diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,76396 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.597107438016529,
+  "eval_steps": 3000,
+  "global_step": 534000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.00043044077134986227,
+      "grad_norm": 71.5,
+      "learning_rate": 9.800000000000001e-06,
+      "loss": 18.2106,
+      "step": 50
+    },
+    {
+      "epoch": 0.0008608815426997245,
+      "grad_norm": 5.375,
+      "learning_rate": 1.9800000000000004e-05,
+      "loss": 12.5337,
+      "step": 100
+    },
+    {
+      "epoch": 0.0012913223140495868,
+      "grad_norm": 7.34375,
+      "learning_rate": 2.98e-05,
+      "loss": 11.224,
+      "step": 150
+    },
+    {
+      "epoch": 0.001721763085399449,
+      "grad_norm": 12.125,
+      "learning_rate": 3.9800000000000005e-05,
+      "loss": 8.4094,
+      "step": 200
+    },
+    {
+      "epoch": 0.0021522038567493114,
+      "grad_norm": 10.375,
+      "learning_rate": 4.9800000000000004e-05,
+      "loss": 7.4048,
+      "step": 250
+    },
+    {
+      "epoch": 0.0025826446280991736,
+      "grad_norm": 6.09375,
+      "learning_rate": 5.9800000000000003e-05,
+      "loss": 7.1413,
+      "step": 300
+    },
+    {
+      "epoch": 0.003013085399449036,
+      "grad_norm": 2.421875,
+      "learning_rate": 6.98e-05,
+      "loss": 6.9154,
+      "step": 350
+    },
+    {
+      "epoch": 0.003443526170798898,
+      "grad_norm": 3.296875,
+      "learning_rate": 7.98e-05,
+      "loss": 6.9162,
+      "step": 400
+    },
+    {
+      "epoch": 0.0038739669421487604,
+      "grad_norm": 1.8671875,
+      "learning_rate": 8.98e-05,
+      "loss": 6.6844,
+      "step": 450
+    },
+    {
+      "epoch": 0.004304407713498623,
+      "grad_norm": 3.375,
+      "learning_rate": 9.98e-05,
+      "loss": 6.5678,
+      "step": 500
+    },
+    {
+      "epoch": 0.004734848484848485,
+      "grad_norm": 6.5,
+      "learning_rate": 0.00010980000000000001,
+      "loss": 7.1726,
+      "step": 550
+    },
+    {
+      "epoch": 0.005165289256198347,
+      "grad_norm": 3.53125,
+      "learning_rate": 0.0001198,
+      "loss": 6.5573,
+      "step": 600
+    },
+    {
+      "epoch": 0.0055957300275482095,
+      "grad_norm": 3.59375,
+      "learning_rate": 0.0001298,
+      "loss": 6.5085,
+      "step": 650
+    },
+    {
+      "epoch": 0.006026170798898072,
+      "grad_norm": 2.984375,
+      "learning_rate": 0.0001398,
+      "loss": 6.0831,
+      "step": 700
+    },
+    {
+      "epoch": 0.006456611570247934,
+      "grad_norm": 1.9921875,
+      "learning_rate": 0.0001498,
+      "loss": 7.2968,
+      "step": 750
+    },
+    {
+      "epoch": 0.006887052341597796,
+      "grad_norm": 3.984375,
+      "learning_rate": 0.0001598,
+      "loss": 6.4507,
+      "step": 800
+    },
+    {
+      "epoch": 0.007317493112947659,
+      "grad_norm": 0.85546875,
+      "learning_rate": 0.0001698,
+      "loss": 6.6401,
+      "step": 850
+    },
+    {
+      "epoch": 0.007747933884297521,
+      "grad_norm": 1.546875,
+      "learning_rate": 0.0001798,
+      "loss": 6.4949,
+      "step": 900
+    },
+    {
+      "epoch": 0.008178374655647382,
+      "grad_norm": 2.625,
+      "learning_rate": 0.0001898,
+      "loss": 6.7644,
+      "step": 950
+    },
+    {
+      "epoch": 0.008608815426997245,
+      "grad_norm": 4.0625,
+      "learning_rate": 0.0001998,
+      "loss": 6.7678,
+      "step": 1000
+    },
+    {
+      "epoch": 0.009039256198347107,
+      "grad_norm": 3.296875,
+      "learning_rate": 0.00019999999647543534,
+      "loss": 6.8504,
+      "step": 1050
+    },
+    {
+      "epoch": 0.00946969696969697,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.00019999998561255402,
+      "loss": 6.9308,
+      "step": 1100
+    },
+    {
+      "epoch": 0.009900137741046831,
+      "grad_norm": 1.5234375,
+      "learning_rate": 0.0001999999674098889,
+      "loss": 6.5577,
+      "step": 1150
+    },
+    {
+      "epoch": 0.010330578512396695,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.00019999994186744137,
+      "loss": 6.6297,
+      "step": 1200
+    },
+    {
+      "epoch": 0.010761019283746556,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.00019999990898521324,
+      "loss": 6.6041,
+      "step": 1250
+    },
+    {
+      "epoch": 0.011191460055096419,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.00019999986876320693,
+      "loss": 6.7533,
+      "step": 1300
+    },
+    {
+      "epoch": 0.01162190082644628,
+      "grad_norm": 0.625,
+      "learning_rate": 0.00019999982120142543,
+      "loss": 6.3455,
+      "step": 1350
+    },
+    {
+      "epoch": 0.012052341597796144,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.0001999997662998722,
+      "loss": 6.4072,
+      "step": 1400
+    },
+    {
+      "epoch": 0.012482782369146005,
+      "grad_norm": 2.21875,
+      "learning_rate": 0.00019999970405855128,
+      "loss": 6.1638,
+      "step": 1450
+    },
+    {
+      "epoch": 0.012913223140495868,
+      "grad_norm": 1.6484375,
+      "learning_rate": 0.00019999963447746722,
+      "loss": 6.7652,
+      "step": 1500
+    },
+    {
+      "epoch": 0.01334366391184573,
+      "grad_norm": 0.953125,
+      "learning_rate": 0.00019999955755662518,
+      "loss": 6.2853,
+      "step": 1550
+    },
+    {
+      "epoch": 0.013774104683195593,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00019999947329603078,
+      "loss": 6.5354,
+      "step": 1600
+    },
+    {
+      "epoch": 0.014204545454545454,
+      "grad_norm": 3.59375,
+      "learning_rate": 0.00019999938169569016,
+      "loss": 6.4226,
+      "step": 1650
+    },
+    {
+      "epoch": 0.014634986225895317,
+      "grad_norm": 1.78125,
+      "learning_rate": 0.00019999928275561007,
+      "loss": 6.2953,
+      "step": 1700
+    },
+    {
+      "epoch": 0.015065426997245179,
+      "grad_norm": 3.46875,
+      "learning_rate": 0.0001999991764757978,
+      "loss": 6.4832,
+      "step": 1750
+    },
+    {
+      "epoch": 0.015495867768595042,
+      "grad_norm": 1.53125,
+      "learning_rate": 0.0001999990628562611,
+      "loss": 6.1984,
+      "step": 1800
+    },
+    {
+      "epoch": 0.015926308539944905,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.0001999989418970084,
+      "loss": 6.6464,
+      "step": 1850
+    },
+    {
+      "epoch": 0.016356749311294765,
+      "grad_norm": 2.5,
+      "learning_rate": 0.00019999881359804848,
+      "loss": 6.5281,
+      "step": 1900
+    },
+    {
+      "epoch": 0.016787190082644628,
+      "grad_norm": 2.265625,
+      "learning_rate": 0.00019999867795939078,
+      "loss": 5.9519,
+      "step": 1950
+    },
+    {
+      "epoch": 0.01721763085399449,
+      "grad_norm": 1.7734375,
+      "learning_rate": 0.00019999853498104528,
+      "loss": 6.4505,
+      "step": 2000
+    },
+    {
+      "epoch": 0.017648071625344354,
+      "grad_norm": 1.9921875,
+      "learning_rate": 0.00019999838466302247,
+      "loss": 7.2526,
+      "step": 2050
+    },
+    {
+      "epoch": 0.018078512396694214,
+      "grad_norm": 1.484375,
+      "learning_rate": 0.0001999982270053334,
+      "loss": 6.2714,
+      "step": 2100
+    },
+    {
+      "epoch": 0.018508953168044077,
+      "grad_norm": 1.8671875,
+      "learning_rate": 0.0001999980620079896,
+      "loss": 6.3783,
+      "step": 2150
+    },
+    {
+      "epoch": 0.01893939393939394,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.0001999978896710032,
+      "loss": 6.0807,
+      "step": 2200
+    },
+    {
+      "epoch": 0.019369834710743803,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00019999770999438684,
+      "loss": 6.0694,
+      "step": 2250
+    },
+    {
+      "epoch": 0.019800275482093663,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.00019999752297815372,
+      "loss": 6.9441,
+      "step": 2300
+    },
+    {
+      "epoch": 0.020230716253443526,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.00019999732862231758,
+      "loss": 6.1499,
+      "step": 2350
+    },
+    {
+      "epoch": 0.02066115702479339,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.00019999712692689265,
+      "loss": 6.6325,
+      "step": 2400
+    },
+    {
+      "epoch": 0.021091597796143252,
+      "grad_norm": 1.203125,
+      "learning_rate": 0.00019999691789189376,
+      "loss": 6.3922,
+      "step": 2450
+    },
+    {
+      "epoch": 0.021522038567493112,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00019999670151733625,
+      "loss": 6.5116,
+      "step": 2500
+    },
+    {
+      "epoch": 0.021952479338842975,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.00019999647780323597,
+      "loss": 6.0381,
+      "step": 2550
+    },
+    {
+      "epoch": 0.022382920110192838,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.0001999962467496094,
+      "loss": 6.0474,
+      "step": 2600
+    },
+    {
+      "epoch": 0.0228133608815427,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.00019999600835647347,
+      "loss": 6.3361,
+      "step": 2650
+    },
+    {
+      "epoch": 0.02324380165289256,
+      "grad_norm": 1.40625,
+      "learning_rate": 0.00019999576262384568,
+      "loss": 6.4502,
+      "step": 2700
+    },
+    {
+      "epoch": 0.023674242424242424,
+      "grad_norm": 1.3359375,
+      "learning_rate": 0.00019999550955174402,
+      "loss": 6.2025,
+      "step": 2750
+    },
+    {
+      "epoch": 0.024104683195592287,
+      "grad_norm": 1.5859375,
+      "learning_rate": 0.00019999524914018712,
+      "loss": 6.3278,
+      "step": 2800
+    },
+    {
+      "epoch": 0.02453512396694215,
+      "grad_norm": 2.296875,
+      "learning_rate": 0.0001999949813891941,
+      "loss": 6.0009,
+      "step": 2850
+    },
+    {
+      "epoch": 0.02496556473829201,
+      "grad_norm": 1.125,
+      "learning_rate": 0.0001999947062987846,
+      "loss": 6.1766,
+      "step": 2900
+    },
+    {
+      "epoch": 0.025396005509641873,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00019999442386897876,
+      "loss": 5.6974,
+      "step": 2950
+    },
+    {
+      "epoch": 0.025826446280991736,
+      "grad_norm": 2.0,
+      "learning_rate": 0.00019999413409979737,
+      "loss": 6.0692,
+      "step": 3000
+    },
+    {
+      "epoch": 0.025826446280991736,
+      "eval_loss": 6.841164588928223,
+      "eval_runtime": 22.2026,
+      "eval_samples_per_second": 28.825,
+      "eval_steps_per_second": 14.413,
+      "eval_tts_loss": 5.814084659717215,
+      "step": 3000
+    },
+    {
+      "epoch": 0.0262568870523416,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00019999383699126168,
+      "loss": 6.3387,
+      "step": 3050
+    },
+    {
+      "epoch": 0.02668732782369146,
+      "grad_norm": 2.265625,
+      "learning_rate": 0.00019999353254339348,
+      "loss": 6.078,
+      "step": 3100
+    },
+    {
+      "epoch": 0.027117768595041322,
+      "grad_norm": 3.21875,
+      "learning_rate": 0.00019999322075621513,
+      "loss": 6.05,
+      "step": 3150
+    },
+    {
+      "epoch": 0.027548209366391185,
+      "grad_norm": 1.890625,
+      "learning_rate": 0.00019999290162974959,
+      "loss": 5.7936,
+      "step": 3200
+    },
+    {
+      "epoch": 0.02797865013774105,
+      "grad_norm": 1.296875,
+      "learning_rate": 0.00019999257516402014,
+      "loss": 5.9983,
+      "step": 3250
+    },
+    {
+      "epoch": 0.028409090909090908,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.00019999224135905082,
+      "loss": 5.9586,
+      "step": 3300
+    },
+    {
+      "epoch": 0.02883953168044077,
+      "grad_norm": 0.890625,
+      "learning_rate": 0.00019999190021486611,
+      "loss": 5.5419,
+      "step": 3350
+    },
+    {
+      "epoch": 0.029269972451790634,
+      "grad_norm": 3.71875,
+      "learning_rate": 0.00019999155173149112,
+      "loss": 5.386,
+      "step": 3400
+    },
+    {
+      "epoch": 0.029700413223140498,
+      "grad_norm": 1.6484375,
+      "learning_rate": 0.0001999911959089513,
+      "loss": 6.187,
+      "step": 3450
+    },
+    {
+      "epoch": 0.030130853994490357,
+      "grad_norm": 1.8203125,
+      "learning_rate": 0.0001999908327472729,
+      "loss": 6.3146,
+      "step": 3500
+    },
+    {
+      "epoch": 0.03056129476584022,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00019999046224648248,
+      "loss": 6.4992,
+      "step": 3550
+    },
+    {
+      "epoch": 0.030991735537190084,
+      "grad_norm": 3.046875,
+      "learning_rate": 0.00019999008440660726,
+      "loss": 5.8337,
+      "step": 3600
+    },
+    {
+      "epoch": 0.03142217630853995,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.000199989699227675,
+      "loss": 5.7526,
+      "step": 3650
+    },
+    {
+      "epoch": 0.03185261707988981,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.0001999893067097139,
+      "loss": 6.2363,
+      "step": 3700
+    },
+    {
+      "epoch": 0.032283057851239666,
+      "grad_norm": 3.078125,
+      "learning_rate": 0.0001999889068527529,
+      "loss": 6.3704,
+      "step": 3750
+    },
+    {
+      "epoch": 0.03271349862258953,
+      "grad_norm": 1.3671875,
+      "learning_rate": 0.00019998849965682122,
+      "loss": 5.7546,
+      "step": 3800
+    },
+    {
+      "epoch": 0.03314393939393939,
+      "grad_norm": 3.796875,
+      "learning_rate": 0.0001999880851219488,
+      "loss": 6.3607,
+      "step": 3850
+    },
+    {
+      "epoch": 0.033574380165289255,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.00019998766324816607,
+      "loss": 6.3463,
+      "step": 3900
+    },
+    {
+      "epoch": 0.03400482093663912,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00019998723403550397,
+      "loss": 5.8813,
+      "step": 3950
+    },
+    {
+      "epoch": 0.03443526170798898,
+      "grad_norm": 1.8203125,
+      "learning_rate": 0.00019998679748399405,
+      "loss": 5.7699,
+      "step": 4000
+    },
+    {
+      "epoch": 0.034865702479338845,
+      "grad_norm": 1.3125,
+      "learning_rate": 0.0001999863535936683,
+      "loss": 5.9576,
+      "step": 4050
+    },
+    {
+      "epoch": 0.03529614325068871,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.00019998590236455934,
+      "loss": 6.3861,
+      "step": 4100
+    },
+    {
+      "epoch": 0.035726584022038564,
+      "grad_norm": 1.40625,
+      "learning_rate": 0.00019998544379670023,
+      "loss": 5.977,
+      "step": 4150
+    },
+    {
+      "epoch": 0.03615702479338843,
+      "grad_norm": 3.390625,
+      "learning_rate": 0.00019998497789012473,
+      "loss": 5.6115,
+      "step": 4200
+    },
+    {
+      "epoch": 0.03658746556473829,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.0001999845046448669,
+      "loss": 5.62,
+      "step": 4250
+    },
+    {
+      "epoch": 0.037017906336088154,
+      "grad_norm": 1.9375,
+      "learning_rate": 0.0001999840240609616,
+      "loss": 5.4771,
+      "step": 4300
+    },
+    {
+      "epoch": 0.03744834710743802,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.00019998353613844406,
+      "loss": 5.2885,
+      "step": 4350
+    },
+    {
+      "epoch": 0.03787878787878788,
+      "grad_norm": 1.2421875,
+      "learning_rate": 0.0001999830408773501,
+      "loss": 5.4431,
+      "step": 4400
+    },
+    {
+      "epoch": 0.03830922865013774,
+      "grad_norm": 1.1875,
+      "learning_rate": 0.00019998253827771602,
+      "loss": 5.2422,
+      "step": 4450
+    },
+    {
+      "epoch": 0.038739669421487606,
+      "grad_norm": 1.28125,
+      "learning_rate": 0.00019998202833957875,
+      "loss": 5.8202,
+      "step": 4500
+    },
+    {
+      "epoch": 0.03917011019283746,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.00019998151106297575,
+      "loss": 5.9591,
+      "step": 4550
+    },
+    {
+      "epoch": 0.039600550964187325,
+      "grad_norm": 0.67578125,
+      "learning_rate": 0.00019998098644794494,
+      "loss": 5.626,
+      "step": 4600
+    },
+    {
+      "epoch": 0.04003099173553719,
+      "grad_norm": 2.234375,
+      "learning_rate": 0.00019998045449452483,
+      "loss": 5.3955,
+      "step": 4650
+    },
+    {
+      "epoch": 0.04046143250688705,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.0001999799152027545,
+      "loss": 5.6826,
+      "step": 4700
+    },
+    {
+      "epoch": 0.040891873278236915,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.00019997936857267344,
+      "loss": 6.3792,
+      "step": 4750
+    },
+    {
+      "epoch": 0.04132231404958678,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.00019997881460432192,
+      "loss": 5.5136,
+      "step": 4800
+    },
+    {
+      "epoch": 0.04175275482093664,
+      "grad_norm": 2.296875,
+      "learning_rate": 0.00019997825329774046,
+      "loss": 5.5313,
+      "step": 4850
+    },
+    {
+      "epoch": 0.042183195592286504,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.00019997768465297036,
+      "loss": 5.5242,
+      "step": 4900
+    },
+    {
+      "epoch": 0.04261363636363636,
+      "grad_norm": 2.234375,
+      "learning_rate": 0.00019997710867005326,
+      "loss": 5.4816,
+      "step": 4950
+    },
+    {
+      "epoch": 0.043044077134986224,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.00019997652534903154,
+      "loss": 5.5978,
+      "step": 5000
+    },
+    {
+      "epoch": 0.04347451790633609,
+      "grad_norm": 1.9921875,
+      "learning_rate": 0.00019997593468994794,
+      "loss": 5.6367,
+      "step": 5050
+    },
+    {
+      "epoch": 0.04390495867768595,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.00019997533669284585,
+      "loss": 5.5829,
+      "step": 5100
+    },
+    {
+      "epoch": 0.04433539944903581,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.00019997473135776914,
+      "loss": 5.4658,
+      "step": 5150
+    },
+    {
+      "epoch": 0.044765840220385676,
+      "grad_norm": 1.8984375,
+      "learning_rate": 0.00019997411868476225,
+      "loss": 5.6926,
+      "step": 5200
+    },
+    {
+      "epoch": 0.04519628099173554,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00019997349867387017,
+      "loss": 5.8453,
+      "step": 5250
+    },
+    {
+      "epoch": 0.0456267217630854,
+      "grad_norm": 1.5078125,
+      "learning_rate": 0.00019997287132513835,
+      "loss": 5.5398,
+      "step": 5300
+    },
+    {
+      "epoch": 0.04605716253443526,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.0001999722366386129,
+      "loss": 5.5623,
+      "step": 5350
+    },
+    {
+      "epoch": 0.04648760330578512,
+      "grad_norm": 1.578125,
+      "learning_rate": 0.00019997159461434034,
+      "loss": 5.9666,
+      "step": 5400
+    },
+    {
+      "epoch": 0.046918044077134985,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00019997094525236786,
+      "loss": 5.7107,
+      "step": 5450
+    },
+    {
+      "epoch": 0.04734848484848485,
+      "grad_norm": 3.265625,
+      "learning_rate": 0.00019997028855274309,
+      "loss": 5.5953,
+      "step": 5500
+    },
+    {
+      "epoch": 0.04777892561983471,
+      "grad_norm": 2.578125,
+      "learning_rate": 0.0001999696245155142,
+      "loss": 5.4392,
+      "step": 5550
+    },
+    {
+      "epoch": 0.048209366391184574,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.00019996895314073002,
+      "loss": 5.6254,
+      "step": 5600
+    },
+    {
+      "epoch": 0.04863980716253444,
+      "grad_norm": 1.8828125,
+      "learning_rate": 0.00019996827442843974,
+      "loss": 5.3142,
+      "step": 5650
+    },
+    {
+      "epoch": 0.0490702479338843,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.0001999675883786932,
+      "loss": 5.831,
+      "step": 5700
+    },
+    {
+      "epoch": 0.04950068870523416,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.00019996689499154075,
+      "loss": 5.4863,
+      "step": 5750
+    },
+    {
+      "epoch": 0.04993112947658402,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.0001999661942670333,
+      "loss": 5.5451,
+      "step": 5800
+    },
+    {
+      "epoch": 0.05036157024793388,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.0001999654862052223,
+      "loss": 6.0222,
+      "step": 5850
+    },
+    {
+      "epoch": 0.050792011019283746,
+      "grad_norm": 2.125,
+      "learning_rate": 0.00019996477080615966,
+      "loss": 5.6131,
+      "step": 5900
+    },
+    {
+      "epoch": 0.05122245179063361,
+      "grad_norm": 3.609375,
+      "learning_rate": 0.00019996404806989793,
+      "loss": 5.7678,
+      "step": 5950
+    },
+    {
+      "epoch": 0.05165289256198347,
+      "grad_norm": 0.89453125,
+      "learning_rate": 0.00019996331799649014,
+      "loss": 5.6118,
+      "step": 6000
+    },
+    {
+      "epoch": 0.05165289256198347,
+      "eval_loss": 6.263210296630859,
+      "eval_runtime": 21.7431,
+      "eval_samples_per_second": 29.435,
+      "eval_steps_per_second": 14.717,
+      "eval_tts_loss": 5.9185563337741085,
+      "step": 6000
+    },
+    {
+      "epoch": 0.052083333333333336,
+      "grad_norm": 3.578125,
+      "learning_rate": 0.00019996258058598988,
+      "loss": 6.1941,
+      "step": 6050
+    },
+    {
+      "epoch": 0.0525137741046832,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.00019996183583845131,
+      "loss": 5.6182,
+      "step": 6100
+    },
+    {
+      "epoch": 0.052944214876033055,
+      "grad_norm": 1.6328125,
+      "learning_rate": 0.00019996108375392905,
+      "loss": 4.9397,
+      "step": 6150
+    },
+    {
+      "epoch": 0.05337465564738292,
+      "grad_norm": 2.296875,
+      "learning_rate": 0.0001999603243324783,
+      "loss": 5.3367,
+      "step": 6200
+    },
+    {
+      "epoch": 0.05380509641873278,
+      "grad_norm": 3.015625,
+      "learning_rate": 0.00019995955757415486,
+      "loss": 5.5759,
+      "step": 6250
+    },
+    {
+      "epoch": 0.054235537190082644,
+      "grad_norm": 0.9765625,
+      "learning_rate": 0.0001999587834790149,
+      "loss": 5.4017,
+      "step": 6300
+    },
+    {
+      "epoch": 0.05466597796143251,
+      "grad_norm": 1.6328125,
+      "learning_rate": 0.00019995800204711538,
+      "loss": 5.6686,
+      "step": 6350
+    },
+    {
+      "epoch": 0.05509641873278237,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.0001999572132785135,
+      "loss": 5.6747,
+      "step": 6400
+    },
+    {
+      "epoch": 0.055526859504132234,
+      "grad_norm": 1.8359375,
+      "learning_rate": 0.00019995641717326725,
+      "loss": 6.1438,
+      "step": 6450
+    },
+    {
+      "epoch": 0.0559573002754821,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00019995561373143506,
+      "loss": 5.768,
+      "step": 6500
+    },
+    {
+      "epoch": 0.05638774104683195,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.00019995480295307587,
+      "loss": 5.3983,
+      "step": 6550
+    },
+    {
+      "epoch": 0.056818181818181816,
+      "grad_norm": 2.234375,
+      "learning_rate": 0.00019995398483824917,
+      "loss": 5.8767,
+      "step": 6600
+    },
+    {
+      "epoch": 0.05724862258953168,
+      "grad_norm": 1.8515625,
+      "learning_rate": 0.0001999531593870151,
+      "loss": 5.8295,
+      "step": 6650
+    },
+    {
+      "epoch": 0.05767906336088154,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.00019995232659943411,
+      "loss": 5.8067,
+      "step": 6700
+    },
+    {
+      "epoch": 0.058109504132231406,
+      "grad_norm": 2.875,
+      "learning_rate": 0.00019995148647556748,
+      "loss": 5.8018,
+      "step": 6750
+    },
+    {
+      "epoch": 0.05853994490358127,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.00019995063901547674,
+      "loss": 5.6411,
+      "step": 6800
+    },
+    {
+      "epoch": 0.05897038567493113,
+      "grad_norm": 1.9140625,
+      "learning_rate": 0.00019994978421922417,
+      "loss": 5.5583,
+      "step": 6850
+    },
+    {
+      "epoch": 0.059400826446280995,
+      "grad_norm": 0.8984375,
+      "learning_rate": 0.0001999489220868725,
+      "loss": 5.5745,
+      "step": 6900
+    },
+    {
+      "epoch": 0.05983126721763085,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00019994805261848496,
+      "loss": 5.3674,
+      "step": 6950
+    },
+    {
+      "epoch": 0.060261707988980714,
+      "grad_norm": 1.9609375,
+      "learning_rate": 0.0001999471758141254,
+      "loss": 4.8269,
+      "step": 7000
+    },
+    {
+      "epoch": 0.06069214876033058,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00019994629167385824,
+      "loss": 5.3259,
+      "step": 7050
+    },
+    {
+      "epoch": 0.06112258953168044,
+      "grad_norm": 2.25,
+      "learning_rate": 0.00019994540019774824,
+      "loss": 5.4702,
+      "step": 7100
+    },
+    {
+      "epoch": 0.061553030303030304,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.0001999445013858609,
+      "loss": 5.229,
+      "step": 7150
+    },
+    {
+      "epoch": 0.06198347107438017,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.00019994359523826225,
+      "loss": 5.6542,
+      "step": 7200
+    },
+    {
+      "epoch": 0.06241391184573003,
+      "grad_norm": 3.34375,
+      "learning_rate": 0.00019994268175501872,
+      "loss": 5.5749,
+      "step": 7250
+    },
+    {
+      "epoch": 0.0628443526170799,
+      "grad_norm": 1.6328125,
+      "learning_rate": 0.00019994176093619736,
+      "loss": 5.0867,
+      "step": 7300
+    },
+    {
+      "epoch": 0.06327479338842976,
+      "grad_norm": 1.4765625,
+      "learning_rate": 0.0001999408327818658,
+      "loss": 5.463,
+      "step": 7350
+    },
+    {
+      "epoch": 0.06370523415977962,
+      "grad_norm": 1.7890625,
+      "learning_rate": 0.00019993989729209216,
+      "loss": 5.404,
+      "step": 7400
+    },
+    {
+      "epoch": 0.06413567493112948,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.00019993895446694505,
+      "loss": 5.2026,
+      "step": 7450
+    },
+    {
+      "epoch": 0.06456611570247933,
+      "grad_norm": 1.4375,
+      "learning_rate": 0.00019993800430649373,
+      "loss": 5.4745,
+      "step": 7500
+    },
+    {
+      "epoch": 0.0649965564738292,
+      "grad_norm": 1.1875,
+      "learning_rate": 0.00019993704681080788,
+      "loss": 5.2587,
+      "step": 7550
+    },
+    {
+      "epoch": 0.06542699724517906,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.00019993608197995785,
+      "loss": 5.2917,
+      "step": 7600
+    },
+    {
+      "epoch": 0.06585743801652892,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.00019993510981401442,
+      "loss": 5.5237,
+      "step": 7650
+    },
+    {
+      "epoch": 0.06628787878787878,
+      "grad_norm": 3.6875,
+      "learning_rate": 0.00019993413031304893,
+      "loss": 5.6417,
+      "step": 7700
+    },
+    {
+      "epoch": 0.06671831955922865,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.0001999331434771333,
+      "loss": 5.3945,
+      "step": 7750
+    },
+    {
+      "epoch": 0.06714876033057851,
+      "grad_norm": 0.984375,
+      "learning_rate": 0.00019993214930633994,
+      "loss": 4.9617,
+      "step": 7800
+    },
+    {
+      "epoch": 0.06757920110192837,
+      "grad_norm": 2.140625,
+      "learning_rate": 0.0001999311478007418,
+      "loss": 5.226,
+      "step": 7850
+    },
+    {
+      "epoch": 0.06800964187327824,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.0001999301389604125,
+      "loss": 5.7361,
+      "step": 7900
+    },
+    {
+      "epoch": 0.0684400826446281,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.00019992912278542595,
+      "loss": 5.7224,
+      "step": 7950
+    },
+    {
+      "epoch": 0.06887052341597796,
+      "grad_norm": 3.296875,
+      "learning_rate": 0.0001999280992758568,
+      "loss": 5.4668,
+      "step": 8000
+    },
+    {
+      "epoch": 0.06930096418732783,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.00019992706843178018,
+      "loss": 5.6261,
+      "step": 8050
+    },
+    {
+      "epoch": 0.06973140495867769,
+      "grad_norm": 2.9375,
+      "learning_rate": 0.0001999260302532717,
+      "loss": 5.6339,
+      "step": 8100
+    },
+    {
+      "epoch": 0.07016184573002755,
+      "grad_norm": 1.3203125,
+      "learning_rate": 0.0001999249847404076,
+      "loss": 5.1293,
+      "step": 8150
+    },
+    {
+      "epoch": 0.07059228650137742,
+      "grad_norm": 1.9296875,
+      "learning_rate": 0.00019992393189326464,
+      "loss": 5.2342,
+      "step": 8200
+    },
+    {
+      "epoch": 0.07102272727272728,
+      "grad_norm": 1.984375,
+      "learning_rate": 0.00019992287171192007,
+      "loss": 5.1224,
+      "step": 8250
+    },
+    {
+      "epoch": 0.07145316804407713,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.0001999218041964517,
+      "loss": 5.6905,
+      "step": 8300
+    },
+    {
+      "epoch": 0.07188360881542699,
+      "grad_norm": 1.8984375,
+      "learning_rate": 0.00019992072934693788,
+      "loss": 5.5532,
+      "step": 8350
+    },
+    {
+      "epoch": 0.07231404958677685,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00019991964716345752,
+      "loss": 5.9902,
+      "step": 8400
+    },
+    {
+      "epoch": 0.07274449035812672,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.00019991855764609005,
+      "loss": 5.455,
+      "step": 8450
+    },
+    {
+      "epoch": 0.07317493112947658,
+      "grad_norm": 1.2109375,
+      "learning_rate": 0.0001999174607949154,
+      "loss": 5.5104,
+      "step": 8500
+    },
+    {
+      "epoch": 0.07360537190082644,
+      "grad_norm": 1.7890625,
+      "learning_rate": 0.00019991635661001413,
+      "loss": 5.6288,
+      "step": 8550
+    },
+    {
+      "epoch": 0.07403581267217631,
+      "grad_norm": 1.21875,
+      "learning_rate": 0.00019991524509146725,
+      "loss": 5.6406,
+      "step": 8600
+    },
+    {
+      "epoch": 0.07446625344352617,
+      "grad_norm": 2.421875,
+      "learning_rate": 0.00019991412623935638,
+      "loss": 5.186,
+      "step": 8650
+    },
+    {
+      "epoch": 0.07489669421487603,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.00019991300005376357,
+      "loss": 5.3564,
+      "step": 8700
+    },
+    {
+      "epoch": 0.0753271349862259,
+      "grad_norm": 3.46875,
+      "learning_rate": 0.00019991186653477155,
+      "loss": 5.5007,
+      "step": 8750
+    },
+    {
+      "epoch": 0.07575757575757576,
+      "grad_norm": 2.0,
+      "learning_rate": 0.0001999107256824635,
+      "loss": 5.2855,
+      "step": 8800
+    },
+    {
+      "epoch": 0.07618801652892562,
+      "grad_norm": 1.890625,
+      "learning_rate": 0.00019990957749692313,
+      "loss": 5.382,
+      "step": 8850
+    },
+    {
+      "epoch": 0.07661845730027549,
+      "grad_norm": 1.8046875,
+      "learning_rate": 0.00019990842197823474,
+      "loss": 5.5644,
+      "step": 8900
+    },
+    {
+      "epoch": 0.07704889807162535,
+      "grad_norm": 2.125,
+      "learning_rate": 0.00019990725912648314,
+      "loss": 5.2274,
+      "step": 8950
+    },
+    {
+      "epoch": 0.07747933884297521,
+      "grad_norm": 1.75,
+      "learning_rate": 0.00019990608894175367,
+      "loss": 5.3645,
+      "step": 9000
+    },
+    {
+      "epoch": 0.07747933884297521,
+      "eval_loss": 6.056007385253906,
+      "eval_runtime": 21.6664,
+      "eval_samples_per_second": 29.539,
+      "eval_steps_per_second": 14.769,
+      "eval_tts_loss": 6.075921739670477,
+      "step": 9000
+    },
+    {
+      "epoch": 0.07790977961432508,
+      "grad_norm": 3.390625,
+      "learning_rate": 0.0001999049114241322,
+      "loss": 5.5031,
+      "step": 9050
+    },
+    {
+      "epoch": 0.07834022038567492,
+      "grad_norm": 2.234375,
+      "learning_rate": 0.00019990372657370523,
+      "loss": 5.5828,
+      "step": 9100
