 31758 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892582696340391, - "loss": 1.6008, - "step": 31759 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018925793140769803, - "loss": 1.5546, - "step": 31760 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018925759318135697, - "loss": 1.5843, - "step": 31761 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892572549550159, - "loss": 1.4437, - "step": 31762 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018925691672867486, - "loss": 1.6611, - "step": 31763 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018925657850233377, - "loss": 1.4438, - "step": 31764 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892562402759927, - "loss": 1.4702, - "step": 31765 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018925590204965165, - "loss": 1.7314, - "step": 31766 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018925556382331057, - "loss": 1.4071, - "step": 31767 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892552255969695, - "loss": 1.497, - "step": 31768 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018925488737062845, - "loss": 1.6196, - "step": 31769 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018925454914428736, - "loss": 1.5656, - "step": 31770 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892542109179463, - "loss": 1.5827, - "step": 31771 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018925387269160524, - "loss": 1.5363, - "step": 31772 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018925353446526416, - "loss": 1.6233, - "step": 31773 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892531962389231, - "loss": 1.3448, - "step": 31774 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018925285801258204, - "loss": 1.4542, - "step": 31775 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018925251978624096, - "loss": 1.5516, - "step": 31776 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892521815598999, - "loss": 1.573, - "step": 31777 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018925184333355884, - "loss": 1.5144, - "step": 31778 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018925150510721775, - "loss": 1.4616, - "step": 31779 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892511668808767, - "loss": 1.4277, - "step": 31780 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018925082865453563, - "loss": 1.4906, - "step": 31781 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018925049042819455, - "loss": 1.5382, - "step": 31782 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892501522018535, - "loss": 1.6317, - "step": 31783 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018924981397551243, - "loss": 1.5903, - "step": 31784 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018924947574917134, - "loss": 1.5119, - "step": 31785 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018924913752283029, - "loss": 1.5288, - "step": 31786 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018924879929648923, - "loss": 1.4337, - "step": 31787 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018924846107014814, - "loss": 1.6091, - "step": 31788 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018924812284380708, - "loss": 1.7148, - "step": 31789 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018924778461746602, - "loss": 1.474, - "step": 31790 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018924744639112494, - "loss": 1.4233, - "step": 31791 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018924710816478388, - "loss": 1.5052, - "step": 31792 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018924676993844282, - "loss": 1.6117, - "step": 31793 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018924643171210173, - "loss": 1.4372, - "step": 31794 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018924609348576067, - "loss": 1.6102, - "step": 31795 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018924575525941962, - "loss": 1.4812, - "step": 31796 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018924541703307853, - "loss": 1.5878, - "step": 31797 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018924507880673747, - "loss": 1.5595, - "step": 31798 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892447405803964, - "loss": 1.4877, - "step": 31799 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018924440235405533, - "loss": 1.5315, - "step": 31800 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018924406412771427, - "loss": 1.3564, - "step": 31801 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892437259013732, - "loss": 1.5834, - "step": 31802 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018924338767503212, - "loss": 1.7174, - "step": 31803 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018924304944869106, - "loss": 1.5305, - "step": 31804 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018924271122235, - "loss": 1.4578, - "step": 31805 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018924237299600895, - "loss": 1.613, - "step": 31806 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018924203476966786, - "loss": 1.6232, - "step": 31807 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892416965433268, - "loss": 1.705, - "step": 31808 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018924135831698574, - "loss": 1.4936, - "step": 31809 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018924102009064466, - "loss": 1.5883, - "step": 31810 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892406818643036, - "loss": 1.6097, - "step": 31811 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018924034363796254, - "loss": 1.5085, - "step": 31812 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018924000541162145, - "loss": 1.5864, - "step": 31813 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892396671852804, - "loss": 1.4192, - "step": 31814 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018923932895893933, - "loss": 1.5335, - "step": 31815 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018923899073259825, - "loss": 1.6831, - "step": 31816 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892386525062572, - "loss": 1.4012, - "step": 31817 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018923831427991613, - "loss": 1.6018, - "step": 31818 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018923797605357507, - "loss": 1.5396, - "step": 31819 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189237637827234, - "loss": 1.5169, - "step": 31820 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018923729960089293, - "loss": 1.4796, - "step": 31821 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018923696137455187, - "loss": 1.6278, - "step": 31822 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892366231482108, - "loss": 1.6598, - "step": 31823 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018923628492186972, - "loss": 1.6399, - "step": 31824 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018923594669552866, - "loss": 1.5752, - "step": 31825 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892356084691876, - "loss": 1.6288, - "step": 31826 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018923527024284652, - "loss": 1.5126, - "step": 31827 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018923493201650546, - "loss": 1.4785, - "step": 31828 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892345937901644, - "loss": 1.5928, - "step": 31829 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018923425556382332, - "loss": 1.5395, - "step": 31830 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018923391733748226, - "loss": 1.5501, - "step": 31831 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892335791111412, - "loss": 1.508, - "step": 31832 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892332408848001, - "loss": 1.506, - "step": 31833 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018923290265845905, - "loss": 1.3069, - "step": 31834 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189232564432118, - "loss": 1.4229, - "step": 31835 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892322262057769, - "loss": 1.5577, - "step": 31836 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018923188797943585, - "loss": 1.6369, - "step": 31837 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892315497530948, - "loss": 1.5667, - "step": 31838 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892312115267537, - "loss": 1.6996, - "step": 31839 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018923087330041265, - "loss": 1.5047, - "step": 31840 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892305350740716, - "loss": 1.6012, - "step": 31841 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892301968477305, - "loss": 1.5353, - "step": 31842 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018922985862138944, - "loss": 1.7464, - "step": 31843 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018922952039504838, - "loss": 1.45, - "step": 31844 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892291821687073, - "loss": 1.4414, - "step": 31845 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018922884394236624, - "loss": 1.4371, - "step": 31846 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018922850571602518, - "loss": 1.4416, - "step": 31847 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892281674896841, - "loss": 1.4673, - "step": 31848 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018922782926334304, - "loss": 1.5806, - "step": 31849 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018922749103700198, - "loss": 1.5728, - "step": 31850 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892271528106609, - "loss": 1.5889, - "step": 31851 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018922681458431983, - "loss": 1.4204, - "step": 31852 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018922647635797877, - "loss": 1.4477, - "step": 31853 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892261381316377, - "loss": 1.5132, - "step": 31854 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018922579990529663, - "loss": 1.5808, - "step": 31855 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018922546167895557, - "loss": 1.6598, - "step": 31856 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018922512345261448, - "loss": 1.675, - "step": 31857 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018922478522627342, - "loss": 1.5193, - "step": 31858 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018922444699993237, - "loss": 1.5719, - "step": 31859 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018922410877359128, - "loss": 1.588, - "step": 31860 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018922377054725022, - "loss": 1.5019, - "step": 31861 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018922343232090916, - "loss": 1.3959, - "step": 31862 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018922309409456808, - "loss": 1.641, - "step": 31863 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018922275586822702, - "loss": 1.6149, - "step": 31864 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018922241764188596, - "loss": 1.4446, - "step": 31865 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018922207941554487, - "loss": 1.494, - "step": 31866 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892217411892038, - "loss": 1.447, - "step": 31867 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018922140296286275, - "loss": 1.4255, - "step": 31868 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018922106473652167, - "loss": 1.5639, - "step": 31869 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892207265101806, - "loss": 1.5785, - "step": 31870 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018922038828383955, - "loss": 1.4878, - "step": 31871 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018922005005749846, - "loss": 1.3642, - "step": 31872 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892197118311574, - "loss": 1.5613, - "step": 31873 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018921937360481635, - "loss": 1.5933, - "step": 31874 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892190353784753, - "loss": 1.5825, - "step": 31875 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018921869715213423, - "loss": 1.5978, - "step": 31876 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018921835892579317, - "loss": 1.4177, - "step": 31877 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018921802069945208, - "loss": 1.6135, - "step": 31878 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018921768247311103, - "loss": 1.6227, - "step": 31879 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018921734424676997, - "loss": 1.4973, - "step": 31880 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018921700602042888, - "loss": 1.5785, - "step": 31881 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018921666779408782, - "loss": 1.5553, - "step": 31882 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018921632956774676, - "loss": 1.4302, - "step": 31883 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018921599134140568, - "loss": 1.6247, - "step": 31884 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018921565311506462, - "loss": 1.5649, - "step": 31885 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018921531488872356, - "loss": 1.4482, - "step": 31886 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018921497666238247, - "loss": 1.6556, - "step": 31887 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018921463843604141, - "loss": 1.4869, - "step": 31888 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018921430020970036, - "loss": 1.5328, - "step": 31889 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018921396198335927, - "loss": 1.481, - "step": 31890 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892136237570182, - "loss": 1.4499, - "step": 31891 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018921328553067715, - "loss": 1.5957, - "step": 31892 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018921294730433607, - "loss": 1.5449, - "step": 31893 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189212609077995, - "loss": 1.421, - "step": 31894 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018921227085165395, - "loss": 1.5271, - "step": 31895 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018921193262531286, - "loss": 1.4819, - "step": 31896 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892115943989718, - "loss": 1.4738, - "step": 31897 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018921125617263074, - "loss": 1.4948, - "step": 31898 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018921091794628966, - "loss": 1.4343, - "step": 31899 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892105797199486, - "loss": 1.4604, - "step": 31900 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018921024149360754, - "loss": 1.4724, - "step": 31901 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018920990326726645, - "loss": 1.4658, - "step": 31902 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892095650409254, - "loss": 1.5632, - "step": 31903 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018920922681458434, - "loss": 1.5853, - "step": 31904 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018920888858824325, - "loss": 1.4746, - "step": 31905 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892085503619022, - "loss": 1.5462, - "step": 31906 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018920821213556113, - "loss": 1.538, - "step": 31907 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018920787390922005, - "loss": 1.397, - "step": 31908 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189207535682879, - "loss": 1.6426, - "step": 31909 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018920719745653793, - "loss": 1.592, - "step": 31910 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018920685923019684, - "loss": 1.6682, - "step": 31911 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018920652100385578, - "loss": 1.7261, - "step": 31912 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018920618277751473, - "loss": 1.363, - "step": 31913 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018920584455117364, - "loss": 