+    },
+    {
+      "epoch": 0.07877066115702479,
+      "grad_norm": 1.8671875,
+      "learning_rate": 0.00019990253439055965,
+      "loss": 5.2726,
+      "step": 9150
+    },
+    {
+      "epoch": 0.07920110192837465,
+      "grad_norm": 2.375,
+      "learning_rate": 0.000199901334874783,
+      "loss": 4.9601,
+      "step": 9200
+    },
+    {
+      "epoch": 0.07963154269972451,
+      "grad_norm": 1.984375,
+      "learning_rate": 0.00019990012802646333,
+      "loss": 5.3289,
+      "step": 9250
+    },
+    {
+      "epoch": 0.08006198347107438,
+      "grad_norm": 0.8828125,
+      "learning_rate": 0.00019989891384568917,
+      "loss": 5.1188,
+      "step": 9300
+    },
+    {
+      "epoch": 0.08049242424242424,
+      "grad_norm": 1.984375,
+      "learning_rate": 0.0001998976923325497,
+      "loss": 5.2473,
+      "step": 9350
+    },
+    {
+      "epoch": 0.0809228650137741,
+      "grad_norm": 1.9921875,
+      "learning_rate": 0.00019989646348713452,
+      "loss": 5.2454,
+      "step": 9400
+    },
+    {
+      "epoch": 0.08135330578512397,
+      "grad_norm": 1.9296875,
+      "learning_rate": 0.00019989522730953386,
+      "loss": 5.3245,
+      "step": 9450
+    },
+    {
+      "epoch": 0.08178374655647383,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.00019989398379983845,
+      "loss": 5.3451,
+      "step": 9500
+    },
+    {
+      "epoch": 0.08221418732782369,
+      "grad_norm": 1.34375,
+      "learning_rate": 0.00019989273295813957,
+      "loss": 5.5107,
+      "step": 9550
+    },
+    {
+      "epoch": 0.08264462809917356,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.000199891474784529,
+      "loss": 5.5401,
+      "step": 9600
+    },
+    {
+      "epoch": 0.08307506887052342,
+      "grad_norm": 2.375,
+      "learning_rate": 0.0001998902092790991,
+      "loss": 5.4677,
+      "step": 9650
+    },
+    {
+      "epoch": 0.08350550964187328,
+      "grad_norm": 3.140625,
+      "learning_rate": 0.00019988893644194277,
+      "loss": 5.4752,
+      "step": 9700
+    },
+    {
+      "epoch": 0.08393595041322315,
+      "grad_norm": 1.96875,
+      "learning_rate": 0.00019988765627315343,
+      "loss": 5.2186,
+      "step": 9750
+    },
+    {
+      "epoch": 0.08436639118457301,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.000199886368772825,
+      "loss": 5.6908,
+      "step": 9800
+    },
+    {
+      "epoch": 0.08479683195592287,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.00019988507394105205,
+      "loss": 4.9511,
+      "step": 9850
+    },
+    {
+      "epoch": 0.08522727272727272,
+      "grad_norm": 1.3359375,
+      "learning_rate": 0.00019988377177792955,
+      "loss": 5.3783,
+      "step": 9900
+    },
+    {
+      "epoch": 0.08565771349862258,
+      "grad_norm": 1.90625,
+      "learning_rate": 0.00019988246228355315,
+      "loss": 5.4935,
+      "step": 9950
+    },
+    {
+      "epoch": 0.08608815426997245,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.00019988114545801887,
+      "loss": 5.4467,
+      "step": 10000
+    },
+    {
+      "epoch": 0.08651859504132231,
+      "grad_norm": 1.984375,
+      "learning_rate": 0.00019987982130142347,
+      "loss": 5.1502,
+      "step": 10050
+    },
+    {
+      "epoch": 0.08694903581267217,
+      "grad_norm": 2.375,
+      "learning_rate": 0.00019987848981386404,
+      "loss": 5.5199,
+      "step": 10100
+    },
+    {
+      "epoch": 0.08737947658402204,
+      "grad_norm": 1.8203125,
+      "learning_rate": 0.00019987715099543835,
+      "loss": 5.4569,
+      "step": 10150
+    },
+    {
+      "epoch": 0.0878099173553719,
+      "grad_norm": 1.609375,
+      "learning_rate": 0.00019987580484624468,
+      "loss": 5.3158,
+      "step": 10200
+    },
+    {
+      "epoch": 0.08824035812672176,
+      "grad_norm": 2.265625,
+      "learning_rate": 0.00019987445136638183,
+      "loss": 5.4302,
+      "step": 10250
+    },
+    {
+      "epoch": 0.08867079889807163,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.00019987309055594912,
+      "loss": 5.4416,
+      "step": 10300
+    },
+    {
+      "epoch": 0.08910123966942149,
+      "grad_norm": 0.46875,
+      "learning_rate": 0.00019987172241504647,
+      "loss": 5.0894,
+      "step": 10350
+    },
+    {
+      "epoch": 0.08953168044077135,
+      "grad_norm": 1.9609375,
+      "learning_rate": 0.00019987034694377424,
+      "loss": 5.3113,
+      "step": 10400
+    },
+    {
+      "epoch": 0.08996212121212122,
+      "grad_norm": 1.796875,
+      "learning_rate": 0.00019986896414223346,
+      "loss": 5.2562,
+      "step": 10450
+    },
+    {
+      "epoch": 0.09039256198347108,
+      "grad_norm": 1.7734375,
+      "learning_rate": 0.00019986757401052553,
+      "loss": 5.2811,
+      "step": 10500
+    },
+    {
+      "epoch": 0.09082300275482094,
+      "grad_norm": 1.9140625,
+      "learning_rate": 0.0001998661765487526,
+      "loss": 5.3296,
+      "step": 10550
+    },
+    {
+      "epoch": 0.0912534435261708,
+      "grad_norm": 0.8828125,
+      "learning_rate": 0.00019986477175701713,
+      "loss": 5.5743,
+      "step": 10600
+    },
+    {
+      "epoch": 0.09168388429752067,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.00019986335963542232,
+      "loss": 5.4655,
+      "step": 10650
+    },
+    {
+      "epoch": 0.09211432506887052,
+      "grad_norm": 1.859375,
+      "learning_rate": 0.00019986194018407173,
+      "loss": 5.0217,
+      "step": 10700
+    },
+    {
+      "epoch": 0.09254476584022038,
+      "grad_norm": 1.640625,
+      "learning_rate": 0.00019986051340306965,
+      "loss": 4.9831,
+      "step": 10750
+    },
+    {
+      "epoch": 0.09297520661157024,
+      "grad_norm": 1.328125,
+      "learning_rate": 0.0001998590792925207,
+      "loss": 5.1353,
+      "step": 10800
+    },
+    {
+      "epoch": 0.0934056473829201,
+      "grad_norm": 0.90234375,
+      "learning_rate": 0.0001998576378525302,
+      "loss": 5.5117,
+      "step": 10850
+    },
+    {
+      "epoch": 0.09383608815426997,
+      "grad_norm": 1.8984375,
+      "learning_rate": 0.0001998561890832039,
+      "loss": 5.5137,
+      "step": 10900
+    },
+    {
+      "epoch": 0.09426652892561983,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.00019985473298464817,
+      "loss": 5.1704,
+      "step": 10950
+    },
+    {
+      "epoch": 0.0946969696969697,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.0001998532695569699,
+      "loss": 5.1635,
+      "step": 11000
+    },
+    {
+      "epoch": 0.09512741046831956,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00019985179880027646,
+      "loss": 5.5797,
+      "step": 11050
+    },
+    {
+      "epoch": 0.09555785123966942,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.00019985032071467588,
+      "loss": 5.4581,
+      "step": 11100
+    },
+    {
+      "epoch": 0.09598829201101929,
+      "grad_norm": 3.015625,
+      "learning_rate": 0.00019984883530027654,
+      "loss": 5.4119,
+      "step": 11150
+    },
+    {
+      "epoch": 0.09641873278236915,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.00019984734255718754,
+      "loss": 5.5263,
+      "step": 11200
+    },
+    {
+      "epoch": 0.09684917355371901,
+      "grad_norm": 1.78125,
+      "learning_rate": 0.00019984584248551842,
+      "loss": 5.3163,
+      "step": 11250
+    },
+    {
+      "epoch": 0.09727961432506887,
+      "grad_norm": 1.9453125,
+      "learning_rate": 0.0001998443350853793,
+      "loss": 5.3459,
+      "step": 11300
+    },
+    {
+      "epoch": 0.09771005509641874,
+      "grad_norm": 1.7109375,
+      "learning_rate": 0.00019984282035688078,
+      "loss": 5.01,
+      "step": 11350
+    },
+    {
+      "epoch": 0.0981404958677686,
+      "grad_norm": 1.84375,
+      "learning_rate": 0.0001998412983001341,
+      "loss": 5.1599,
+      "step": 11400
+    },
+    {
+      "epoch": 0.09857093663911845,
+      "grad_norm": 3.125,
+      "learning_rate": 0.0001998397689152509,
+      "loss": 5.3925,
+      "step": 11450
+    },
+    {
+      "epoch": 0.09900137741046831,
+      "grad_norm": 1.484375,
+      "learning_rate": 0.0001998382322023435,
+      "loss": 5.4049,
+      "step": 11500
+    },
+    {
+      "epoch": 0.09943181818181818,
+      "grad_norm": 2.578125,
+      "learning_rate": 0.00019983668816152462,
+      "loss": 5.404,
+      "step": 11550
+    },
+    {
+      "epoch": 0.09986225895316804,
+      "grad_norm": 1.8984375,
+      "learning_rate": 0.0001998351367929077,
+      "loss": 5.8699,
+      "step": 11600
+    },
+    {
+      "epoch": 0.1002926997245179,
+      "grad_norm": 1.921875,
+      "learning_rate": 0.00019983357809660648,
+      "loss": 5.0001,
+      "step": 11650
+    },
+    {
+      "epoch": 0.10072314049586777,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.00019983201207273546,
+      "loss": 5.2185,
+      "step": 11700
+    },
+    {
+      "epoch": 0.10115358126721763,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.0001998304387214095,
+      "loss": 5.5458,
+      "step": 11750
+    },
+    {
+      "epoch": 0.10158402203856749,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.00019982885804274418,
+      "loss": 5.4759,
+      "step": 11800
+    },
+    {
+      "epoch": 0.10201446280991736,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.00019982727003685548,
+      "loss": 4.8674,
+      "step": 11850
+    },
+    {
+      "epoch": 0.10244490358126722,
+      "grad_norm": 3.34375,
+      "learning_rate": 0.0001998256747038599,
+      "loss": 5.4233,
+      "step": 11900
+    },
+    {
+      "epoch": 0.10287534435261708,
+      "grad_norm": 1.7890625,
+      "learning_rate": 0.0001998240720438746,
+      "loss": 5.2329,
+      "step": 11950
+    },
+    {
+      "epoch": 0.10330578512396695,
+      "grad_norm": 1.953125,
+      "learning_rate": 0.00019982246205701719,
+      "loss": 5.3219,
+      "step": 12000
+    },
+    {
+      "epoch": 0.10330578512396695,
+      "eval_loss": 5.914431095123291,
+      "eval_runtime": 21.8067,
+      "eval_samples_per_second": 29.349,
+      "eval_steps_per_second": 14.674,
+      "eval_tts_loss": 6.132826582490758,
+      "step": 12000
+    },
+    {
+      "epoch": 0.10373622589531681,
+      "grad_norm": 1.8125,
+      "learning_rate": 0.00019982084474340584,
+      "loss": 5.1588,
+      "step": 12050
+    },
+    {
+      "epoch": 0.10416666666666667,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00019981922010315923,
+      "loss": 5.4076,
+      "step": 12100
+    },
+    {
+      "epoch": 0.10459710743801653,
+      "grad_norm": 2.125,
+      "learning_rate": 0.00019981758813639665,
+      "loss": 5.226,
+      "step": 12150
+    },
+    {
+      "epoch": 0.1050275482093664,
+      "grad_norm": 1.203125,
+      "learning_rate": 0.00019981594884323786,
+      "loss": 5.5601,
+      "step": 12200
+    },
+    {
+      "epoch": 0.10545798898071625,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.00019981430222380318,
+      "loss": 5.1232,
+      "step": 12250
+    },
+    {
+      "epoch": 0.10588842975206611,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.00019981264827821348,
+      "loss": 5.0244,
+      "step": 12300
+    },
+    {
+      "epoch": 0.10631887052341597,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.00019981098700659015,
+      "loss": 5.4677,
+      "step": 12350
+    },
+    {
+      "epoch": 0.10674931129476584,
+      "grad_norm": 1.9375,
+      "learning_rate": 0.00019980931840905513,
+      "loss": 4.7182,
+      "step": 12400
+    },
+    {
+      "epoch": 0.1071797520661157,
+      "grad_norm": 3.234375,
+      "learning_rate": 0.00019980764248573088,
+      "loss": 5.2571,
+      "step": 12450
+    },
+    {
+      "epoch": 0.10761019283746556,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.00019980595923674042,
+      "loss": 5.2264,
+      "step": 12500
+    },
+    {
+      "epoch": 0.10804063360881543,
+      "grad_norm": 1.3984375,
+      "learning_rate": 0.00019980426866220724,
+      "loss": 5.1868,
+      "step": 12550
+    },
+    {
+      "epoch": 0.10847107438016529,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.00019980257076225553,
+      "loss": 5.0848,
+      "step": 12600
+    },
+    {
+      "epoch": 0.10890151515151515,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.00019980086553700983,
+      "loss": 5.2408,
+      "step": 12650
+    },
+    {
+      "epoch": 0.10933195592286502,
+      "grad_norm": 3.46875,
+      "learning_rate": 0.00019979915298659533,
+      "loss": 5.0821,
+      "step": 12700
+    },
+    {
+      "epoch": 0.10976239669421488,
+      "grad_norm": 3.328125,
+      "learning_rate": 0.00019979743311113773,
+      "loss": 5.2788,
+      "step": 12750
+    },
+    {
+      "epoch": 0.11019283746556474,
+      "grad_norm": 1.4765625,
+      "learning_rate": 0.00019979570591076325,
+      "loss": 4.6787,
+      "step": 12800
+    },
+    {
+      "epoch": 0.1106232782369146,
+      "grad_norm": 1.9453125,
+      "learning_rate": 0.0001997939713855987,
+      "loss": 5.2985,
+      "step": 12850
+    },
+    {
+      "epoch": 0.11105371900826447,
+      "grad_norm": 1.6796875,
+      "learning_rate": 0.00019979222953577135,
+      "loss": 5.2763,
+      "step": 12900
+    },
+    {
+      "epoch": 0.11148415977961433,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.00019979048036140905,
+      "loss": 4.944,
+      "step": 12950
+    },
+    {
+      "epoch": 0.1119146005509642,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.00019978872386264022,
+      "loss": 5.0702,
+      "step": 13000
+    },
+    {
+      "epoch": 0.11234504132231404,
+      "grad_norm": 1.7734375,
+      "learning_rate": 0.00019978696003959372,
+      "loss": 4.8859,
+      "step": 13050
+    },
+    {
+      "epoch": 0.1127754820936639,
+      "grad_norm": 3.015625,
+      "learning_rate": 0.00019978518889239908,
+      "loss": 5.0737,
+      "step": 13100
+    },
+    {
+      "epoch": 0.11320592286501377,
+      "grad_norm": 1.9765625,
+      "learning_rate": 0.00019978341042118625,
+      "loss": 4.9093,
+      "step": 13150
+    },
+    {
+      "epoch": 0.11363636363636363,
+      "grad_norm": 1.921875,
+      "learning_rate": 0.00019978162462608578,
+      "loss": 5.3666,
+      "step": 13200
+    },
+    {
+      "epoch": 0.1140668044077135,
+      "grad_norm": 1.7890625,
+      "learning_rate": 0.0001997798315072288,
+      "loss": 5.7781,
+      "step": 13250
+    },
+    {
+      "epoch": 0.11449724517906336,
+      "grad_norm": 3.078125,
+      "learning_rate": 0.00019977803106474681,
+      "loss": 5.461,
+      "step": 13300
+    },
+    {
+      "epoch": 0.11492768595041322,
+      "grad_norm": 2.296875,
+      "learning_rate": 0.00019977622329877206,
+      "loss": 5.0734,
+      "step": 13350
+    },
+    {
+      "epoch": 0.11535812672176309,
+      "grad_norm": 1.96875,
+      "learning_rate": 0.00019977440820943715,
+      "loss": 5.5124,
+      "step": 13400
+    },
+    {
+      "epoch": 0.11578856749311295,
+      "grad_norm": 2.625,
+      "learning_rate": 0.0001997725857968754,
+      "loss": 5.1575,
+      "step": 13450
+    },
+    {
+      "epoch": 0.11621900826446281,
+      "grad_norm": 1.2734375,
+      "learning_rate": 0.00019977075606122046,
+      "loss": 4.9595,
+      "step": 13500
+    },
+    {
+      "epoch": 0.11664944903581267,
+      "grad_norm": 2.578125,
+      "learning_rate": 0.00019976891900260672,
+      "loss": 5.2753,
+      "step": 13550
+    },
+    {
+      "epoch": 0.11707988980716254,
+      "grad_norm": 2.234375,
+      "learning_rate": 0.00019976707462116897,
+      "loss": 5.3432,
+      "step": 13600
+    },
+    {
+      "epoch": 0.1175103305785124,
+      "grad_norm": 2.921875,
+      "learning_rate": 0.00019976522291704262,
+      "loss": 5.3526,
+      "step": 13650
+    },
+    {
+      "epoch": 0.11794077134986226,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.00019976336389036355,
+      "loss": 5.0664,
+      "step": 13700
+    },
+    {
+      "epoch": 0.11837121212121213,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.00019976149754126824,
+      "loss": 5.2789,
+      "step": 13750
+    },
+    {
+      "epoch": 0.11880165289256199,
+      "grad_norm": 1.671875,
+      "learning_rate": 0.0001997596238698936,
+      "loss": 4.8826,
+      "step": 13800
+    },
+    {
+      "epoch": 0.11923209366391184,
+      "grad_norm": 1.9375,
+      "learning_rate": 0.00019975774287637723,
+      "loss": 5.2363,
+      "step": 13850
+    },
+    {
+      "epoch": 0.1196625344352617,
+      "grad_norm": 1.7734375,
+      "learning_rate": 0.00019975585456085716,
+      "loss": 5.1902,
+      "step": 13900
+    },
+    {
+      "epoch": 0.12009297520661157,
+      "grad_norm": 1.921875,
+      "learning_rate": 0.000199753958923472,
+      "loss": 5.099,
+      "step": 13950
+    },
+    {
+      "epoch": 0.12052341597796143,
+      "grad_norm": 1.375,
+      "learning_rate": 0.00019975205596436092,
+      "loss": 5.3852,
+      "step": 14000
+    },
+    {
+      "epoch": 0.12095385674931129,
+      "grad_norm": 3.0625,
+      "learning_rate": 0.00019975014568366351,
+      "loss": 4.9896,
+      "step": 14050
+    },
+    {
+      "epoch": 0.12138429752066116,
+      "grad_norm": 2.625,
+      "learning_rate": 0.00019974822808152003,
+      "loss": 5.0135,
+      "step": 14100
+    },
+    {
+      "epoch": 0.12181473829201102,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.0001997463031580712,
+      "loss": 5.3912,
+      "step": 14150
+    },
+    {
+      "epoch": 0.12224517906336088,
+      "grad_norm": 0.609375,
+      "learning_rate": 0.00019974437091345836,
+      "loss": 5.245,
+      "step": 14200
+    },
+    {
+      "epoch": 0.12267561983471074,
+      "grad_norm": 1.4140625,
+      "learning_rate": 0.0001997424313478233,
+      "loss": 4.9216,
+      "step": 14250
+    },
+    {
+      "epoch": 0.12310606060606061,
+      "grad_norm": 1.3671875,
+      "learning_rate": 0.0001997404844613084,
+      "loss": 4.7488,
+      "step": 14300
+    },
+    {
+      "epoch": 0.12353650137741047,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.00019973853025405652,
+      "loss": 4.8481,
+      "step": 14350
+    },
+    {
+      "epoch": 0.12396694214876033,
+      "grad_norm": 2.578125,
+      "learning_rate": 0.0001997365687262111,
+      "loss": 4.7889,
+      "step": 14400
+    },
+    {
+      "epoch": 0.1243973829201102,
+      "grad_norm": 0.51953125,
+      "learning_rate": 0.00019973459987791614,
+      "loss": 5.1318,
+      "step": 14450
+    },
+    {
+      "epoch": 0.12482782369146006,
+      "grad_norm": 1.7734375,
+      "learning_rate": 0.00019973262370931612,
+      "loss": 5.2832,
+      "step": 14500
+    },
+    {
+      "epoch": 0.1252582644628099,
+      "grad_norm": 1.9921875,
+      "learning_rate": 0.0001997306402205561,
+      "loss": 5.3014,
+      "step": 14550
+    },
+    {
+      "epoch": 0.1256887052341598,
+      "grad_norm": 1.9765625,
+      "learning_rate": 0.00019972864941178166,
+      "loss": 5.1253,
+      "step": 14600
+    },
+    {
+      "epoch": 0.12611914600550964,
+      "grad_norm": 1.90625,
+      "learning_rate": 0.00019972665128313892,
+      "loss": 5.3287,
+      "step": 14650
+    },
+    {
+      "epoch": 0.1265495867768595,
+      "grad_norm": 1.578125,
+      "learning_rate": 0.00019972464583477458,
+      "loss": 5.1724,
+      "step": 14700
+    },
+    {
+      "epoch": 0.12698002754820936,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00019972263306683577,
+      "loss": 5.1616,
+      "step": 14750
+    },
+    {
+      "epoch": 0.12741046831955924,
+      "grad_norm": 2.375,
+      "learning_rate": 0.00019972061297947023,
+      "loss": 5.0238,
+      "step": 14800
+    },
+    {
+      "epoch": 0.1278409090909091,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0001997185855728263,
+      "loss": 5.0968,
+      "step": 14850
+    },
+    {
+      "epoch": 0.12827134986225897,
+      "grad_norm": 1.4765625,
+      "learning_rate": 0.00019971655084705271,
+      "loss": 5.3728,
+      "step": 14900
+    },
+    {
+      "epoch": 0.12870179063360881,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.00019971450880229886,
+      "loss": 4.845,
+      "step": 14950
+    },
+    {
+      "epoch": 0.12913223140495866,
+      "grad_norm": 1.8984375,
+      "learning_rate": 0.00019971245943871456,
+      "loss": 4.943,
+      "step": 15000
+    },
+    {
+      "epoch": 0.12913223140495866,
+      "eval_loss": 5.828057765960693,
+      "eval_runtime": 22.6847,
+      "eval_samples_per_second": 28.213,
+      "eval_steps_per_second": 14.106,
+      "eval_tts_loss": 6.258981663526243,
+      "step": 15000
+    },
+    {
+      "epoch": 0.12956267217630854,
+      "grad_norm": 0.8359375,
+      "learning_rate": 0.0001997104027564503,
+      "loss": 4.6706,
+      "step": 15050
+    },
+    {
+      "epoch": 0.1299931129476584,
+      "grad_norm": 1.984375,
+      "learning_rate": 0.00019970833875565704,
+      "loss": 5.4813,
+      "step": 15100
+    },
+    {
+      "epoch": 0.13042355371900827,
+      "grad_norm": 1.59375,
+      "learning_rate": 0.0001997062674364862,
+      "loss": 4.9743,
+      "step": 15150
+    },
+    {
+      "epoch": 0.13085399449035812,
+      "grad_norm": 3.53125,
+      "learning_rate": 0.00019970418879908987,
+      "loss": 5.1961,
+      "step": 15200
+    },
+    {
+      "epoch": 0.131284435261708,
+      "grad_norm": 1.5390625,
+      "learning_rate": 0.00019970210284362062,
+      "loss": 4.988,
+      "step": 15250
+    },
+    {
+      "epoch": 0.13171487603305784,
+      "grad_norm": 1.1875,
+      "learning_rate": 0.0001997000095702315,
+      "loss": 5.1799,
+      "step": 15300
+    },
+    {
+      "epoch": 0.13214531680440772,
+      "grad_norm": 1.5390625,
+      "learning_rate": 0.0001996979089790762,
+      "loss": 5.1497,
+      "step": 15350
+    },
+    {
+      "epoch": 0.13257575757575757,
+      "grad_norm": 1.2734375,
+      "learning_rate": 0.0001996958010703089,
+      "loss": 5.6535,
+      "step": 15400
+    },
+    {
+      "epoch": 0.13300619834710745,
+      "grad_norm": 1.5546875,
+      "learning_rate": 0.00019969368584408427,
+      "loss": 5.2685,
+      "step": 15450
+    },
+    {
+      "epoch": 0.1334366391184573,
+      "grad_norm": 1.7734375,
+      "learning_rate": 0.00019969156330055762,
+      "loss": 5.1259,
+      "step": 15500
+    },
+    {
+      "epoch": 0.13386707988980717,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.0001996894334398847,
+      "loss": 5.5331,
+      "step": 15550
+    },
+    {
+      "epoch": 0.13429752066115702,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.00019968729626222185,
+      "loss": 5.2926,
+      "step": 15600
+    },
+    {
+      "epoch": 0.1347279614325069,
+      "grad_norm": 2.125,
+      "learning_rate": 0.00019968515176772597,
+      "loss": 5.5779,
+      "step": 15650
+    },
+    {
+      "epoch": 0.13515840220385675,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.00019968299995655443,
+      "loss": 4.4885,
+      "step": 15700
+    },
+    {
+      "epoch": 0.13558884297520662,
+      "grad_norm": 3.109375,
+      "learning_rate": 0.00019968084082886515,
+      "loss": 5.2712,
+      "step": 15750
+    },
+    {
+      "epoch": 0.13601928374655647,
+      "grad_norm": 2.875,
+      "learning_rate": 0.0001996786743848166,
+      "loss": 5.0487,
+      "step": 15800
+    },
+    {
+      "epoch": 0.13644972451790632,
+      "grad_norm": 1.984375,
+      "learning_rate": 0.00019967650062456787,
+      "loss": 5.0838,
+      "step": 15850
+    },
+    {
+      "epoch": 0.1368801652892562,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.0001996743195482784,
+      "loss": 5.327,
+      "step": 15900
+    },
+    {
+      "epoch": 0.13731060606060605,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.00019967213115610835,
+      "loss": 4.9537,
+      "step": 15950
+    },
+    {
+      "epoch": 0.13774104683195593,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00019966993544821832,
+      "loss": 5.4254,
+      "step": 16000
+    },
+    {
+      "epoch": 0.13817148760330578,
+      "grad_norm": 1.984375,
+      "learning_rate": 0.00019966773242476946,
+      "loss": 5.1223,
+      "step": 16050
+    },
+    {
+      "epoch": 0.13860192837465565,
+      "grad_norm": 1.6328125,
+      "learning_rate": 0.0001996655220859235,
+      "loss": 5.2873,
+      "step": 16100
+    },
+    {
+      "epoch": 0.1390323691460055,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.00019966330443184265,
+      "loss": 5.3693,
+      "step": 16150
+    },
+    {
+      "epoch": 0.13946280991735538,
+      "grad_norm": 1.8828125,
+      "learning_rate": 0.0001996610794626897,
+      "loss": 5.4051,
+      "step": 16200
+    },
+    {
+      "epoch": 0.13989325068870523,
+      "grad_norm": 3.390625,
+      "learning_rate": 0.00019965884717862794,
+      "loss": 5.1706,
+      "step": 16250
+    },
+    {
+      "epoch": 0.1403236914600551,
+      "grad_norm": 3.8125,
+      "learning_rate": 0.00019965660757982119,
+      "loss": 4.7145,
+      "step": 16300
+    },
+    {
+      "epoch": 0.14075413223140495,
+      "grad_norm": 2.25,
+      "learning_rate": 0.00019965436066643388,
+      "loss": 5.057,
+      "step": 16350
+    },
+    {
+      "epoch": 0.14118457300275483,
+      "grad_norm": 2.625,
+      "learning_rate": 0.0001996521064386309,
+      "loss": 5.2056,
+      "step": 16400
+    },
+    {
+      "epoch": 0.14161501377410468,
+      "grad_norm": 1.9375,
+      "learning_rate": 0.00019964984489657773,
+      "loss": 5.1436,
+      "step": 16450
+    },
+    {
+      "epoch": 0.14204545454545456,
+      "grad_norm": 1.46875,
+      "learning_rate": 0.00019964757604044034,
+      "loss": 5.2476,
+      "step": 16500
+    },
+    {
+      "epoch": 0.1424758953168044,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00019964529987038528,
+      "loss": 5.2183,
+      "step": 16550
+    },
+    {
+      "epoch": 0.14290633608815426,
+      "grad_norm": 2.125,
+      "learning_rate": 0.0001996430163865796,
+      "loss": 5.3169,
+      "step": 16600
+    },
+    {
+      "epoch": 0.14333677685950413,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.00019964072558919087,
+      "loss": 5.2249,
+      "step": 16650
+    },
+    {
+      "epoch": 0.14376721763085398,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00019963842747838727,
+      "loss": 5.4004,
+      "step": 16700
+    },
+    {
+      "epoch": 0.14419765840220386,
+      "grad_norm": 1.2578125,
+      "learning_rate": 0.0001996361220543375,
+      "loss": 5.2227,
+      "step": 16750
+    },
+    {
+      "epoch": 0.1446280991735537,
+      "grad_norm": 1.9453125,
+      "learning_rate": 0.00019963380931721072,
+      "loss": 5.0179,
+      "step": 16800
+    },
+    {
+      "epoch": 0.1450585399449036,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.00019963148926717671,
+      "loss": 4.8251,
+      "step": 16850
+    },
+    {
+      "epoch": 0.14548898071625344,
+      "grad_norm": 3.546875,
+      "learning_rate": 0.00019962916190440576,
+      "loss": 5.2592,
+      "step": 16900
+    },
+    {
+      "epoch": 0.1459194214876033,
+      "grad_norm": 2.296875,
+      "learning_rate": 0.00019962682722906867,
+      "loss": 4.9115,
+      "step": 16950
+    },
+    {
+      "epoch": 0.14634986225895316,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.00019962448524133683,
+      "loss": 5.3176,
+      "step": 17000
+    },
+    {
+      "epoch": 0.14678030303030304,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.00019962213594138211,
+      "loss": 5.3436,
+      "step": 17050
+    },
+    {
+      "epoch": 0.1472107438016529,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.00019961977932937697,
+      "loss": 4.6579,
+      "step": 17100
+    },
+    {
+      "epoch": 0.14764118457300276,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00019961741540549434,
+      "loss": 4.5179,
+      "step": 17150
+    },
+    {
+      "epoch": 0.14807162534435261,
+      "grad_norm": 1.8125,
+      "learning_rate": 0.00019961504416990777,
+      "loss": 5.3218,
+      "step": 17200
+    },
+    {
+      "epoch": 0.1485020661157025,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.00019961266562279129,
+      "loss": 5.0783,
+      "step": 17250
+    },
+    {
+      "epoch": 0.14893250688705234,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.00019961027976431945,
+      "loss": 5.2808,
+      "step": 17300
+    },
+    {
+      "epoch": 0.14936294765840222,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.00019960788659466742,
+      "loss": 4.6248,
+      "step": 17350
+    },
+    {
+      "epoch": 0.14979338842975207,
+      "grad_norm": 1.71875,
+      "learning_rate": 0.00019960548611401083,
+      "loss": 5.1576,
+      "step": 17400
+    },
+    {
+      "epoch": 0.15022382920110192,
+      "grad_norm": 1.53125,
+      "learning_rate": 0.00019960307832252584,
+      "loss": 5.4072,
+      "step": 17450
+    },
+    {
+      "epoch": 0.1506542699724518,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.0001996006632203892,
+      "loss": 5.401,
+      "step": 17500
+    },
+    {
+      "epoch": 0.15108471074380164,
+      "grad_norm": 0.7890625,
+      "learning_rate": 0.00019959824080777817,
+      "loss": 5.081,
+      "step": 17550
+    },
+    {
+      "epoch": 0.15151515151515152,
+      "grad_norm": 1.890625,
+      "learning_rate": 0.00019959581108487057,
+      "loss": 5.4486,
+      "step": 17600
+    },
+    {
+      "epoch": 0.15194559228650137,
+      "grad_norm": 1.96875,
+      "learning_rate": 0.00019959337405184475,
+      "loss": 5.3831,
+      "step": 17650
+    },
+    {
+      "epoch": 0.15237603305785125,
+      "grad_norm": 1.71875,
+      "learning_rate": 0.00019959092970887952,
+      "loss": 5.0445,
+      "step": 17700
+    },
+    {
+      "epoch": 0.1528064738292011,
+      "grad_norm": 1.578125,
+      "learning_rate": 0.00019958847805615432,
+      "loss": 5.1443,
+      "step": 17750
+    },
+    {
+      "epoch": 0.15323691460055097,
+      "grad_norm": 1.453125,
+      "learning_rate": 0.00019958601909384914,
+      "loss": 5.1912,
+      "step": 17800
+    },
+    {
+      "epoch": 0.15366735537190082,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.0001995835528221444,
+      "loss": 5.4637,
+      "step": 17850
+    },
+    {
+      "epoch": 0.1540977961432507,
+      "grad_norm": 2.21875,
+      "learning_rate": 0.00019958107924122114,
+      "loss": 5.22,
+      "step": 17900
+    },
+    {
+      "epoch": 0.15452823691460055,
+      "grad_norm": 1.2109375,
+      "learning_rate": 0.0001995785983512609,
+      "loss": 4.6821,
+      "step": 17950
+    },
+    {
+      "epoch": 0.15495867768595042,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00019957611015244581,
+      "loss": 5.3205,
+      "step": 18000
+    },
+    {
+      "epoch": 0.15495867768595042,
+      "eval_loss": 5.764292240142822,
+      "eval_runtime": 51.9481,
+      "eval_samples_per_second": 12.32,
+      "eval_steps_per_second": 6.16,
+      "eval_tts_loss": 6.31840074806535,
+      "step": 18000
+    },
+    {
+      "epoch": 0.15538911845730027,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.0001995736146449585,
+      "loss": 5.047,
+      "step": 18050
+    },
+    {
+      "epoch": 0.15581955922865015,
+      "grad_norm": 3.890625,
+      "learning_rate": 0.0001995711118289821,
+      "loss": 5.2397,
+      "step": 18100
+    },
+    {
+      "epoch": 0.15625,
+      "grad_norm": 1.9296875,
+      "learning_rate": 0.00019956860170470032,
+      "loss": 5.2703,
+      "step": 18150
+    },
+    {
+      "epoch": 0.15668044077134985,
+      "grad_norm": 1.78125,
+      "learning_rate": 0.00019956608427229737,
+      "loss": 5.3346,
+      "step": 18200
+    },
+    {
+      "epoch": 0.15711088154269973,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.00019956355953195808,
+      "loss": 5.038,
+      "step": 18250
+    },
+    {
+      "epoch": 0.15754132231404958,
+      "grad_norm": 3.453125,
+      "learning_rate": 0.00019956102748386776,
+      "loss": 4.848,
+      "step": 18300
+    },
+    {
+      "epoch": 0.15797176308539945,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.0001995584881282122,
+      "loss": 5.0021,
+      "step": 18350
+    },
+    {
+      "epoch": 0.1584022038567493,
+      "grad_norm": 1.4609375,
+      "learning_rate": 0.00019955594146517784,
+      "loss": 5.301,
+      "step": 18400
+    },
+    {
+      "epoch": 0.15883264462809918,
+      "grad_norm": 1.8515625,
+      "learning_rate": 0.0001995533874949516,
+      "loss": 5.2125,
+      "step": 18450
+    },
+    {
+      "epoch": 0.15926308539944903,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00019955082621772087,
+      "loss": 4.8894,
+      "step": 18500
+    },
+    {
+      "epoch": 0.1596935261707989,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.0001995482576336737,
+      "loss": 4.9082,
+      "step": 18550
+    },
+    {
+      "epoch": 0.16012396694214875,
+      "grad_norm": 1.875,
+      "learning_rate": 0.0001995456817429986,
+      "loss": 5.1639,
+      "step": 18600
+    },
+    {
+      "epoch": 0.16055440771349863,
+      "grad_norm": 3.125,
+      "learning_rate": 0.0001995430985458846,
+      "loss": 4.9199,
+      "step": 18650
+    },
+    {
+      "epoch": 0.16098484848484848,
+      "grad_norm": 1.34375,
+      "learning_rate": 0.0001995405080425214,
+      "loss": 5.023,
+      "step": 18700
+    },
+    {
+      "epoch": 0.16141528925619836,
+      "grad_norm": 1.2734375,
+      "learning_rate": 0.00019953791023309906,
+      "loss": 5.451,
+      "step": 18750
+    },
+    {
+      "epoch": 0.1618457300275482,
+      "grad_norm": 1.65625,
+      "learning_rate": 0.00019953530511780825,
+      "loss": 5.2853,
+      "step": 18800
+    },
+    {
+      "epoch": 0.16227617079889808,
+      "grad_norm": 1.9140625,
+      "learning_rate": 0.00019953269269684022,
+      "loss": 5.2157,
+      "step": 18850
+    },
+    {
+      "epoch": 0.16270661157024793,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.00019953007297038671,
+      "loss": 5.4332,
+      "step": 18900
+    },
+    {
+      "epoch": 0.16313705234159778,
+      "grad_norm": 1.7578125,
+      "learning_rate": 0.00019952744593863995,
+      "loss": 5.0494,
+      "step": 18950
+    },
+    {
+      "epoch": 0.16356749311294766,
+      "grad_norm": 1.3125,
+      "learning_rate": 0.00019952481160179282,
+      "loss": 5.0298,
+      "step": 19000
+    },
+    {
+      "epoch": 0.1639979338842975,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00019952216996003866,
+      "loss": 4.9961,
+      "step": 19050
+    },
+    {
+      "epoch": 0.16442837465564739,
+      "grad_norm": 1.5859375,
+      "learning_rate": 0.00019951952101357131,
+      "loss": 5.2907,
+      "step": 19100
+    },
+    {
+      "epoch": 0.16485881542699724,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.00019951686476258528,
+      "loss": 5.0851,
+      "step": 19150
+    },
+    {
+      "epoch": 0.1652892561983471,
+      "grad_norm": 1.65625,
+      "learning_rate": 0.00019951420120727548,
+      "loss": 4.832,
+      "step": 19200
+    },
+    {
+      "epoch": 0.16571969696969696,
+      "grad_norm": 1.625,
+      "learning_rate": 0.00019951153034783743,
+      "loss": 5.2386,
+      "step": 19250
+    },
+    {
+      "epoch": 0.16615013774104684,
+      "grad_norm": 1.34375,
+      "learning_rate": 0.00019950885218446716,
+      "loss": 4.9833,
+      "step": 19300
+    },
+    {
+      "epoch": 0.1665805785123967,
+      "grad_norm": 2.375,
+      "learning_rate": 0.00019950616671736123,
+      "loss": 5.1157,
+      "step": 19350
+    },
+    {
+      "epoch": 0.16701101928374656,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.00019950347394671677,
+      "loss": 5.1355,
+      "step": 19400
+    },
+    {
+      "epoch": 0.1674414600550964,
+      "grad_norm": 0.8203125,
+      "learning_rate": 0.00019950077387273136,
+      "loss": 4.8928,
+      "step": 19450
+    },
+    {
+      "epoch": 0.1678719008264463,
+      "grad_norm": 2.421875,
+      "learning_rate": 0.00019949806649560326,
+      "loss": 4.9164,
+      "step": 19500
+    },
+    {
+      "epoch": 0.16830234159779614,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.00019949535181553116,
+      "loss": 4.7298,
+      "step": 19550
+    },
+    {
+      "epoch": 0.16873278236914602,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.0001994926298327143,
+      "loss": 5.0695,
+      "step": 19600
+    },
+    {
+      "epoch": 0.16916322314049587,
+      "grad_norm": 3.0625,
+      "learning_rate": 0.0001994899005473525,
+      "loss": 4.9886,
+      "step": 19650
+    },
+    {
+      "epoch": 0.16959366391184574,
+      "grad_norm": 1.8046875,
+      "learning_rate": 0.00019948716395964603,
+      "loss": 4.9026,
+      "step": 19700
+    },
+    {
+      "epoch": 0.1700241046831956,
+      "grad_norm": 2.75,
+      "learning_rate": 0.0001994844200697958,
+      "loss": 4.9833,
+      "step": 19750
+    },
+    {
+      "epoch": 0.17045454545454544,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.00019948166887800314,
+      "loss": 4.8935,
+      "step": 19800
+    },
+    {
+      "epoch": 0.17088498622589532,
+      "grad_norm": 1.765625,
+      "learning_rate": 0.00019947891038447007,
+      "loss": 5.2164,
+      "step": 19850
+    },
+    {
+      "epoch": 0.17131542699724517,
+      "grad_norm": 0.953125,
+      "learning_rate": 0.000199476144589399,
+      "loss": 5.0647,
+      "step": 19900
+    },
+    {
+      "epoch": 0.17174586776859505,
+      "grad_norm": 1.453125,
+      "learning_rate": 0.00019947337149299292,
+      "loss": 4.9461,
+      "step": 19950
+    },
+    {
+      "epoch": 0.1721763085399449,
+      "grad_norm": 1.234375,
+      "learning_rate": 0.00019947059109545542,
+      "loss": 5.3985,
+      "step": 20000
+    },
+    {
+      "epoch": 0.17260674931129477,
+      "grad_norm": 1.9140625,
+      "learning_rate": 0.00019946780339699056,
+      "loss": 4.8297,
+      "step": 20050
+    },
+    {
+      "epoch": 0.17303719008264462,
+      "grad_norm": 1.625,
+      "learning_rate": 0.00019946500839780292,
+      "loss": 5.1055,
+      "step": 20100
+    },
+    {
+      "epoch": 0.1734676308539945,
+      "grad_norm": 0.64453125,
+      "learning_rate": 0.00019946220609809765,
+      "loss": 5.3082,
+      "step": 20150
+    },
+    {
+      "epoch": 0.17389807162534435,
+      "grad_norm": 2.578125,
+      "learning_rate": 0.00019945939649808046,
+      "loss": 5.178,
+      "step": 20200
+    },
+    {
+      "epoch": 0.17432851239669422,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.0001994565795979576,
+      "loss": 5.0136,
+      "step": 20250
+    },
+    {
+      "epoch": 0.17475895316804407,
+      "grad_norm": 1.484375,
+      "learning_rate": 0.00019945375539793576,
+      "loss": 5.0176,
+      "step": 20300
+    },
+    {
+      "epoch": 0.17518939393939395,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.00019945092389822225,
+      "loss": 4.9571,
+      "step": 20350
+    },
+    {
+      "epoch": 0.1756198347107438,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.0001994480850990249,
+      "loss": 5.1597,
+      "step": 20400
+    },
+    {
+      "epoch": 0.17605027548209368,
+      "grad_norm": 1.5703125,
+      "learning_rate": 0.00019944523900055207,
+      "loss": 5.1464,
+      "step": 20450
+    },
+    {
+      "epoch": 0.17648071625344353,
+      "grad_norm": 3.125,
+      "learning_rate": 0.00019944238560301266,
+      "loss": 4.6427,
+      "step": 20500
+    },
+    {
+      "epoch": 0.17691115702479338,
+      "grad_norm": 3.46875,
+      "learning_rate": 0.0001994395249066161,
+      "loss": 5.2441,
+      "step": 20550
+    },
+    {
+      "epoch": 0.17734159779614325,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00019943665691157239,
+      "loss": 4.9263,
+      "step": 20600
+    },
+    {
+      "epoch": 0.1777720385674931,
+      "grad_norm": 1.8203125,
+      "learning_rate": 0.000199433781618092,
+      "loss": 5.0064,
+      "step": 20650
+    },
+    {
+      "epoch": 0.17820247933884298,
+      "grad_norm": 1.7109375,
+      "learning_rate": 0.00019943089902638595,
+      "loss": 5.1541,
+      "step": 20700
+    },
+    {
+      "epoch": 0.17863292011019283,
+      "grad_norm": 1.828125,
+      "learning_rate": 0.00019942800913666587,
+      "loss": 5.2126,
+      "step": 20750
+    },
+    {
+      "epoch": 0.1790633608815427,
+      "grad_norm": 1.8359375,
+      "learning_rate": 0.00019942511194914378,
+      "loss": 5.073,
+      "step": 20800
+    },
+    {
+      "epoch": 0.17949380165289255,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.00019942220746403246,
+      "loss": 5.1939,
+      "step": 20850
+    },
+    {
+      "epoch": 0.17992424242424243,
+      "grad_norm": 1.7109375,
+      "learning_rate": 0.000199419295681545,
+      "loss": 4.937,
+      "step": 20900
+    },
+    {
+      "epoch": 0.18035468319559228,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.00019941637660189513,
+      "loss": 5.2756,
+      "step": 20950
+    },
+    {
+      "epoch": 0.18078512396694216,
+      "grad_norm": 0.93359375,
+      "learning_rate": 0.00019941345022529716,
+      "loss": 4.8813,
+      "step": 21000
+    },
+    {
+      "epoch": 0.18078512396694216,
+      "eval_loss": 5.706298828125,
+      "eval_runtime": 21.9517,
+      "eval_samples_per_second": 29.155,
+      "eval_steps_per_second": 14.577,
+      "eval_tts_loss": 6.426198953232276,
+      "step": 21000
+    },
+    {
+      "epoch": 0.181215564738292,
+      "grad_norm": 1.4375,
+      "learning_rate": 0.0001994105165519658,
+      "loss": 4.9124,
+      "step": 21050
+    },
+    {
+      "epoch": 0.18164600550964188,
+      "grad_norm": 1.7890625,
+      "learning_rate": 0.00019940757558211642,
+      "loss": 4.8005,
+      "step": 21100
+    },
+    {
+      "epoch": 0.18207644628099173,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.00019940462731596488,
+      "loss": 5.1707,
+      "step": 21150
+    },
+    {
+      "epoch": 0.1825068870523416,
+      "grad_norm": 1.3046875,
+      "learning_rate": 0.00019940167175372758,
+      "loss": 5.1448,
+      "step": 21200
+    },
+    {
+      "epoch": 0.18293732782369146,
+      "grad_norm": 3.21875,
+      "learning_rate": 0.00019939870889562142,
+      "loss": 4.6402,
+      "step": 21250
+    },
+    {
+      "epoch": 0.18336776859504134,
+      "grad_norm": 2.5,
+      "learning_rate": 0.0001993957387418639,
+      "loss": 4.7229,
+      "step": 21300
+    },
+    {
+      "epoch": 0.18379820936639119,
+      "grad_norm": 1.90625,
+      "learning_rate": 0.000199392761292673,
+      "loss": 5.0102,
+      "step": 21350
+    },
+    {
+      "epoch": 0.18422865013774103,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.0001993897765482673,
+      "loss": 5.0873,
+      "step": 21400
+    },
+    {
+      "epoch": 0.1846590909090909,
+      "grad_norm": 1.9609375,
+      "learning_rate": 0.00019938678450886582,
+      "loss": 4.853,
+      "step": 21450
+    },
+    {
+      "epoch": 0.18508953168044076,
+      "grad_norm": 1.3359375,
+      "learning_rate": 0.00019938378517468818,
+      "loss": 4.9831,
+      "step": 21500
+    },
+    {
+      "epoch": 0.18551997245179064,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.00019938077854595454,
+      "loss": 4.881,
+      "step": 21550
+    },
+    {
+      "epoch": 0.1859504132231405,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.0001993777646228856,
+      "loss": 5.1419,
+      "step": 21600
+    },
+    {
+      "epoch": 0.18638085399449036,
+      "grad_norm": 1.5078125,
+      "learning_rate": 0.00019937474340570255,
+      "loss": 5.1254,
+      "step": 21650
+    },
+    {
+      "epoch": 0.1868112947658402,
+      "grad_norm": 1.609375,
+      "learning_rate": 0.00019937171489462714,
+      "loss": 4.7089,
+      "step": 21700
+    },
+    {
+      "epoch": 0.1872417355371901,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.00019936867908988166,
+      "loss": 5.0335,
+      "step": 21750
+    },
+    {
+      "epoch": 0.18767217630853994,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.00019936563599168893,
+      "loss": 4.7765,
+      "step": 21800
+    },
+    {
+      "epoch": 0.18810261707988982,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.0001993625856002723,
+      "loss": 4.9271,
+      "step": 21850
+    },
+    {
+      "epoch": 0.18853305785123967,
+      "grad_norm": 1.421875,
+      "learning_rate": 0.00019935952791585567,
+      "loss": 5.09,
+      "step": 21900
+    },
+    {
+      "epoch": 0.18896349862258954,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.0001993564629386635,
+      "loss": 5.1261,
+      "step": 21950
+    },
+    {
+      "epoch": 0.1893939393939394,
+      "grad_norm": 1.875,
+      "learning_rate": 0.0001993533906689207,
+      "loss": 5.0418,
+      "step": 22000
+    },
+    {
+      "epoch": 0.18982438016528927,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.00019935031110685278,
+      "loss": 5.4802,
+      "step": 22050
+    },
+    {
+      "epoch": 0.19025482093663912,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00019934722425268578,
+      "loss": 5.5118,
+      "step": 22100
+    },
+    {
+      "epoch": 0.19068526170798897,
+      "grad_norm": 1.7578125,
+      "learning_rate": 0.0001993441301066463,
+      "loss": 5.4153,
+      "step": 22150
+    },
+    {
+      "epoch": 0.19111570247933884,
+      "grad_norm": 1.8359375,
+      "learning_rate": 0.0001993410286689614,
+      "loss": 4.8482,
+      "step": 22200
+    },
+    {
+      "epoch": 0.1915461432506887,
+      "grad_norm": 1.9296875,
+      "learning_rate": 0.00019933791993985873,
+      "loss": 5.3068,
+      "step": 22250
+    },
+    {
+      "epoch": 0.19197658402203857,
+      "grad_norm": 1.2265625,
+      "learning_rate": 0.00019933480391956648,
+      "loss": 4.8431,
+      "step": 22300
+    },
+    {
+      "epoch": 0.19240702479338842,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.00019933168060831333,
+      "loss": 5.0834,
+      "step": 22350
+    },
+    {
+      "epoch": 0.1928374655647383,
+      "grad_norm": 3.125,
+      "learning_rate": 0.00019932855000632854,
+      "loss": 5.0882,
+      "step": 22400
+    },
+    {
+      "epoch": 0.19326790633608815,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.0001993254121138419,
+      "loss": 4.9763,
+      "step": 22450
+    },
+    {
+      "epoch": 0.19369834710743802,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.00019932226693108372,
+      "loss": 5.2509,
+      "step": 22500
+    },
+    {
+      "epoch": 0.19412878787878787,
+      "grad_norm": 0.87890625,
+      "learning_rate": 0.00019931911445828483,
+      "loss": 5.0617,
+      "step": 22550
+    },
+    {
+      "epoch": 0.19455922865013775,
+      "grad_norm": 1.8984375,
+      "learning_rate": 0.00019931595469567664,
+      "loss": 5.1761,
+      "step": 22600
+    },
+    {
+      "epoch": 0.1949896694214876,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.00019931278764349103,
+      "loss": 4.7677,
+      "step": 22650
+    },
+    {
+      "epoch": 0.19542011019283748,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.0001993096133019605,
+      "loss": 4.954,
+      "step": 22700
+    },
+    {
+      "epoch": 0.19585055096418733,
+      "grad_norm": 1.8203125,
+      "learning_rate": 0.00019930643167131805,
+      "loss": 5.566,
+      "step": 22750
+    },
+    {
+      "epoch": 0.1962809917355372,
+      "grad_norm": 1.9453125,
+      "learning_rate": 0.00019930324275179716,
+      "loss": 4.9606,
+      "step": 22800
+    },
+    {
+      "epoch": 0.19671143250688705,
+      "grad_norm": 1.7109375,
+      "learning_rate": 0.00019930004654363192,
+      "loss": 5.078,
+      "step": 22850
+    },
+    {
+      "epoch": 0.1971418732782369,
+      "grad_norm": 1.125,
+      "learning_rate": 0.00019929684304705693,
+      "loss": 4.8986,
+      "step": 22900
+    },
+    {
+      "epoch": 0.19757231404958678,
+      "grad_norm": 1.9609375,
+      "learning_rate": 0.00019929363226230725,
+      "loss": 4.6425,
+      "step": 22950
+    },
+    {
+      "epoch": 0.19800275482093663,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00019929041418961864,
+      "loss": 4.8875,
+      "step": 23000
+    },
+    {
+      "epoch": 0.1984331955922865,
+      "grad_norm": 1.78125,
+      "learning_rate": 0.00019928718882922724,
+      "loss": 4.9262,
+      "step": 23050
+    },
+    {
+      "epoch": 0.19886363636363635,
+      "grad_norm": 1.8984375,
+      "learning_rate": 0.00019928395618136978,
+      "loss": 4.9604,
+      "step": 23100
+    },
+    {
+      "epoch": 0.19929407713498623,
+      "grad_norm": 1.375,
+      "learning_rate": 0.00019928071624628359,
+      "loss": 4.9066,
+      "step": 23150
+    },
+    {
+      "epoch": 0.19972451790633608,
+      "grad_norm": 2.0,
+      "learning_rate": 0.00019927746902420642,
+      "loss": 4.7944,
+      "step": 23200
+    },
+    {
+      "epoch": 0.20015495867768596,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.00019927421451537662,
+      "loss": 5.0945,
+      "step": 23250
+    },
+    {
+      "epoch": 0.2005853994490358,
+      "grad_norm": 3.703125,
+      "learning_rate": 0.00019927095272003305,
+      "loss": 5.3205,
+      "step": 23300
+    },
+    {
+      "epoch": 0.20101584022038568,
+      "grad_norm": 2.25,
+      "learning_rate": 0.00019926768363841518,
+      "loss": 5.0747,
+      "step": 23350
+    },
+    {
+      "epoch": 0.20144628099173553,
+      "grad_norm": 3.421875,
+      "learning_rate": 0.0001992644072707629,
+      "loss": 4.8941,
+      "step": 23400
+    },
+    {
+      "epoch": 0.2018767217630854,
+      "grad_norm": 1.734375,
+      "learning_rate": 0.00019926112361731672,
+      "loss": 4.9982,
+      "step": 23450
+    },
+    {
+      "epoch": 0.20230716253443526,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.00019925783267831757,
+      "loss": 5.2764,
+      "step": 23500
+    },
+    {
+      "epoch": 0.20273760330578514,
+      "grad_norm": 1.7109375,
+      "learning_rate": 0.0001992545344540071,
+      "loss": 4.7052,
+      "step": 23550
+    },
+    {
+      "epoch": 0.20316804407713498,
+      "grad_norm": 1.578125,
+      "learning_rate": 0.00019925122894462734,
+      "loss": 4.958,
+      "step": 23600
+    },
+    {
+      "epoch": 0.20359848484848486,
+      "grad_norm": 1.421875,
+      "learning_rate": 0.0001992479161504209,
+      "loss": 5.0593,
+      "step": 23650
+    },
+    {
+      "epoch": 0.2040289256198347,
+      "grad_norm": 2.25,
+      "learning_rate": 0.00019924459607163102,
+      "loss": 5.4678,
+      "step": 23700
+    },
+    {
+      "epoch": 0.20445936639118456,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.00019924126870850126,
+      "loss": 4.9233,
+      "step": 23750
+    },
+    {
+      "epoch": 0.20488980716253444,
+      "grad_norm": 2.921875,
+      "learning_rate": 0.00019923793406127592,
+      "loss": 5.0045,
+      "step": 23800
+    },
+    {
+      "epoch": 0.2053202479338843,
+      "grad_norm": 1.6171875,
+      "learning_rate": 0.00019923459213019973,
+      "loss": 5.1675,
+      "step": 23850
+    },
+    {
+      "epoch": 0.20575068870523416,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.000199231242915518,
+      "loss": 5.1394,
+      "step": 23900
+    },
+    {
+      "epoch": 0.206181129476584,
+      "grad_norm": 1.9921875,
+      "learning_rate": 0.0001992278864174765,
+      "loss": 5.2315,
+      "step": 23950
+    },
+    {
+      "epoch": 0.2066115702479339,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.0001992245226363217,
+      "loss": 5.1691,
+      "step": 24000
+    },
+    {
+      "epoch": 0.2066115702479339,
+      "eval_loss": 5.665122032165527,
+      "eval_runtime": 21.7307,
+      "eval_samples_per_second": 29.451,
+      "eval_steps_per_second": 14.726,
+      "eval_tts_loss": 6.368378734417094,
+      "step": 24000
+    },
+    {
+      "epoch": 0.20704201101928374,
+      "grad_norm": 2.875,
+      "learning_rate": 0.00019922115157230036,
+      "loss": 4.621,
+      "step": 24050
+    },
+    {
+      "epoch": 0.20747245179063362,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.00019921777322566003,
+      "loss": 4.8669,
+      "step": 24100
+    },
+    {
+      "epoch": 0.20790289256198347,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.00019921438759664856,
+      "loss": 5.0948,
+      "step": 24150
+    },
+    {
+      "epoch": 0.20833333333333334,
+      "grad_norm": 3.015625,
+      "learning_rate": 0.00019921099468551455,
+      "loss": 5.3568,
+      "step": 24200
+    },
+    {
+      "epoch": 0.2087637741046832,
+      "grad_norm": 1.7734375,
+      "learning_rate": 0.00019920759449250697,
+      "loss": 5.0607,
+      "step": 24250
+    },
+    {
+      "epoch": 0.20919421487603307,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.0001992041870178754,
+      "loss": 4.8101,
+      "step": 24300
+    },
+    {
+      "epoch": 0.20962465564738292,
+      "grad_norm": 1.25,
+      "learning_rate": 0.00019920077226186995,
+      "loss": 5.199,
+      "step": 24350
+    },
+    {
+      "epoch": 0.2100550964187328,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.00019919735022474124,
+      "loss": 5.0901,
+      "step": 24400
+    },
+    {
+      "epoch": 0.21048553719008264,
+      "grad_norm": 2.140625,
+      "learning_rate": 0.00019919392090674047,
+      "loss": 5.2075,
+      "step": 24450
+    },
+    {
+      "epoch": 0.2109159779614325,
+      "grad_norm": 2.375,
+      "learning_rate": 0.00019919048430811932,
+      "loss": 4.9512,
+      "step": 24500
+    },
+    {
+      "epoch": 0.21134641873278237,
+      "grad_norm": 1.734375,
+      "learning_rate": 0.00019918704042913004,
+      "loss": 4.9328,
+      "step": 24550
+    },
+    {
+      "epoch": 0.21177685950413222,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.00019918358927002543,
+      "loss": 5.1121,
+      "step": 24600
+    },
+    {
+      "epoch": 0.2122073002754821,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.00019918013083105872,
+      "loss": 5.1803,
+      "step": 24650
+    },
+    {
+      "epoch": 0.21263774104683195,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.00019917666511248382,
+      "loss": 4.8848,
+      "step": 24700
+    },
+    {
+      "epoch": 0.21306818181818182,
+      "grad_norm": 1.734375,
+      "learning_rate": 0.00019917319211455508,
+      "loss": 5.0794,
+      "step": 24750
+    },
+    {
+      "epoch": 0.21349862258953167,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.00019916971183752742,
+      "loss": 5.141,
+      "step": 24800
+    },
+    {
+      "epoch": 0.21392906336088155,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00019916622428165629,
+      "loss": 5.0965,
+      "step": 24850
+    },
+    {
+      "epoch": 0.2143595041322314,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.0001991627294471976,
+      "loss": 5.1589,
+      "step": 24900
+    },
+    {
+      "epoch": 0.21478994490358128,
+      "grad_norm": 1.171875,
+      "learning_rate": 0.00019915922733440797,
+      "loss": 4.7798,
+      "step": 24950
+    },
+    {
+      "epoch": 0.21522038567493113,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.00019915571794354442,
+      "loss": 5.0545,
+      "step": 25000
+    },
+    {
+      "epoch": 0.215650826446281,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.00019915220127486447,
+      "loss": 4.9558,
+      "step": 25050
+    },
+    {
+      "epoch": 0.21608126721763085,
+      "grad_norm": 1.59375,
+      "learning_rate": 0.00019914867732862633,
+      "loss": 4.7492,
+      "step": 25100
+    },
+    {
+      "epoch": 0.21651170798898073,
+      "grad_norm": 1.625,
+      "learning_rate": 0.00019914514610508855,
+      "loss": 4.8764,
+      "step": 25150
+    },
+    {
+      "epoch": 0.21694214876033058,
+      "grad_norm": 1.265625,
+      "learning_rate": 0.00019914160760451036,
+      "loss": 4.8477,
+      "step": 25200
+    },
+    {
+      "epoch": 0.21737258953168045,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.00019913806182715152,
+      "loss": 5.0519,
+      "step": 25250
+    },
+    {
+      "epoch": 0.2178030303030303,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.0001991345087732722,
+      "loss": 5.1269,
+      "step": 25300
+    },
+    {
+      "epoch": 0.21823347107438015,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.00019913094844313327,
+      "loss": 4.9829,
+      "step": 25350
+    },
+    {
+      "epoch": 0.21866391184573003,
+      "grad_norm": 2.265625,
+      "learning_rate": 0.00019912738083699598,
+      "loss": 5.1944,
+      "step": 25400
+    },
+    {
+      "epoch": 0.21909435261707988,
+      "grad_norm": 1.796875,
+      "learning_rate": 0.00019912380595512225,
+      "loss": 4.7664,
+      "step": 25450
+    },
+    {
+      "epoch": 0.21952479338842976,
+      "grad_norm": 3.03125,
+      "learning_rate": 0.00019912022379777437,
+      "loss": 5.0875,
+      "step": 25500
+    },
+    {
+      "epoch": 0.2199552341597796,
+      "grad_norm": 3.359375,
+      "learning_rate": 0.00019911663436521538,
+      "loss": 4.8201,
+      "step": 25550
+    },
+    {
+      "epoch": 0.22038567493112948,
+      "grad_norm": 1.984375,
+      "learning_rate": 0.00019911303765770866,
+      "loss": 5.0903,
+      "step": 25600
+    },
+    {
+      "epoch": 0.22081611570247933,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.00019910943367551823,
+      "loss": 4.6082,
+      "step": 25650
+    },
+    {
+      "epoch": 0.2212465564738292,
+      "grad_norm": 1.46875,
+      "learning_rate": 0.0001991058224189086,
+      "loss": 5.0438,
+      "step": 25700
+    },
+    {
+      "epoch": 0.22167699724517906,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.00019910220388814486,
+      "loss": 5.5352,
+      "step": 25750
+    },
+    {
+      "epoch": 0.22210743801652894,
+      "grad_norm": 1.1484375,
+      "learning_rate": 0.00019909857808349254,
+      "loss": 4.7568,
+      "step": 25800
+    },
+    {
+      "epoch": 0.22253787878787878,
+      "grad_norm": 1.765625,
+      "learning_rate": 0.00019909494500521782,
+      "loss": 5.0173,
+      "step": 25850
+    },
+    {
+      "epoch": 0.22296831955922866,
+      "grad_norm": 2.25,
+      "learning_rate": 0.0001990913046535874,
+      "loss": 4.6996,
+      "step": 25900
+    },
+    {
+      "epoch": 0.2233987603305785,
+      "grad_norm": 2.25,
+      "learning_rate": 0.00019908765702886832,
+      "loss": 5.0819,
+      "step": 25950
+    },
+    {
+      "epoch": 0.2238292011019284,
+      "grad_norm": 0.8125,
+      "learning_rate": 0.00019908400213132849,
+      "loss": 5.0021,
+      "step": 26000
+    },
+    {
+      "epoch": 0.22425964187327824,
+      "grad_norm": 1.5078125,
+      "learning_rate": 0.00019908033996123603,
+      "loss": 4.7302,
+      "step": 26050
+    },
+    {
+      "epoch": 0.2246900826446281,
+      "grad_norm": 1.8828125,
+      "learning_rate": 0.00019907667051885983,
+      "loss": 5.0459,
+      "step": 26100
+    },
+    {
+      "epoch": 0.22512052341597796,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.00019907299380446917,
+      "loss": 5.4258,
+      "step": 26150
+    },
+    {
+      "epoch": 0.2255509641873278,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.00019906930981833393,
+      "loss": 4.7663,
+      "step": 26200
+    },
+    {
+      "epoch": 0.2259814049586777,
+      "grad_norm": 2.234375,
+      "learning_rate": 0.0001990656185607245,
+      "loss": 4.9105,
+      "step": 26250
+    },
+    {
+      "epoch": 0.22641184573002754,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.0001990619200319118,
+      "loss": 4.9766,
+      "step": 26300
+    },
+    {
+      "epoch": 0.22684228650137742,
+      "grad_norm": 1.4375,
+      "learning_rate": 0.00019905821423216733,
+      "loss": 4.5605,
+      "step": 26350
+    },
+    {
+      "epoch": 0.22727272727272727,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.00019905450116176304,
+      "loss": 5.0108,
+      "step": 26400
+    },
+    {
+      "epoch": 0.22770316804407714,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.00019905078082097153,
+      "loss": 5.0003,
+      "step": 26450
+    },
+    {
+      "epoch": 0.228133608815427,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.00019904705321006577,
+      "loss": 5.2441,
+      "step": 26500
+    },
+    {
+      "epoch": 0.22856404958677687,
+      "grad_norm": 1.5625,
+      "learning_rate": 0.00019904331832931943,
+      "loss": 4.8511,
+      "step": 26550
+    },
+    {
+      "epoch": 0.22899449035812672,
+      "grad_norm": 1.796875,
+      "learning_rate": 0.0001990395761790066,
+      "loss": 5.0509,
+      "step": 26600
+    },
+    {
+      "epoch": 0.2294249311294766,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.000199035826759402,
+      "loss": 5.0985,
+      "step": 26650
+    },
+    {
+      "epoch": 0.22985537190082644,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.0001990320700707808,
+      "loss": 4.9686,
+      "step": 26700
+    },
+    {
+      "epoch": 0.23028581267217632,
+      "grad_norm": 1.7890625,
+      "learning_rate": 0.00019902830611341873,
+      "loss": 5.3253,
+      "step": 26750
+    },
+    {
+      "epoch": 0.23071625344352617,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.00019902453488759205,
+      "loss": 5.1206,
+      "step": 26800
+    },
+    {
+      "epoch": 0.23114669421487602,
+      "grad_norm": 4.15625,
+      "learning_rate": 0.00019902075639357752,
+      "loss": 4.9917,
+      "step": 26850
+    },
+    {
+      "epoch": 0.2315771349862259,
+      "grad_norm": 3.5,
+      "learning_rate": 0.0001990169706316526,
+      "loss": 4.4503,
+      "step": 26900
+    },
+    {
+      "epoch": 0.23200757575757575,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.000199013177602095,
+      "loss": 5.2462,
+      "step": 26950
+    },
+    {
+      "epoch": 0.23243801652892562,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.00019900937730518326,
+      "loss": 4.8377,
+      "step": 27000
+    },
+    {
+      "epoch": 0.23243801652892562,
+      "eval_loss": 5.625140190124512,
+      "eval_runtime": 21.8035,
+      "eval_samples_per_second": 29.353,
+      "eval_steps_per_second": 14.677,
+      "eval_tts_loss": 6.439033366911451,
+      "step": 27000
+    },
+    {
+      "epoch": 0.23286845730027547,
+      "grad_norm": 2.75,
+      "learning_rate": 0.0001990055697411962,
+      "loss": 5.2129,
+      "step": 27050
+    },
+    {
+      "epoch": 0.23329889807162535,
+      "grad_norm": 1.6640625,
+      "learning_rate": 0.00019900175491041337,
+      "loss": 5.4218,
+      "step": 27100
+    },
+    {
+      "epoch": 0.2337293388429752,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00019899793281311472,
+      "loss": 4.7001,
+      "step": 27150
+    },
+    {
+      "epoch": 0.23415977961432508,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.0001989941034495808,
+      "loss": 5.4653,
+      "step": 27200
+    },
+    {
+      "epoch": 0.23459022038567492,
+      "grad_norm": 1.9921875,
+      "learning_rate": 0.0001989902668200927,
+      "loss": 4.8957,
+      "step": 27250
+    },
+    {
+      "epoch": 0.2350206611570248,