1.6095, - "step": 31914 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018920550632483258, - "loss": 1.537, - "step": 31915 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018920516809849152, - "loss": 1.4433, - "step": 31916 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018920482987215044, - "loss": 1.4404, - "step": 31917 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018920449164580938, - "loss": 1.3563, - "step": 31918 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018920415341946832, - "loss": 1.6191, - "step": 31919 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018920381519312723, - "loss": 1.4754, - "step": 31920 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018920347696678617, - "loss": 1.5041, - "step": 31921 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018920313874044511, - "loss": 1.3752, - "step": 31922 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018920280051410403, - "loss": 1.8334, - "step": 31923 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018920246228776297, - "loss": 1.5395, - "step": 31924 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892021240614219, - "loss": 1.4428, - "step": 31925 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018920178583508083, - "loss": 1.4658, - "step": 31926 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018920144760873977, - "loss": 1.5904, - "step": 31927 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001892011093823987, - "loss": 1.4172, - "step": 31928 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018920077115605762, - "loss": 1.5423, - "step": 31929 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018920043292971656, - "loss": 1.4514, - "step": 31930 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018920009470337553, - "loss": 1.6433, - "step": 31931 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018919975647703444, - "loss": 1.4581, - "step": 31932 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018919941825069339, - "loss": 1.5385, - "step": 31933 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018919908002435233, - "loss": 1.5624, - "step": 31934 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018919874179801124, - "loss": 1.5896, - "step": 31935 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018919840357167018, - "loss": 1.3554, - "step": 31936 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018919806534532912, - "loss": 1.5475, - "step": 31937 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018919772711898804, - "loss": 1.6344, - "step": 31938 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018919738889264698, - "loss": 1.6923, - "step": 31939 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018919705066630592, - "loss": 1.6492, - "step": 31940 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018919671243996483, - "loss": 1.5089, - "step": 31941 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018919637421362377, - "loss": 1.4832, - "step": 31942 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018919603598728272, - "loss": 1.5377, - "step": 31943 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018919569776094163, - "loss": 1.5747, - "step": 31944 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018919535953460057, - "loss": 1.5525, - "step": 31945 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891950213082595, - "loss": 1.4864, - "step": 31946 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018919468308191843, - "loss": 1.5742, - "step": 31947 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018919434485557737, - "loss": 1.5936, - "step": 31948 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891940066292363, - "loss": 1.4781, - "step": 31949 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018919366840289522, - "loss": 1.4832, - "step": 31950 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018919333017655416, - "loss": 1.6876, - "step": 31951 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891929919502131, - "loss": 1.6041, - "step": 31952 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018919265372387202, - "loss": 1.6831, - "step": 31953 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018919231549753096, - "loss": 1.428, - "step": 31954 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891919772711899, - "loss": 1.4787, - "step": 31955 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018919163904484882, - "loss": 1.6074, - "step": 31956 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018919130081850776, - "loss": 1.6222, - "step": 31957 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891909625921667, - "loss": 1.4473, - "step": 31958 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891906243658256, - "loss": 1.5335, - "step": 31959 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018919028613948455, - "loss": 1.5535, - "step": 31960 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891899479131435, - "loss": 1.3922, - "step": 31961 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891896096868024, - "loss": 1.5127, - "step": 31962 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018918927146046135, - "loss": 1.6306, - "step": 31963 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891889332341203, - "loss": 1.6138, - "step": 31964 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891885950077792, - "loss": 1.5552, - "step": 31965 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018918825678143815, - "loss": 1.5709, - "step": 31966 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891879185550971, - "loss": 1.5071, - "step": 31967 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189187580328756, - "loss": 1.4646, - "step": 31968 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018918724210241494, - "loss": 1.4079, - "step": 31969 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018918690387607388, - "loss": 1.6523, - "step": 31970 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891865656497328, - "loss": 1.3907, - "step": 31971 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018918622742339174, - "loss": 1.5612, - "step": 31972 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018918588919705068, - "loss": 1.5007, - "step": 31973 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891855509707096, - "loss": 1.5194, - "step": 31974 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018918521274436853, - "loss": 1.515, - "step": 31975 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018918487451802748, - "loss": 1.596, - "step": 31976 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891845362916864, - "loss": 1.417, - "step": 31977 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018918419806534533, - "loss": 1.3332, - "step": 31978 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018918385983900427, - "loss": 1.3897, - "step": 31979 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018918352161266319, - "loss": 1.6057, - "step": 31980 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018918318338632213, - "loss": 1.4617, - "step": 31981 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018918284515998107, - "loss": 1.5427, - "step": 31982 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018918250693363998, - "loss": 1.6437, - "step": 31983 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018918216870729892, - "loss": 1.5699, - "step": 31984 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018918183048095786, - "loss": 1.5672, - "step": 31985 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018918149225461678, - "loss": 1.574, - "step": 31986 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018918115402827575, - "loss": 1.5494, - "step": 31987 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018918081580193466, - "loss": 1.5353, - "step": 31988 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891804775755936, - "loss": 1.5727, - "step": 31989 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018918013934925254, - "loss": 1.7708, - "step": 31990 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018917980112291148, - "loss": 1.454, - "step": 31991 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891794628965704, - "loss": 1.5341, - "step": 31992 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018917912467022934, - "loss": 1.6185, - "step": 31993 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018917878644388828, - "loss": 1.4133, - "step": 31994 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891784482175472, - "loss": 1.7388, - "step": 31995 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018917810999120614, - "loss": 1.595, - "step": 31996 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018917777176486508, - "loss": 1.694, - "step": 31997 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189177433538524, - "loss": 1.4177, - "step": 31998 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018917709531218293, - "loss": 1.4455, - "step": 31999 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018917675708584187, - "loss": 1.503, - "step": 32000 - }, - { - "epoch": 0.11, - "eval_loss": 1.3671499399038463, - "mrc_em": 19.0, - "mrc_f1": 27.01688311688313, - "nli_acc": 0.18, - "step": 32000, - "sts_f1": 0.6292134831460675, - "ynat_f1": 0.7736295869967943 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891764188595008, - "loss": 1.4585, - "step": 32001 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018917608063315973, - "loss": 1.5863, - "step": 32002 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018917574240681867, - "loss": 1.587, - "step": 32003 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018917540418047758, - "loss": 1.5155, - "step": 32004 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018917506595413652, - "loss": 1.3883, - "step": 32005 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018917472772779547, - "loss": 1.5358, - "step": 32006 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018917438950145438, - "loss": 1.5618, - "step": 32007 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018917405127511332, - "loss": 1.4675, - "step": 32008 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018917371304877226, - "loss": 1.5223, - "step": 32009 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018917337482243118, - "loss": 1.4319, - "step": 32010 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018917303659609012, - "loss": 1.6935, - "step": 32011 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018917269836974906, - "loss": 1.5621, - "step": 32012 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018917236014340797, - "loss": 1.4504, - "step": 32013 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018917202191706691, - "loss": 1.613, - "step": 32014 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018917168369072585, - "loss": 1.3799, - "step": 32015 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018917134546438477, - "loss": 1.4389, - "step": 32016 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891710072380437, - "loss": 1.582, - "step": 32017 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018917066901170265, - "loss": 1.5564, - "step": 32018 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018917033078536157, - "loss": 1.6323, - "step": 32019 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891699925590205, - "loss": 1.4896, - "step": 32020 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018916965433267945, - "loss": 1.5817, - "step": 32021 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018916931610633836, - "loss": 1.3562, - "step": 32022 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891689778799973, - "loss": 1.5419, - "step": 32023 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018916863965365624, - "loss": 1.6247, - "step": 32024 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018916830142731516, - "loss": 1.4245, - "step": 32025 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891679632009741, - "loss": 1.6272, - "step": 32026 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018916762497463304, - "loss": 1.5976, - "step": 32027 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018916728674829195, - "loss": 1.6543, - "step": 32028 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891669485219509, - "loss": 1.5979, - "step": 32029 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018916661029560984, - "loss": 1.3808, - "step": 32030 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018916627206926875, - "loss": 1.5859, - "step": 32031 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891659338429277, - "loss": 1.3502, - "step": 32032 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018916559561658663, - "loss": 1.5568, - "step": 32033 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018916525739024555, - "loss": 1.7328, - "step": 32034 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891649191639045, - "loss": 1.5264, - "step": 32035 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018916458093756343, - "loss": 1.576, - "step": 32036 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018916424271122234, - "loss": 1.6025, - "step": 32037 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018916390448488128, - "loss": 1.5729, - "step": 32038 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018916356625854023, - "loss": 1.6306, - "step": 32039 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018916322803219914, - "loss": 1.5514, - "step": 32040 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018916288980585808, - "loss": 1.4373, - "step": 32041 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018916255157951702, - "loss": 1.5183, - "step": 32042 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018916221335317596, - "loss": 1.6292, - "step": 32043 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891618751268349, - "loss": 1.4934, - "step": 32044 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018916153690049382, - "loss": 1.5732, - "step": 32045 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018916119867415276, - "loss": 1.4758, - "step": 32046 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891608604478117, - "loss": 1.5434, - "step": 32047 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018916052222147061, - "loss": 1.6982, - "step": 32048 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018916018399512956, - "loss": 1.824, - "step": 32049 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891598457687885, - "loss": 1.6664, - "step": 32050 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891595075424474, - "loss": 1.5697, - "step": 32051 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018915916931610635, - "loss": 1.5097, - "step": 32052 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891588310897653, - "loss": 1.5835, - "step": 32053 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891584928634242, - "loss": 1.4707, - "step": 32054 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018915815463708315, - "loss": 1.5322, - "step": 32055 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891578164107421, - "loss": 1.4405, - "step": 32056 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189157478184401, - "loss": 1.5834, - "step": 32057 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018915713995805994, - "loss": 1.4561, - "step": 32058 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018915680173171889, - "loss": 1.5089, - "step": 32059 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891564635053778, - "loss": 1.5043, - "step": 32060 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018915612527903674, - "loss": 1.5424, - "step": 32061 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018915578705269568, - "loss": 1.4264, - "step": 32062 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891554488263546, - "loss": 1.53, - "step": 32063 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018915511060001354, - "loss": 1.7065, - "step": 32064 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018915477237367248, - "loss": 1.3556, - "step": 32065 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891544341473314, - "loss": 1.4405, - "step": 32066 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018915409592099033, - "loss": 1.582, - "step": 32067 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018915375769464927, - "loss": 