+      "grad_norm": 1.875,
+      "learning_rate": 0.00019898642292493198,
+      "loss": 5.0856,
+      "step": 27300
+    },
+    {
+      "epoch": 0.23545110192837465,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00019898257176438077,
+      "loss": 5.1142,
+      "step": 27350
+    },
+    {
+      "epoch": 0.23588154269972453,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.0001989787133387218,
+      "loss": 5.0284,
+      "step": 27400
+    },
+    {
+      "epoch": 0.23631198347107438,
+      "grad_norm": 1.3359375,
+      "learning_rate": 0.0001989748476482382,
+      "loss": 5.4126,
+      "step": 27450
+    },
+    {
+      "epoch": 0.23674242424242425,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.00019897097469321375,
+      "loss": 5.0992,
+      "step": 27500
+    },
+    {
+      "epoch": 0.2371728650137741,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.00019896709447393273,
+      "loss": 5.2379,
+      "step": 27550
+    },
+    {
+      "epoch": 0.23760330578512398,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.00019896320699067986,
+      "loss": 4.486,
+      "step": 27600
+    },
+    {
+      "epoch": 0.23803374655647383,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.00019895931224374054,
+      "loss": 4.7783,
+      "step": 27650
+    },
+    {
+      "epoch": 0.23846418732782368,
+      "grad_norm": 1.796875,
+      "learning_rate": 0.0001989554102334006,
+      "loss": 4.9106,
+      "step": 27700
+    },
+    {
+      "epoch": 0.23889462809917356,
+      "grad_norm": 3.296875,
+      "learning_rate": 0.00019895150095994649,
+      "loss": 5.1797,
+      "step": 27750
+    },
+    {
+      "epoch": 0.2393250688705234,
+      "grad_norm": 1.703125,
+      "learning_rate": 0.00019894758442366505,
+      "loss": 5.3304,
+      "step": 27800
+    },
+    {
+      "epoch": 0.23975550964187328,
+      "grad_norm": 1.84375,
+      "learning_rate": 0.00019894366062484387,
+      "loss": 5.0072,
+      "step": 27850
+    },
+    {
+      "epoch": 0.24018595041322313,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.00019893972956377086,
+      "loss": 5.0721,
+      "step": 27900
+    },
+    {
+      "epoch": 0.240616391184573,
+      "grad_norm": 2.21875,
+      "learning_rate": 0.00019893579124073457,
+      "loss": 5.1727,
+      "step": 27950
+    },
+    {
+      "epoch": 0.24104683195592286,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.00019893184565602409,
+      "loss": 4.9496,
+      "step": 28000
+    },
+    {
+      "epoch": 0.24147727272727273,
+      "grad_norm": 3.09375,
+      "learning_rate": 0.00019892789280992896,
+      "loss": 5.1087,
+      "step": 28050
+    },
+    {
+      "epoch": 0.24190771349862258,
+      "grad_norm": 1.671875,
+      "learning_rate": 0.00019892393270273934,
+      "loss": 4.8786,
+      "step": 28100
+    },
+    {
+      "epoch": 0.24233815426997246,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.00019891996533474593,
+      "loss": 4.4389,
+      "step": 28150
+    },
+    {
+      "epoch": 0.2427685950413223,
+      "grad_norm": 2.21875,
+      "learning_rate": 0.0001989159907062399,
+      "loss": 4.5667,
+      "step": 28200
+    },
+    {
+      "epoch": 0.2431990358126722,
+      "grad_norm": 2.0,
+      "learning_rate": 0.00019891200881751295,
+      "loss": 4.7674,
+      "step": 28250
+    },
+    {
+      "epoch": 0.24362947658402204,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.0001989080196688574,
+      "loss": 4.8112,
+      "step": 28300
+    },
+    {
+      "epoch": 0.2440599173553719,
+      "grad_norm": 2.21875,
+      "learning_rate": 0.000198904023260566,
+      "loss": 5.1912,
+      "step": 28350
+    },
+    {
+      "epoch": 0.24449035812672176,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.00019890001959293208,
+      "loss": 4.9896,
+      "step": 28400
+    },
+    {
+      "epoch": 0.2449207988980716,
+      "grad_norm": 3.09375,
+      "learning_rate": 0.00019889600866624952,
+      "loss": 4.7702,
+      "step": 28450
+    },
+    {
+      "epoch": 0.2453512396694215,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.0001988919904808127,
+      "loss": 4.751,
+      "step": 28500
+    },
+    {
+      "epoch": 0.24578168044077134,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.0001988879650369165,
+      "loss": 5.1838,
+      "step": 28550
+    },
+    {
+      "epoch": 0.24621212121212122,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.0001988839323348565,
+      "loss": 5.2764,
+      "step": 28600
+    },
+    {
+      "epoch": 0.24664256198347106,
+      "grad_norm": 1.421875,
+      "learning_rate": 0.00019887989237492862,
+      "loss": 5.0927,
+      "step": 28650
+    },
+    {
+      "epoch": 0.24707300275482094,
+      "grad_norm": 2.921875,
+      "learning_rate": 0.00019887584515742933,
+      "loss": 4.6307,
+      "step": 28700
+    },
+    {
+      "epoch": 0.2475034435261708,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.00019887179068265578,
+      "loss": 5.0863,
+      "step": 28750
+    },
+    {
+      "epoch": 0.24793388429752067,
+      "grad_norm": 3.3125,
+      "learning_rate": 0.00019886772895090553,
+      "loss": 4.9633,
+      "step": 28800
+    },
+    {
+      "epoch": 0.24836432506887052,
+      "grad_norm": 3.25,
+      "learning_rate": 0.0001988636599624767,
+      "loss": 5.2618,
+      "step": 28850
+    },
+    {
+      "epoch": 0.2487947658402204,
+      "grad_norm": 3.046875,
+      "learning_rate": 0.0001988595837176679,
+      "loss": 4.9674,
+      "step": 28900
+    },
+    {
+      "epoch": 0.24922520661157024,
+      "grad_norm": 1.8125,
+      "learning_rate": 0.00019885550021677838,
+      "loss": 4.7908,
+      "step": 28950
+    },
+    {
+      "epoch": 0.24965564738292012,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.00019885140946010785,
+      "loss": 5.018,
+      "step": 29000
+    },
+    {
+      "epoch": 0.25008608815426997,
+      "grad_norm": 0.703125,
+      "learning_rate": 0.00019884731144795655,
+      "loss": 5.1767,
+      "step": 29050
+    },
+    {
+      "epoch": 0.2505165289256198,
+      "grad_norm": 2.375,
+      "learning_rate": 0.0001988432061806252,
+      "loss": 4.9314,
+      "step": 29100
+    },
+    {
+      "epoch": 0.2509469696969697,
+      "grad_norm": 2.234375,
+      "learning_rate": 0.00019883909365841528,
+      "loss": 4.8966,
+      "step": 29150
+    },
+    {
+      "epoch": 0.2513774104683196,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.00019883497388162847,
+      "loss": 4.9233,
+      "step": 29200
+    },
+    {
+      "epoch": 0.2518078512396694,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.00019883084685056728,
+      "loss": 5.1622,
+      "step": 29250
+    },
+    {
+      "epoch": 0.25223829201101927,
+      "grad_norm": 2.265625,
+      "learning_rate": 0.00019882671256553453,
+      "loss": 5.2549,
+      "step": 29300
+    },
+    {
+      "epoch": 0.2526687327823691,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.00019882257102683376,
+      "loss": 4.8635,
+      "step": 29350
+    },
+    {
+      "epoch": 0.253099173553719,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.00019881842223476884,
+      "loss": 4.8018,
+      "step": 29400
+    },
+    {
+      "epoch": 0.2535296143250689,
+      "grad_norm": 1.8671875,
+      "learning_rate": 0.00019881426618964437,
+      "loss": 5.2167,
+      "step": 29450
+    },
+    {
+      "epoch": 0.2539600550964187,
+      "grad_norm": 2.234375,
+      "learning_rate": 0.00019881010289176537,
+      "loss": 4.8911,
+      "step": 29500
+    },
+    {
+      "epoch": 0.2543904958677686,
+      "grad_norm": 1.5390625,
+      "learning_rate": 0.00019880593234143738,
+      "loss": 5.5701,
+      "step": 29550
+    },
+    {
+      "epoch": 0.2548209366391185,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00019880175453896657,
+      "loss": 4.6705,
+      "step": 29600
+    },
+    {
+      "epoch": 0.2552513774104683,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.00019879756948465956,
+      "loss": 4.6333,
+      "step": 29650
+    },
+    {
+      "epoch": 0.2556818181818182,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.00019879337717882353,
+      "loss": 4.9866,
+      "step": 29700
+    },
+    {
+      "epoch": 0.256112258953168,
+      "grad_norm": 1.8125,
+      "learning_rate": 0.00019878917762176612,
+      "loss": 5.3702,
+      "step": 29750
+    },
+    {
+      "epoch": 0.25654269972451793,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.00019878497081379565,
+      "loss": 4.6589,
+      "step": 29800
+    },
+    {
+      "epoch": 0.2569731404958678,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.00019878075675522087,
+      "loss": 4.8626,
+      "step": 29850
+    },
+    {
+      "epoch": 0.25740358126721763,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00019877653544635109,
+      "loss": 5.1007,
+      "step": 29900
+    },
+    {
+      "epoch": 0.2578340220385675,
+      "grad_norm": 1.765625,
+      "learning_rate": 0.00019877230688749613,
+      "loss": 4.7402,
+      "step": 29950
+    },
+    {
+      "epoch": 0.25826446280991733,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.00019876807107896633,
+      "loss": 4.6641,
+      "step": 30000
+    },
+    {
+      "epoch": 0.25826446280991733,
+      "eval_loss": 5.591824531555176,
+      "eval_runtime": 22.0839,
+      "eval_samples_per_second": 28.98,
+      "eval_steps_per_second": 14.49,
+      "eval_tts_loss": 6.428915743431203,
+      "step": 30000
+    },
+    {
+      "epoch": 0.25869490358126723,
+      "grad_norm": 1.3984375,
+      "learning_rate": 0.00019876382802107263,
+      "loss": 4.726,
+      "step": 30050
+    },
+    {
+      "epoch": 0.2591253443526171,
+      "grad_norm": 2.125,
+      "learning_rate": 0.00019875957771412644,
+      "loss": 5.2665,
+      "step": 30100
+    },
+    {
+      "epoch": 0.25955578512396693,
+      "grad_norm": 4.28125,
+      "learning_rate": 0.00019875532015843978,
+      "loss": 4.9599,
+      "step": 30150
+    },
+    {
+      "epoch": 0.2599862258953168,
+      "grad_norm": 2.21875,
+      "learning_rate": 0.00019875105535432507,
+      "loss": 4.8781,
+      "step": 30200
+    },
+    {
+      "epoch": 0.2604166666666667,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.00019874678330209533,
+      "loss": 5.1677,
+      "step": 30250
+    },
+    {
+      "epoch": 0.26084710743801653,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.0001987425040020642,
+      "loss": 5.4665,
+      "step": 30300
+    },
+    {
+      "epoch": 0.2612775482093664,
+      "grad_norm": 2.25,
+      "learning_rate": 0.0001987382174545457,
+      "loss": 5.0117,
+      "step": 30350
+    },
+    {
+      "epoch": 0.26170798898071623,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.00019873392365985448,
+      "loss": 4.5659,
+      "step": 30400
+    },
+    {
+      "epoch": 0.26213842975206614,
+      "grad_norm": 1.5234375,
+      "learning_rate": 0.0001987296226183057,
+      "loss": 4.8937,
+      "step": 30450
+    },
+    {
+      "epoch": 0.262568870523416,
+      "grad_norm": 1.3671875,
+      "learning_rate": 0.00019872531433021506,
+      "loss": 5.4825,
+      "step": 30500
+    },
+    {
+      "epoch": 0.26299931129476584,
+      "grad_norm": 1.953125,
+      "learning_rate": 0.00019872099879589877,
+      "loss": 5.0801,
+      "step": 30550
+    },
+    {
+      "epoch": 0.2634297520661157,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.00019871667601567352,
+      "loss": 5.1016,
+      "step": 30600
+    },
+    {
+      "epoch": 0.2638601928374656,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.0001987123459898567,
+      "loss": 5.0152,
+      "step": 30650
+    },
+    {
+      "epoch": 0.26429063360881544,
+      "grad_norm": 3.984375,
+      "learning_rate": 0.00019870800871876605,
+      "loss": 4.9308,
+      "step": 30700
+    },
+    {
+      "epoch": 0.2647210743801653,
+      "grad_norm": 0.7734375,
+      "learning_rate": 0.0001987036642027199,
+      "loss": 4.6529,
+      "step": 30750
+    },
+    {
+      "epoch": 0.26515151515151514,
+      "grad_norm": 1.8984375,
+      "learning_rate": 0.0001986993124420372,
+      "loss": 4.8347,
+      "step": 30800
+    },
+    {
+      "epoch": 0.265581955922865,
+      "grad_norm": 1.9921875,
+      "learning_rate": 0.00019869495343703732,
+      "loss": 4.3844,
+      "step": 30850
+    },
+    {
+      "epoch": 0.2660123966942149,
+      "grad_norm": 3.03125,
+      "learning_rate": 0.0001986905871880402,
+      "loss": 4.7447,
+      "step": 30900
+    },
+    {
+      "epoch": 0.26644283746556474,
+      "grad_norm": 2.140625,
+      "learning_rate": 0.00019868621369536634,
+      "loss": 4.6668,
+      "step": 30950
+    },
+    {
+      "epoch": 0.2668732782369146,
+      "grad_norm": 2.140625,
+      "learning_rate": 0.00019868183295933668,
+      "loss": 4.811,
+      "step": 31000
+    },
+    {
+      "epoch": 0.26730371900826444,
+      "grad_norm": 3.125,
+      "learning_rate": 0.00019867744498027287,
+      "loss": 5.2151,
+      "step": 31050
+    },
+    {
+      "epoch": 0.26773415977961434,
+      "grad_norm": 1.9921875,
+      "learning_rate": 0.00019867304975849684,
+      "loss": 5.2058,
+      "step": 31100
+    },
+    {
+      "epoch": 0.2681646005509642,
+      "grad_norm": 4.34375,
+      "learning_rate": 0.00019866864729433127,
+      "loss": 5.1636,
+      "step": 31150
+    },
+    {
+      "epoch": 0.26859504132231404,
+      "grad_norm": 1.6796875,
+      "learning_rate": 0.0001986642375880993,
+      "loss": 5.0151,
+      "step": 31200
+    },
+    {
+      "epoch": 0.2690254820936639,
+      "grad_norm": 1.328125,
+      "learning_rate": 0.00019865982064012455,
+      "loss": 4.6903,
+      "step": 31250
+    },
+    {
+      "epoch": 0.2694559228650138,
+      "grad_norm": 2.140625,
+      "learning_rate": 0.00019865539645073127,
+      "loss": 5.1308,
+      "step": 31300
+    },
+    {
+      "epoch": 0.26988636363636365,
+      "grad_norm": 1.7890625,
+      "learning_rate": 0.0001986509650202441,
+      "loss": 4.9051,
+      "step": 31350
+    },
+    {
+      "epoch": 0.2703168044077135,
+      "grad_norm": 1.5703125,
+      "learning_rate": 0.00019864652634898842,
+      "loss": 4.8119,
+      "step": 31400
+    },
+    {
+      "epoch": 0.27074724517906334,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.0001986420804372899,
+      "loss": 5.0894,
+      "step": 31450
+    },
+    {
+      "epoch": 0.27117768595041325,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.00019863762728547494,
+      "loss": 5.0927,
+      "step": 31500
+    },
+    {
+      "epoch": 0.2716081267217631,
+      "grad_norm": 2.140625,
+      "learning_rate": 0.00019863316689387034,
+      "loss": 5.1256,
+      "step": 31550
+    },
+    {
+      "epoch": 0.27203856749311295,
+      "grad_norm": 2.9375,
+      "learning_rate": 0.00019862869926280351,
+      "loss": 4.9328,
+      "step": 31600
+    },
+    {
+      "epoch": 0.2724690082644628,
+      "grad_norm": 1.921875,
+      "learning_rate": 0.00019862422439260237,
+      "loss": 4.9261,
+      "step": 31650
+    },
+    {
+      "epoch": 0.27289944903581265,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00019861974228359535,
+      "loss": 4.8875,
+      "step": 31700
+    },
+    {
+      "epoch": 0.27332988980716255,
+      "grad_norm": 3.125,
+      "learning_rate": 0.00019861525293611142,
+      "loss": 4.849,
+      "step": 31750
+    },
+    {
+      "epoch": 0.2737603305785124,
+      "grad_norm": 1.8203125,
+      "learning_rate": 0.00019861075635048013,
+      "loss": 5.115,
+      "step": 31800
+    },
+    {
+      "epoch": 0.27419077134986225,
+      "grad_norm": 1.9140625,
+      "learning_rate": 0.00019860625252703145,
+      "loss": 4.9396,
+      "step": 31850
+    },
+    {
+      "epoch": 0.2746212121212121,
+      "grad_norm": 1.5703125,
+      "learning_rate": 0.000198601741466096,
+      "loss": 4.9381,
+      "step": 31900
+    },
+    {
+      "epoch": 0.275051652892562,
+      "grad_norm": 1.4609375,
+      "learning_rate": 0.0001985972231680049,
+      "loss": 4.8725,
+      "step": 31950
+    },
+    {
+      "epoch": 0.27548209366391185,
+      "grad_norm": 3.4375,
+      "learning_rate": 0.0001985926976330897,
+      "loss": 4.9087,
+      "step": 32000
+    },
+    {
+      "epoch": 0.2759125344352617,
+      "grad_norm": 1.2890625,
+      "learning_rate": 0.0001985881648616827,
+      "loss": 4.9087,
+      "step": 32050
+    },
+    {
+      "epoch": 0.27634297520661155,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.00019858362485411645,
+      "loss": 5.0181,
+      "step": 32100
+    },
+    {
+      "epoch": 0.27677341597796146,
+      "grad_norm": 2.125,
+      "learning_rate": 0.0001985790776107243,
+      "loss": 5.1998,
+      "step": 32150
+    },
+    {
+      "epoch": 0.2772038567493113,
+      "grad_norm": 1.3203125,
+      "learning_rate": 0.00019857452313183993,
+      "loss": 4.4733,
+      "step": 32200
+    },
+    {
+      "epoch": 0.27763429752066116,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.00019856996141779764,
+      "loss": 5.2766,
+      "step": 32250
+    },
+    {
+      "epoch": 0.278064738292011,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00019856539246893227,
+      "loss": 5.408,
+      "step": 32300
+    },
+    {
+      "epoch": 0.2784951790633609,
+      "grad_norm": 1.734375,
+      "learning_rate": 0.00019856081628557914,
+      "loss": 4.8507,
+      "step": 32350
+    },
+    {
+      "epoch": 0.27892561983471076,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.00019855623286807418,
+      "loss": 5.0983,
+      "step": 32400
+    },
+    {
+      "epoch": 0.2793560606060606,
+      "grad_norm": 1.828125,
+      "learning_rate": 0.00019855164221675375,
+      "loss": 4.8606,
+      "step": 32450
+    },
+    {
+      "epoch": 0.27978650137741046,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.00019854704433195486,
+      "loss": 4.9416,
+      "step": 32500
+    },
+    {
+      "epoch": 0.2802169421487603,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00019854243921401491,
+      "loss": 4.8396,
+      "step": 32550
+    },
+    {
+      "epoch": 0.2806473829201102,
+      "grad_norm": 1.9765625,
+      "learning_rate": 0.00019853782686327196,
+      "loss": 5.062,
+      "step": 32600
+    },
+    {
+      "epoch": 0.28107782369146006,
+      "grad_norm": 2.265625,
+      "learning_rate": 0.00019853320728006454,
+      "loss": 4.5996,
+      "step": 32650
+    },
+    {
+      "epoch": 0.2815082644628099,
+      "grad_norm": 3.296875,
+      "learning_rate": 0.00019852858046473165,
+      "loss": 5.3079,
+      "step": 32700
+    },
+    {
+      "epoch": 0.28193870523415976,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.00019852394641761302,
+      "loss": 4.8414,
+      "step": 32750
+    },
+    {
+      "epoch": 0.28236914600550966,
+      "grad_norm": 1.4140625,
+      "learning_rate": 0.00019851930513904863,
+      "loss": 5.0199,
+      "step": 32800
+    },
+    {
+      "epoch": 0.2827995867768595,
+      "grad_norm": 1.3515625,
+      "learning_rate": 0.00019851465662937926,
+      "loss": 4.7016,
+      "step": 32850
+    },
+    {
+      "epoch": 0.28323002754820936,
+      "grad_norm": 1.21875,
+      "learning_rate": 0.00019851000088894606,
+      "loss": 5.0567,
+      "step": 32900
+    },
+    {
+      "epoch": 0.2836604683195592,
+      "grad_norm": 1.125,
+      "learning_rate": 0.0001985053379180907,
+      "loss": 4.6057,
+      "step": 32950
+    },
+    {
+      "epoch": 0.2840909090909091,
+      "grad_norm": 2.234375,
+      "learning_rate": 0.00019850066771715552,
+      "loss": 5.2001,
+      "step": 33000
+    },
+    {
+      "epoch": 0.2840909090909091,
+      "eval_loss": 5.571471214294434,
+      "eval_runtime": 21.7382,
+      "eval_samples_per_second": 29.441,
+      "eval_steps_per_second": 14.721,
+      "eval_tts_loss": 6.407271942785936,
+      "step": 33000
+    },
+    {
+      "epoch": 0.28452134986225897,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00019849599028648322,
+      "loss": 4.811,
+      "step": 33050
+    },
+    {
+      "epoch": 0.2849517906336088,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00019849130562641718,
+      "loss": 4.6642,
+      "step": 33100
+    },
+    {
+      "epoch": 0.28538223140495866,
+      "grad_norm": 1.8359375,
+      "learning_rate": 0.0001984866137373012,
+      "loss": 4.4969,
+      "step": 33150
+    },
+    {
+      "epoch": 0.2858126721763085,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.00019848191461947968,
+      "loss": 4.6382,
+      "step": 33200
+    },
+    {
+      "epoch": 0.2862431129476584,
+      "grad_norm": 2.125,
+      "learning_rate": 0.00019847720827329752,
+      "loss": 5.1926,
+      "step": 33250
+    },
+    {
+      "epoch": 0.28667355371900827,
+      "grad_norm": 1.6796875,
+      "learning_rate": 0.00019847249469910017,
+      "loss": 5.0532,
+      "step": 33300
+    },
+    {
+      "epoch": 0.2871039944903581,
+      "grad_norm": 1.9765625,
+      "learning_rate": 0.00019846777389723354,
+      "loss": 4.789,
+      "step": 33350
+    },
+    {
+      "epoch": 0.28753443526170797,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.00019846304586804419,
+      "loss": 4.7021,
+      "step": 33400
+    },
+    {
+      "epoch": 0.28796487603305787,
+      "grad_norm": 1.6796875,
+      "learning_rate": 0.00019845831061187912,
+      "loss": 4.7884,
+      "step": 33450
+    },
+    {
+      "epoch": 0.2883953168044077,
+      "grad_norm": 3.015625,
+      "learning_rate": 0.00019845356812908587,
+      "loss": 4.9174,
+      "step": 33500
+    },
+    {
+      "epoch": 0.28882575757575757,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.00019844881842001258,
+      "loss": 5.1589,
+      "step": 33550
+    },
+    {
+      "epoch": 0.2892561983471074,
+      "grad_norm": 1.9140625,
+      "learning_rate": 0.0001984440614850078,
+      "loss": 4.7724,
+      "step": 33600
+    },
+    {
+      "epoch": 0.2896866391184573,
+      "grad_norm": 1.6875,
+      "learning_rate": 0.00019843929732442071,
+      "loss": 5.1374,
+      "step": 33650
+    },
+    {
+      "epoch": 0.2901170798898072,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.00019843452593860104,
+      "loss": 4.3952,
+      "step": 33700
+    },
+    {
+      "epoch": 0.290547520661157,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.0001984297473278989,
+      "loss": 5.0098,
+      "step": 33750
+    },
+    {
+      "epoch": 0.29097796143250687,
+      "grad_norm": 2.140625,
+      "learning_rate": 0.00019842496149266513,
+      "loss": 4.8159,
+      "step": 33800
+    },
+    {
+      "epoch": 0.2914084022038568,
+      "grad_norm": 1.9765625,
+      "learning_rate": 0.00019842016843325095,
+      "loss": 5.1525,
+      "step": 33850
+    },
+    {
+      "epoch": 0.2918388429752066,
+      "grad_norm": 1.921875,
+      "learning_rate": 0.00019841536815000814,
+      "loss": 4.7047,
+      "step": 33900
+    },
+    {
+      "epoch": 0.2922692837465565,
+      "grad_norm": 1.3515625,
+      "learning_rate": 0.00019841056064328906,
+      "loss": 5.241,
+      "step": 33950
+    },
+    {
+      "epoch": 0.2926997245179063,
+      "grad_norm": 2.234375,
+      "learning_rate": 0.0001984057459134466,
+      "loss": 5.0182,
+      "step": 34000
+    },
+    {
+      "epoch": 0.2931301652892562,
+      "grad_norm": 1.6171875,
+      "learning_rate": 0.00019840092396083408,
+      "loss": 4.9981,
+      "step": 34050
+    },
+    {
+      "epoch": 0.2935606060606061,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.00019839609478580544,
+      "loss": 5.21,
+      "step": 34100
+    },
+    {
+      "epoch": 0.2939910468319559,
+      "grad_norm": 2.21875,
+      "learning_rate": 0.00019839125838871517,
+      "loss": 4.8529,
+      "step": 34150
+    },
+    {
+      "epoch": 0.2944214876033058,
+      "grad_norm": 1.203125,
+      "learning_rate": 0.00019838641476991825,
+      "loss": 4.7993,
+      "step": 34200
+    },
+    {
+      "epoch": 0.2948519283746556,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00019838156392977012,
+      "loss": 4.9721,
+      "step": 34250
+    },
+    {
+      "epoch": 0.29528236914600553,
+      "grad_norm": 1.7109375,
+      "learning_rate": 0.00019837670586862692,
+      "loss": 4.8963,
+      "step": 34300
+    },
+    {
+      "epoch": 0.2957128099173554,
+      "grad_norm": 1.578125,
+      "learning_rate": 0.0001983718405868451,
+      "loss": 5.0663,
+      "step": 34350
+    },
+    {
+      "epoch": 0.29614325068870523,
+      "grad_norm": 1.203125,
+      "learning_rate": 0.0001983669680847819,
+      "loss": 4.7316,
+      "step": 34400
+    },
+    {
+      "epoch": 0.2965736914600551,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.0001983620883627949,
+      "loss": 5.0639,
+      "step": 34450
+    },
+    {
+      "epoch": 0.297004132231405,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.00019835720142124218,
+      "loss": 4.6894,
+      "step": 34500
+    },
+    {
+      "epoch": 0.29743457300275483,
+      "grad_norm": 0.828125,
+      "learning_rate": 0.00019835230726048252,
+      "loss": 4.9307,
+      "step": 34550
+    },
+    {
+      "epoch": 0.2978650137741047,
+      "grad_norm": 2.25,
+      "learning_rate": 0.00019834740588087514,
+      "loss": 4.8107,
+      "step": 34600
+    },
+    {
+      "epoch": 0.29829545454545453,
+      "grad_norm": 3.40625,
+      "learning_rate": 0.00019834249728277974,
+      "loss": 4.5763,
+      "step": 34650
+    },
+    {
+      "epoch": 0.29872589531680444,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.00019833758146655666,
+      "loss": 5.0326,
+      "step": 34700
+    },
+    {
+      "epoch": 0.2991563360881543,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00019833265843256665,
+      "loss": 4.803,
+      "step": 34750
+    },
+    {
+      "epoch": 0.29958677685950413,
+      "grad_norm": 1.984375,
+      "learning_rate": 0.0001983277281811711,
+      "loss": 4.6619,
+      "step": 34800
+    },
+    {
+      "epoch": 0.300017217630854,
+      "grad_norm": 1.25,
+      "learning_rate": 0.00019832279071273185,
+      "loss": 4.7168,
+      "step": 34850
+    },
+    {
+      "epoch": 0.30044765840220383,
+      "grad_norm": 3.734375,
+      "learning_rate": 0.0001983178460276113,
+      "loss": 4.6914,
+      "step": 34900
+    },
+    {
+      "epoch": 0.30087809917355374,
+      "grad_norm": 1.2890625,
+      "learning_rate": 0.0001983128941261724,
+      "loss": 5.0874,
+      "step": 34950
+    },
+    {
+      "epoch": 0.3013085399449036,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.0001983079350087786,
+      "loss": 5.0653,
+      "step": 35000
+    },
+    {
+      "epoch": 0.30173898071625344,
+      "grad_norm": 1.8671875,
+      "learning_rate": 0.00019830296867579387,
+      "loss": 5.085,
+      "step": 35050
+    },
+    {
+      "epoch": 0.3021694214876033,
+      "grad_norm": 3.53125,
+      "learning_rate": 0.00019829799512758277,
+      "loss": 5.2666,
+      "step": 35100
+    },
+    {
+      "epoch": 0.3025998622589532,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.0001982930143645103,
+      "loss": 5.2022,
+      "step": 35150
+    },
+    {
+      "epoch": 0.30303030303030304,
+      "grad_norm": 3.578125,
+      "learning_rate": 0.00019828802638694205,
+      "loss": 4.8083,
+      "step": 35200
+    },
+    {
+      "epoch": 0.3034607438016529,
+      "grad_norm": 1.859375,
+      "learning_rate": 0.00019828303119524417,
+      "loss": 4.9576,
+      "step": 35250
+    },
+    {
+      "epoch": 0.30389118457300274,
+      "grad_norm": 1.4765625,
+      "learning_rate": 0.00019827802878978322,
+      "loss": 4.984,
+      "step": 35300
+    },
+    {
+      "epoch": 0.30432162534435264,
+      "grad_norm": 2.21875,
+      "learning_rate": 0.00019827301917092643,
+      "loss": 4.5902,
+      "step": 35350
+    },
+    {
+      "epoch": 0.3047520661157025,
+      "grad_norm": 3.9375,
+      "learning_rate": 0.00019826800233904148,
+      "loss": 4.8497,
+      "step": 35400
+    },
+    {
+      "epoch": 0.30518250688705234,
+      "grad_norm": 1.2734375,
+      "learning_rate": 0.00019826297829449657,
+      "loss": 5.0224,
+      "step": 35450
+    },
+    {
+      "epoch": 0.3056129476584022,
+      "grad_norm": 3.21875,
+      "learning_rate": 0.00019825794703766047,
+      "loss": 5.0513,
+      "step": 35500
+    },
+    {
+      "epoch": 0.30604338842975204,
+      "grad_norm": 1.171875,
+      "learning_rate": 0.00019825290856890249,
+      "loss": 4.9505,
+      "step": 35550
+    },
+    {
+      "epoch": 0.30647382920110194,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.0001982478628885924,
+      "loss": 4.6933,
+      "step": 35600
+    },
+    {
+      "epoch": 0.3069042699724518,
+      "grad_norm": 3.21875,
+      "learning_rate": 0.00019824280999710054,
+      "loss": 4.8053,
+      "step": 35650
+    },
+    {
+      "epoch": 0.30733471074380164,
+      "grad_norm": 3.265625,
+      "learning_rate": 0.00019823774989479782,
+      "loss": 4.5851,
+      "step": 35700
+    },
+    {
+      "epoch": 0.3077651515151515,
+      "grad_norm": 1.4609375,
+      "learning_rate": 0.00019823268258205564,
+      "loss": 4.8727,
+      "step": 35750
+    },
+    {
+      "epoch": 0.3081955922865014,
+      "grad_norm": 0.61328125,
+      "learning_rate": 0.0001982276080592459,
+      "loss": 4.9059,
+      "step": 35800
+    },
+    {
+      "epoch": 0.30862603305785125,
+      "grad_norm": 1.9140625,
+      "learning_rate": 0.00019822252632674105,
+      "loss": 5.0789,
+      "step": 35850
+    },
+    {
+      "epoch": 0.3090564738292011,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.0001982174373849141,
+      "loss": 5.0924,
+      "step": 35900
+    },
+    {
+      "epoch": 0.30948691460055094,
+      "grad_norm": 2.5,
+      "learning_rate": 0.00019821234123413862,
+      "loss": 5.1402,
+      "step": 35950
+    },
+    {
+      "epoch": 0.30991735537190085,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.00019820723787478855,
+      "loss": 5.0153,
+      "step": 36000
+    },
+    {
+      "epoch": 0.30991735537190085,
+      "eval_loss": 5.54570198059082,
+      "eval_runtime": 21.7411,
+      "eval_samples_per_second": 29.437,
+      "eval_steps_per_second": 14.719,
+      "eval_tts_loss": 6.579320323344557,
+      "step": 36000
+    },
+    {
+      "epoch": 0.3103477961432507,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.00019820212730723853,
+      "loss": 4.8887,
+      "step": 36050
+    },
+    {
+      "epoch": 0.31077823691460055,
+      "grad_norm": 0.98828125,
+      "learning_rate": 0.00019819700953186366,
+      "loss": 5.1352,
+      "step": 36100
+    },
+    {
+      "epoch": 0.3112086776859504,
+      "grad_norm": 3.484375,
+      "learning_rate": 0.00019819188454903955,
+      "loss": 5.1846,
+      "step": 36150
+    },
+    {
+      "epoch": 0.3116391184573003,
+      "grad_norm": 0.94921875,
+      "learning_rate": 0.0001981867523591424,
+      "loss": 5.0842,
+      "step": 36200
+    },
+    {
+      "epoch": 0.31206955922865015,
+      "grad_norm": 1.671875,
+      "learning_rate": 0.00019818161296254887,
+      "loss": 4.758,
+      "step": 36250
+    },
+    {
+      "epoch": 0.3125,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.0001981764663596362,
+      "loss": 4.8793,
+      "step": 36300
+    },
+    {
+      "epoch": 0.31293044077134985,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00019817131255078213,
+      "loss": 4.7948,
+      "step": 36350
+    },
+    {
+      "epoch": 0.3133608815426997,