1.5482, - "step": 32068 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891534194683082, - "loss": 1.4394, - "step": 32069 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018915308124196713, - "loss": 1.7414, - "step": 32070 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018915274301562607, - "loss": 1.5341, - "step": 32071 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018915240478928498, - "loss": 1.4924, - "step": 32072 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018915206656294393, - "loss": 1.3957, - "step": 32073 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018915172833660287, - "loss": 1.556, - "step": 32074 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891513901102618, - "loss": 1.4816, - "step": 32075 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018915105188392072, - "loss": 1.5363, - "step": 32076 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018915071365757966, - "loss": 1.4721, - "step": 32077 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891503754312386, - "loss": 1.4186, - "step": 32078 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018915003720489752, - "loss": 1.5836, - "step": 32079 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018914969897855646, - "loss": 1.5277, - "step": 32080 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891493607522154, - "loss": 1.5379, - "step": 32081 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018914902252587431, - "loss": 1.7284, - "step": 32082 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018914868429953326, - "loss": 1.5829, - "step": 32083 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891483460731922, - "loss": 1.5662, - "step": 32084 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891480078468511, - "loss": 1.4756, - "step": 32085 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018914766962051005, - "loss": 1.6023, - "step": 32086 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189147331394169, - "loss": 1.4799, - "step": 32087 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891469931678279, - "loss": 1.4082, - "step": 32088 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018914665494148685, - "loss": 1.3033, - "step": 32089 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891463167151458, - "loss": 1.521, - "step": 32090 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891459784888047, - "loss": 1.4799, - "step": 32091 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018914564026246364, - "loss": 1.5483, - "step": 32092 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018914530203612259, - "loss": 1.4608, - "step": 32093 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891449638097815, - "loss": 1.3771, - "step": 32094 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018914462558344044, - "loss": 1.4486, - "step": 32095 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018914428735709938, - "loss": 1.3438, - "step": 32096 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891439491307583, - "loss": 1.6226, - "step": 32097 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018914361090441724, - "loss": 1.3776, - "step": 32098 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018914327267807618, - "loss": 1.7147, - "step": 32099 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018914293445173512, - "loss": 1.5817, - "step": 32100 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018914259622539406, - "loss": 1.4982, - "step": 32101 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018914225799905297, - "loss": 1.5155, - "step": 32102 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018914191977271192, - "loss": 1.4772, - "step": 32103 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018914158154637086, - "loss": 1.5476, - "step": 32104 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018914124332002977, - "loss": 1.4707, - "step": 32105 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891409050936887, - "loss": 1.5125, - "step": 32106 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018914056686734765, - "loss": 1.6152, - "step": 32107 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018914022864100657, - "loss": 1.3785, - "step": 32108 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891398904146655, - "loss": 1.55, - "step": 32109 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018913955218832445, - "loss": 1.6745, - "step": 32110 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018913921396198336, - "loss": 1.6323, - "step": 32111 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891388757356423, - "loss": 1.4741, - "step": 32112 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018913853750930125, - "loss": 1.4352, - "step": 32113 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018913819928296016, - "loss": 1.631, - "step": 32114 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891378610566191, - "loss": 1.5877, - "step": 32115 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018913752283027804, - "loss": 1.5675, - "step": 32116 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018913718460393696, - "loss": 1.517, - "step": 32117 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891368463775959, - "loss": 1.5639, - "step": 32118 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018913650815125484, - "loss": 1.5223, - "step": 32119 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018913616992491375, - "loss": 1.5776, - "step": 32120 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891358316985727, - "loss": 1.4099, - "step": 32121 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018913549347223164, - "loss": 1.4803, - "step": 32122 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018913515524589055, - "loss": 1.5562, - "step": 32123 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891348170195495, - "loss": 1.579, - "step": 32124 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018913447879320843, - "loss": 1.5829, - "step": 32125 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018913414056686735, - "loss": 1.5796, - "step": 32126 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891338023405263, - "loss": 1.4923, - "step": 32127 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018913346411418523, - "loss": 1.6461, - "step": 32128 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018913312588784414, - "loss": 1.3965, - "step": 32129 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018913278766150308, - "loss": 1.2614, - "step": 32130 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018913244943516202, - "loss": 1.5304, - "step": 32131 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018913211120882094, - "loss": 1.7462, - "step": 32132 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018913177298247988, - "loss": 1.6086, - "step": 32133 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018913143475613882, - "loss": 1.4542, - "step": 32134 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018913109652979773, - "loss": 1.5947, - "step": 32135 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018913075830345668, - "loss": 1.6519, - "step": 32136 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018913042007711562, - "loss": 1.5452, - "step": 32137 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018913008185077453, - "loss": 1.716, - "step": 32138 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018912974362443347, - "loss": 1.5166, - "step": 32139 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891294053980924, - "loss": 1.7306, - "step": 32140 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018912906717175133, - "loss": 1.5207, - "step": 32141 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018912872894541027, - "loss": 1.5304, - "step": 32142 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891283907190692, - "loss": 1.4866, - "step": 32143 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018912805249272812, - "loss": 1.6279, - "step": 32144 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018912771426638706, - "loss": 1.686, - "step": 32145 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189127376040046, - "loss": 1.4793, - "step": 32146 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018912703781370492, - "loss": 1.5777, - "step": 32147 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018912669958736386, - "loss": 1.4328, - "step": 32148 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891263613610228, - "loss": 1.5551, - "step": 32149 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018912602313468172, - "loss": 1.5597, - "step": 32150 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018912568490834066, - "loss": 1.5507, - "step": 32151 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891253466819996, - "loss": 1.6335, - "step": 32152 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891250084556585, - "loss": 1.6847, - "step": 32153 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018912467022931745, - "loss": 1.5935, - "step": 32154 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891243320029764, - "loss": 1.5885, - "step": 32155 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018912399377663534, - "loss": 1.7383, - "step": 32156 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018912365555029428, - "loss": 1.5693, - "step": 32157 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018912331732395322, - "loss": 1.7545, - "step": 32158 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018912297909761213, - "loss": 1.5161, - "step": 32159 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018912264087127107, - "loss": 1.5956, - "step": 32160 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018912230264493001, - "loss": 1.6063, - "step": 32161 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018912196441858893, - "loss": 1.5658, - "step": 32162 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018912162619224787, - "loss": 1.6476, - "step": 32163 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891212879659068, - "loss": 1.643, - "step": 32164 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018912094973956572, - "loss": 1.6536, - "step": 32165 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018912061151322467, - "loss": 1.5227, - "step": 32166 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891202732868836, - "loss": 1.6922, - "step": 32167 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018911993506054252, - "loss": 1.5163, - "step": 32168 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018911959683420146, - "loss": 1.5758, - "step": 32169 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891192586078604, - "loss": 1.4911, - "step": 32170 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018911892038151932, - "loss": 1.5322, - "step": 32171 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018911858215517826, - "loss": 1.617, - "step": 32172 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891182439288372, - "loss": 1.4885, - "step": 32173 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018911790570249611, - "loss": 1.5662, - "step": 32174 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018911756747615505, - "loss": 1.5555, - "step": 32175 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189117229249814, - "loss": 1.5463, - "step": 32176 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891168910234729, - "loss": 1.6838, - "step": 32177 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018911655279713185, - "loss": 1.6138, - "step": 32178 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891162145707908, - "loss": 1.5231, - "step": 32179 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891158763444497, - "loss": 1.5735, - "step": 32180 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018911553811810865, - "loss": 1.5358, - "step": 32181 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891151998917676, - "loss": 1.555, - "step": 32182 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891148616654265, - "loss": 1.48, - "step": 32183 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018911452343908544, - "loss": 1.5004, - "step": 32184 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018911418521274438, - "loss": 1.651, - "step": 32185 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891138469864033, - "loss": 1.3978, - "step": 32186 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018911350876006224, - "loss": 1.5302, - "step": 32187 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018911317053372118, - "loss": 1.6089, - "step": 32188 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891128323073801, - "loss": 1.4527, - "step": 32189 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018911249408103904, - "loss": 1.6157, - "step": 32190 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018911215585469798, - "loss": 1.5373, - "step": 32191 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891118176283569, - "loss": 1.6953, - "step": 32192 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018911147940201583, - "loss": 1.5051, - "step": 32193 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018911114117567477, - "loss": 1.443, - "step": 32194 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891108029493337, - "loss": 1.7463, - "step": 32195 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018911046472299263, - "loss": 1.514, - "step": 32196 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018911012649665157, - "loss": 1.3287, - "step": 32197 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018910978827031048, - "loss": 1.487, - "step": 32198 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018910945004396943, - "loss": 1.5116, - "step": 32199 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018910911181762837, - "loss": 1.6004, - "step": 32200 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018910877359128728, - "loss": 1.4158, - "step": 32201 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018910843536494622, - "loss": 1.5267, - "step": 32202 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018910809713860516, - "loss": 1.5233, - "step": 32203 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018910775891226408, - "loss": 1.5906, - "step": 32204 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018910742068592302, - "loss": 1.7295, - "step": 32205 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018910708245958196, - "loss": 1.4544, - "step": 32206 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018910674423324087, - "loss": 1.4657, - "step": 32207 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018910640600689981, - "loss": 1.5568, - "step": 32208 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018910606778055876, - "loss": 1.5529, - "step": 32209 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018910572955421767, - "loss": 1.5035, - "step": 32210 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891053913278766, - "loss": 1.5919, - "step": 32211 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018910505310153558, - "loss": 1.5775, - "step": 32212 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891047148751945, - "loss": 1.6856, - "step": 32213 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018910437664885343, - "loss": 1.5248, - "step": 32214 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018910403842251238, - "loss": 1.4604, - "step": 32215 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001891037001961713, - "loss": 1.4571, - "step": 32216 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018910336196983023, - "loss": 1.395, - "step": 32217 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018910302374348917, - "loss": 1.6022, - "step": 32218 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018910268551714809, - "loss": 1.5668, - "step": 32219 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018910234729080703, - "loss": 1.5302, - "step": 32220 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018910200906446597, - "loss": 1.3272, - "step": 32221 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018910167083812488, - "loss": 1.4897, - "step": 32222 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018910133261178382, - "loss": 1.4681, - "step": 32223 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018910099438544276, - "loss": 1.5384, - "step": 32224 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018910065615910168, - "loss": 1.5322, - "step": 32225 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018910031793276062, - "loss": 1.5509, - "step": 32226 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018909997970641956, - "loss": 1.5819, - "step": 32227 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018909964148007847, - "loss": 1.6641, - "step": 32228 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018909930325373742, - "loss": 1.5999, - "step": 32229 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018909896502739636, - "loss": 1.5739, - "step": 32230 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018909862680105527, - "loss": 1.5275, - "step": 32231 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890982885747142, - "loss": 1.5649, - "step": 32232 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018909795034837315, - "loss": 1.5704, - "step": 32233 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018909761212203207, - "loss": 1.449, - "step": 32234 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189097273895691, - "loss": 1.5431, - "step": 32235 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018909693566934995, - "loss": 1.4867, - "step": 32236 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018909659744300886, - "loss": 1.4791, - "step": 32237 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890962592166678, - "loss": 1.6077, - "step": 32238 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018909592099032675, - "loss": 1.6066, - "step": 32239 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018909558276398566, - "loss": 1.4112, - "step": 32240 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890952445376446, - "loss": 1.5126, - "step": 32241 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018909490631130354, - "loss": 1.5039, - "step": 32242 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018909456808496246, - "loss": 1.5158, - "step": 32243 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890942298586214, - "loss": 1.6481, - "step": 32244 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018909389163228034, - "loss": 1.5185, - "step": 32245 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018909355340593925, - "loss": 1.447, - "step": 32246 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890932151795982, - "loss": 1.4638, - "step": 32247 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018909287695325713, - "loss": 1.6958, - "step": 32248 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018909253872691605, - "loss": 1.5155, - "step": 32249 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189092200500575, - "loss": 1.5083, - "step": 32250 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018909186227423393, - "loss": 1.5551, - "step": 32251 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018909152404789284, - "loss": 1.5106, - "step": 32252 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018909118582155179, - "loss": 1.5634, - "step": 32253 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018909084759521073, - "loss": 1.6376, - "step": 32254 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018909050936886964, - "loss": 1.5963, - "step": 32255 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018909017114252858, - "loss": 1.5499, - "step": 32256 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018908983291618752, - "loss": 1.4967, - "step": 32257 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018908949468984644, - "loss": 1.485, - "step": 32258 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018908915646350538, - "loss": 1.5405, - "step": 32259 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018908881823716432, - "loss": 1.7496, - "step": 32260 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018908848001082323, - "loss": 1.4002, - "step": 32261 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018908814178448217, - "loss": 1.6917, - "step": 32262 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018908780355814112, - "loss": 1.6311, - "step": 32263 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018908746533180003, - "loss": 1.5634, - "step": 32264 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018908712710545897, - "loss": 1.5189, - "step": 32265 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890867888791179, - "loss": 1.6194, - "step": 32266 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018908645065277683, - "loss": 1.5654, - "step": 32267 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890861124264358, - "loss": 1.4868, - "step": 32268 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018908577420009474, - "loss": 1.6734, - "step": 32269 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018908543597375365, - "loss": 1.5444, - "step": 32270 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890850977474126, - "loss": 1.4921, - "step": 32271 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018908475952107153, - "loss": 1.6735, - "step": 32272 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018908442129473045, - "loss": 1.627, - "step": 32273 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890840830683894, - "loss": 1.553, - "step": 32274 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018908374484204833, - "loss": 1.5411, - "step": 32275 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018908340661570724, - "loss": 1.3459, - "step": 32276 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018908306838936618, - "loss": 1.4525, - "step": 32277 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018908273016302512, - "loss": 1.4971, - "step": 32278 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018908239193668404, - "loss": 1.4122, - "step": 32279 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018908205371034298, - "loss": 1.5692, - "step": 32280 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018908171548400192, - "loss": 1.5688, - "step": 32281 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018908137725766084, - "loss": 1.5436, - "step": 32282 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018908103903131978, - "loss": 1.6877, - "step": 32283 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018908070080497872, - "loss": 1.5152, - "step": 32284 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018908036257863763, - "loss": 1.4947, - "step": 32285 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018908002435229657, - "loss": 1.4857, - "step": 32286 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018907968612595551, - "loss": 1.5528, - "step": 32287 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018907934789961443, - "loss": 1.4684, - "step": 32288 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018907900967327337, - "loss": 1.5528, - "step": 32289 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890786714469323, - "loss": 1.4486, - "step": 32290 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018907833322059122, - "loss": 1.4224, - "step": 32291 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018907799499425017, - "loss": 1.6486, - "step": 32292 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890776567679091, - "loss": 1.7366, - "step": 32293 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018907731854156802, - "loss": 1.2254, - "step": 32294 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018907698031522696, - "loss": 1.5528, - "step": 32295 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890766420888859, - "loss": 1.5659, - "step": 32296 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018907630386254482, - "loss": 1.515, - "step": 32297 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018907596563620376, - "loss": 1.5786, - "step": 32298 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890756274098627, - "loss": 1.5456, - "step": 32299 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890752891835216, - "loss": 1.4011, - "step": 32300 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018907495095718055, - "loss": 1.5317, - "step": 32301 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890746127308395, - "loss": 1.5544, - "step": 32302 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890742745044984, - "loss": 1.5587, - "step": 32303 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018907393627815735, - "loss": 1.4493, - "step": 32304 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890735980518163, - "loss": 1.5979, - "step": 32305 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890732598254752, - "loss": 1.5659, - "step": 32306 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018907292159913415, - "loss": 1.6828, - "step": 32307 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890725833727931, - "loss": 1.5025, - "step": 32308 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189072245146452, - "loss": 1.5949, - "step": 32309 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018907190692011094, - "loss": 1.5841, - "step": 32310 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018907156869376988, - "loss": 1.5587, - "step": 32311 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890712304674288, - "loss": 1.6187, - "step": 32312 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018907089224108774, - "loss": 1.5158, - "step": 32313 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018907055401474668, - "loss": 1.6392, - "step": 32314 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890702157884056, - "loss": 1.4181, - "step": 32315 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018906987756206454, - "loss": 1.7613, - "step": 32316 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018906953933572348, - "loss": 1.5457, - "step": 32317 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890692011093824, - "loss": 1.5905, - "step": 32318 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018906886288304133, - "loss": 1.455, - "step": 32319 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018906852465670027, - "loss": 1.5046, - "step": 32320 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890681864303592, - "loss": 1.5744, - "step": 32321 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018906784820401813, - "loss": 1.6028, - "step": 32322 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018906750997767707, - "loss": 1.6976, - "step": 32323 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189067171751336, - "loss": 1.5462, - "step": 32324 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018906683352499495, - "loss": 1.4165, - "step": 32325 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018906649529865387, - "loss": 1.5299, - "step": 32326 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890661570723128, - "loss": 1.5999, - "step": 32327 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018906581884597175, - "loss": 1.755, - "step": 32328 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018906548061963066, - "loss": 1.5062, - "step": 32329 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890651423932896, - "loss": 1.3108, - "step": 32330 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018906480416694854, - "loss": 1.6234, - "step": 32331 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018906446594060746, - "loss": 1.6729, - "step": 32332 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890641277142664, - "loss": 1.6118, - "step": 32333 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018906378948792534, - "loss": 1.4681, - "step": 32334 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018906345126158425, - "loss": 1.5765, - "step": 32335 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890631130352432, - "loss": 1.5411, - "step": 32336 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018906277480890214, - "loss": 1.6357, - "step": 32337 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018906243658256105, - "loss": 1.4328, - "step": 32338 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018906209835622, - "loss": 1.5469, - "step": 32339 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018906176012987893, - "loss": 1.8024, - "step": 32340 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018906142190353787, - "loss": 1.4223, - "step": 32341 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890610836771968, - "loss": 1.385, - "step": 32342 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018906074545085573, - "loss": 1.3636, - "step": 32343 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018906040722451467, - "loss": 1.3848, - "step": 32344 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018906006899817358, - "loss": 1.4813, - "step": 32345 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018905973077183253, - "loss": 1.6273, - "step": 32346 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018905939254549147, - "loss": 1.5579, - "step": 32347 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018905905431915038, - "loss": 1.5473, - "step": 32348 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018905871609280932, - "loss": 1.4673, - "step": 32349 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018905837786646826, - "loss": 1.3586, - "step": 32350 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018905803964012718, - "loss": 1.6095, - "step": 32351 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018905770141378612, - "loss": 1.5013, - "step": 32352 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018905736318744506, - "loss": 1.6787, - "step": 32353 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018905702496110397, - "loss": 1.6132, - "step": 32354 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018905668673476291, - "loss": 1.5642, - "step": 32355 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018905634850842186, - "loss": 1.5079, - "step": 32356 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018905601028208077, - "loss": 1.6203, - "step": 32357 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890556720557397, - "loss": 1.5216, - "step": 32358 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018905533382939865, - "loss": 1.576, - "step": 32359 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018905499560305757, - "loss": 1.4085, - "step": 32360 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890546573767165, - "loss": 1.6134, - "step": 32361 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018905431915037545, - "loss": 1.6414, - "step": 32362 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018905398092403436, - "loss": 1.5316, - "step": 32363 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890536426976933, - "loss": 1.6037, - "step": 32364 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018905330447135224, - "loss": 1.4853, - "step": 32365 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018905296624501116, - "loss": 1.4762, - "step": 32366 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890526280186701, - "loss": 1.4985, - "step": 32367 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018905228979232904, - "loss": 1.5638, - "step": 32368 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018905195156598796, - "loss": 1.582, - "step": 32369 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890516133396469, - "loss": 1.5109, - "step": 32370 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018905127511330584, - "loss": 1.7093, - "step": 32371 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018905093688696475, - "loss": 1.5141, - "step": 32372 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890505986606237, - "loss": 1.5296, - "step": 32373 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018905026043428263, - "loss": 1.4532, - "step": 32374 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018904992220794155, - "loss": 1.4403, - "step": 32375 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890495839816005, - "loss": 1.51, - "step": 32376 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018904924575525943, - "loss": 1.6146, - "step": 32377 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018904890752891834, - "loss": 1.6001, - "step": 32378 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018904856930257729, - "loss": 1.7649, - "step": 32379 