+      "grad_norm": 3.40625,
+      "learning_rate": 0.00019816615153636492,
+      "loss": 4.7607,
+      "step": 36400
+    },
+    {
+      "epoch": 0.3137913223140496,
+      "grad_norm": 0.9375,
+      "learning_rate": 0.0001981609833167634,
+      "loss": 4.9049,
+      "step": 36450
+    },
+    {
+      "epoch": 0.31422176308539945,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.00019815580789235692,
+      "loss": 4.8472,
+      "step": 36500
+    },
+    {
+      "epoch": 0.3146522038567493,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00019815062526352533,
+      "loss": 4.8835,
+      "step": 36550
+    },
+    {
+      "epoch": 0.31508264462809915,
+      "grad_norm": 2.0,
+      "learning_rate": 0.000198145435430649,
+      "loss": 4.9046,
+      "step": 36600
+    },
+    {
+      "epoch": 0.31551308539944906,
+      "grad_norm": 3.40625,
+      "learning_rate": 0.0001981402383941089,
+      "loss": 4.6263,
+      "step": 36650
+    },
+    {
+      "epoch": 0.3159435261707989,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00019813503415428644,
+      "loss": 5.2997,
+      "step": 36700
+    },
+    {
+      "epoch": 0.31637396694214875,
+      "grad_norm": 2.375,
+      "learning_rate": 0.00019812982271156363,
+      "loss": 5.2157,
+      "step": 36750
+    },
+    {
+      "epoch": 0.3168044077134986,
+      "grad_norm": 1.7734375,
+      "learning_rate": 0.00019812460406632297,
+      "loss": 5.0089,
+      "step": 36800
+    },
+    {
+      "epoch": 0.3172348484848485,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.00019811937821894747,
+      "loss": 4.4919,
+      "step": 36850
+    },
+    {
+      "epoch": 0.31766528925619836,
+      "grad_norm": 1.828125,
+      "learning_rate": 0.0001981141451698207,
+      "loss": 4.6964,
+      "step": 36900
+    },
+    {
+      "epoch": 0.3180957300275482,
+      "grad_norm": 1.53125,
+      "learning_rate": 0.00019810890491932682,
+      "loss": 4.7849,
+      "step": 36950
+    },
+    {
+      "epoch": 0.31852617079889806,
+      "grad_norm": 1.7109375,
+      "learning_rate": 0.0001981036574678504,
+      "loss": 4.8493,
+      "step": 37000
+    },
+    {
+      "epoch": 0.31895661157024796,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.0001980984028157766,
+      "loss": 5.0103,
+      "step": 37050
+    },
+    {
+      "epoch": 0.3193870523415978,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00019809314096349106,
+      "loss": 4.7065,
+      "step": 37100
+    },
+    {
+      "epoch": 0.31981749311294766,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.00019808787191138004,
+      "loss": 4.8871,
+      "step": 37150
+    },
+    {
+      "epoch": 0.3202479338842975,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00019808259565983028,
+      "loss": 4.7011,
+      "step": 37200
+    },
+    {
+      "epoch": 0.32067837465564736,
+      "grad_norm": 1.171875,
+      "learning_rate": 0.00019807731220922902,
+      "loss": 5.0205,
+      "step": 37250
+    },
+    {
+      "epoch": 0.32110881542699726,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.00019807202155996406,
+      "loss": 5.2195,
+      "step": 37300
+    },
+    {
+      "epoch": 0.3215392561983471,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.00019806672371242373,
+      "loss": 5.2518,
+      "step": 37350
+    },
+    {
+      "epoch": 0.32196969696969696,
+      "grad_norm": 2.921875,
+      "learning_rate": 0.00019806141866699685,
+      "loss": 4.9155,
+      "step": 37400
+    },
+    {
+      "epoch": 0.3224001377410468,
+      "grad_norm": 1.1953125,
+      "learning_rate": 0.00019805610642407287,
+      "loss": 4.7224,
+      "step": 37450
+    },
+    {
+      "epoch": 0.3228305785123967,
+      "grad_norm": 0.984375,
+      "learning_rate": 0.0001980507869840416,
+      "loss": 4.8169,
+      "step": 37500
+    },
+    {
+      "epoch": 0.32326101928374656,
+      "grad_norm": 1.4921875,
+      "learning_rate": 0.00019804546034729356,
+      "loss": 4.7099,
+      "step": 37550
+    },
+    {
+      "epoch": 0.3236914600550964,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.0001980401265142197,
+      "loss": 5.2715,
+      "step": 37600
+    },
+    {
+      "epoch": 0.32412190082644626,
+      "grad_norm": 1.2109375,
+      "learning_rate": 0.00019803478548521147,
+      "loss": 4.791,
+      "step": 37650
+    },
+    {
+      "epoch": 0.32455234159779617,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.0001980294372606609,
+      "loss": 4.7433,
+      "step": 37700
+    },
+    {
+      "epoch": 0.324982782369146,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.00019802408184096056,
+      "loss": 4.679,
+      "step": 37750
+    },
+    {
+      "epoch": 0.32541322314049587,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.0001980187192265035,
+      "loss": 4.6584,
+      "step": 37800
+    },
+    {
+      "epoch": 0.3258436639118457,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00019801334941768338,
+      "loss": 4.847,
+      "step": 37850
+    },
+    {
+      "epoch": 0.32627410468319556,
+      "grad_norm": 1.4453125,
+      "learning_rate": 0.00019800797241489426,
+      "loss": 4.9452,
+      "step": 37900
+    },
+    {
+      "epoch": 0.32670454545454547,
+      "grad_norm": 4.1875,
+      "learning_rate": 0.00019800258821853084,
+      "loss": 5.0311,
+      "step": 37950
+    },
+    {
+      "epoch": 0.3271349862258953,
+      "grad_norm": 2.125,
+      "learning_rate": 0.00019799719682898832,
+      "loss": 5.1609,
+      "step": 38000
+    },
+    {
+      "epoch": 0.32756542699724517,
+      "grad_norm": 1.96875,
+      "learning_rate": 0.00019799179824666238,
+      "loss": 4.9125,
+      "step": 38050
+    },
+    {
+      "epoch": 0.327995867768595,
+      "grad_norm": 1.5859375,
+      "learning_rate": 0.0001979863924719493,
+      "loss": 5.2746,
+      "step": 38100
+    },
+    {
+      "epoch": 0.3284263085399449,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.00019798097950524585,
+      "loss": 4.5547,
+      "step": 38150
+    },
+    {
+      "epoch": 0.32885674931129477,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.00019797555934694927,
+      "loss": 5.2883,
+      "step": 38200
+    },
+    {
+      "epoch": 0.3292871900826446,
+      "grad_norm": 1.4296875,
+      "learning_rate": 0.00019797013199745746,
+      "loss": 5.2175,
+      "step": 38250
+    },
+    {
+      "epoch": 0.32971763085399447,
+      "grad_norm": 1.9765625,
+      "learning_rate": 0.00019796469745716878,
+      "loss": 4.9056,
+      "step": 38300
+    },
+    {
+      "epoch": 0.3301480716253444,
+      "grad_norm": 2.265625,
+      "learning_rate": 0.00019795925572648203,
+      "loss": 5.3339,
+      "step": 38350
+    },
+    {
+      "epoch": 0.3305785123966942,
+      "grad_norm": 1.3515625,
+      "learning_rate": 0.00019795380680579668,
+      "loss": 4.4741,
+      "step": 38400
+    },
+    {
+      "epoch": 0.3310089531680441,
+      "grad_norm": 1.984375,
+      "learning_rate": 0.0001979483506955127,
+      "loss": 4.9175,
+      "step": 38450
+    },
+    {
+      "epoch": 0.3314393939393939,
+      "grad_norm": 0.71875,
+      "learning_rate": 0.0001979428873960305,
+      "loss": 4.7652,
+      "step": 38500
+    },
+    {
+      "epoch": 0.3318698347107438,
+      "grad_norm": 3.015625,
+      "learning_rate": 0.00019793741690775108,
+      "loss": 5.0055,
+      "step": 38550
+    },
+    {
+      "epoch": 0.3323002754820937,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.00019793193923107598,
+      "loss": 5.0159,
+      "step": 38600
+    },
+    {
+      "epoch": 0.3327307162534435,
+      "grad_norm": 2.125,
+      "learning_rate": 0.00019792645436640725,
+      "loss": 4.5909,
+      "step": 38650
+    },
+    {
+      "epoch": 0.3331611570247934,
+      "grad_norm": 1.6796875,
+      "learning_rate": 0.0001979209623141475,
+      "loss": 4.8598,
+      "step": 38700
+    },
+    {
+      "epoch": 0.3335915977961432,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.00019791546307469977,
+      "loss": 4.8634,
+      "step": 38750
+    },
+    {
+      "epoch": 0.33402203856749313,
+      "grad_norm": 1.6640625,
+      "learning_rate": 0.0001979099566484677,
+      "loss": 5.0617,
+      "step": 38800
+    },
+    {
+      "epoch": 0.334452479338843,
+      "grad_norm": 2.0,
+      "learning_rate": 0.0001979044430358555,
+      "loss": 4.7291,
+      "step": 38850
+    },
+    {
+      "epoch": 0.3348829201101928,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.00019789892223726782,
+      "loss": 4.8745,
+      "step": 38900
+    },
+    {
+      "epoch": 0.3353133608815427,
+      "grad_norm": 1.5546875,
+      "learning_rate": 0.00019789339425310988,
+      "loss": 4.8053,
+      "step": 38950
+    },
+    {
+      "epoch": 0.3357438016528926,
+      "grad_norm": 1.7734375,
+      "learning_rate": 0.00019788785908378745,
+      "loss": 4.6575,
+      "step": 39000
+    },
+    {
+      "epoch": 0.3357438016528926,
+      "eval_loss": 5.525368690490723,
+      "eval_runtime": 22.016,
+      "eval_samples_per_second": 29.07,
+      "eval_steps_per_second": 14.535,
+      "eval_tts_loss": 6.599256973241367,
+      "step": 39000
+    },
+    {
+      "epoch": 0.33617424242424243,
+      "grad_norm": 2.234375,
+      "learning_rate": 0.00019788231672970675,
+      "loss": 5.2571,
+      "step": 39050
+    },
+    {
+      "epoch": 0.3366046831955923,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00019787676719127458,
+      "loss": 5.0282,
+      "step": 39100
+    },
+    {
+      "epoch": 0.33703512396694213,
+      "grad_norm": 1.6953125,
+      "learning_rate": 0.00019787121046889832,
+      "loss": 4.8208,
+      "step": 39150
+    },
+    {
+      "epoch": 0.33746556473829203,
+      "grad_norm": 3.203125,
+      "learning_rate": 0.00019786564656298577,
+      "loss": 5.3306,
+      "step": 39200
+    },
+    {
+      "epoch": 0.3378960055096419,
+      "grad_norm": 0.83984375,
+      "learning_rate": 0.00019786007547394535,
+      "loss": 5.0133,
+      "step": 39250
+    },
+    {
+      "epoch": 0.33832644628099173,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.00019785449720218588,
+      "loss": 4.7902,
+      "step": 39300
+    },
+    {
+      "epoch": 0.3387568870523416,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.00019784891174811693,
+      "loss": 4.8296,
+      "step": 39350
+    },
+    {
+      "epoch": 0.3391873278236915,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.00019784331911214836,
+      "loss": 4.6619,
+      "step": 39400
+    },
+    {
+      "epoch": 0.33961776859504134,
+      "grad_norm": 1.53125,
+      "learning_rate": 0.00019783771929469065,
+      "loss": 4.2706,
+      "step": 39450
+    },
+    {
+      "epoch": 0.3400482093663912,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00019783211229615488,
+      "loss": 4.7978,
+      "step": 39500
+    },
+    {
+      "epoch": 0.34047865013774103,
+      "grad_norm": 1.703125,
+      "learning_rate": 0.00019782649811695256,
+      "loss": 4.92,
+      "step": 39550
+    },
+    {
+      "epoch": 0.3409090909090909,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.00019782087675749573,
+      "loss": 5.1624,
+      "step": 39600
+    },
+    {
+      "epoch": 0.3413395316804408,
+      "grad_norm": 1.53125,
+      "learning_rate": 0.00019781524821819706,
+      "loss": 4.7182,
+      "step": 39650
+    },
+    {
+      "epoch": 0.34176997245179064,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.0001978096124994696,
+      "loss": 5.1676,
+      "step": 39700
+    },
+    {
+      "epoch": 0.3422004132231405,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.000197803969601727,
+      "loss": 4.8049,
+      "step": 39750
+    },
+    {
+      "epoch": 0.34263085399449034,
+      "grad_norm": 3.578125,
+      "learning_rate": 0.0001977983195253835,
+      "loss": 4.7238,
+      "step": 39800
+    },
+    {
+      "epoch": 0.34306129476584024,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00019779266227085374,
+      "loss": 4.5067,
+      "step": 39850
+    },
+    {
+      "epoch": 0.3434917355371901,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00019778699783855298,
+      "loss": 5.2104,
+      "step": 39900
+    },
+    {
+      "epoch": 0.34392217630853994,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.000197781326228897,
+      "loss": 4.7217,
+      "step": 39950
+    },
+    {
+      "epoch": 0.3443526170798898,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.00019777564744230202,
+      "loss": 5.0633,
+      "step": 40000
+    },
+    {
+      "epoch": 0.3447830578512397,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.0001977699614791849,
+      "loss": 4.6669,
+      "step": 40050
+    },
+    {
+      "epoch": 0.34521349862258954,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00019776426833996297,
+      "loss": 5.3334,
+      "step": 40100
+    },
+    {
+      "epoch": 0.3456439393939394,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.00019775856802505408,
+      "loss": 4.8324,
+      "step": 40150
+    },
+    {
+      "epoch": 0.34607438016528924,
+      "grad_norm": 1.1875,
+      "learning_rate": 0.00019775286053487663,
+      "loss": 5.0114,
+      "step": 40200
+    },
+    {
+      "epoch": 0.3465048209366391,
+      "grad_norm": 1.96875,
+      "learning_rate": 0.00019774714586984955,
+      "loss": 5.1082,
+      "step": 40250
+    },
+    {
+      "epoch": 0.346935261707989,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.00019774142403039226,
+      "loss": 4.7672,
+      "step": 40300
+    },
+    {
+      "epoch": 0.34736570247933884,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.00019773569501692475,
+      "loss": 4.8024,
+      "step": 40350
+    },
+    {
+      "epoch": 0.3477961432506887,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.0001977299588298675,
+      "loss": 4.7896,
+      "step": 40400
+    },
+    {
+      "epoch": 0.34822658402203854,
+      "grad_norm": 1.5234375,
+      "learning_rate": 0.00019772421546964157,
+      "loss": 4.5769,
+      "step": 40450
+    },
+    {
+      "epoch": 0.34865702479338845,
+      "grad_norm": 1.1953125,
+      "learning_rate": 0.00019771846493666842,
+      "loss": 4.7507,
+      "step": 40500
+    },
+    {
+      "epoch": 0.3490874655647383,
+      "grad_norm": 1.3046875,
+      "learning_rate": 0.00019771270723137026,
+      "loss": 4.4713,
+      "step": 40550
+    },
+    {
+      "epoch": 0.34951790633608815,
+      "grad_norm": 0.98828125,
+      "learning_rate": 0.0001977069423541696,
+      "loss": 5.0715,
+      "step": 40600
+    },
+    {
+      "epoch": 0.349948347107438,
+      "grad_norm": 3.328125,
+      "learning_rate": 0.00019770117030548957,
+      "loss": 5.2178,
+      "step": 40650
+    },
+    {
+      "epoch": 0.3503787878787879,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00019769539108575388,
+      "loss": 5.0733,
+      "step": 40700
+    },
+    {
+      "epoch": 0.35080922865013775,
+      "grad_norm": 2.375,
+      "learning_rate": 0.00019768960469538667,
+      "loss": 4.7981,
+      "step": 40750
+    },
+    {
+      "epoch": 0.3512396694214876,
+      "grad_norm": 1.640625,
+      "learning_rate": 0.00019768381113481268,
+      "loss": 5.2429,
+      "step": 40800
+    },
+    {
+      "epoch": 0.35167011019283745,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.0001976780104044571,
+      "loss": 5.2152,
+      "step": 40850
+    },
+    {
+      "epoch": 0.35210055096418735,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.00019767220250474572,
+      "loss": 5.3687,
+      "step": 40900
+    },
+    {
+      "epoch": 0.3525309917355372,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.00019766638743610483,
+      "loss": 4.721,
+      "step": 40950
+    },
+    {
+      "epoch": 0.35296143250688705,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.00019766056519896123,
+      "loss": 4.9641,
+      "step": 41000
+    },
+    {
+      "epoch": 0.3533918732782369,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00019765473579374225,
+      "loss": 4.9655,
+      "step": 41050
+    },
+    {
+      "epoch": 0.35382231404958675,
+      "grad_norm": 1.921875,
+      "learning_rate": 0.00019764889922087582,
+      "loss": 4.7365,
+      "step": 41100
+    },
+    {
+      "epoch": 0.35425275482093666,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.00019764305548079024,
+      "loss": 4.5452,
+      "step": 41150
+    },
+    {
+      "epoch": 0.3546831955922865,
+      "grad_norm": 1.75,
+      "learning_rate": 0.0001976372045739145,
+      "loss": 4.9724,
+      "step": 41200
+    },
+    {
+      "epoch": 0.35511363636363635,
+      "grad_norm": 1.90625,
+      "learning_rate": 0.000197631346500678,
+      "loss": 5.1451,
+      "step": 41250
+    },
+    {
+      "epoch": 0.3555440771349862,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.00019762548126151072,
+      "loss": 4.6637,
+      "step": 41300
+    },
+    {
+      "epoch": 0.3559745179063361,
+      "grad_norm": 5.1875,
+      "learning_rate": 0.00019761960885684315,
+      "loss": 4.7258,
+      "step": 41350
+    },
+    {
+      "epoch": 0.35640495867768596,
+      "grad_norm": 2.234375,
+      "learning_rate": 0.00019761372928710637,
+      "loss": 4.8053,
+      "step": 41400
+    },
+    {
+      "epoch": 0.3568353994490358,
+      "grad_norm": 1.828125,
+      "learning_rate": 0.00019760784255273185,
+      "loss": 4.8886,
+      "step": 41450
+    },
+    {
+      "epoch": 0.35726584022038566,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.0001976019486541517,
+      "loss": 4.6134,
+      "step": 41500
+    },
+    {
+      "epoch": 0.35769628099173556,
+      "grad_norm": 1.3828125,
+      "learning_rate": 0.00019759604759179853,
+      "loss": 5.0279,
+      "step": 41550
+    },
+    {
+      "epoch": 0.3581267217630854,
+      "grad_norm": 0.9609375,
+      "learning_rate": 0.00019759013936610545,
+      "loss": 4.3883,
+      "step": 41600
+    },
+    {
+      "epoch": 0.35855716253443526,
+      "grad_norm": 3.84375,
+      "learning_rate": 0.00019758422397750608,
+      "loss": 4.7281,
+      "step": 41650
+    },
+    {
+      "epoch": 0.3589876033057851,
+      "grad_norm": 1.421875,
+      "learning_rate": 0.00019757830142643466,
+      "loss": 4.8342,
+      "step": 41700
+    },
+    {
+      "epoch": 0.359418044077135,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.00019757237171332584,
+      "loss": 4.6784,
+      "step": 41750
+    },
+    {
+      "epoch": 0.35984848484848486,
+      "grad_norm": 2.25,
+      "learning_rate": 0.00019756643483861486,
+      "loss": 5.204,
+      "step": 41800
+    },
+    {
+      "epoch": 0.3602789256198347,
+      "grad_norm": 1.421875,
+      "learning_rate": 0.0001975604908027375,
+      "loss": 4.9922,
+      "step": 41850
+    },
+    {
+      "epoch": 0.36070936639118456,
+      "grad_norm": 1.84375,
+      "learning_rate": 0.00019755453960613003,
+      "loss": 5.0035,
+      "step": 41900
+    },
+    {
+      "epoch": 0.3611398071625344,
+      "grad_norm": 2.125,
+      "learning_rate": 0.00019754858124922925,
+      "loss": 4.8395,
+      "step": 41950
+    },
+    {
+      "epoch": 0.3615702479338843,
+      "grad_norm": 1.8828125,
+      "learning_rate": 0.00019754261573247246,
+      "loss": 4.9879,
+      "step": 42000
+    },
+    {
+      "epoch": 0.3615702479338843,
+      "eval_loss": 5.509988784790039,
+      "eval_runtime": 21.6827,
+      "eval_samples_per_second": 29.517,
+      "eval_steps_per_second": 14.758,
+      "eval_tts_loss": 6.5986155310177885,
+      "step": 42000
+    },
+    {
+      "epoch": 0.36200068870523416,
+      "grad_norm": 1.6484375,
+      "learning_rate": 0.00019753664305629756,
+      "loss": 4.7334,
+      "step": 42050
+    },
+    {
+      "epoch": 0.362431129476584,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.0001975306632211429,
+      "loss": 4.9475,
+      "step": 42100
+    },
+    {
+      "epoch": 0.36286157024793386,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.0001975246762274474,
+      "loss": 4.9682,
+      "step": 42150
+    },
+    {
+      "epoch": 0.36329201101928377,
+      "grad_norm": 2.140625,
+      "learning_rate": 0.00019751868207565053,
+      "loss": 4.6935,
+      "step": 42200
+    },
+    {
+      "epoch": 0.3637224517906336,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.00019751268076619217,
+      "loss": 4.8626,
+      "step": 42250
+    },
+    {
+      "epoch": 0.36415289256198347,
+      "grad_norm": 1.8359375,
+      "learning_rate": 0.0001975066722995129,
+      "loss": 5.2555,
+      "step": 42300
+    },
+    {
+      "epoch": 0.3645833333333333,
+      "grad_norm": 0.98046875,
+      "learning_rate": 0.00019750065667605365,
+      "loss": 4.9685,
+      "step": 42350
+    },
+    {
+      "epoch": 0.3650137741046832,
+      "grad_norm": 1.984375,
+      "learning_rate": 0.00019749463389625595,
+      "loss": 5.1185,
+      "step": 42400
+    },
+    {
+      "epoch": 0.36544421487603307,
+      "grad_norm": 2.0,
+      "learning_rate": 0.0001974886039605619,
+      "loss": 4.749,
+      "step": 42450
+    },
+    {
+      "epoch": 0.3658746556473829,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.00019748256686941408,
+      "loss": 4.8788,
+      "step": 42500
+    },
+    {
+      "epoch": 0.36630509641873277,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.0001974765226232556,
+      "loss": 5.0896,
+      "step": 42550
+    },
+    {
+      "epoch": 0.36673553719008267,
+      "grad_norm": 3.078125,
+      "learning_rate": 0.00019747047122253006,
+      "loss": 4.7231,
+      "step": 42600
+    },
+    {
+      "epoch": 0.3671659779614325,
+      "grad_norm": 3.1875,
+      "learning_rate": 0.0001974644126676817,
+      "loss": 4.7672,
+      "step": 42650
+    },
+    {
+      "epoch": 0.36759641873278237,
+      "grad_norm": 3.75,
+      "learning_rate": 0.0001974583469591551,
+      "loss": 5.1181,
+      "step": 42700
+    },
+    {
+      "epoch": 0.3680268595041322,
+      "grad_norm": 1.7265625,
+      "learning_rate": 0.0001974522740973955,
+      "loss": 5.2168,
+      "step": 42750
+    },
+    {
+      "epoch": 0.36845730027548207,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.0001974461940828487,
+      "loss": 4.9432,
+      "step": 42800
+    },
+    {
+      "epoch": 0.368887741046832,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.0001974401069159609,
+      "loss": 4.9133,
+      "step": 42850
+    },
+    {
+      "epoch": 0.3693181818181818,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.0001974340125971789,
+      "loss": 4.5893,
+      "step": 42900
+    },
+    {
+      "epoch": 0.3697486225895317,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.00019742791112695003,
+      "loss": 5.0533,
+      "step": 42950
+    },
+    {
+      "epoch": 0.3701790633608815,
+      "grad_norm": 1.3359375,
+      "learning_rate": 0.0001974218025057221,
+      "loss": 4.8669,
+      "step": 43000
+    },
+    {
+      "epoch": 0.3706095041322314,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.00019741568673394344,
+      "loss": 4.7203,
+      "step": 43050
+    },
+    {
+      "epoch": 0.3710399449035813,
+      "grad_norm": 1.8125,
+      "learning_rate": 0.000197409563812063,
+      "loss": 4.8914,
+      "step": 43100
+    },
+    {
+      "epoch": 0.3714703856749311,
+      "grad_norm": 1.7734375,
+      "learning_rate": 0.00019740343374053014,
+      "loss": 4.6111,
+      "step": 43150
+    },
+    {
+      "epoch": 0.371900826446281,
+      "grad_norm": 1.6796875,
+      "learning_rate": 0.0001973972965197948,
+      "loss": 4.9909,
+      "step": 43200
+    },
+    {
+      "epoch": 0.3723312672176309,
+      "grad_norm": 1.9765625,
+      "learning_rate": 0.00019739115215030748,
+      "loss": 4.5076,
+      "step": 43250
+    },
+    {
+      "epoch": 0.37276170798898073,
+      "grad_norm": 3.296875,
+      "learning_rate": 0.0001973850006325191,
+      "loss": 5.1815,
+      "step": 43300
+    },
+    {
+      "epoch": 0.3731921487603306,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.00019737884196688126,
+      "loss": 4.8593,
+      "step": 43350
+    },
+    {
+      "epoch": 0.3736225895316804,
+      "grad_norm": 0.765625,
+      "learning_rate": 0.0001973726761538459,
+      "loss": 5.1336,
+      "step": 43400
+    },
+    {
+      "epoch": 0.3740530303030303,
+      "grad_norm": 1.6015625,
+      "learning_rate": 0.00019736650319386562,
+      "loss": 4.9215,
+      "step": 43450
+    },
+    {
+      "epoch": 0.3744834710743802,
+      "grad_norm": 1.734375,
+      "learning_rate": 0.0001973603230873935,
+      "loss": 4.9942,
+      "step": 43500
+    },
+    {
+      "epoch": 0.37491391184573003,
+      "grad_norm": 1.859375,
+      "learning_rate": 0.00019735413583488314,
+      "loss": 4.55,
+      "step": 43550
+    },
+    {
+      "epoch": 0.3753443526170799,
+      "grad_norm": 3.5625,
+      "learning_rate": 0.00019734794143678864,
+      "loss": 4.8435,
+      "step": 43600
+    },
+    {
+      "epoch": 0.37577479338842973,
+      "grad_norm": 1.5859375,
+      "learning_rate": 0.00019734173989356473,
+      "loss": 4.8101,
+      "step": 43650
+    },
+    {
+      "epoch": 0.37620523415977963,
+      "grad_norm": 1.921875,
+      "learning_rate": 0.00019733553120566654,
+      "loss": 4.5615,
+      "step": 43700
+    },
+    {
+      "epoch": 0.3766356749311295,
+      "grad_norm": 1.9765625,
+      "learning_rate": 0.00019732931537354978,
+      "loss": 4.6809,
+      "step": 43750
+    },
+    {
+      "epoch": 0.37706611570247933,
+      "grad_norm": 3.125,
+      "learning_rate": 0.00019732309239767069,
+      "loss": 5.0443,
+      "step": 43800
+    },
+    {
+      "epoch": 0.3774965564738292,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.000197316862278486,
+      "loss": 4.5044,
+      "step": 43850
+    },
+    {
+      "epoch": 0.3779269972451791,
+      "grad_norm": 1.9921875,
+      "learning_rate": 0.000197310625016453,
+      "loss": 4.8331,
+      "step": 43900
+    },
+    {
+      "epoch": 0.37835743801652894,
+      "grad_norm": 1.90625,
+      "learning_rate": 0.00019730438061202948,
+      "loss": 5.0114,
+      "step": 43950
+    },
+    {
+      "epoch": 0.3787878787878788,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0001972981290656738,
+      "loss": 5.1022,
+      "step": 44000
+    },
+    {
+      "epoch": 0.37921831955922863,
+      "grad_norm": 0.76953125,
+      "learning_rate": 0.00019729187037784477,
+      "loss": 4.5445,
+      "step": 44050
+    },
+    {
+      "epoch": 0.37964876033057854,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.0001972856045490018,
+      "loss": 4.8187,
+      "step": 44100
+    },
+    {
+      "epoch": 0.3800792011019284,
+      "grad_norm": 2.421875,
+      "learning_rate": 0.00019727933157960478,
+      "loss": 5.0647,
+      "step": 44150
+    },
+    {
+      "epoch": 0.38050964187327824,
+      "grad_norm": 3.09375,
+      "learning_rate": 0.0001972730514701141,
+      "loss": 5.378,
+      "step": 44200
+    },
+    {
+      "epoch": 0.3809400826446281,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00019726676422099075,
+      "loss": 5.1012,
+      "step": 44250
+    },
+    {
+      "epoch": 0.38137052341597794,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.00019726046983269618,
+      "loss": 4.9323,
+      "step": 44300
+    },
+    {
+      "epoch": 0.38180096418732784,
+      "grad_norm": 1.765625,
+      "learning_rate": 0.00019725416830569236,
+      "loss": 5.0225,
+      "step": 44350
+    },
+    {
+      "epoch": 0.3822314049586777,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.00019724785964044187,
+      "loss": 4.6229,
+      "step": 44400
+    },
+    {
+      "epoch": 0.38266184573002754,
+      "grad_norm": 0.98046875,
+      "learning_rate": 0.0001972415438374077,
+      "loss": 5.0021,
+      "step": 44450
+    },
+    {
+      "epoch": 0.3830922865013774,
+      "grad_norm": 1.40625,
+      "learning_rate": 0.00019723522089705342,
+      "loss": 4.5124,
+      "step": 44500
+    },
+    {
+      "epoch": 0.3835227272727273,
+      "grad_norm": 1.125,
+      "learning_rate": 0.00019722889081984316,
+      "loss": 4.9181,
+      "step": 44550
+    },
+    {
+      "epoch": 0.38395316804407714,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.0001972225536062415,
+      "loss": 4.9511,
+      "step": 44600
+    },
+    {
+      "epoch": 0.384383608815427,
+      "grad_norm": 4.0,
+      "learning_rate": 0.00019721620925671355,
+      "loss": 4.7302,
+      "step": 44650
+    },
+    {
+      "epoch": 0.38481404958677684,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.00019720985777172504,
+      "loss": 5.0943,
+      "step": 44700
+    },
+    {
+      "epoch": 0.38524449035812675,
+      "grad_norm": 3.375,
+      "learning_rate": 0.0001972034991517421,
+      "loss": 4.7611,
+      "step": 44750
+    },
+    {
+      "epoch": 0.3856749311294766,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.0001971971333972315,
+      "loss": 4.7436,
+      "step": 44800
+    },
+    {
+      "epoch": 0.38610537190082644,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.0001971907605086604,
+      "loss": 4.4675,
+      "step": 44850
+    },
+    {
+      "epoch": 0.3865358126721763,
+      "grad_norm": 1.8359375,
+      "learning_rate": 0.0001971843804864966,
+      "loss": 4.7749,
+      "step": 44900
+    },
+    {
+      "epoch": 0.3869662534435262,
+      "grad_norm": 2.5,
+      "learning_rate": 0.0001971779933312084,
+      "loss": 5.3325,
+      "step": 44950
+    },
+    {
+      "epoch": 0.38739669421487605,
+      "grad_norm": 1.328125,
+      "learning_rate": 0.0001971715990432645,
+      "loss": 4.9475,
+      "step": 45000
+    },
+    {
+      "epoch": 0.38739669421487605,
+      "eval_loss": 5.486639976501465,
+      "eval_runtime": 22.0151,
+      "eval_samples_per_second": 29.071,
+      "eval_steps_per_second": 14.536,
+      "eval_tts_loss": 6.651565814050026,
+      "step": 45000
+    },
+    {
+      "epoch": 0.3878271349862259,
+      "grad_norm": 3.03125,
+      "learning_rate": 0.00019716519762313434,
+      "loss": 5.0611,
+      "step": 45050
+    },
+    {
+      "epoch": 0.38825757575757575,
+      "grad_norm": 2.296875,
+      "learning_rate": 0.00019715878907128776,
+      "loss": 5.0095,
+      "step": 45100
+    },
+    {
+      "epoch": 0.3886880165289256,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.0001971523733881951,
+      "loss": 4.7267,
+      "step": 45150
+    },
+    {
+      "epoch": 0.3891184573002755,
+      "grad_norm": 1.421875,
+      "learning_rate": 0.0001971459505743272,
+      "loss": 4.7917,
+      "step": 45200
+    },
+    {
+      "epoch": 0.38954889807162535,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.0001971395206301556,
+      "loss": 4.4387,
+      "step": 45250
+    },
+    {
+      "epoch": 0.3899793388429752,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00019713308355615218,
+      "loss": 5.0246,
+      "step": 45300
+    },
+    {
+      "epoch": 0.39040977961432505,
+      "grad_norm": 1.203125,
+      "learning_rate": 0.00019712663935278938,
+      "loss": 4.9227,
+      "step": 45350
+    },
+    {
+      "epoch": 0.39084022038567495,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.00019712018802054026,
+      "loss": 4.914,
+      "step": 45400
+    },
+    {
+      "epoch": 0.3912706611570248,
+      "grad_norm": 1.75,
+      "learning_rate": 0.0001971137295598783,
+      "loss": 5.1644,
+      "step": 45450
+    },
+    {
+      "epoch": 0.39170110192837465,
+      "grad_norm": 3.796875,