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018904823107623623, - "loss": 1.6438, - "step": 32380 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018904789284989517, - "loss": 1.4882, - "step": 32381 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890475546235541, - "loss": 1.6948, - "step": 32382 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018904721639721302, - "loss": 1.48, - "step": 32383 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018904687817087196, - "loss": 1.764, - "step": 32384 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890465399445309, - "loss": 1.555, - "step": 32385 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018904620171818982, - "loss": 1.3819, - "step": 32386 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018904586349184876, - "loss": 1.3214, - "step": 32387 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890455252655077, - "loss": 1.6504, - "step": 32388 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018904518703916662, - "loss": 1.6042, - "step": 32389 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018904484881282556, - "loss": 1.4733, - "step": 32390 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890445105864845, - "loss": 1.5931, - "step": 32391 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890441723601434, - "loss": 1.4551, - "step": 32392 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018904383413380235, - "loss": 1.5808, - "step": 32393 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890434959074613, - "loss": 1.7961, - "step": 32394 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890431576811202, - "loss": 1.4702, - "step": 32395 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018904281945477915, - "loss": 1.5833, - "step": 32396 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890424812284381, - "loss": 1.4976, - "step": 32397 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189042143002097, - "loss": 1.5585, - "step": 32398 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018904180477575595, - "loss": 1.4807, - "step": 32399 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890414665494149, - "loss": 1.5289, - "step": 32400 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890411283230738, - "loss": 1.4854, - "step": 32401 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018904079009673274, - "loss": 1.5431, - "step": 32402 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018904045187039168, - "loss": 1.4338, - "step": 32403 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890401136440506, - "loss": 1.5698, - "step": 32404 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018903977541770954, - "loss": 1.6079, - "step": 32405 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018903943719136848, - "loss": 1.5396, - "step": 32406 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890390989650274, - "loss": 1.4495, - "step": 32407 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018903876073868633, - "loss": 1.5598, - "step": 32408 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018903842251234528, - "loss": 1.7089, - "step": 32409 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890380842860042, - "loss": 1.4789, - "step": 32410 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018903774605966313, - "loss": 1.7171, - "step": 32411 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018903740783332207, - "loss": 1.4935, - "step": 32412 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018903706960698099, - "loss": 1.4501, - "step": 32413 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018903673138063993, - "loss": 1.557, - "step": 32414 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018903639315429887, - "loss": 1.6032, - "step": 32415 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018903605492795778, - "loss": 1.4906, - "step": 32416 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018903571670161672, - "loss": 1.597, - "step": 32417 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018903537847527566, - "loss": 1.4408, - "step": 32418 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018903504024893458, - "loss": 1.6024, - "step": 32419 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018903470202259352, - "loss": 1.5166, - "step": 32420 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018903436379625246, - "loss": 1.4742, - "step": 32421 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018903402556991137, - "loss": 1.6168, - "step": 32422 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018903368734357032, - "loss": 1.4789, - "step": 32423 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018903334911722926, - "loss": 1.5175, - "step": 32424 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018903301089088817, - "loss": 1.4984, - "step": 32425 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890326726645471, - "loss": 1.765, - "step": 32426 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018903233443820605, - "loss": 1.5005, - "step": 32427 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189031996211865, - "loss": 1.6223, - "step": 32428 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890316579855239, - "loss": 1.5935, - "step": 32429 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018903131975918285, - "loss": 1.5814, - "step": 32430 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890309815328418, - "loss": 1.4431, - "step": 32431 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890306433065007, - "loss": 1.5665, - "step": 32432 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018903030508015965, - "loss": 1.5011, - "step": 32433 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890299668538186, - "loss": 1.5063, - "step": 32434 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890296286274775, - "loss": 1.4574, - "step": 32435 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018902929040113644, - "loss": 1.5934, - "step": 32436 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018902895217479538, - "loss": 1.5174, - "step": 32437 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018902861394845432, - "loss": 1.3901, - "step": 32438 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018902827572211327, - "loss": 1.6667, - "step": 32439 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018902793749577218, - "loss": 1.4831, - "step": 32440 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018902759926943112, - "loss": 1.3969, - "step": 32441 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018902726104309006, - "loss": 1.5913, - "step": 32442 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018902692281674898, - "loss": 1.5456, - "step": 32443 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018902658459040792, - "loss": 1.6931, - "step": 32444 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018902624636406686, - "loss": 1.4449, - "step": 32445 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018902590813772577, - "loss": 1.4573, - "step": 32446 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018902556991138471, - "loss": 1.4692, - "step": 32447 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018902523168504365, - "loss": 1.6896, - "step": 32448 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018902489345870257, - "loss": 1.3987, - "step": 32449 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890245552323615, - "loss": 1.4965, - "step": 32450 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018902421700602045, - "loss": 1.5197, - "step": 32451 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018902387877967937, - "loss": 1.4973, - "step": 32452 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890235405533383, - "loss": 1.6951, - "step": 32453 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018902320232699725, - "loss": 1.3938, - "step": 32454 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018902286410065616, - "loss": 1.4531, - "step": 32455 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890225258743151, - "loss": 1.5505, - "step": 32456 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018902218764797404, - "loss": 1.6358, - "step": 32457 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018902184942163296, - "loss": 1.5445, - "step": 32458 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890215111952919, - "loss": 1.4785, - "step": 32459 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018902117296895084, - "loss": 1.5151, - "step": 32460 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018902083474260975, - "loss": 1.477, - "step": 32461 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890204965162687, - "loss": 1.6762, - "step": 32462 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018902015828992764, - "loss": 1.6129, - "step": 32463 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018901982006358655, - "loss": 1.3558, - "step": 32464 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890194818372455, - "loss": 1.7112, - "step": 32465 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018901914361090443, - "loss": 1.6276, - "step": 32466 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018901880538456335, - "loss": 1.5015, - "step": 32467 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890184671582223, - "loss": 1.6844, - "step": 32468 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018901812893188123, - "loss": 1.7748, - "step": 32469 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018901779070554014, - "loss": 1.4, - "step": 32470 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018901745247919908, - "loss": 1.4374, - "step": 32471 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018901711425285803, - "loss": 1.6031, - "step": 32472 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018901677602651694, - "loss": 1.4219, - "step": 32473 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018901643780017588, - "loss": 1.5083, - "step": 32474 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018901609957383482, - "loss": 1.4869, - "step": 32475 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018901576134749374, - "loss": 1.5851, - "step": 32476 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018901542312115268, - "loss": 1.4837, - "step": 32477 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018901508489481162, - "loss": 1.5047, - "step": 32478 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018901474666847053, - "loss": 1.5051, - "step": 32479 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018901440844212947, - "loss": 1.2725, - "step": 32480 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018901407021578841, - "loss": 1.4284, - "step": 32481 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018901373198944733, - "loss": 1.3685, - "step": 32482 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018901339376310627, - "loss": 1.6344, - "step": 32483 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890130555367652, - "loss": 1.6744, - "step": 32484 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018901271731042412, - "loss": 1.6471, - "step": 32485 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018901237908408307, - "loss": 1.4959, - "step": 32486 - }, - { - "epoch": 0.11, - "learning_rate": 0.000189012040857742, - "loss": 1.5321, - "step": 32487 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018901170263140092, - "loss": 1.5673, - "step": 32488 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018901136440505986, - "loss": 1.4758, - "step": 32489 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890110261787188, - "loss": 1.5703, - "step": 32490 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018901068795237772, - "loss": 1.6125, - "step": 32491 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018901034972603666, - "loss": 1.5623, - "step": 32492 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018901001149969563, - "loss": 1.5036, - "step": 32493 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018900967327335454, - "loss": 1.6299, - "step": 32494 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018900933504701348, - "loss": 1.6366, - "step": 32495 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018900899682067242, - "loss": 1.5221, - "step": 32496 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018900865859433134, - "loss": 1.6282, - "step": 32497 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018900832036799028, - "loss": 1.4639, - "step": 32498 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018900798214164922, - "loss": 1.5621, - "step": 32499 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018900764391530813, - "loss": 1.6516, - "step": 32500 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018900730568896707, - "loss": 1.5059, - "step": 32501 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018900696746262602, - "loss": 1.5689, - "step": 32502 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018900662923628493, - "loss": 1.4477, - "step": 32503 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018900629100994387, - "loss": 1.4202, - "step": 32504 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890059527836028, - "loss": 1.6317, - "step": 32505 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018900561455726173, - "loss": 1.6992, - "step": 32506 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018900527633092067, - "loss": 1.5279, - "step": 32507 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890049381045796, - "loss": 1.6063, - "step": 32508 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018900459987823852, - "loss": 1.7192, - "step": 32509 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018900426165189746, - "loss": 1.5384, - "step": 32510 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890039234255564, - "loss": 1.539, - "step": 32511 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018900358519921532, - "loss": 1.5805, - "step": 32512 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018900324697287426, - "loss": 1.447, - "step": 32513 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890029087465332, - "loss": 1.6198, - "step": 32514 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018900257052019211, - "loss": 1.5504, - "step": 32515 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018900223229385106, - "loss": 1.3851, - "step": 32516 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018900189406751, - "loss": 1.484, - "step": 32517 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890015558411689, - "loss": 1.4253, - "step": 32518 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018900121761482785, - "loss": 1.4754, - "step": 32519 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890008793884868, - "loss": 1.4743, - "step": 32520 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001890005411621457, - "loss": 1.5147, - "step": 32521 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018900020293580465, - "loss": 1.4752, - "step": 32522 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889998647094636, - "loss": 1.5423, - "step": 32523 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889995264831225, - "loss": 1.5695, - "step": 32524 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018899918825678144, - "loss": 1.5621, - "step": 32525 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018899885003044039, - "loss": 1.5349, - "step": 32526 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889985118040993, - "loss": 1.4877, - "step": 32527 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018899817357775824, - "loss": 1.5769, - "step": 32528 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018899783535141718, - "loss": 1.4383, - "step": 32529 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889974971250761, - "loss": 1.5255, - "step": 32530 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018899715889873504, - "loss": 1.6341, - "step": 32531 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018899682067239398, - "loss": 1.7243, - "step": 32532 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889964824460529, - "loss": 1.5119, - "step": 32533 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018899614421971183, - "loss": 1.3157, - "step": 32534 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018899580599337077, - "loss": 1.5794, - "step": 32535 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889954677670297, - "loss": 1.2913, - "step": 32536 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018899512954068863, - "loss": 1.4621, - "step": 32537 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018899479131434757, - "loss": 1.4653, - "step": 32538 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018899445308800649, - "loss": 1.5036, - "step": 32539 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018899411486166543, - "loss": 1.5135, - "step": 32540 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018899377663532437, - "loss": 1.5596, - "step": 32541 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018899343840898328, - "loss": 1.5863, - "step": 32542 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018899310018264222, - "loss": 1.5096, - "step": 32543 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018899276195630116, - "loss": 1.4801, - "step": 32544 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018899242372996008, - "loss": 1.5135, - "step": 32545 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018899208550361902, - "loss": 1.7342, - "step": 32546 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018899174727727796, - "loss": 1.5643, - "step": 32547 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018899140905093687, - "loss": 1.7257, - "step": 32548 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018899107082459584, - "loss": 1.6271, - "step": 32549 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018899073259825478, - "loss": 1.4939, - "step": 32550 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889903943719137, - "loss": 1.5476, - "step": 32551 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018899005614557264, - "loss": 1.5004, - "step": 32552 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018898971791923158, - "loss": 1.477, - "step": 32553 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889893796928905, - "loss": 1.5066, - "step": 32554 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018898904146654944, - "loss": 1.4436, - "step": 32555 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018898870324020838, - "loss": 1.5898, - "step": 32556 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889883650138673, - "loss": 1.5723, - "step": 32557 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018898802678752623, - "loss": 1.5755, - "step": 32558 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018898768856118517, - "loss": 1.5037, - "step": 32559 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889873503348441, - "loss": 1.2562, - "step": 32560 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018898701210850303, - "loss": 1.6879, - "step": 32561 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018898667388216197, - "loss": 1.5036, - "step": 32562 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018898633565582088, - "loss": 1.4709, - "step": 32563 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018898599742947982, - "loss": 1.6449, - "step": 32564 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018898565920313877, - "loss": 1.4823, - "step": 32565 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018898532097679768, - "loss": 1.5151, - "step": 32566 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018898498275045662, - "loss": 1.4011, - "step": 32567 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018898464452411556, - "loss": 1.6333, - "step": 32568 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018898430629777448, - "loss": 1.3298, - "step": 32569 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018898396807143342, - "loss": 1.5371, - "step": 32570 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018898362984509236, - "loss": 1.4929, - "step": 32571 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018898329161875127, - "loss": 1.6849, - "step": 32572 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889829533924102, - "loss": 1.3305, - "step": 32573 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018898261516606915, - "loss": 1.4668, - "step": 32574 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018898227693972807, - "loss": 1.6106, - "step": 32575 - }, - { - "epoch": 0.11, - "learning_rate": 0.000188981938713387, - "loss": 1.502, - "step": 32576 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018898160048704595, - "loss": 1.5064, - "step": 32577 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018898126226070486, - "loss": 1.563, - "step": 32578 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889809240343638, - "loss": 1.488, - "step": 32579 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018898058580802275, - "loss": 1.4401, - "step": 32580 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018898024758168166, - "loss": 1.4187, - "step": 32581 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889799093553406, - "loss": 1.58, - "step": 32582 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018897957112899954, - "loss": 1.5761, - "step": 32583 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018897923290265846, - "loss": 1.3127, - "step": 32584 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889788946763174, - "loss": 1.5683, - "step": 32585 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018897855644997634, - "loss": 1.6495, - "step": 32586 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018897821822363525, - "loss": 1.578, - "step": 32587 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889778799972942, - "loss": 1.6465, - "step": 32588 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018897754177095314, - "loss": 1.7077, - "step": 32589 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018897720354461205, - "loss": 1.3443, - "step": 32590 - }, - { - "epoch": 0.11, - "learning_rate": 0.000188976865318271, - "loss": 1.6564, - "step": 32591 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018897652709192993, - "loss": 1.523, - "step": 32592 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018897618886558885, - "loss": 1.4277, - "step": 32593 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889758506392478, - "loss": 1.7395, - "step": 32594 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018897551241290673, - "loss": 1.569, - "step": 32595 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018897517418656564, - "loss": 1.5977, - "step": 32596 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018897483596022458, - "loss": 1.52, - "step": 32597 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018897449773388352, - "loss": 1.5188, - "step": 32598 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018897415950754244, - "loss": 1.6926, - "step": 32599 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018897382128120138, - "loss": 1.5518, - "step": 32600 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018897348305486032, - "loss": 1.5116, - "step": 32601 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018897314482851924, - "loss": 1.4199, - "step": 32602 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018897280660217818, - "loss": 1.573, - "step": 32603 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018897246837583712, - "loss": 1.6216, - "step": 32604 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018897213014949606, - "loss": 1.7047, - "step": 32605 - }, - { - "epoch": 0.11, - "learning_rate": 0.000188971791923155, - "loss": 1.5095, - "step": 32606 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018897145369681391, - "loss": 1.5624, - "step": 32607 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018897111547047285, - "loss": 1.5792, - "step": 32608 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889707772441318, - "loss": 1.557, - "step": 32609 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018897043901779074, - "loss": 1.4951, - "step": 32610 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018897010079144965, - "loss": 1.6497, - "step": 32611 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889697625651086, - "loss": 1.4479, - "step": 32612 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018896942433876753, - "loss": 1.6678, - "step": 32613 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018896908611242645, - "loss": 1.5786, - "step": 32614 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889687478860854, - "loss": 1.6098, - "step": 32615 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018896840965974433, - "loss": 1.4887, - "step": 32616 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018896807143340324, - "loss": 1.3981, - "step": 32617 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018896773320706218, - "loss": 1.527, - "step": 32618 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018896739498072113, - "loss": 1.5968, - "step": 32619 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018896705675438004, - "loss": 1.4777, - "step": 32620 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018896671852803898, - "loss": 1.6482, - "step": 32621 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018896638030169792, - "loss": 1.6428, - "step": 32622 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018896604207535684, - "loss": 1.4912, - "step": 32623 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018896570384901578, - "loss": 1.5403, - "step": 32624 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018896536562267472, - "loss": 1.4108, - "step": 32625 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018896502739633363, - "loss": 1.6351, - "step": 32626 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018896468916999257, - "loss": 1.6185, - "step": 32627 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018896435094365151, - "loss": 1.5497, - "step": 32628 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018896401271731043, - "loss": 1.4949, - "step": 32629 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018896367449096937, - "loss": 1.5121, - "step": 32630 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889633362646283, - "loss": 1.5744, - "step": 32631 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018896299803828723, - "loss": 1.327, - "step": 32632 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018896265981194617, - "loss": 1.3324, - "step": 32633 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889623215856051, - "loss": 1.8043, - "step": 32634 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018896198335926402, - "loss": 1.4222, - "step": 32635 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018896164513292296, - "loss": 1.4415, - "step": 32636 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889613069065819, - "loss": 1.4401, - "step": 32637 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018896096868024082, - "loss": 1.4179, - "step": 32638 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018896063045389976, - "loss": 1.4561, - "step": 32639 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889602922275587, - "loss": 1.4764, - "step": 32640 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018895995400121761, - "loss": 1.6543, - "step": 32641 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018895961577487656, - "loss": 1.6625, - "step": 32642 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889592775485355, - "loss": 1.5895, - "step": 32643 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889589393221944, - "loss": 1.5591, - "step": 32644 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018895860109585335, - "loss": 1.3947, - "step": 32645 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889582628695123, - "loss": 1.5339, - "step": 32646 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889579246431712, - "loss": 1.4653, - "step": 32647 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018895758641683015, - "loss": 1.4141, - "step": 32648 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889572481904891, - "loss": 1.6345, - "step": 32649 - }, - { - "epoch": 0.11, - "learning_rate": 0.000188956909964148, - "loss": 1.5803, - "step": 32650 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018895657173780694, - "loss": 1.4508, - "step": 32651 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018895623351146589, - "loss": 1.5936, - "step": 32652 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889558952851248, - "loss": 1.3567, - "step": 32653 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018895555705878374, - "loss": 1.4341, - "step": 32654 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018895521883244268, - "loss": 1.5089, - "step": 32655 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889548806061016, - "loss": 1.3124, - "step": 32656 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018895454237976054, - "loss": 1.4482, - "step": 32657 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018895420415341948, - "loss": 1.526, - "step": 32658 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889538659270784, - "loss": 1.5368, - "step": 32659 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018895352770073733, - "loss": 1.5698, - "step": 32660 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018895318947439627, - "loss": 1.5132, - "step": 32661 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018895285124805522, - "loss": 1.4938, - "step": 32662 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018895251302171416, - "loss": 1.5471, - "step": 32663 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018895217479537307, - "loss": 1.5928, - "step": 32664 - }, - { - "epoch": 0.11, - "learning_rate": 0.000188951836569032, - "loss": 1.7462, - "step": 32665 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018895149834269095, - "loss": 1.5271, - "step": 32666 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018895116011634987, - "loss": 1.5709, - "step": 32667 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889508218900088, - "loss": 1.4039, - "step": 32668 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018895048366366775, - "loss": 1.4216, - "step": 32669 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018895014543732666, - "loss": 1.7163, - "step": 32670 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889498072109856, - "loss": 1.426, - "step": 32671 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018894946898464455, - "loss": 1.5097, - "step": 32672 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018894913075830346, - "loss": 1.4543, - "step": 32673 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889487925319624, - "loss": 1.5991, - "step": 32674 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018894845430562134, - "loss": 1.5035, - "step": 32675 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018894811607928026, - "loss": 1.6229, - "step": 32676 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889477778529392, - "loss": 1.4756, - "step": 32677 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018894743962659814, - "loss": 1.4891, - "step": 32678 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018894710140025705, - "loss": 1.6096, - "step": 32679 - }, - { - "epoch": 0.11, - "learning_rate": 0.000188946763173916, - "loss": 1.461, - "step": 32680 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018894642494757493, - "loss": 1.5894, - "step": 32681 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018894608672123385, - "loss": 1.7348, - "step": 32682 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889457484948928, - "loss": 1.4465, - "step": 32683 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018894541026855173, - "loss": 1.5072, - "step": 32684 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018894507204221064, - "loss": 1.5021, - "step": 32685 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018894473381586959, - "loss": 1.4991, - "step": 32686 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018894439558952853, - "loss": 1.4546, - "step": 32687 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018894405736318744, - "loss": 1.5112, - "step": 32688 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018894371913684638, - "loss": 1.5405, - "step": 32689 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018894338091050532, - "loss": 1.5945, - "step": 32690 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018894304268416424, - "loss": 1.5182, - "step": 32691 