+      "learning_rate": 0.00019710726397127753,
+      "loss": 4.3398,
+      "step": 45500
+    },
+    {
+      "epoch": 0.3921315426997245,
+      "grad_norm": 3.046875,
+      "learning_rate": 0.0001971007912552125,
+      "loss": 4.9143,
+      "step": 45550
+    },
+    {
+      "epoch": 0.3925619834710744,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.0001970943114121583,
+      "loss": 4.5839,
+      "step": 45600
+    },
+    {
+      "epoch": 0.39299242424242425,
+      "grad_norm": 2.5,
+      "learning_rate": 0.00019708782444259056,
+      "loss": 5.107,
+      "step": 45650
+    },
+    {
+      "epoch": 0.3934228650137741,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00019708133034698544,
+      "loss": 4.8543,
+      "step": 45700
+    },
+    {
+      "epoch": 0.39385330578512395,
+      "grad_norm": 1.8984375,
+      "learning_rate": 0.0001970748291258195,
+      "loss": 5.2229,
+      "step": 45750
+    },
+    {
+      "epoch": 0.3942837465564738,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00019706832077956996,
+      "loss": 4.8235,
+      "step": 45800
+    },
+    {
+      "epoch": 0.3947141873278237,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00019706180530871455,
+      "loss": 4.9943,
+      "step": 45850
+    },
+    {
+      "epoch": 0.39514462809917356,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.00019705528271373146,
+      "loss": 4.9746,
+      "step": 45900
+    },
+    {
+      "epoch": 0.3955750688705234,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00019704875299509942,
+      "loss": 5.3528,
+      "step": 45950
+    },
+    {
+      "epoch": 0.39600550964187325,
+      "grad_norm": 1.6484375,
+      "learning_rate": 0.00019704221615329774,
+      "loss": 5.0962,
+      "step": 46000
+    },
+    {
+      "epoch": 0.39643595041322316,
+      "grad_norm": 1.9765625,
+      "learning_rate": 0.00019703567218880616,
+      "loss": 5.0475,
+      "step": 46050
+    },
+    {
+      "epoch": 0.396866391184573,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.00019702912110210503,
+      "loss": 4.435,
+      "step": 46100
+    },
+    {
+      "epoch": 0.39729683195592286,
+      "grad_norm": 1.7265625,
+      "learning_rate": 0.00019702256289367518,
+      "loss": 5.1928,
+      "step": 46150
+    },
+    {
+      "epoch": 0.3977272727272727,
+      "grad_norm": 3.578125,
+      "learning_rate": 0.00019701599756399797,
+      "loss": 4.7587,
+      "step": 46200
+    },
+    {
+      "epoch": 0.3981577134986226,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00019700942511355525,
+      "loss": 4.5542,
+      "step": 46250
+    },
+    {
+      "epoch": 0.39858815426997246,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.00019700284554282946,
+      "loss": 5.2381,
+      "step": 46300
+    },
+    {
+      "epoch": 0.3990185950413223,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.0001969962588523035,
+      "loss": 5.0197,
+      "step": 46350
+    },
+    {
+      "epoch": 0.39944903581267216,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.0001969896650424609,
+      "loss": 4.9794,
+      "step": 46400
+    },
+    {
+      "epoch": 0.39987947658402206,
+      "grad_norm": 1.3828125,
+      "learning_rate": 0.00019698306411378548,
+      "loss": 5.0362,
+      "step": 46450
+    },
+    {
+      "epoch": 0.4003099173553719,
+      "grad_norm": 1.921875,
+      "learning_rate": 0.00019697645606676185,
+      "loss": 4.4054,
+      "step": 46500
+    },
+    {
+      "epoch": 0.40074035812672176,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.00019696984090187498,
+      "loss": 5.0158,
+      "step": 46550
+    },
+    {
+      "epoch": 0.4011707988980716,
+      "grad_norm": 1.984375,
+      "learning_rate": 0.00019696321861961046,
+      "loss": 4.9003,
+      "step": 46600
+    },
+    {
+      "epoch": 0.40160123966942146,
+      "grad_norm": 1.7109375,
+      "learning_rate": 0.0001969565892204543,
+      "loss": 4.7631,
+      "step": 46650
+    },
+    {
+      "epoch": 0.40203168044077137,
+      "grad_norm": 1.5078125,
+      "learning_rate": 0.00019694995270489306,
+      "loss": 4.7591,
+      "step": 46700
+    },
+    {
+      "epoch": 0.4024621212121212,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.00019694330907341393,
+      "loss": 4.4943,
+      "step": 46750
+    },
+    {
+      "epoch": 0.40289256198347106,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.0001969366583265045,
+      "loss": 4.812,
+      "step": 46800
+    },
+    {
+      "epoch": 0.4033230027548209,
+      "grad_norm": 1.0,
+      "learning_rate": 0.0001969300004646529,
+      "loss": 4.7527,
+      "step": 46850
+    },
+    {
+      "epoch": 0.4037534435261708,
+      "grad_norm": 1.5390625,
+      "learning_rate": 0.0001969233354883478,
+      "loss": 4.9423,
+      "step": 46900
+    },
+    {
+      "epoch": 0.40418388429752067,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.0001969166633980784,
+      "loss": 4.8899,
+      "step": 46950
+    },
+    {
+      "epoch": 0.4046143250688705,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.00019690998419433448,
+      "loss": 4.7378,
+      "step": 47000
+    },
+    {
+      "epoch": 0.40504476584022037,
+      "grad_norm": 1.8359375,
+      "learning_rate": 0.0001969032978776062,
+      "loss": 4.3138,
+      "step": 47050
+    },
+    {
+      "epoch": 0.40547520661157027,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00019689660444838434,
+      "loss": 4.676,
+      "step": 47100
+    },
+    {
+      "epoch": 0.4059056473829201,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.0001968899039071602,
+      "loss": 4.9377,
+      "step": 47150
+    },
+    {
+      "epoch": 0.40633608815426997,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00019688319625442557,
+      "loss": 4.7513,
+      "step": 47200
+    },
+    {
+      "epoch": 0.4067665289256198,
+      "grad_norm": 1.7265625,
+      "learning_rate": 0.0001968764814906728,
+      "loss": 4.5865,
+      "step": 47250
+    },
+    {
+      "epoch": 0.4071969696969697,
+      "grad_norm": 4.15625,
+      "learning_rate": 0.00019686975961639468,
+      "loss": 4.9636,
+      "step": 47300
+    },
+    {
+      "epoch": 0.4076274104683196,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.00019686303063208465,
+      "loss": 4.5476,
+      "step": 47350
+    },
+    {
+      "epoch": 0.4080578512396694,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.00019685629453823663,
+      "loss": 5.0127,
+      "step": 47400
+    },
+    {
+      "epoch": 0.40848829201101927,
+      "grad_norm": 1.5625,
+      "learning_rate": 0.00019684955133534492,
+      "loss": 4.731,
+      "step": 47450
+    },
+    {
+      "epoch": 0.4089187327823691,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.00019684280102390457,
+      "loss": 4.8993,
+      "step": 47500
+    },
+    {
+      "epoch": 0.409349173553719,
+      "grad_norm": 3.40625,
+      "learning_rate": 0.00019683604360441092,
+      "loss": 5.4113,
+      "step": 47550
+    },
+    {
+      "epoch": 0.4097796143250689,
+      "grad_norm": 4.5625,
+      "learning_rate": 0.00019682927907736005,
+      "loss": 4.563,
+      "step": 47600
+    },
+    {
+      "epoch": 0.4102100550964187,
+      "grad_norm": 3.59375,
+      "learning_rate": 0.00019682250744324843,
+      "loss": 4.8675,
+      "step": 47650
+    },
+    {
+      "epoch": 0.4106404958677686,
+      "grad_norm": 1.84375,
+      "learning_rate": 0.0001968157287025731,
+      "loss": 4.8817,
+      "step": 47700
+    },
+    {
+      "epoch": 0.4110709366391185,
+      "grad_norm": 1.2265625,
+      "learning_rate": 0.00019680894285583155,
+      "loss": 5.0027,
+      "step": 47750
+    },
+    {
+      "epoch": 0.4115013774104683,
+      "grad_norm": 2.21875,
+      "learning_rate": 0.00019680214990352193,
+      "loss": 4.8271,
+      "step": 47800
+    },
+    {
+      "epoch": 0.4119318181818182,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.00019679534984614274,
+      "loss": 4.8948,
+      "step": 47850
+    },
+    {
+      "epoch": 0.412362258953168,
+      "grad_norm": 2.296875,
+      "learning_rate": 0.00019678854268419314,
+      "loss": 4.8333,
+      "step": 47900
+    },
+    {
+      "epoch": 0.41279269972451793,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.0001967817284181728,
+      "loss": 4.8425,
+      "step": 47950
+    },
+    {
+      "epoch": 0.4132231404958678,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.00019677490704858176,
+      "loss": 4.9557,
+      "step": 48000
+    },
+    {
+      "epoch": 0.4132231404958678,
+      "eval_loss": 5.4774956703186035,
+      "eval_runtime": 21.8718,
+      "eval_samples_per_second": 29.261,
+      "eval_steps_per_second": 14.631,
+      "eval_tts_loss": 6.593891539484386,
+      "step": 48000
+    },
+    {
+      "epoch": 0.41365358126721763,
+      "grad_norm": 3.765625,
+      "learning_rate": 0.0001967680785759208,
+      "loss": 4.9363,
+      "step": 48050
+    },
+    {
+      "epoch": 0.4140840220385675,
+      "grad_norm": 1.3359375,
+      "learning_rate": 0.00019676124300069104,
+      "loss": 4.9799,
+      "step": 48100
+    },
+    {
+      "epoch": 0.41451446280991733,
+      "grad_norm": 1.4296875,
+      "learning_rate": 0.00019675440032339427,
+      "loss": 4.9791,
+      "step": 48150
+    },
+    {
+      "epoch": 0.41494490358126723,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.0001967475505445327,
+      "loss": 4.8335,
+      "step": 48200
+    },
+    {
+      "epoch": 0.4153753443526171,
+      "grad_norm": 1.7890625,
+      "learning_rate": 0.00019674069366460903,
+      "loss": 4.8282,
+      "step": 48250
+    },
+    {
+      "epoch": 0.41580578512396693,
+      "grad_norm": 1.7265625,
+      "learning_rate": 0.00019673382968412662,
+      "loss": 4.7353,
+      "step": 48300
+    },
+    {
+      "epoch": 0.4162362258953168,
+      "grad_norm": 1.8515625,
+      "learning_rate": 0.00019672695860358923,
+      "loss": 4.3679,
+      "step": 48350
+    },
+    {
+      "epoch": 0.4166666666666667,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.0001967200804235012,
+      "loss": 4.8819,
+      "step": 48400
+    },
+    {
+      "epoch": 0.41709710743801653,
+      "grad_norm": 1.8828125,
+      "learning_rate": 0.00019671319514436735,
+      "loss": 4.6303,
+      "step": 48450
+    },
+    {
+      "epoch": 0.4175275482093664,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.00019670630276669306,
+      "loss": 4.6477,
+      "step": 48500
+    },
+    {
+      "epoch": 0.41795798898071623,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.0001966994032909842,
+      "loss": 4.6415,
+      "step": 48550
+    },
+    {
+      "epoch": 0.41838842975206614,
+      "grad_norm": 3.625,
+      "learning_rate": 0.00019669249671774722,
+      "loss": 5.1527,
+      "step": 48600
+    },
+    {
+      "epoch": 0.418818870523416,
+      "grad_norm": 2.875,
+      "learning_rate": 0.000196685583047489,
+      "loss": 4.9671,
+      "step": 48650
+    },
+    {
+      "epoch": 0.41924931129476584,
+      "grad_norm": 1.453125,
+      "learning_rate": 0.00019667866228071702,
+      "loss": 4.906,
+      "step": 48700
+    },
+    {
+      "epoch": 0.4196797520661157,
+      "grad_norm": 0.73828125,
+      "learning_rate": 0.00019667173441793927,
+      "loss": 4.6206,
+      "step": 48750
+    },
+    {
+      "epoch": 0.4201101928374656,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.00019666479945966417,
+      "loss": 4.8596,
+      "step": 48800
+    },
+    {
+      "epoch": 0.42054063360881544,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.00019665785740640077,
+      "loss": 4.8986,
+      "step": 48850
+    },
+    {
+      "epoch": 0.4209710743801653,
+      "grad_norm": 1.3046875,
+      "learning_rate": 0.0001966509082586586,
+      "loss": 4.6868,
+      "step": 48900
+    },
+    {
+      "epoch": 0.42140151515151514,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.00019664395201694773,
+      "loss": 5.3235,
+      "step": 48950
+    },
+    {
+      "epoch": 0.421831955922865,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.0001966369886817787,
+      "loss": 4.627,
+      "step": 49000
+    },
+    {
+      "epoch": 0.4222623966942149,
+      "grad_norm": 0.890625,
+      "learning_rate": 0.00019663001825366265,
+      "loss": 4.6956,
+      "step": 49050
+    },
+    {
+      "epoch": 0.42269283746556474,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.00019662304073311116,
+      "loss": 4.9834,
+      "step": 49100
+    },
+    {
+      "epoch": 0.4231232782369146,
+      "grad_norm": 1.7265625,
+      "learning_rate": 0.00019661605612063634,
+      "loss": 5.1055,
+      "step": 49150
+    },
+    {
+      "epoch": 0.42355371900826444,
+      "grad_norm": 3.203125,
+      "learning_rate": 0.0001966090644167509,
+      "loss": 4.7372,
+      "step": 49200
+    },
+    {
+      "epoch": 0.42398415977961434,
+      "grad_norm": 1.703125,
+      "learning_rate": 0.000196602065621968,
+      "loss": 4.866,
+      "step": 49250
+    },
+    {
+      "epoch": 0.4244146005509642,
+      "grad_norm": 0.8671875,
+      "learning_rate": 0.0001965950597368013,
+      "loss": 4.77,
+      "step": 49300
+    },
+    {
+      "epoch": 0.42484504132231404,
+      "grad_norm": 1.53125,
+      "learning_rate": 0.00019658804676176504,
+      "loss": 5.0861,
+      "step": 49350
+    },
+    {
+      "epoch": 0.4252754820936639,
+      "grad_norm": 1.265625,
+      "learning_rate": 0.000196581026697374,
+      "loss": 4.5885,
+      "step": 49400
+    },
+    {
+      "epoch": 0.4257059228650138,
+      "grad_norm": 1.96875,
+      "learning_rate": 0.00019657399954414335,
+      "loss": 5.1443,
+      "step": 49450
+    },
+    {
+      "epoch": 0.42613636363636365,
+      "grad_norm": 1.5078125,
+      "learning_rate": 0.00019656696530258896,
+      "loss": 4.2604,
+      "step": 49500
+    },
+    {
+      "epoch": 0.4265668044077135,
+      "grad_norm": 2.375,
+      "learning_rate": 0.00019655992397322707,
+      "loss": 4.8881,
+      "step": 49550
+    },
+    {
+      "epoch": 0.42699724517906334,
+      "grad_norm": 2.21875,
+      "learning_rate": 0.00019655287555657452,
+      "loss": 4.8634,
+      "step": 49600
+    },
+    {
+      "epoch": 0.42742768595041325,
+      "grad_norm": 1.5234375,
+      "learning_rate": 0.00019654582005314863,
+      "loss": 5.0439,
+      "step": 49650
+    },
+    {
+      "epoch": 0.4278581267217631,
+      "grad_norm": 3.015625,
+      "learning_rate": 0.00019653875746346728,
+      "loss": 5.151,
+      "step": 49700
+    },
+    {
+      "epoch": 0.42828856749311295,
+      "grad_norm": 1.2109375,
+      "learning_rate": 0.00019653168778804885,
+      "loss": 4.8142,
+      "step": 49750
+    },
+    {
+      "epoch": 0.4287190082644628,
+      "grad_norm": 1.171875,
+      "learning_rate": 0.0001965246110274122,
+      "loss": 5.2469,
+      "step": 49800
+    },
+    {
+      "epoch": 0.42914944903581265,
+      "grad_norm": 1.640625,
+      "learning_rate": 0.0001965175271820768,
+      "loss": 4.9468,
+      "step": 49850
+    },
+    {
+      "epoch": 0.42957988980716255,
+      "grad_norm": 3.828125,
+      "learning_rate": 0.00019651043625256259,
+      "loss": 5.138,
+      "step": 49900
+    },
+    {
+      "epoch": 0.4300103305785124,
+      "grad_norm": 1.5078125,
+      "learning_rate": 0.00019650333823938997,
+      "loss": 4.9163,
+      "step": 49950
+    },
+    {
+      "epoch": 0.43044077134986225,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.00019649623314307995,
+      "loss": 4.7608,
+      "step": 50000
+    },
+    {
+      "epoch": 0.4308712121212121,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00019648912096415408,
+      "loss": 5.1712,
+      "step": 50050
+    },
+    {
+      "epoch": 0.431301652892562,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.00019648200170313432,
+      "loss": 5.0902,
+      "step": 50100
+    },
+    {
+      "epoch": 0.43173209366391185,
+      "grad_norm": 3.484375,
+      "learning_rate": 0.0001964748753605432,
+      "loss": 5.0191,
+      "step": 50150
+    },
+    {
+      "epoch": 0.4321625344352617,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.0001964677419369038,
+      "loss": 4.9103,
+      "step": 50200
+    },
+    {
+      "epoch": 0.43259297520661155,
+      "grad_norm": 3.03125,
+      "learning_rate": 0.0001964606014327397,
+      "loss": 4.9049,
+      "step": 50250
+    },
+    {
+      "epoch": 0.43302341597796146,
+      "grad_norm": 3.359375,
+      "learning_rate": 0.000196453453848575,
+      "loss": 4.5552,
+      "step": 50300
+    },
+    {
+      "epoch": 0.4334538567493113,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.0001964462991849343,
+      "loss": 4.4175,
+      "step": 50350
+    },
+    {
+      "epoch": 0.43388429752066116,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.0001964391374423428,
+      "loss": 5.0406,
+      "step": 50400
+    },
+    {
+      "epoch": 0.434314738292011,
+      "grad_norm": 1.71875,
+      "learning_rate": 0.00019643196862132605,
+      "loss": 4.4845,
+      "step": 50450
+    },
+    {
+      "epoch": 0.4347451790633609,
+      "grad_norm": 3.28125,
+      "learning_rate": 0.0001964247927224103,
+      "loss": 4.7683,
+      "step": 50500
+    },
+    {
+      "epoch": 0.43517561983471076,
+      "grad_norm": 1.640625,
+      "learning_rate": 0.00019641760974612222,
+      "loss": 4.9496,
+      "step": 50550
+    },
+    {
+      "epoch": 0.4356060606060606,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.00019641041969298905,
+      "loss": 4.7391,
+      "step": 50600
+    },
+    {
+      "epoch": 0.43603650137741046,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.0001964032225635385,
+      "loss": 5.0805,
+      "step": 50650
+    },
+    {
+      "epoch": 0.4364669421487603,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00019639601835829882,
+      "loss": 4.4855,
+      "step": 50700
+    },
+    {
+      "epoch": 0.4368973829201102,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.00019638880707779878,
+      "loss": 5.1064,
+      "step": 50750
+    },
+    {
+      "epoch": 0.43732782369146006,
+      "grad_norm": 1.9296875,
+      "learning_rate": 0.00019638158872256774,
+      "loss": 4.5543,
+      "step": 50800
+    },
+    {
+      "epoch": 0.4377582644628099,
+      "grad_norm": 2.234375,
+      "learning_rate": 0.0001963743632931354,
+      "loss": 4.9067,
+      "step": 50850
+    },
+    {
+      "epoch": 0.43818870523415976,
+      "grad_norm": 1.7265625,
+      "learning_rate": 0.0001963671307900322,
+      "loss": 4.9135,
+      "step": 50900
+    },
+    {
+      "epoch": 0.43861914600550966,
+      "grad_norm": 0.734375,
+      "learning_rate": 0.0001963598912137889,
+      "loss": 5.1316,
+      "step": 50950
+    },
+    {
+      "epoch": 0.4390495867768595,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00019635264456493694,
+      "loss": 5.1975,
+      "step": 51000
+    },
+    {
+      "epoch": 0.4390495867768595,
+      "eval_loss": 5.461007595062256,
+      "eval_runtime": 21.9788,
+      "eval_samples_per_second": 29.119,
+      "eval_steps_per_second": 14.559,
+      "eval_tts_loss": 6.650621942901377,
+      "step": 51000
+    },
+    {
+      "epoch": 0.43948002754820936,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00019634539084400815,
+      "loss": 4.9518,
+      "step": 51050
+    },
+    {
+      "epoch": 0.4399104683195592,
+      "grad_norm": 1.7265625,
+      "learning_rate": 0.00019633813005153497,
+      "loss": 4.6406,
+      "step": 51100
+    },
+    {
+      "epoch": 0.4403409090909091,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00019633086218805032,
+      "loss": 4.8793,
+      "step": 51150
+    },
+    {
+      "epoch": 0.44077134986225897,
+      "grad_norm": 1.296875,
+      "learning_rate": 0.00019632358725408765,
+      "loss": 5.1101,
+      "step": 51200
+    },
+    {
+      "epoch": 0.4412017906336088,
+      "grad_norm": 3.328125,
+      "learning_rate": 0.00019631630525018094,
+      "loss": 5.1925,
+      "step": 51250
+    },
+    {
+      "epoch": 0.44163223140495866,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.00019630901617686462,
+      "loss": 5.0485,
+      "step": 51300
+    },
+    {
+      "epoch": 0.4420626721763085,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.00019630172003467375,
+      "loss": 4.6684,
+      "step": 51350
+    },
+    {
+      "epoch": 0.4424931129476584,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.00019629441682414382,
+      "loss": 5.0334,
+      "step": 51400
+    },
+    {
+      "epoch": 0.44292355371900827,
+      "grad_norm": 2.375,
+      "learning_rate": 0.0001962871065458109,
+      "loss": 4.949,
+      "step": 51450
+    },
+    {
+      "epoch": 0.4433539944903581,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.00019627978920021147,
+      "loss": 4.9946,
+      "step": 51500
+    },
+    {
+      "epoch": 0.44378443526170797,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.00019627246478788274,
+      "loss": 4.7547,
+      "step": 51550
+    },
+    {
+      "epoch": 0.44421487603305787,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.00019626513330936218,
+      "loss": 4.5459,
+      "step": 51600
+    },
+    {
+      "epoch": 0.4446453168044077,
+      "grad_norm": 1.9453125,
+      "learning_rate": 0.00019625779476518796,
+      "loss": 5.0765,
+      "step": 51650
+    },
+    {
+      "epoch": 0.44507575757575757,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.0001962504491558987,
+      "loss": 4.6943,
+      "step": 51700
+    },
+    {
+      "epoch": 0.4455061983471074,
+      "grad_norm": 1.703125,
+      "learning_rate": 0.0001962430964820336,
+      "loss": 4.7206,
+      "step": 51750
+    },
+    {
+      "epoch": 0.4459366391184573,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.00019623573674413223,
+      "loss": 4.7757,
+      "step": 51800
+    },
+    {
+      "epoch": 0.4463670798898072,
+      "grad_norm": 1.734375,
+      "learning_rate": 0.00019622836994273487,
+      "loss": 4.5886,
+      "step": 51850
+    },
+    {
+      "epoch": 0.446797520661157,
+      "grad_norm": 3.296875,
+      "learning_rate": 0.00019622099607838223,
+      "loss": 4.6718,
+      "step": 51900
+    },
+    {
+      "epoch": 0.44722796143250687,
+      "grad_norm": 1.421875,
+      "learning_rate": 0.00019621361515161548,
+      "loss": 4.576,
+      "step": 51950
+    },
+    {
+      "epoch": 0.4476584022038568,
+      "grad_norm": 1.296875,
+      "learning_rate": 0.00019620622716297638,
+      "loss": 4.7298,
+      "step": 52000
+    },
+    {
+      "epoch": 0.4480888429752066,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.00019619883211300718,
+      "loss": 4.7639,
+      "step": 52050
+    },
+    {
+      "epoch": 0.4485192837465565,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.0001961914300022507,
+      "loss": 4.7986,
+      "step": 52100
+    },
+    {
+      "epoch": 0.4489497245179063,
+      "grad_norm": 1.6484375,
+      "learning_rate": 0.00019618402083125024,
+      "loss": 4.9084,
+      "step": 52150
+    },
+    {
+      "epoch": 0.4493801652892562,
+      "grad_norm": 1.75,
+      "learning_rate": 0.00019617660460054956,
+      "loss": 5.2089,
+      "step": 52200
+    },
+    {
+      "epoch": 0.4498106060606061,
+      "grad_norm": 1.9609375,
+      "learning_rate": 0.00019616918131069305,
+      "loss": 5.0402,
+      "step": 52250
+    },
+    {
+      "epoch": 0.4502410468319559,
+      "grad_norm": 3.515625,
+      "learning_rate": 0.00019616175096222554,
+      "loss": 5.2387,
+      "step": 52300
+    },
+    {
+      "epoch": 0.4506714876033058,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.0001961543135556924,
+      "loss": 5.2524,
+      "step": 52350
+    },
+    {
+      "epoch": 0.4511019283746556,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00019614686909163952,
+      "loss": 4.9647,
+      "step": 52400
+    },
+    {
+      "epoch": 0.45153236914600553,
+      "grad_norm": 1.4765625,
+      "learning_rate": 0.00019613941757061335,
+      "loss": 4.9352,
+      "step": 52450
+    },
+    {
+      "epoch": 0.4519628099173554,
+      "grad_norm": 2.25,
+      "learning_rate": 0.00019613195899316076,
+      "loss": 4.6792,
+      "step": 52500
+    },
+    {
+      "epoch": 0.45239325068870523,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00019612449335982922,
+      "loss": 4.7962,
+      "step": 52550
+    },
+    {
+      "epoch": 0.4528236914600551,
+      "grad_norm": 2.375,
+      "learning_rate": 0.00019611702067116668,
+      "loss": 4.75,
+      "step": 52600
+    },
+    {
+      "epoch": 0.453254132231405,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.0001961095409277216,
+      "loss": 4.6786,
+      "step": 52650
+    },
+    {
+      "epoch": 0.45368457300275483,
+      "grad_norm": 1.25,
+      "learning_rate": 0.00019610205413004303,
+      "loss": 4.2868,
+      "step": 52700
+    },
+    {
+      "epoch": 0.4541150137741047,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00019609456027868045,
+      "loss": 4.8512,
+      "step": 52750
+    },
+    {
+      "epoch": 0.45454545454545453,
+      "grad_norm": 0.9140625,
+      "learning_rate": 0.00019608705937418388,
+      "loss": 5.1612,
+      "step": 52800
+    },
+    {
+      "epoch": 0.45497589531680444,
+      "grad_norm": 0.70703125,
+      "learning_rate": 0.00019607955141710392,
+      "loss": 4.6844,
+      "step": 52850
+    },
+    {
+      "epoch": 0.4554063360881543,
+      "grad_norm": 1.7890625,
+      "learning_rate": 0.0001960720364079916,
+      "loss": 4.8468,
+      "step": 52900
+    },
+    {
+      "epoch": 0.45583677685950413,
+      "grad_norm": 0.462890625,
+      "learning_rate": 0.00019606451434739847,
+      "loss": 5.0051,
+      "step": 52950
+    },
+    {
+      "epoch": 0.456267217630854,
+      "grad_norm": 1.328125,
+      "learning_rate": 0.00019605698523587672,
+      "loss": 4.7107,
+      "step": 53000
+    },
+    {
+      "epoch": 0.45669765840220383,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.0001960494490739789,
+      "loss": 4.8353,
+      "step": 53050
+    },
+    {
+      "epoch": 0.45712809917355374,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00019604190586225819,
+      "loss": 4.9876,
+      "step": 53100
+    },
+    {
+      "epoch": 0.4575585399449036,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00019603435560126823,
+      "loss": 4.9235,
+      "step": 53150
+    },
+    {
+      "epoch": 0.45798898071625344,
+      "grad_norm": 1.375,
+      "learning_rate": 0.00019602679829156322,
+      "loss": 4.6953,
+      "step": 53200
+    },
+    {
+      "epoch": 0.4584194214876033,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.00019601923393369777,
+      "loss": 5.0293,
+      "step": 53250
+    },
+    {
+      "epoch": 0.4588498622589532,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.00019601166252822716,
+      "loss": 4.8473,
+      "step": 53300
+    },
+    {
+      "epoch": 0.45928030303030304,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.0001960040840757071,
+      "loss": 4.9234,
+      "step": 53350
+    },
+    {
+      "epoch": 0.4597107438016529,
+      "grad_norm": 1.59375,
+      "learning_rate": 0.00019599649857669383,
+      "loss": 4.9429,
+      "step": 53400
+    },
+    {
+      "epoch": 0.46014118457300274,
+      "grad_norm": 1.8671875,
+      "learning_rate": 0.0001959889060317441,
+      "loss": 4.5721,
+      "step": 53450
+    },
+    {
+      "epoch": 0.46057162534435264,
+      "grad_norm": 1.9765625,
+      "learning_rate": 0.00019598130644141518,
+      "loss": 4.5661,
+      "step": 53500
+    },
+    {
+      "epoch": 0.4610020661157025,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.0001959736998062649,
+      "loss": 4.886,
+      "step": 53550
+    },
+    {
+      "epoch": 0.46143250688705234,
+      "grad_norm": 2.375,
+      "learning_rate": 0.00019596608612685153,
+      "loss": 4.9356,
+      "step": 53600
+    },
+    {
+      "epoch": 0.4618629476584022,
+      "grad_norm": 1.5859375,
+      "learning_rate": 0.0001959584654037339,
+      "loss": 4.7174,
+      "step": 53650
+    },
+    {
+      "epoch": 0.46229338842975204,
+      "grad_norm": 4.03125,
+      "learning_rate": 0.0001959508376374714,
+      "loss": 4.8117,
+      "step": 53700
+    },
+    {
+      "epoch": 0.46272382920110194,
+      "grad_norm": 2.5,
+      "learning_rate": 0.00019594320282862384,
+      "loss": 4.8152,
+      "step": 53750
+    },
+    {
+      "epoch": 0.4631542699724518,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00019593556097775164,
+      "loss": 4.7632,
+      "step": 53800
+    },
+    {
+      "epoch": 0.46358471074380164,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.00019592791208541567,
+      "loss": 4.5068,
+      "step": 53850
+    },
+    {
+      "epoch": 0.4640151515151515,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.00019592025615217733,
+      "loss": 4.7496,
+      "step": 53900
+    },
+    {
+      "epoch": 0.4644455922865014,
+      "grad_norm": 1.96875,
+      "learning_rate": 0.00019591259317859857,
+      "loss": 4.8606,
+      "step": 53950
+    },
+    {
+      "epoch": 0.46487603305785125,
+      "grad_norm": 3.46875,
+      "learning_rate": 0.00019590492316524184,
+      "loss": 4.969,
+      "step": 54000
+    },
+    {
+      "epoch": 0.46487603305785125,
+      "eval_loss": 5.447500705718994,
+      "eval_runtime": 21.7174,
+      "eval_samples_per_second": 29.469,
+      "eval_steps_per_second": 14.735,
+      "eval_tts_loss": 6.713471357580835,
+      "step": 54000
+    },
+    {
+      "epoch": 0.4653064738292011,
+      "grad_norm": 1.875,
+      "learning_rate": 0.00019589724611267012,
+      "loss": 4.9705,
+      "step": 54050
+    },
+    {
+      "epoch": 0.46573691460055094,
+      "grad_norm": 0.671875,
+      "learning_rate": 0.0001958895620214468,
+      "loss": 4.7252,
+      "step": 54100
+    },
+    {
+      "epoch": 0.46616735537190085,
+      "grad_norm": 2.5,
+      "learning_rate": 0.00019588187089213598,
+      "loss": 4.8425,
+      "step": 54150
+    },
+    {
+      "epoch": 0.4665977961432507,
+      "grad_norm": 1.8046875,
+      "learning_rate": 0.00019587417272530212,
+      "loss": 4.8882,
+      "step": 54200
+    },
+    {
+      "epoch": 0.46702823691460055,
+      "grad_norm": 3.78125,
+      "learning_rate": 0.00019586646752151027,
+      "loss": 4.8315,
+      "step": 54250
+    },
+    {
+      "epoch": 0.4674586776859504,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.000195858755281326,
+      "loss": 4.743,
+      "step": 54300
+    },
+    {
+      "epoch": 0.4678891184573003,
+      "grad_norm": 1.3046875,
+      "learning_rate": 0.00019585103600531527,
+      "loss": 4.571,
+      "step": 54350
+    },
+    {
+      "epoch": 0.46831955922865015,
+      "grad_norm": 3.328125,
+      "learning_rate": 0.00019584330969404476,
+      "loss": 5.2712,
+      "step": 54400
+    },
+    {
+      "epoch": 0.46875,
+      "grad_norm": 1.875,
+      "learning_rate": 0.00019583557634808154,
+      "loss": 4.2545,
+      "step": 54450
+    },
+    {
+      "epoch": 0.46918044077134985,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.00019582783596799321,
+      "loss": 4.5995,
+      "step": 54500
+    },
+    {
+      "epoch": 0.4696108815426997,
+      "grad_norm": 1.3125,
+      "learning_rate": 0.0001958200885543479,
+      "loss": 4.4244,
+      "step": 54550
+    },
+    {
+      "epoch": 0.4700413223140496,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.00019581233410771422,