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018894270445782318, - "loss": 1.5823, - "step": 32692 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018894236623148212, - "loss": 1.5458, - "step": 32693 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018894202800514106, - "loss": 1.5453, - "step": 32694 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018894168977879997, - "loss": 1.632, - "step": 32695 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018894135155245892, - "loss": 1.4491, - "step": 32696 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018894101332611786, - "loss": 1.3889, - "step": 32697 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018894067509977677, - "loss": 1.4883, - "step": 32698 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889403368734357, - "loss": 1.4994, - "step": 32699 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018893999864709465, - "loss": 1.5828, - "step": 32700 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018893966042075357, - "loss": 1.619, - "step": 32701 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889393221944125, - "loss": 1.4151, - "step": 32702 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018893898396807145, - "loss": 1.4996, - "step": 32703 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018893864574173036, - "loss": 1.6274, - "step": 32704 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889383075153893, - "loss": 1.6601, - "step": 32705 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018893796928904825, - "loss": 1.4181, - "step": 32706 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018893763106270716, - "loss": 1.5339, - "step": 32707 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889372928363661, - "loss": 1.4651, - "step": 32708 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018893695461002504, - "loss": 1.5839, - "step": 32709 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018893661638368396, - "loss": 1.562, - "step": 32710 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889362781573429, - "loss": 1.5475, - "step": 32711 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018893593993100184, - "loss": 1.5227, - "step": 32712 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018893560170466075, - "loss": 1.3542, - "step": 32713 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889352634783197, - "loss": 1.2833, - "step": 32714 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018893492525197864, - "loss": 1.3436, - "step": 32715 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018893458702563755, - "loss": 1.605, - "step": 32716 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018893424879929652, - "loss": 1.5216, - "step": 32717 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018893391057295543, - "loss": 1.7181, - "step": 32718 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018893357234661437, - "loss": 1.4696, - "step": 32719 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018893323412027331, - "loss": 1.5188, - "step": 32720 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018893289589393223, - "loss": 1.5379, - "step": 32721 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018893255766759117, - "loss": 1.4935, - "step": 32722 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889322194412501, - "loss": 1.4649, - "step": 32723 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018893188121490902, - "loss": 1.5616, - "step": 32724 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018893154298856797, - "loss": 1.5893, - "step": 32725 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889312047622269, - "loss": 1.4984, - "step": 32726 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018893086653588582, - "loss": 1.4086, - "step": 32727 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018893052830954476, - "loss": 1.3125, - "step": 32728 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889301900832037, - "loss": 1.5254, - "step": 32729 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018892985185686262, - "loss": 1.6828, - "step": 32730 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018892951363052156, - "loss": 1.5215, - "step": 32731 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889291754041805, - "loss": 1.5807, - "step": 32732 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889288371778394, - "loss": 1.5719, - "step": 32733 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018892849895149835, - "loss": 1.5076, - "step": 32734 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889281607251573, - "loss": 1.5656, - "step": 32735 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889278224988162, - "loss": 1.571, - "step": 32736 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018892748427247515, - "loss": 1.4775, - "step": 32737 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889271460461341, - "loss": 1.5614, - "step": 32738 - }, - { - "epoch": 0.11, - "learning_rate": 0.000188926807819793, - "loss": 1.7073, - "step": 32739 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018892646959345195, - "loss": 1.4579, - "step": 32740 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889261313671109, - "loss": 1.5983, - "step": 32741 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889257931407698, - "loss": 1.4378, - "step": 32742 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018892545491442874, - "loss": 1.5143, - "step": 32743 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018892511668808768, - "loss": 1.3418, - "step": 32744 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889247784617466, - "loss": 1.4669, - "step": 32745 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018892444023540554, - "loss": 1.6336, - "step": 32746 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018892410200906448, - "loss": 1.6796, - "step": 32747 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889237637827234, - "loss": 1.3735, - "step": 32748 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018892342555638234, - "loss": 1.5012, - "step": 32749 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018892308733004128, - "loss": 1.5593, - "step": 32750 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889227491037002, - "loss": 1.6487, - "step": 32751 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018892241087735913, - "loss": 1.4682, - "step": 32752 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018892207265101807, - "loss": 1.4706, - "step": 32753 - }, - { - "epoch": 0.11, - "learning_rate": 0.000188921734424677, - "loss": 1.5227, - "step": 32754 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018892139619833593, - "loss": 1.4807, - "step": 32755 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018892105797199487, - "loss": 1.474, - "step": 32756 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018892071974565378, - "loss": 1.52, - "step": 32757 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018892038151931272, - "loss": 1.548, - "step": 32758 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018892004329297167, - "loss": 1.6903, - "step": 32759 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018891970506663058, - "loss": 1.5828, - "step": 32760 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018891936684028952, - "loss": 1.7545, - "step": 32761 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018891902861394846, - "loss": 1.4383, - "step": 32762 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018891869038760738, - "loss": 1.5933, - "step": 32763 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018891835216126632, - "loss": 1.5088, - "step": 32764 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018891801393492526, - "loss": 1.5262, - "step": 32765 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018891767570858417, - "loss": 1.614, - "step": 32766 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018891733748224311, - "loss": 1.5517, - "step": 32767 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018891699925590205, - "loss": 1.6355, - "step": 32768 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018891666102956097, - "loss": 1.3146, - "step": 32769 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889163228032199, - "loss": 1.6275, - "step": 32770 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018891598457687885, - "loss": 1.5648, - "step": 32771 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018891564635053777, - "loss": 1.6472, - "step": 32772 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889153081241967, - "loss": 1.6716, - "step": 32773 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018891496989785567, - "loss": 1.3819, - "step": 32774 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889146316715146, - "loss": 1.3752, - "step": 32775 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018891429344517353, - "loss": 1.5832, - "step": 32776 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018891395521883247, - "loss": 1.6502, - "step": 32777 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018891361699249138, - "loss": 1.5577, - "step": 32778 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018891327876615033, - "loss": 1.6469, - "step": 32779 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018891294053980927, - "loss": 1.5015, - "step": 32780 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018891260231346818, - "loss": 1.5617, - "step": 32781 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018891226408712712, - "loss": 1.6015, - "step": 32782 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018891192586078606, - "loss": 1.6284, - "step": 32783 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018891158763444498, - "loss": 1.666, - "step": 32784 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018891124940810392, - "loss": 1.5611, - "step": 32785 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018891091118176286, - "loss": 1.497, - "step": 32786 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018891057295542177, - "loss": 1.5382, - "step": 32787 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018891023472908071, - "loss": 1.5698, - "step": 32788 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018890989650273966, - "loss": 1.5101, - "step": 32789 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018890955827639857, - "loss": 1.4673, - "step": 32790 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889092200500575, - "loss": 1.7341, - "step": 32791 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018890888182371645, - "loss": 1.4427, - "step": 32792 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018890854359737537, - "loss": 1.3795, - "step": 32793 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889082053710343, - "loss": 1.4342, - "step": 32794 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018890786714469325, - "loss": 1.5813, - "step": 32795 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018890752891835216, - "loss": 1.6444, - "step": 32796 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889071906920111, - "loss": 1.635, - "step": 32797 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018890685246567004, - "loss": 1.4888, - "step": 32798 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018890651423932896, - "loss": 1.6075, - "step": 32799 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889061760129879, - "loss": 1.5621, - "step": 32800 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018890583778664684, - "loss": 1.4194, - "step": 32801 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018890549956030576, - "loss": 1.731, - "step": 32802 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889051613339647, - "loss": 1.5968, - "step": 32803 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018890482310762364, - "loss": 1.4112, - "step": 32804 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018890448488128255, - "loss": 1.5228, - "step": 32805 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889041466549415, - "loss": 1.5705, - "step": 32806 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018890380842860043, - "loss": 1.5344, - "step": 32807 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018890347020225935, - "loss": 1.6223, - "step": 32808 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001889031319759183, - "loss": 1.438, - "step": 32809 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018890279374957723, - "loss": 1.4654, - "step": 32810 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018890245552323614, - "loss": 1.4473, - "step": 32811 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018890211729689509, - "loss": 1.5431, - "step": 32812 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018890177907055403, - "loss": 1.5934, - "step": 32813 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018890144084421294, - "loss": 1.4272, - "step": 32814 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018890110261787188, - "loss": 1.5875, - "step": 32815 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018890076439153082, - "loss": 1.557, - "step": 32816 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018890042616518974, - "loss": 1.5566, - "step": 32817 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018890008793884868, - "loss": 1.5366, - "step": 32818 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018889974971250762, - "loss": 1.3605, - "step": 32819 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018889941148616653, - "loss": 1.5268, - "step": 32820 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018889907325982547, - "loss": 1.5211, - "step": 32821 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018889873503348442, - "loss": 1.6375, - "step": 32822 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018889839680714333, - "loss": 1.5921, - "step": 32823 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018889805858080227, - "loss": 1.5897, - "step": 32824 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888977203544612, - "loss": 1.5489, - "step": 32825 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018889738212812013, - "loss": 1.5759, - "step": 32826 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018889704390177907, - "loss": 1.392, - "step": 32827 - }, - { - "epoch": 0.11, - "learning_rate": 0.000188896705675438, - "loss": 1.4839, - "step": 32828 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018889636744909692, - "loss": 1.6071, - "step": 32829 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888960292227559, - "loss": 1.4756, - "step": 32830 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018889569099641483, - "loss": 1.594, - "step": 32831 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018889535277007375, - "loss": 1.6241, - "step": 32832 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888950145437327, - "loss": 1.671, - "step": 32833 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018889467631739163, - "loss": 1.4215, - "step": 32834 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018889433809105054, - "loss": 1.5799, - "step": 32835 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018889399986470948, - "loss": 1.3379, - "step": 32836 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018889366163836842, - "loss": 1.643, - "step": 32837 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018889332341202734, - "loss": 1.5077, - "step": 32838 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018889298518568628, - "loss": 1.5817, - "step": 32839 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018889264695934522, - "loss": 1.4422, - "step": 32840 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018889230873300413, - "loss": 1.6028, - "step": 32841 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018889197050666308, - "loss": 1.4542, - "step": 32842 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018889163228032202, - "loss": 1.5989, - "step": 32843 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018889129405398093, - "loss": 1.6308, - "step": 32844 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018889095582763987, - "loss": 1.5178, - "step": 32845 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888906176012988, - "loss": 1.5671, - "step": 32846 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018889027937495773, - "loss": 1.6054, - "step": 32847 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018888994114861667, - "loss": 1.5485, - "step": 32848 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888896029222756, - "loss": 1.6324, - "step": 32849 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018888926469593452, - "loss": 1.5109, - "step": 32850 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018888892646959346, - "loss": 1.4074, - "step": 32851 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888885882432524, - "loss": 1.3958, - "step": 32852 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018888825001691132, - "loss": 1.674, - "step": 32853 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018888791179057026, - "loss": 1.2944, - "step": 32854 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888875735642292, - "loss": 1.4698, - "step": 32855 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018888723533788812, - "loss": 1.5847, - "step": 32856 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018888689711154706, - "loss": 1.6296, - "step": 32857 - }, - { - "epoch": 0.11, - "learning_rate": 0.000188886558885206, - "loss": 1.5278, - "step": 32858 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888862206588649, - "loss": 1.6879, - "step": 32859 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018888588243252385, - "loss": 1.6073, - "step": 32860 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888855442061828, - "loss": 1.6143, - "step": 32861 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888852059798417, - "loss": 1.5206, - "step": 32862 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018888486775350065, - "loss": 1.679, - "step": 32863 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888845295271596, - "loss": 1.7002, - "step": 32864 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888841913008185, - "loss": 1.6036, - "step": 32865 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018888385307447745, - "loss": 1.4054, - "step": 32866 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888835148481364, - "loss": 1.4938, - "step": 32867 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888831766217953, - "loss": 1.6096, - "step": 32868 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018888283839545424, - "loss": 1.5658, - "step": 32869 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018888250016911318, - "loss": 1.4833, - "step": 32870 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888821619427721, - "loss": 1.7462, - "step": 32871 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018888182371643104, - "loss": 1.7243, - "step": 32872 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018888148549008998, - "loss": 1.4538, - "step": 32873 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888811472637489, - "loss": 1.5408, - "step": 32874 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018888080903740784, - "loss": 1.5915, - "step": 32875 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018888047081106678, - "loss": 1.701, - "step": 32876 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888801325847257, - "loss": 1.573, - "step": 32877 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018887979435838463, - "loss": 1.3683, - "step": 32878 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018887945613204357, - "loss": 1.5531, - "step": 32879 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888791179057025, - "loss": 1.4994, - "step": 32880 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018887877967936143, - "loss": 1.5959, - "step": 32881 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018887844145302037, - "loss": 1.716, - "step": 32882 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018887810322667928, - "loss": 1.5469, - "step": 32883 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018887776500033822, - "loss": 1.7052, - "step": 32884 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018887742677399717, - "loss": 1.6625, - "step": 32885 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888770885476561, - "loss": 1.5357, - "step": 32886 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018887675032131505, - "loss": 1.5548, - "step": 32887 - }, - { - "epoch": 0.11, - "learning_rate": 0.000188876412094974, - "loss": 1.4873, - "step": 32888 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888760738686329, - "loss": 1.6604, - "step": 32889 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018887573564229184, - "loss": 1.6461, - "step": 32890 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018887539741595078, - "loss": 1.6343, - "step": 32891 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888750591896097, - "loss": 1.472, - "step": 32892 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018887472096326864, - "loss": 1.6009, - "step": 32893 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018887438273692758, - "loss": 1.4084, - "step": 32894 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888740445105865, - "loss": 1.6008, - "step": 32895 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018887370628424544, - "loss": 1.6674, - "step": 32896 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018887336805790438, - "loss": 1.5343, - "step": 32897 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888730298315633, - "loss": 1.4965, - "step": 32898 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018887269160522223, - "loss": 1.3628, - "step": 32899 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018887235337888117, - "loss": 1.5617, - "step": 32900 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888720151525401, - "loss": 1.533, - "step": 32901 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018887167692619903, - "loss": 1.5366, - "step": 32902 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018887133869985797, - "loss": 1.614, - "step": 32903 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018887100047351688, - "loss": 1.5403, - "step": 32904 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018887066224717583, - "loss": 1.6974, - "step": 32905 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018887032402083477, - "loss": 1.5805, - "step": 32906 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018886998579449368, - "loss": 1.4798, - "step": 32907 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018886964756815262, - "loss": 1.4308, - "step": 32908 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018886930934181156, - "loss": 1.6661, - "step": 32909 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018886897111547048, - "loss": 1.3852, - "step": 32910 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018886863288912942, - "loss": 1.7219, - "step": 32911 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018886829466278836, - "loss": 1.5663, - "step": 32912 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018886795643644727, - "loss": 1.6363, - "step": 32913 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018886761821010621, - "loss": 1.4457, - "step": 32914 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018886727998376516, - "loss": 1.3958, - "step": 32915 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018886694175742407, - "loss": 1.5457, - "step": 32916 - }, - { - "epoch": 0.11, - "learning_rate": 0.000188866603531083, - "loss": 1.6133, - "step": 32917 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018886626530474195, - "loss": 1.5148, - "step": 32918 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018886592707840087, - "loss": 1.5365, - "step": 32919 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888655888520598, - "loss": 1.4479, - "step": 32920 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018886525062571875, - "loss": 1.5469, - "step": 32921 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018886491239937766, - "loss": 1.6263, - "step": 32922 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888645741730366, - "loss": 1.4536, - "step": 32923 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018886423594669554, - "loss": 1.5886, - "step": 32924 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018886389772035446, - "loss": 1.6548, - "step": 32925 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888635594940134, - "loss": 1.5508, - "step": 32926 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018886322126767234, - "loss": 1.4358, - "step": 32927 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018886288304133125, - "loss": 1.7068, - "step": 32928 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888625448149902, - "loss": 1.5984, - "step": 32929 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018886220658864914, - "loss": 1.4956, - "step": 32930 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018886186836230805, - "loss": 1.5899, - "step": 32931 - }, - { - "epoch": 0.11, - "learning_rate": 0.000188861530135967, - "loss": 1.3834, - "step": 32932 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018886119190962593, - "loss": 1.6749, - "step": 32933 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018886085368328485, - "loss": 1.5852, - "step": 32934 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888605154569438, - "loss": 1.5737, - "step": 32935 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018886017723060273, - "loss": 1.6709, - "step": 32936 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018885983900426164, - "loss": 1.6729, - "step": 32937 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018885950077792058, - "loss": 1.6722, - "step": 32938 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018885916255157953, - "loss": 1.7208, - "step": 32939 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018885882432523844, - "loss": 1.588, - "step": 32940 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018885848609889738, - "loss": 1.5724, - "step": 32941 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018885814787255632, - "loss": 1.7149, - "step": 32942 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018885780964621526, - "loss": 1.5308, - "step": 32943 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888574714198742, - "loss": 1.4637, - "step": 32944 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018885713319353312, - "loss": 1.5622, - "step": 32945 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018885679496719206, - "loss": 1.4696, - "step": 32946 - }, - { - "epoch": 0.11, - "learning_rate": 0.000188856456740851, - "loss": 1.5925, - "step": 32947 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018885611851450991, - "loss": 1.549, - "step": 32948 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018885578028816886, - "loss": 1.4483, - "step": 32949 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888554420618278, - "loss": 1.4719, - "step": 32950 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888551038354867, - "loss": 1.3653, - "step": 32951 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018885476560914565, - "loss": 1.5534, - "step": 32952 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888544273828046, - "loss": 1.6044, - "step": 32953 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888540891564635, - "loss": 1.4624, - "step": 32954 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018885375093012245, - "loss": 1.4443, - "step": 32955 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888534127037814, - "loss": 1.5099, - "step": 32956 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888530744774403, - "loss": 1.4987, - "step": 32957 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018885273625109924, - "loss": 1.568, - "step": 32958 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018885239802475819, - "loss": 1.732, - "step": 32959 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888520597984171, - "loss": 1.5771, - "step": 32960 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018885172157207604, - "loss": 1.4584, - "step": 32961 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018885138334573498, - "loss": 1.4729, - "step": 32962 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018885104511939392, - "loss": 1.6039, - "step": 32963 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018885070689305284, - "loss": 1.6033, - "step": 32964 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018885036866671178, - "loss": 1.724, - "step": 32965 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018885003044037072, - "loss": 1.8527, - "step": 32966 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018884969221402963, - "loss": 1.5637, - "step": 32967 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018884935398768857, - "loss": 1.4037, - "step": 32968 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018884901576134752, - "loss": 1.4476, - "step": 32969 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018884867753500643, - "loss": 1.5122, - "step": 32970 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018884833930866537, - "loss": 1.7093, - "step": 32971 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888480010823243, - "loss": 1.7083, - "step": 32972 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018884766285598323, - "loss": 1.5427, - "step": 32973 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018884732462964217, - "loss": 1.481, - "step": 32974 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888469864033011, - "loss": 1.5534, - "step": 32975 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018884664817696002, - "loss": 1.7012, - "step": 32976 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018884630995061896, - "loss": 1.6328, - "step": 32977 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888459717242779, - "loss": 1.5845, - "step": 32978 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018884563349793682, - "loss": 1.6226, - "step": 32979 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018884529527159576, - "loss": 1.6419, - "step": 32980 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888449570452547, - "loss": 1.639, - "step": 32981 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018884461881891362, - "loss": 1.5041, - "step": 32982 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018884428059257256, - "loss": 1.5973, - "step": 32983 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888439423662315, - "loss": 1.6322, - "step": 32984 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888436041398904, - "loss": 1.5042, - "step": 32985 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018884326591354935, - "loss": 1.5252, - "step": 32986 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888429276872083, - "loss": 1.5875, - "step": 32987 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888425894608672, - "loss": 1.5755, - "step": 32988 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018884225123452615, - "loss": 1.5562, - "step": 32989 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888419130081851, - "loss": 1.7098, - "step": 32990 - }, - { - "epoch": 0.11, - "learning_rate": 0.000188841574781844, - "loss": 1.4144, - "step": 32991 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018884123655550295, - "loss": 1.6513, - "step": 32992 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888408983291619, - "loss": 1.5667, - "step": 32993 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888405601028208, - "loss": 1.4718, - "step": 32994 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018884022187647974, - "loss": 1.465, - "step": 32995 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018883988365013868, - "loss": 1.4332, - "step": 32996 - }, - { - "epoch": 0.11, - "learning_rate": 0.0001888395454237976, - "loss": 1.659, - "step": 32997 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018883920719745657, - "loss": 1.5151, - "step": 32998 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018883886897111548, - "loss": 1.461, - "step": 32999 - }, - { - "epoch": 0.11, - "learning_rate": 0.00018883853074477442, - "loss": 1.5465, - "step": 33000 - }, - { - "epoch": 0.11, - "eval_loss": 1.4923377403846154, - "mrc_em": 22.0, - "mrc_f1": 28.33650793650794, - "nli_acc": 0.08, - "step": 33000, - "sts_f1": 0.7008547008547008, - "ynat_f1": 0.7436445133813555 - } - ], - "max_steps": 591320, - "num_train_epochs": 2, - "total_flos": 4.788419682468102e+18, - "trial_name": null, - "trial_params": null -}