+      "loss": 4.6487,
+      "step": 54600
+    },
+    {
+      "epoch": 0.47047176308539945,
+      "grad_norm": 3.3125,
+      "learning_rate": 0.0001958045726286614,
+      "loss": 4.7865,
+      "step": 54650
+    },
+    {
+      "epoch": 0.4709022038567493,
+      "grad_norm": 1.84375,
+      "learning_rate": 0.00019579680411775904,
+      "loss": 4.8535,
+      "step": 54700
+    },
+    {
+      "epoch": 0.47133264462809915,
+      "grad_norm": 2.234375,
+      "learning_rate": 0.0001957890285755774,
+      "loss": 4.7888,
+      "step": 54750
+    },
+    {
+      "epoch": 0.47176308539944906,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.00019578124600268715,
+      "loss": 4.4223,
+      "step": 54800
+    },
+    {
+      "epoch": 0.4721935261707989,
+      "grad_norm": 3.390625,
+      "learning_rate": 0.00019577345639965952,
+      "loss": 4.9533,
+      "step": 54850
+    },
+    {
+      "epoch": 0.47262396694214875,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.00019576565976706624,
+      "loss": 5.1331,
+      "step": 54900
+    },
+    {
+      "epoch": 0.4730544077134986,
+      "grad_norm": 0.953125,
+      "learning_rate": 0.00019575785610547957,
+      "loss": 5.0388,
+      "step": 54950
+    },
+    {
+      "epoch": 0.4734848484848485,
+      "grad_norm": 1.953125,
+      "learning_rate": 0.00019575004541547234,
+      "loss": 4.9406,
+      "step": 55000
+    },
+    {
+      "epoch": 0.47391528925619836,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.00019574222769761776,
+      "loss": 4.7529,
+      "step": 55050
+    },
+    {
+      "epoch": 0.4743457300275482,
+      "grad_norm": 2.296875,
+      "learning_rate": 0.0001957344029524897,
+      "loss": 4.5646,
+      "step": 55100
+    },
+    {
+      "epoch": 0.47477617079889806,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.00019572657118066235,
+      "loss": 4.8436,
+      "step": 55150
+    },
+    {
+      "epoch": 0.47520661157024796,
+      "grad_norm": 1.46875,
+      "learning_rate": 0.0001957187323827107,
+      "loss": 4.5989,
+      "step": 55200
+    },
+    {
+      "epoch": 0.4756370523415978,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.00019571088655921003,
+      "loss": 4.8347,
+      "step": 55250
+    },
+    {
+      "epoch": 0.47606749311294766,
+      "grad_norm": 1.9765625,
+      "learning_rate": 0.00019570303371073622,
+      "loss": 4.9279,
+      "step": 55300
+    },
+    {
+      "epoch": 0.4764979338842975,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.00019569517383786561,
+      "loss": 4.7051,
+      "step": 55350
+    },
+    {
+      "epoch": 0.47692837465564736,
+      "grad_norm": 3.78125,
+      "learning_rate": 0.0001956873069411752,
+      "loss": 5.0896,
+      "step": 55400
+    },
+    {
+      "epoch": 0.47735881542699726,
+      "grad_norm": 3.359375,
+      "learning_rate": 0.00019567943302124227,
+      "loss": 4.6047,
+      "step": 55450
+    },
+    {
+      "epoch": 0.4777892561983471,
+      "grad_norm": 0.7421875,
+      "learning_rate": 0.00019567155207864486,
+      "loss": 4.6413,
+      "step": 55500
+    },
+    {
+      "epoch": 0.47821969696969696,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.00019566366411396136,
+      "loss": 4.5122,
+      "step": 55550
+    },
+    {
+      "epoch": 0.4786501377410468,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.00019565576912777073,
+      "loss": 4.851,
+      "step": 55600
+    },
+    {
+      "epoch": 0.4790805785123967,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00019564786712065243,
+      "loss": 4.8688,
+      "step": 55650
+    },
+    {
+      "epoch": 0.47951101928374656,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.00019563995809318652,
+      "loss": 4.8937,
+      "step": 55700
+    },
+    {
+      "epoch": 0.4799414600550964,
+      "grad_norm": 3.078125,
+      "learning_rate": 0.00019563204204595344,
+      "loss": 4.8586,
+      "step": 55750
+    },
+    {
+      "epoch": 0.48037190082644626,
+      "grad_norm": 1.6796875,
+      "learning_rate": 0.00019562411897953424,
+      "loss": 4.4797,
+      "step": 55800
+    },
+    {
+      "epoch": 0.48080234159779617,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.00019561618889451046,
+      "loss": 4.846,
+      "step": 55850
+    },
+    {
+      "epoch": 0.481232782369146,
+      "grad_norm": 1.34375,
+      "learning_rate": 0.0001956082517914641,
+      "loss": 4.6205,
+      "step": 55900
+    },
+    {
+      "epoch": 0.48166322314049587,
+      "grad_norm": 3.5,
+      "learning_rate": 0.0001956003076709778,
+      "loss": 4.8843,
+      "step": 55950
+    },
+    {
+      "epoch": 0.4820936639118457,
+      "grad_norm": 1.65625,
+      "learning_rate": 0.00019559235653363456,
+      "loss": 5.009,
+      "step": 56000
+    },
+    {
+      "epoch": 0.48252410468319556,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.00019558439838001808,
+      "loss": 4.5382,
+      "step": 56050
+    },
+    {
+      "epoch": 0.48295454545454547,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.0001955764332107124,
+      "loss": 4.1105,
+      "step": 56100
+    },
+    {
+      "epoch": 0.4833849862258953,
+      "grad_norm": 3.5625,
+      "learning_rate": 0.00019556846102630212,
+      "loss": 4.8291,
+      "step": 56150
+    },
+    {
+      "epoch": 0.48381542699724517,
+      "grad_norm": 1.9296875,
+      "learning_rate": 0.00019556048182737247,
+      "loss": 4.5713,
+      "step": 56200
+    },
+    {
+      "epoch": 0.484245867768595,
+      "grad_norm": 1.7890625,
+      "learning_rate": 0.00019555249561450903,
+      "loss": 5.1687,
+      "step": 56250
+    },
+    {
+      "epoch": 0.4846763085399449,
+      "grad_norm": 0.984375,
+      "learning_rate": 0.00019554450238829803,
+      "loss": 4.69,
+      "step": 56300
+    },
+    {
+      "epoch": 0.48510674931129477,
+      "grad_norm": 2.25,
+      "learning_rate": 0.0001955365021493261,
+      "loss": 4.8855,
+      "step": 56350
+    },
+    {
+      "epoch": 0.4855371900826446,
+      "grad_norm": 3.375,
+      "learning_rate": 0.00019552849489818045,
+      "loss": 5.0094,
+      "step": 56400
+    },
+    {
+      "epoch": 0.48596763085399447,
+      "grad_norm": 2.875,
+      "learning_rate": 0.00019552048063544884,
+      "loss": 4.6932,
+      "step": 56450
+    },
+    {
+      "epoch": 0.4863980716253444,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.00019551245936171943,
+      "loss": 4.9196,
+      "step": 56500
+    },
+    {
+      "epoch": 0.4868285123966942,
+      "grad_norm": 1.8984375,
+      "learning_rate": 0.00019550443107758104,
+      "loss": 5.2379,
+      "step": 56550
+    },
+    {
+      "epoch": 0.4872589531680441,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.0001954963957836229,
+      "loss": 4.4327,
+      "step": 56600
+    },
+    {
+      "epoch": 0.4876893939393939,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.00019548835348043474,
+      "loss": 5.1293,
+      "step": 56650
+    },
+    {
+      "epoch": 0.4881198347107438,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.0001954803041686069,
+      "loss": 4.3975,
+      "step": 56700
+    },
+    {
+      "epoch": 0.4885502754820937,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.00019547224784873015,
+      "loss": 4.7219,
+      "step": 56750
+    },
+    {
+      "epoch": 0.4889807162534435,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.00019546418452139584,
+      "loss": 5.1161,
+      "step": 56800
+    },
+    {
+      "epoch": 0.4894111570247934,
+      "grad_norm": 1.828125,
+      "learning_rate": 0.00019545611418719577,
+      "loss": 5.2622,
+      "step": 56850
+    },
+    {
+      "epoch": 0.4898415977961432,
+      "grad_norm": 1.4375,
+      "learning_rate": 0.0001954480368467223,
+      "loss": 5.0309,
+      "step": 56900
+    },
+    {
+      "epoch": 0.49027203856749313,
+      "grad_norm": 1.953125,
+      "learning_rate": 0.0001954399525005683,
+      "loss": 5.0479,
+      "step": 56950
+    },
+    {
+      "epoch": 0.490702479338843,
+      "grad_norm": 2.625,
+      "learning_rate": 0.0001954318611493271,
+      "loss": 4.9872,
+      "step": 57000
+    },
+    {
+      "epoch": 0.490702479338843,
+      "eval_loss": 5.437960147857666,
+      "eval_runtime": 21.7735,
+      "eval_samples_per_second": 29.394,
+      "eval_steps_per_second": 14.697,
+      "eval_tts_loss": 6.678905297852088,
+      "step": 57000
+    },
+    {
+      "epoch": 0.4911329201101928,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.00019542376279359266,
+      "loss": 5.0722,
+      "step": 57050
+    },
+    {
+      "epoch": 0.4915633608815427,
+      "grad_norm": 1.8046875,
+      "learning_rate": 0.00019541565743395932,
+      "loss": 4.6575,
+      "step": 57100
+    },
+    {
+      "epoch": 0.4919938016528926,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.00019540754507102202,
+      "loss": 5.1254,
+      "step": 57150
+    },
+    {
+      "epoch": 0.49242424242424243,
+      "grad_norm": 1.2265625,
+      "learning_rate": 0.00019539942570537616,
+      "loss": 4.8798,
+      "step": 57200
+    },
+    {
+      "epoch": 0.4928546831955923,
+      "grad_norm": 3.265625,
+      "learning_rate": 0.00019539129933761776,
+      "loss": 5.0643,
+      "step": 57250
+    },
+    {
+      "epoch": 0.49328512396694213,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.00019538316596834322,
+      "loss": 4.7154,
+      "step": 57300
+    },
+    {
+      "epoch": 0.49371556473829203,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.00019537502559814946,
+      "loss": 4.5703,
+      "step": 57350
+    },
+    {
+      "epoch": 0.4941460055096419,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.00019536687822763407,
+      "loss": 5.1492,
+      "step": 57400
+    },
+    {
+      "epoch": 0.49457644628099173,
+      "grad_norm": 2.375,
+      "learning_rate": 0.00019535872385739503,
+      "loss": 4.771,
+      "step": 57450
+    },
+    {
+      "epoch": 0.4950068870523416,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.0001953505624880308,
+      "loss": 4.6707,
+      "step": 57500
+    },
+    {
+      "epoch": 0.4954373278236915,
+      "grad_norm": 0.9140625,
+      "learning_rate": 0.00019534239412014043,
+      "loss": 5.1132,
+      "step": 57550
+    },
+    {
+      "epoch": 0.49586776859504134,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.00019533421875432346,
+      "loss": 4.7795,
+      "step": 57600
+    },
+    {
+      "epoch": 0.4962982093663912,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00019532603639117998,
+      "loss": 5.437,
+      "step": 57650
+    },
+    {
+      "epoch": 0.49672865013774103,
+      "grad_norm": 2.140625,
+      "learning_rate": 0.00019531784703131054,
+      "loss": 4.8702,
+      "step": 57700
+    },
+    {
+      "epoch": 0.4971590909090909,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00019530965067531618,
+      "loss": 4.9499,
+      "step": 57750
+    },
+    {
+      "epoch": 0.4975895316804408,
+      "grad_norm": 1.2421875,
+      "learning_rate": 0.0001953014473237985,
+      "loss": 4.9733,
+      "step": 57800
+    },
+    {
+      "epoch": 0.49801997245179064,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.00019529323697735967,
+      "loss": 4.667,
+      "step": 57850
+    },
+    {
+      "epoch": 0.4984504132231405,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.00019528501963660225,
+      "loss": 4.7631,
+      "step": 57900
+    },
+    {
+      "epoch": 0.49888085399449034,
+      "grad_norm": 3.09375,
+      "learning_rate": 0.00019527679530212942,
+      "loss": 4.8465,
+      "step": 57950
+    },
+    {
+      "epoch": 0.49931129476584024,
+      "grad_norm": 1.9453125,
+      "learning_rate": 0.00019526856397454481,
+      "loss": 4.6451,
+      "step": 58000
+    },
+    {
+      "epoch": 0.4997417355371901,
+      "grad_norm": 3.515625,
+      "learning_rate": 0.00019526032565445257,
+      "loss": 4.7026,
+      "step": 58050
+    },
+    {
+      "epoch": 0.5001721763085399,
+      "grad_norm": 1.8359375,
+      "learning_rate": 0.00019525208034245738,
+      "loss": 4.8911,
+      "step": 58100
+    },
+    {
+      "epoch": 0.5006026170798898,
+      "grad_norm": 1.953125,
+      "learning_rate": 0.00019524382803916444,
+      "loss": 4.4846,
+      "step": 58150
+    },
+    {
+      "epoch": 0.5010330578512396,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.00019523556874517946,
+      "loss": 4.8332,
+      "step": 58200
+    },
+    {
+      "epoch": 0.5014634986225895,
+      "grad_norm": 1.921875,
+      "learning_rate": 0.00019522730246110861,
+      "loss": 4.3156,
+      "step": 58250
+    },
+    {
+      "epoch": 0.5018939393939394,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.00019521902918755865,
+      "loss": 5.1603,
+      "step": 58300
+    },
+    {
+      "epoch": 0.5023243801652892,
+      "grad_norm": 2.875,
+      "learning_rate": 0.00019521074892513683,
+      "loss": 5.1468,
+      "step": 58350
+    },
+    {
+      "epoch": 0.5027548209366391,
+      "grad_norm": 1.609375,
+      "learning_rate": 0.0001952024616744509,
+      "loss": 4.638,
+      "step": 58400
+    },
+    {
+      "epoch": 0.5031852617079889,
+      "grad_norm": 1.375,
+      "learning_rate": 0.00019519416743610907,
+      "loss": 4.8574,
+      "step": 58450
+    },
+    {
+      "epoch": 0.5036157024793388,
+      "grad_norm": 2.5,
+      "learning_rate": 0.0001951858662107202,
+      "loss": 4.7201,
+      "step": 58500
+    },
+    {
+      "epoch": 0.5040461432506887,
+      "grad_norm": 1.2890625,
+      "learning_rate": 0.00019517755799889356,
+      "loss": 4.8031,
+      "step": 58550
+    },
+    {
+      "epoch": 0.5044765840220385,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.00019516924280123891,
+      "loss": 5.17,
+      "step": 58600
+    },
+    {
+      "epoch": 0.5049070247933884,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00019516092061836664,
+      "loss": 5.0124,
+      "step": 58650
+    },
+    {
+      "epoch": 0.5053374655647382,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.0001951525914508875,
+      "loss": 4.7327,
+      "step": 58700
+    },
+    {
+      "epoch": 0.5057679063360881,
+      "grad_norm": 1.4453125,
+      "learning_rate": 0.0001951442552994129,
+      "loss": 4.8515,
+      "step": 58750
+    },
+    {
+      "epoch": 0.506198347107438,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.00019513591216455467,
+      "loss": 4.973,
+      "step": 58800
+    },
+    {
+      "epoch": 0.5066287878787878,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.00019512756204692516,
+      "loss": 4.7377,
+      "step": 58850
+    },
+    {
+      "epoch": 0.5070592286501377,
+      "grad_norm": 2.21875,
+      "learning_rate": 0.00019511920494713728,
+      "loss": 4.7408,
+      "step": 58900
+    },
+    {
+      "epoch": 0.5074896694214877,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.0001951108408658044,
+      "loss": 5.0065,
+      "step": 58950
+    },
+    {
+      "epoch": 0.5079201101928374,
+      "grad_norm": 1.390625,
+      "learning_rate": 0.00019510246980354045,
+      "loss": 4.4652,
+      "step": 59000
+    },
+    {
+      "epoch": 0.5083505509641874,
+      "grad_norm": 2.21875,
+      "learning_rate": 0.00019509409176095982,
+      "loss": 4.676,
+      "step": 59050
+    },
+    {
+      "epoch": 0.5087809917355371,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.0001950857067386775,
+      "loss": 4.7497,
+      "step": 59100
+    },
+    {
+      "epoch": 0.509211432506887,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.00019507731473730882,
+      "loss": 4.8371,
+      "step": 59150
+    },
+    {
+      "epoch": 0.509641873278237,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.00019506891575746984,
+      "loss": 4.4019,
+      "step": 59200
+    },
+    {
+      "epoch": 0.5100723140495868,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.00019506050979977698,
+      "loss": 4.7386,
+      "step": 59250
+    },
+    {
+      "epoch": 0.5105027548209367,
+      "grad_norm": 3.9375,
+      "learning_rate": 0.00019505209686484725,
+      "loss": 4.9919,
+      "step": 59300
+    },
+    {
+      "epoch": 0.5109331955922864,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.00019504367695329807,
+      "loss": 4.4062,
+      "step": 59350
+    },
+    {
+      "epoch": 0.5113636363636364,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.00019503525006574757,
+      "loss": 4.5798,
+      "step": 59400
+    },
+    {
+      "epoch": 0.5117940771349863,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.0001950268162028141,
+      "loss": 4.8806,
+      "step": 59450
+    },
+    {
+      "epoch": 0.512224517906336,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.00019501837536511685,
+      "loss": 4.6123,
+      "step": 59500
+    },
+    {
+      "epoch": 0.512654958677686,
+      "grad_norm": 1.296875,
+      "learning_rate": 0.00019500992755327524,
+      "loss": 4.9083,
+      "step": 59550
+    },
+    {
+      "epoch": 0.5130853994490359,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.0001950014727679094,
+      "loss": 5.2236,
+      "step": 59600
+    },
+    {
+      "epoch": 0.5135158402203857,
+      "grad_norm": 1.546875,
+      "learning_rate": 0.00019499301100963984,
+      "loss": 4.8901,
+      "step": 59650
+    },
+    {
+      "epoch": 0.5139462809917356,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.00019498454227908764,
+      "loss": 5.1904,
+      "step": 59700
+    },
+    {
+      "epoch": 0.5143767217630854,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.0001949760665768744,
+      "loss": 5.1363,
+      "step": 59750
+    },
+    {
+      "epoch": 0.5148071625344353,
+      "grad_norm": 3.28125,
+      "learning_rate": 0.00019496758390362228,
+      "loss": 4.4957,
+      "step": 59800
+    },
+    {
+      "epoch": 0.5152376033057852,
+      "grad_norm": 2.0,
+      "learning_rate": 0.00019495909425995377,
+      "loss": 4.9785,
+      "step": 59850
+    },
+    {
+      "epoch": 0.515668044077135,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00019495059764649208,
+      "loss": 4.8777,
+      "step": 59900
+    },
+    {
+      "epoch": 0.5160984848484849,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.0001949420940638608,
+      "loss": 4.6085,
+      "step": 59950
+    },
+    {
+      "epoch": 0.5165289256198347,
+      "grad_norm": 3.1875,
+      "learning_rate": 0.00019493358351268408,
+      "loss": 4.7669,
+      "step": 60000
+    },
+    {
+      "epoch": 0.5165289256198347,
+      "eval_loss": 5.428589820861816,
+      "eval_runtime": 22.0407,
+      "eval_samples_per_second": 29.037,
+      "eval_steps_per_second": 14.519,
+      "eval_tts_loss": 6.65994587151647,
+      "step": 60000
+    },
+    {
+      "epoch": 0.5169593663911846,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.0001949250659935866,
+      "loss": 4.7544,
+      "step": 60050
+    },
+    {
+      "epoch": 0.5173898071625345,
+      "grad_norm": 2.375,
+      "learning_rate": 0.00019491654150719353,
+      "loss": 5.1902,
+      "step": 60100
+    },
+    {
+      "epoch": 0.5178202479338843,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.0001949080100541305,
+      "loss": 4.8204,
+      "step": 60150
+    },
+    {
+      "epoch": 0.5182506887052342,
+      "grad_norm": 2.265625,
+      "learning_rate": 0.00019489947163502374,
+      "loss": 4.2814,
+      "step": 60200
+    },
+    {
+      "epoch": 0.5186811294765841,
+      "grad_norm": 3.65625,
+      "learning_rate": 0.00019489092625049996,
+      "loss": 4.8777,
+      "step": 60250
+    },
+    {
+      "epoch": 0.5191115702479339,
+      "grad_norm": 2.578125,
+      "learning_rate": 0.00019488237390118633,
+      "loss": 5.1736,
+      "step": 60300
+    },
+    {
+      "epoch": 0.5195420110192838,
+      "grad_norm": 3.25,
+      "learning_rate": 0.0001948738145877106,
+      "loss": 4.9632,
+      "step": 60350
+    },
+    {
+      "epoch": 0.5199724517906336,
+      "grad_norm": 1.2265625,
+      "learning_rate": 0.00019486524831070103,
+      "loss": 4.6269,
+      "step": 60400
+    },
+    {
+      "epoch": 0.5204028925619835,
+      "grad_norm": 1.21875,
+      "learning_rate": 0.00019485667507078633,
+      "loss": 5.2265,
+      "step": 60450
+    },
+    {
+      "epoch": 0.5208333333333334,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.0001948480948685958,
+      "loss": 5.2691,
+      "step": 60500
+    },
+    {
+      "epoch": 0.5212637741046832,
+      "grad_norm": 1.296875,
+      "learning_rate": 0.00019483950770475913,
+      "loss": 4.9249,
+      "step": 60550
+    },
+    {
+      "epoch": 0.5216942148760331,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.00019483091357990666,
+      "loss": 4.324,
+      "step": 60600
+    },
+    {
+      "epoch": 0.522124655647383,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.00019482231249466918,
+      "loss": 4.9368,
+      "step": 60650
+    },
+    {
+      "epoch": 0.5225550964187328,
+      "grad_norm": 3.515625,
+      "learning_rate": 0.00019481370444967796,
+      "loss": 4.5046,
+      "step": 60700
+    },
+    {
+      "epoch": 0.5229855371900827,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.00019480508944556485,
+      "loss": 4.9599,
+      "step": 60750
+    },
+    {
+      "epoch": 0.5234159779614325,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00019479646748296215,
+      "loss": 5.1871,
+      "step": 60800
+    },
+    {
+      "epoch": 0.5238464187327824,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.00019478783856250267,
+      "loss": 4.8656,
+      "step": 60850
+    },
+    {
+      "epoch": 0.5242768595041323,
+      "grad_norm": 1.6328125,
+      "learning_rate": 0.00019477920268481981,
+      "loss": 4.9236,
+      "step": 60900
+    },
+    {
+      "epoch": 0.5247073002754821,
+      "grad_norm": 1.9765625,
+      "learning_rate": 0.00019477055985054738,
+      "loss": 4.9892,
+      "step": 60950
+    },
+    {
+      "epoch": 0.525137741046832,
+      "grad_norm": 1.4375,
+      "learning_rate": 0.00019476191006031977,
+      "loss": 4.7424,
+      "step": 61000
+    },
+    {
+      "epoch": 0.5255681818181818,
+      "grad_norm": 1.96875,
+      "learning_rate": 0.00019475325331477184,
+      "loss": 5.054,
+      "step": 61050
+    },
+    {
+      "epoch": 0.5259986225895317,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.00019474458961453897,
+      "loss": 4.6265,
+      "step": 61100
+    },
+    {
+      "epoch": 0.5264290633608816,
+      "grad_norm": 1.953125,
+      "learning_rate": 0.00019473591896025712,
+      "loss": 5.1348,
+      "step": 61150
+    },
+    {
+      "epoch": 0.5268595041322314,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.00019472724135256262,
+      "loss": 4.3568,
+      "step": 61200
+    },
+    {
+      "epoch": 0.5272899449035813,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.0001947185567920924,
+      "loss": 4.5678,
+      "step": 61250
+    },
+    {
+      "epoch": 0.5277203856749312,
+      "grad_norm": 2.265625,
+      "learning_rate": 0.0001947098652794839,
+      "loss": 4.8169,
+      "step": 61300
+    },
+    {
+      "epoch": 0.528150826446281,
+      "grad_norm": 2.125,
+      "learning_rate": 0.0001947011668153751,
+      "loss": 4.7715,
+      "step": 61350
+    },
+    {
+      "epoch": 0.5285812672176309,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.00019469246140040437,
+      "loss": 5.2144,
+      "step": 61400
+    },
+    {
+      "epoch": 0.5290117079889807,
+      "grad_norm": 1.015625,
+      "learning_rate": 0.00019468374903521075,
+      "loss": 4.3464,
+      "step": 61450
+    },
+    {
+      "epoch": 0.5294421487603306,
+      "grad_norm": 1.84375,
+      "learning_rate": 0.00019467502972043364,
+      "loss": 4.5928,
+      "step": 61500
+    },
+    {
+      "epoch": 0.5298725895316805,
+      "grad_norm": 2.375,
+      "learning_rate": 0.00019466630345671304,
+      "loss": 5.1056,
+      "step": 61550
+    },
+    {
+      "epoch": 0.5303030303030303,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00019465757024468946,
+      "loss": 4.2716,
+      "step": 61600
+    },
+    {
+      "epoch": 0.5307334710743802,
+      "grad_norm": 3.140625,
+      "learning_rate": 0.00019464883008500386,
+      "loss": 4.6047,
+      "step": 61650
+    },
+    {
+      "epoch": 0.53116391184573,
+      "grad_norm": 1.7421875,
+      "learning_rate": 0.0001946400829782978,
+      "loss": 5.0897,
+      "step": 61700
+    },
+    {
+      "epoch": 0.5315943526170799,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.0001946313289252133,
+      "loss": 4.6044,
+      "step": 61750
+    },
+    {
+      "epoch": 0.5320247933884298,
+      "grad_norm": 2.0,
+      "learning_rate": 0.00019462256792639285,
+      "loss": 5.1399,
+      "step": 61800
+    },
+    {
+      "epoch": 0.5324552341597796,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.00019461379998247946,
+      "loss": 5.0757,
+      "step": 61850
+    },
+    {
+      "epoch": 0.5328856749311295,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.00019460502509411676,
+      "loss": 4.7783,
+      "step": 61900
+    },
+    {
+      "epoch": 0.5333161157024794,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.00019459624326194873,
+      "loss": 4.5562,
+      "step": 61950
+    },
+    {
+      "epoch": 0.5337465564738292,
+      "grad_norm": 1.734375,
+      "learning_rate": 0.00019458745448662,
+      "loss": 4.5624,
+      "step": 62000
+    },
+    {
+      "epoch": 0.5341769972451791,
+      "grad_norm": 1.2421875,
+      "learning_rate": 0.00019457865876877558,
+      "loss": 4.9627,
+      "step": 62050
+    },
+    {
+      "epoch": 0.5346074380165289,
+      "grad_norm": 1.78125,
+      "learning_rate": 0.00019456985610906115,
+      "loss": 4.5935,
+      "step": 62100
+    },
+    {
+      "epoch": 0.5350378787878788,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.00019456104650812273,
+      "loss": 4.8025,
+      "step": 62150
+    },
+    {
+      "epoch": 0.5354683195592287,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.00019455222996660695,
+      "loss": 5.0809,
+      "step": 62200
+    },
+    {
+      "epoch": 0.5358987603305785,
+      "grad_norm": 1.8984375,
+      "learning_rate": 0.00019454340648516092,
+      "loss": 4.8076,
+      "step": 62250
+    },
+    {
+      "epoch": 0.5363292011019284,
+      "grad_norm": 1.3125,
+      "learning_rate": 0.00019453457606443227,
+      "loss": 4.9953,
+      "step": 62300
+    },
+    {
+      "epoch": 0.5367596418732782,
+      "grad_norm": 1.8046875,
+      "learning_rate": 0.00019452573870506915,
+      "loss": 4.9248,
+      "step": 62350
+    },
+    {
+      "epoch": 0.5371900826446281,
+      "grad_norm": 2.140625,
+      "learning_rate": 0.00019451689440772016,
+      "loss": 4.9168,
+      "step": 62400
+    },
+    {
+      "epoch": 0.537620523415978,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.00019450804317303454,
+      "loss": 4.3827,
+      "step": 62450
+    },
+    {
+      "epoch": 0.5380509641873278,
+      "grad_norm": 3.5625,
+      "learning_rate": 0.00019449918500166183,
+      "loss": 5.1639,
+      "step": 62500
+    },
+    {
+      "epoch": 0.5384814049586777,
+      "grad_norm": 3.1875,
+      "learning_rate": 0.00019449031989425227,
+      "loss": 4.911,
+      "step": 62550
+    },
+    {
+      "epoch": 0.5389118457300276,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.00019448144785145653,
+      "loss": 5.0495,
+      "step": 62600
+    },
+    {
+      "epoch": 0.5393422865013774,
+      "grad_norm": 3.125,
+      "learning_rate": 0.0001944725688739258,
+      "loss": 4.989,
+      "step": 62650
+    },
+    {
+      "epoch": 0.5397727272727273,
+      "grad_norm": 4.90625,
+      "learning_rate": 0.00019446368296231178,
+      "loss": 4.8458,
+      "step": 62700
+    },
+    {
+      "epoch": 0.5402031680440771,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.0001944547901172667,
+      "loss": 5.0329,
+      "step": 62750
+    },
+    {
+      "epoch": 0.540633608815427,
+      "grad_norm": 3.40625,
+      "learning_rate": 0.00019444589033944322,
+      "loss": 4.6332,
+      "step": 62800
+    },
+    {
+      "epoch": 0.5410640495867769,
+      "grad_norm": 2.25,
+      "learning_rate": 0.0001944369836294946,
+      "loss": 5.0479,
+      "step": 62850
+    },
+    {
+      "epoch": 0.5414944903581267,
+      "grad_norm": 1.5234375,
+      "learning_rate": 0.0001944280699880746,
+      "loss": 4.6948,
+      "step": 62900
+    },
+    {
+      "epoch": 0.5419249311294766,
+      "grad_norm": 1.578125,
+      "learning_rate": 0.00019441914941583738,
+      "loss": 4.8837,
+      "step": 62950
+    },
+    {
+      "epoch": 0.5423553719008265,
+      "grad_norm": 1.6015625,
+      "learning_rate": 0.00019441022191343777,
+      "loss": 4.8535,
+      "step": 63000
+    },
+    {
+      "epoch": 0.5423553719008265,
+      "eval_loss": 5.41285514831543,
+      "eval_runtime": 21.855,
+      "eval_samples_per_second": 29.284,
+      "eval_steps_per_second": 14.642,
+      "eval_tts_loss": 6.7312361713261595,
+      "step": 63000
+    },
+    {
+      "epoch": 0.5427858126721763,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.00019440128748153098,
+      "loss": 4.6159,
+      "step": 63050
+    },
+    {
+      "epoch": 0.5432162534435262,
+      "grad_norm": 2.21875,
+      "learning_rate": 0.00019439234612077282,
+      "loss": 5.0116,
+      "step": 63100
+    },
+    {
+      "epoch": 0.543646694214876,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.00019438339783181953,
+      "loss": 5.0751,
+      "step": 63150
+    },
+    {
+      "epoch": 0.5440771349862259,
+      "grad_norm": 3.0625,
+      "learning_rate": 0.0001943744426153279,
+      "loss": 4.9378,
+      "step": 63200
+    },
+    {
+      "epoch": 0.5445075757575758,
+      "grad_norm": 1.28125,
+      "learning_rate": 0.0001943654804719553,
+      "loss": 4.4809,
+      "step": 63250
+    },
+    {
+      "epoch": 0.5449380165289256,
+      "grad_norm": 1.3671875,
+      "learning_rate": 0.00019435651140235936,
+      "loss": 4.7089,
+      "step": 63300
+    },
+    {
+      "epoch": 0.5453684573002755,
+      "grad_norm": 0.8046875,
+      "learning_rate": 0.0001943475354071986,
+      "loss": 4.6986,
+      "step": 63350
+    },
+    {
+      "epoch": 0.5457988980716253,
+      "grad_norm": 3.671875,
+      "learning_rate": 0.00019433855248713166,
+      "loss": 4.7862,
+      "step": 63400
+    },
+    {
+      "epoch": 0.5462293388429752,
+      "grad_norm": 1.2890625,
+      "learning_rate": 0.00019432956264281797,
+      "loss": 4.6603,
+      "step": 63450
+    },
+    {
+      "epoch": 0.5466597796143251,
+      "grad_norm": 1.875,
+      "learning_rate": 0.00019432056587491734,
+      "loss": 5.0196,
+      "step": 63500
+    },
+    {
+      "epoch": 0.5470902203856749,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.00019431156218409012,
+      "loss": 4.8541,
+      "step": 63550
+    },
+    {
+      "epoch": 0.5475206611570248,
+      "grad_norm": 3.03125,
+      "learning_rate": 0.00019430255157099714,
+      "loss": 5.0201,
+      "step": 63600
+    },
+    {
+      "epoch": 0.5479511019283747,
+      "grad_norm": 3.3125,
+      "learning_rate": 0.00019429353403629978,
+      "loss": 4.8323,
+      "step": 63650
+    },
+    {
+      "epoch": 0.5483815426997245,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.0001942845095806599,
+      "loss": 4.8278,
+      "step": 63700
+    },
+    {
+      "epoch": 0.5488119834710744,
+      "grad_norm": 0.6640625,
+      "learning_rate": 0.00019427547820473982,
+      "loss": 4.5837,
+      "step": 63750
+    },
+    {
+      "epoch": 0.5492424242424242,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.00019426643990920256,
+      "loss": 4.949,
+      "step": 63800
+    },
+    {
+      "epoch": 0.5496728650137741,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.0001942573946947114,
+      "loss": 4.9298,
+      "step": 63850
+    },
+    {
+      "epoch": 0.550103305785124,
+      "grad_norm": 2.140625,
+      "learning_rate": 0.00019424834256193024,
+      "loss": 5.0459,
+      "step": 63900
+    },
+    {
+      "epoch": 0.5505337465564738,
+      "grad_norm": 2.421875,
+      "learning_rate": 0.00019423928351152355,
+      "loss": 5.2147,
+      "step": 63950
+    },
+    {
+      "epoch": 0.5509641873278237,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.0001942302175441562,
+      "loss": 4.8054,
+      "step": 64000
+    },
+    {
+      "epoch": 0.5513946280991735,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00019422114466049364,
+      "loss": 4.829,
+      "step": 64050
+    },
+    {
+      "epoch": 0.5518250688705234,
+      "grad_norm": 1.671875,
+      "learning_rate": 0.00019421206486120175,
+      "loss": 4.9328,
+      "step": 64100
+    },
+    {
+      "epoch": 0.5522555096418733,
+      "grad_norm": 1.5,
+      "learning_rate": 0.00019420297814694705,
+      "loss": 4.6441,
+      "step": 64150
+    },
+    {
+      "epoch": 0.5526859504132231,
+      "grad_norm": 1.9765625,
+      "learning_rate": 0.0001941938845183964,
+      "loss": 4.9526,
+      "step": 64200
+    },
+    {
+      "epoch": 0.553116391184573,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.0001941847839762173,
+      "loss": 5.1159,
+      "step": 64250
+    },
+    {
+      "epoch": 0.5535468319559229,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.0001941756765210777,
+      "loss": 4.7393,
+      "step": 64300
+    },
+    {
+      "epoch": 0.5539772727272727,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00019416656215364605,
+      "loss": 4.9843,
+      "step": 64350
+    },
+    {
+      "epoch": 0.5544077134986226,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.00019415744087459137,
+      "loss": 4.694,
+      "step": 64400
+    },
+    {
+      "epoch": 0.5548381542699724,
+      "grad_norm": 1.90625,
+      "learning_rate": 0.0001941483126845831,
+      "loss": 4.915,
+      "step": 64450
+    },
+    {
+      "epoch": 0.5552685950413223,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.00019413917758429126,
+      "loss": 4.4675,
+      "step": 64500
+    },
+    {
+      "epoch": 0.5556990358126722,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.0001941300355743863,
+      "loss": 4.8105,
+      "step": 64550
+    },
+    {
+      "epoch": 0.556129476584022,
+      "grad_norm": 1.9765625,
+      "learning_rate": 0.0001941208866555393,
+      "loss": 4.9246,
+      "step": 64600
+    },
+    {
+      "epoch": 0.5565599173553719,
+      "grad_norm": 1.2578125,
+      "learning_rate": 0.0001941117308284217,
+      "loss": 5.0171,
+      "step": 64650
+    },
+    {
+      "epoch": 0.5569903581267218,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.00019410256809370552,
+      "loss": 4.8394,
+      "step": 64700
+    },
+    {
+      "epoch": 0.5574207988980716,
+      "grad_norm": 3.015625,
+      "learning_rate": 0.00019409339845206334,
+      "loss": 4.8064,
+      "step": 64750
+    },
+    {
+      "epoch": 0.5578512396694215,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.00019408422190416818,
+      "loss": 4.8562,
+      "step": 64800
+    },
+    {
+      "epoch": 0.5582816804407713,
+      "grad_norm": 1.625,
+      "learning_rate": 0.00019407503845069352,
+      "loss": 5.2035,
+      "step": 64850
+    },
+    {
+      "epoch": 0.5587121212121212,
+      "grad_norm": 1.9765625,
+      "learning_rate": 0.00019406584809231346,
+      "loss": 4.5825,
+      "step": 64900
+    },
+    {
+      "epoch": 0.5591425619834711,
+      "grad_norm": 4.15625,
+      "learning_rate": 0.00019405665082970254,
+      "loss": 5.1706,
+      "step": 64950
+    },
+    {
+      "epoch": 0.5595730027548209,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.0001940474466635358,
+      "loss": 4.7249,
+      "step": 65000
+    },
+    {
+      "epoch": 0.5600034435261708,
+      "grad_norm": 4.1875,
+      "learning_rate": 0.00019403823559448882,
+      "loss": 4.4334,
+      "step": 65050
+    },
+    {
+      "epoch": 0.5604338842975206,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.00019402901762323772,
+      "loss": 5.2081,
+      "step": 65100
+    },
+    {
+      "epoch": 0.5608643250688705,
+      "grad_norm": 1.453125,
+      "learning_rate": 0.000194019792750459,
+      "loss": 4.5441,
+      "step": 65150
+    },
+    {
+      "epoch": 0.5612947658402204,
+      "grad_norm": 1.7265625,
+      "learning_rate": 0.00019401056097682976,
+      "loss": 4.7629,
+      "step": 65200
+    },
+    {
+      "epoch": 0.5617252066115702,
+      "grad_norm": 1.3203125,
+      "learning_rate": 0.00019400132230302765,
+      "loss": 4.8572,
+      "step": 65250
+    },
+    {
+      "epoch": 0.5621556473829201,
+      "grad_norm": 1.2265625,
+      "learning_rate": 0.00019399207672973073,
+      "loss": 4.9189,
+      "step": 65300
+    },
+    {
+      "epoch": 0.56258608815427,
+      "grad_norm": 2.21875,
+      "learning_rate": 0.0001939828242576176,
+      "loss": 4.6595,
+      "step": 65350
+    },
+    {
+      "epoch": 0.5630165289256198,
+      "grad_norm": 1.8359375,
+      "learning_rate": 0.00019397356488736739,
+      "loss": 4.8636,
+      "step": 65400
+    },
+    {
+      "epoch": 0.5634469696969697,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.0001939642986196597,
+      "loss": 4.6871,
+      "step": 65450
+    },
+    {
+      "epoch": 0.5638774104683195,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.00019395502545517467,
+      "loss": 4.6867,
+      "step": 65500
+    },
+    {
+      "epoch": 0.5643078512396694,
+      "grad_norm": 0.6328125,
+      "learning_rate": 0.0001939457453945929,
+      "loss": 4.7733,
+      "step": 65550
+    },
+    {
+      "epoch": 0.5647382920110193,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.00019393645843859556,
+      "loss": 4.6338,
+      "step": 65600
+    },
+    {
+      "epoch": 0.5651687327823691,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.00019392716458786427,
+      "loss": 4.6369,
+      "step": 65650
+    },
+    {
+      "epoch": 0.565599173553719,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00019391786384308118,
+      "loss": 4.8525,
+      "step": 65700
+    },
+    {
+      "epoch": 0.5660296143250688,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.00019390855620492897,
+      "loss": 4.3963,
+      "step": 65750
+    },
+    {
+      "epoch": 0.5664600550964187,
+      "grad_norm": 1.4375,
+      "learning_rate": 0.0001938992416740908,
+      "loss": 4.6911,
+      "step": 65800
+    },
+    {
+      "epoch": 0.5668904958677686,
+      "grad_norm": 1.3359375,
+      "learning_rate": 0.0001938899202512503,
+      "loss": 4.6394,
+      "step": 65850
+    },
+    {
+      "epoch": 0.5673209366391184,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.00019388059193709166,
+      "loss": 4.9945,
+      "step": 65900
+    },
+    {
+      "epoch": 0.5677513774104683,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.00019387125673229956,
+      "loss": 4.6664,
+      "step": 65950
+    },
+    {
+      "epoch": 0.5681818181818182,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.0001938619146375592,
+      "loss": 4.8697,
+      "step": 66000
+    },
+    {
+      "epoch": 0.5681818181818182,
+      "eval_loss": 5.408427715301514,
+      "eval_runtime": 22.0911,
+      "eval_samples_per_second": 28.971,
+      "eval_steps_per_second": 14.485,
+      "eval_tts_loss": 6.665151977341963,
+      "step": 66000
+    },
+    {
+      "epoch": 0.568612258953168,
+      "grad_norm": 3.359375,
+      "learning_rate": 0.00019385256565355622,
+      "loss": 5.2537,
+      "step": 66050
+    },
+    {
+      "epoch": 0.5690426997245179,
+      "grad_norm": 2.5,
+      "learning_rate": 0.0001938432097809769,
+      "loss": 5.1396,
+      "step": 66100
+    },
+    {
+      "epoch": 0.5694731404958677,
+      "grad_norm": 2.265625,
+      "learning_rate": 0.00019383384702050788,
+      "loss": 4.7616,
+      "step": 66150
+    },
+    {
+      "epoch": 0.5699035812672176,
+      "grad_norm": 2.125,
+      "learning_rate": 0.00019382447737283632,
+      "loss": 4.814,
+      "step": 66200
+    },
+    {
+      "epoch": 0.5703340220385675,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00019381510083865005,
+      "loss": 5.0777,
+      "step": 66250
+    },
+    {
+      "epoch": 0.5707644628099173,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.0001938057174186372,
+      "loss": 4.7864,
+      "step": 66300
+    },
+    {
+      "epoch": 0.5711949035812672,
+      "grad_norm": 2.875,
+      "learning_rate": 0.00019379632711348657,
+      "loss": 4.9105,
+      "step": 66350
+    },
+    {
+      "epoch": 0.571625344352617,
+      "grad_norm": 2.140625,
+      "learning_rate": 0.0001937869299238873,
+      "loss": 4.7395,
+      "step": 66400
+    },
+    {
+      "epoch": 0.5720557851239669,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.00019377752585052912,
+      "loss": 4.8076,
+      "step": 66450
+    },
+    {
+      "epoch": 0.5724862258953168,
+      "grad_norm": 2.625,
+      "learning_rate": 0.00019376811489410235,
+      "loss": 4.6751,
+      "step": 66500
+    },
+    {
+      "epoch": 0.5729166666666666,
+      "grad_norm": 2.125,
+      "learning_rate": 0.00019375869705529772,
+      "loss": 5.0308,
+      "step": 66550
+    },
+    {
+      "epoch": 0.5733471074380165,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.00019374927233480643,
+      "loss": 4.6561,
+      "step": 66600
+    },
+    {
+      "epoch": 0.5737775482093664,
+      "grad_norm": 2.421875,
+      "learning_rate": 0.00019373984073332024,
+      "loss": 4.8788,
+      "step": 66650
+    },
+    {
+      "epoch": 0.5742079889807162,
+      "grad_norm": 4.0,
+      "learning_rate": 0.00019373040225153146,
+      "loss": 4.4974,
+      "step": 66700
+    },
+    {
+      "epoch": 0.5746384297520661,
+      "grad_norm": 2.265625,
+      "learning_rate": 0.00019372095689013279,
+      "loss": 4.7771,
+      "step": 66750
+    },
+    {
+      "epoch": 0.5750688705234159,
+      "grad_norm": 2.9375,
+      "learning_rate": 0.00019371150464981754,
+      "loss": 4.8839,
+      "step": 66800
+    },
+    {
+      "epoch": 0.5754993112947658,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00019370204553127948,
+      "loss": 4.8058,
+      "step": 66850
+    },
+    {
+      "epoch": 0.5759297520661157,
+      "grad_norm": 2.421875,
+      "learning_rate": 0.00019369257953521287,
+      "loss": 4.8367,
+      "step": 66900
+    },
+    {
+      "epoch": 0.5763601928374655,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00019368310666231252,
+      "loss": 4.6082,
+      "step": 66950
+    },
+    {
+      "epoch": 0.5767906336088154,
+      "grad_norm": 1.421875,
+      "learning_rate": 0.0001936736269132737,
+      "loss": 5.2713,
+      "step": 67000
+    },
+    {
+      "epoch": 0.5772210743801653,
+      "grad_norm": 4.375,
+      "learning_rate": 0.00019366414028879221,
+      "loss": 5.1557,
+      "step": 67050
+    },
+    {
+      "epoch": 0.5776515151515151,
+      "grad_norm": 1.453125,
+      "learning_rate": 0.00019365464678956434,
+      "loss": 4.841,
+      "step": 67100
+    },
+    {
+      "epoch": 0.578081955922865,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00019364514641628694,
+      "loss": 4.9736,
+      "step": 67150
+    },
+    {
+      "epoch": 0.5785123966942148,
+      "grad_norm": 4.03125,
+      "learning_rate": 0.00019363563916965722,
+      "loss": 4.921,
+      "step": 67200
+    },
+    {
+      "epoch": 0.5789428374655647,
+      "grad_norm": 3.078125,
+      "learning_rate": 0.00019362612505037307,
+      "loss": 5.083,
+      "step": 67250
+    },
+    {
+      "epoch": 0.5793732782369146,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.0001936166040591328,
+      "loss": 5.2489,
+      "step": 67300
+    },
+    {
+      "epoch": 0.5798037190082644,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.00019360707619663517,
+      "loss": 4.8065,
+      "step": 67350
+    },
+    {
+      "epoch": 0.5802341597796143,
+      "grad_norm": 3.0625,
+      "learning_rate": 0.0001935975414635796,
+      "loss": 4.9339,
+      "step": 67400
+    },
+    {
+      "epoch": 0.5806646005509641,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.00019358799986066584,
+      "loss": 4.7002,
+      "step": 67450
+    },
+    {
+      "epoch": 0.581095041322314,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.00019357845138859422,
+      "loss": 4.8721,
+      "step": 67500
+    },
+    {
+      "epoch": 0.581525482093664,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.00019356889604806566,
+      "loss": 4.8496,
+      "step": 67550
+    },
+    {
+      "epoch": 0.5819559228650137,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.00019355933383978142,
+      "loss": 4.9558,
+      "step": 67600
+    },
+    {
+      "epoch": 0.5823863636363636,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.00019354976476444337,
+      "loss": 5.0111,
+      "step": 67650
+    },
+    {
+      "epoch": 0.5828168044077136,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.0001935401888227539,
+      "loss": 4.5638,
+      "step": 67700
+    },
+    {
+      "epoch": 0.5832472451790633,
+      "grad_norm": 0.765625,
+      "learning_rate": 0.0001935306060154158,
+      "loss": 4.6319,
+      "step": 67750
+    },
+    {
+      "epoch": 0.5836776859504132,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00019352101634313243,
+      "loss": 4.7251,
+      "step": 67800
+    },
+    {
+      "epoch": 0.584108126721763,
+      "grad_norm": 3.5625,
+      "learning_rate": 0.0001935114198066077,
+      "loss": 4.4962,
+      "step": 67850
+    },
+    {
+      "epoch": 0.584538567493113,
+      "grad_norm": 3.796875,
+      "learning_rate": 0.00019350181640654597,
+      "loss": 4.8778,
+      "step": 67900
+    },
+    {
+      "epoch": 0.5849690082644629,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00019349220614365206,
+      "loss": 4.6719,
+      "step": 67950
+    },
+    {
+      "epoch": 0.5853994490358126,
+      "grad_norm": 3.296875,
+      "learning_rate": 0.0001934825890186314,
+      "loss": 4.5005,
+      "step": 68000
+    },
+    {
+      "epoch": 0.5858298898071626,
+      "grad_norm": 1.515625,
+      "learning_rate": 0.0001934729650321898,
+      "loss": 5.1956,
+      "step": 68050
+    },
+    {
+      "epoch": 0.5862603305785123,
+      "grad_norm": 1.7421875,
+      "learning_rate": 0.00019346333418503373,
+      "loss": 4.8757,
+      "step": 68100
+    },
+    {
+      "epoch": 0.5866907713498623,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00019345369647786998,
+      "loss": 4.4,
+      "step": 68150
+    },
+    {
+      "epoch": 0.5871212121212122,
+      "grad_norm": 3.375,
+      "learning_rate": 0.00019344405191140603,
+      "loss": 4.8859,
+      "step": 68200
+    },
+    {
+      "epoch": 0.587551652892562,
+      "grad_norm": 2.875,
+      "learning_rate": 0.00019343440048634965,
+      "loss": 4.8281,
+      "step": 68250
+    },
+    {
+      "epoch": 0.5879820936639119,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00019342474220340936,
+      "loss": 5.076,
+      "step": 68300
+    },
+    {
+      "epoch": 0.5884125344352618,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.00019341507706329396,
+      "loss": 4.6841,
+      "step": 68350
+    },
+    {
+      "epoch": 0.5888429752066116,
+      "grad_norm": 1.625,
+      "learning_rate": 0.00019340540506671293,
+      "loss": 4.8212,
+      "step": 68400
+    },
+    {
+      "epoch": 0.5892734159779615,
+      "grad_norm": 1.90625,
+      "learning_rate": 0.00019339572621437608,
+      "loss": 4.6378,
+      "step": 68450
+    },
+    {
+      "epoch": 0.5897038567493113,
+      "grad_norm": 4.03125,
+      "learning_rate": 0.00019338604050699393,
+      "loss": 4.9896,
+      "step": 68500
+    },
+    {
+      "epoch": 0.5901342975206612,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.0001933763479452773,
+      "loss": 5.0799,
+      "step": 68550
+    },
+    {
+      "epoch": 0.5905647382920111,
+      "grad_norm": 3.5625,
+      "learning_rate": 0.00019336664852993767,
+      "loss": 4.815,
+      "step": 68600
+    },
+    {
+      "epoch": 0.5909951790633609,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.0001933569422616869,
+      "loss": 4.7069,
+      "step": 68650
+    },
+    {
+      "epoch": 0.5914256198347108,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.00019334722914123741,
+      "loss": 4.6069,
+      "step": 68700
+    },
+    {
+      "epoch": 0.5918560606060606,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.00019333750916930215,
+      "loss": 4.9758,
+      "step": 68750
+    },
+    {
+      "epoch": 0.5922865013774105,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.00019332778234659457,
+      "loss": 4.4766,
+      "step": 68800
+    },
+    {
+      "epoch": 0.5927169421487604,
+      "grad_norm": 1.8828125,
+      "learning_rate": 0.00019331804867382855,
+      "loss": 4.9831,
+      "step": 68850
+    },
+    {
+      "epoch": 0.5931473829201102,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.00019330830815171852,
+      "loss": 4.6332,
+      "step": 68900
+    },
+    {
+      "epoch": 0.5935778236914601,
+      "grad_norm": 1.9296875,
+      "learning_rate": 0.00019329856078097946,
+      "loss": 4.5474,
+      "step": 68950
+    },
+    {
+      "epoch": 0.59400826446281,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.00019328880656232676,
+      "loss": 4.7471,
+      "step": 69000
+    },
+    {
+      "epoch": 0.59400826446281,
+      "eval_loss": 5.3918561935424805,
+      "eval_runtime": 21.9308,
+      "eval_samples_per_second": 29.183,
+      "eval_steps_per_second": 14.591,
+      "eval_tts_loss": 6.719209646573399,
+      "step": 69000
+    },
+    {
+      "epoch": 0.5944387052341598,
+      "grad_norm": 1.2421875,
+      "learning_rate": 0.00019327904549647635,
+      "loss": 4.8058,
+      "step": 69050
+    },
+    {
+      "epoch": 0.5948691460055097,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.0001932692775841447,
+      "loss": 4.8438,
+      "step": 69100
+    },
+    {
+      "epoch": 0.5952995867768595,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.0001932595028260488,
+      "loss": 4.9355,
+      "step": 69150
+    },
+    {
+      "epoch": 0.5957300275482094,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.000193249721222906,
+      "loss": 4.7622,
+      "step": 69200
+    },
+    {
+      "epoch": 0.5961604683195593,
+      "grad_norm": 0.62109375,
+      "learning_rate": 0.0001932399327754343,
+      "loss": 4.7971,
+      "step": 69250
+    },
+    {
+      "epoch": 0.5965909090909091,
+      "grad_norm": 1.5390625,
+      "learning_rate": 0.00019323013748435215,
+      "loss": 4.6563,
+      "step": 69300
+    },
+    {
+      "epoch": 0.597021349862259,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.00019322033535037849,
+      "loss": 4.7125,
+      "step": 69350
+    },
+    {
+      "epoch": 0.5974517906336089,
+      "grad_norm": 2.296875,
+      "learning_rate": 0.0001932105263742328,
+      "loss": 4.9237,
+      "step": 69400
+    },
+    {
+      "epoch": 0.5978822314049587,
+      "grad_norm": 1.390625,
+      "learning_rate": 0.000193200710556635,
+      "loss": 4.6762,
+      "step": 69450
+    },
+    {
+      "epoch": 0.5983126721763086,
+      "grad_norm": 1.8828125,
+      "learning_rate": 0.0001931908878983056,
+      "loss": 5.0598,
+      "step": 69500
+    },
+    {
+      "epoch": 0.5987431129476584,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.0001931810583999655,
+      "loss": 4.9928,
+      "step": 69550
+    },
+    {
+      "epoch": 0.5991735537190083,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00019317122206233625,
+      "loss": 4.9787,
+      "step": 69600
+    },
+    {
+      "epoch": 0.5996039944903582,
+      "grad_norm": 3.0625,
+      "learning_rate": 0.0001931613788861397,
+      "loss": 4.8721,
+      "step": 69650
+    },
+    {
+      "epoch": 0.600034435261708,
+      "grad_norm": 3.296875,
+      "learning_rate": 0.00019315152887209843,
+      "loss": 4.7348,
+      "step": 69700
+    },
+    {
+      "epoch": 0.6004648760330579,
+      "grad_norm": 3.703125,
+      "learning_rate": 0.00019314167202093535,
+      "loss": 5.0721,
+      "step": 69750
+    },
+    {
+      "epoch": 0.6008953168044077,
+      "grad_norm": 1.421875,
+      "learning_rate": 0.00019313180833337393,
+      "loss": 4.7295,
+      "step": 69800
+    },
+    {
+      "epoch": 0.6013257575757576,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00019312193781013815,
+      "loss": 4.2876,
+      "step": 69850
+    },
+    {
+      "epoch": 0.6017561983471075,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.0001931120604519525,
+      "loss": 5.0536,
+      "step": 69900
+    },
+    {
+      "epoch": 0.6021866391184573,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.00019310217625954196,
+      "loss": 4.6664,
+      "step": 69950
+    },
+    {
+      "epoch": 0.6026170798898072,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.000193092285233632,
+      "loss": 4.6545,
+      "step": 70000
+    },
+    {
+      "epoch": 0.6030475206611571,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.0001930823873749486,
+      "loss": 5.2324,
+      "step": 70050
+    },
+    {
+      "epoch": 0.6034779614325069,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.0001930724826842182,
+      "loss": 4.2899,
+      "step": 70100
+    },
+    {
+      "epoch": 0.6039084022038568,
+      "grad_norm": 2.234375,
+      "learning_rate": 0.00019306257116216786,
+      "loss": 5.1086,
+      "step": 70150
+    },
+    {
+      "epoch": 0.6043388429752066,
+      "grad_norm": 3.625,
+      "learning_rate": 0.00019305265280952503,
+      "loss": 4.5098,
+      "step": 70200
+    },
+    {
+      "epoch": 0.6047692837465565,
+      "grad_norm": 1.4609375,
+      "learning_rate": 0.00019304272762701765,
+      "loss": 4.8431,
+      "step": 70250
+    },
+    {
+      "epoch": 0.6051997245179064,
+      "grad_norm": 1.84375,
+      "learning_rate": 0.00019303279561537427,
+      "loss": 4.9329,
+      "step": 70300
+    },
+    {
+      "epoch": 0.6056301652892562,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.0001930228567753239,
+      "loss": 4.8364,
+      "step": 70350
+    },
+    {
+      "epoch": 0.6060606060606061,
+      "grad_norm": 3.078125,
+      "learning_rate": 0.00019301291110759592,
+      "loss": 4.9719,
+      "step": 70400
+    },
+    {
+      "epoch": 0.6064910468319559,
+      "grad_norm": 1.578125,
+      "learning_rate": 0.0001930029586129204,
+      "loss": 4.6531,
+      "step": 70450
+    },
+    {
+      "epoch": 0.6069214876033058,
+      "grad_norm": 2.0,
+      "learning_rate": 0.00019299299929202785,
+      "loss": 4.7328,
+      "step": 70500
+    },
+    {
+      "epoch": 0.6073519283746557,
+      "grad_norm": 4.0,
+      "learning_rate": 0.0001929830331456492,
+      "loss": 4.7255,
+      "step": 70550
+    },
+    {
+      "epoch": 0.6077823691460055,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.00019297306017451602,
+      "loss": 4.7601,
+      "step": 70600
+    },
+    {
+      "epoch": 0.6082128099173554,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.00019296308037936025,
+      "loss": 4.934,
+      "step": 70650
+    },
+    {
+      "epoch": 0.6086432506887053,
+      "grad_norm": 3.453125,
+      "learning_rate": 0.00019295309376091434,
+      "loss": 4.3747,
+      "step": 70700
+    },
+    {
+      "epoch": 0.6090736914600551,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.0001929431003199114,
+      "loss": 4.5453,
+      "step": 70750
+    },
+    {
+      "epoch": 0.609504132231405,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.00019293310005708483,
+      "loss": 4.6482,
+      "step": 70800
+    },
+    {
+      "epoch": 0.6099345730027548,
+      "grad_norm": 2.875,
+      "learning_rate": 0.0001929230929731687,
+      "loss": 5.0405,
+      "step": 70850
+    },
+    {
+      "epoch": 0.6103650137741047,
+      "grad_norm": 2.125,
+      "learning_rate": 0.00019291307906889747,
+      "loss": 4.8022,
+      "step": 70900
+    },
+    {
+      "epoch": 0.6107954545454546,
+      "grad_norm": 3.265625,
+      "learning_rate": 0.0001929030583450061,
+      "loss": 4.6024,
+      "step": 70950
+    },
+    {
+      "epoch": 0.6112258953168044,
+      "grad_norm": 2.25,
+      "learning_rate": 0.00019289303080223018,
+      "loss": 4.329,
+      "step": 71000
+    },
+    {
+      "epoch": 0.6116563360881543,
+      "grad_norm": 1.75,
+      "learning_rate": 0.00019288299644130565,
+      "loss": 4.7542,
+      "step": 71050
+    },
+    {
+      "epoch": 0.6120867768595041,
+      "grad_norm": 1.953125,
+      "learning_rate": 0.00019287295526296903,
+      "loss": 4.7364,
+      "step": 71100
+    },
+    {
+      "epoch": 0.612517217630854,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.0001928629072679573,
+      "loss": 5.1028,
+      "step": 71150
+    },
+    {
+      "epoch": 0.6129476584022039,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00019285285245700798,
+      "loss": 4.7884,
+      "step": 71200
+    },
+    {
+      "epoch": 0.6133780991735537,
+      "grad_norm": 1.8203125,
+      "learning_rate": 0.00019284279083085909,
+      "loss": 5.159,
+      "step": 71250
+    },
+    {
+      "epoch": 0.6138085399449036,
+      "grad_norm": 1.5859375,
+      "learning_rate": 0.0001928327223902491,
+      "loss": 4.3514,
+      "step": 71300
+    },
+    {
+      "epoch": 0.6142389807162535,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.000192822647135917,
+      "loss": 5.1681,
+      "step": 71350
+    },
+    {
+      "epoch": 0.6146694214876033,
+      "grad_norm": 1.8515625,
+      "learning_rate": 0.0001928125650686023,
+      "loss": 4.4608,
+      "step": 71400
+    },
+    {
+      "epoch": 0.6150998622589532,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00019280247618904507,
+      "loss": 4.9095,
+      "step": 71450
+    },
+    {
+      "epoch": 0.615530303030303,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.00019279238049798572,
+      "loss": 4.8419,
+      "step": 71500
+    },
+    {
+      "epoch": 0.6159607438016529,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.0001927822779961653,
+      "loss": 4.6071,
+      "step": 71550
+    },
+    {
+      "epoch": 0.6163911845730028,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.00019277216868432531,
+      "loss": 4.6845,
+      "step": 71600
+    },
+    {
+      "epoch": 0.6168216253443526,
+      "grad_norm": 2.25,
+      "learning_rate": 0.00019276205256320771,
+      "loss": 5.2177,
+      "step": 71650
+    },
+    {
+      "epoch": 0.6172520661157025,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.00019275192963355507,
+      "loss": 4.3372,
+      "step": 71700
+    },
+    {
+      "epoch": 0.6176825068870524,
+      "grad_norm": 0.98046875,
+      "learning_rate": 0.00019274179989611033,
+      "loss": 4.8568,
+      "step": 71750
+    },
+    {
+      "epoch": 0.6181129476584022,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.00019273166335161707,
+      "loss": 4.987,
+      "step": 71800
+    },
+    {
+      "epoch": 0.6185433884297521,
+      "grad_norm": 2.578125,
+      "learning_rate": 0.00019272152000081918,
+      "loss": 4.6778,
+      "step": 71850
+    },
+    {
+      "epoch": 0.6189738292011019,
+      "grad_norm": 1.5234375,
+      "learning_rate": 0.00019271136984446124,
+      "loss": 4.6203,
+      "step": 71900
+    },
+    {
+      "epoch": 0.6194042699724518,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.0001927012128832882,
+      "loss": 4.6888,
+      "step": 71950
+    },
+    {
+      "epoch": 0.6198347107438017,
+      "grad_norm": 2.21875,
+      "learning_rate": 0.00019269104911804564,
+      "loss": 4.9668,
+      "step": 72000
+    },
+    {
+      "epoch": 0.6198347107438017,
+      "eval_loss": 5.390753746032715,
+      "eval_runtime": 22.0586,
+      "eval_samples_per_second": 29.014,
+      "eval_steps_per_second": 14.507,
+      "eval_tts_loss": 6.741954032634387,
+      "step": 72000
+    },
+    {
+      "epoch": 0.6202651515151515,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.00019268087854947945,
+      "loss": 5.1815,
+      "step": 72050
+    },
+    {
+      "epoch": 0.6206955922865014,
+      "grad_norm": 2.140625,
+      "learning_rate": 0.0001926707011783362,
+      "loss": 4.65,
+      "step": 72100
+    },
+    {
+      "epoch": 0.6211260330578512,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.0001926605170053629,
+      "loss": 4.8605,
+      "step": 72150
+    },
+    {
+      "epoch": 0.6215564738292011,
+      "grad_norm": 1.2109375,
+      "learning_rate": 0.00019265032603130698,
+      "loss": 4.9785,
+      "step": 72200
+    },
+    {
+      "epoch": 0.621986914600551,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.00019264012825691648,
+      "loss": 4.4861,
+      "step": 72250
+    },
+    {
+      "epoch": 0.6224173553719008,
+      "grad_norm": 3.265625,
+      "learning_rate": 0.00019262992368293988,
+      "loss": 4.8425,
+      "step": 72300
+    },
+    {
+      "epoch": 0.6228477961432507,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.0001926197123101262,
+      "loss": 5.143,
+      "step": 72350
+    },
+    {
+      "epoch": 0.6232782369146006,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.0001926094941392249,
+      "loss": 5.0026,
+      "step": 72400
+    },
+    {
+      "epoch": 0.6237086776859504,
+      "grad_norm": 1.859375,
+      "learning_rate": 0.00019259926917098596,
+      "loss": 4.5678,
+      "step": 72450
+    },
+    {
+      "epoch": 0.6241391184573003,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.00019258903740615994,
+      "loss": 4.6358,
+      "step": 72500
+    },
+    {
+      "epoch": 0.6245695592286501,
+      "grad_norm": 1.2578125,
+      "learning_rate": 0.00019257879884549775,
+      "loss": 4.7183,
+      "step": 72550
+    },
+    {
+      "epoch": 0.625,
+      "grad_norm": 2.296875,
+      "learning_rate": 0.00019256855348975097,
+      "loss": 4.7642,
+      "step": 72600
+    },
+    {
+      "epoch": 0.6254304407713499,
+      "grad_norm": 2.421875,
+      "learning_rate": 0.00019255830133967146,
+      "loss": 4.8014,
+      "step": 72650
+    },
+    {
+      "epoch": 0.6258608815426997,
+      "grad_norm": 1.9296875,
+      "learning_rate": 0.00019254804239601182,
+      "loss": 4.5654,
+      "step": 72700
+    },
+    {
+      "epoch": 0.6262913223140496,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00019253777665952497,
+      "loss": 4.8976,
+      "step": 72750
+    },
+    {
+      "epoch": 0.6267217630853994,
+      "grad_norm": 1.4375,
+      "learning_rate": 0.00019252750413096444,
+      "loss": 5.0384,
+      "step": 72800
+    },
+    {
+      "epoch": 0.6271522038567493,
+      "grad_norm": 3.1875,
+      "learning_rate": 0.00019251722481108415,
+      "loss": 4.4505,
+      "step": 72850
+    },
+    {
+      "epoch": 0.6275826446280992,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.00019250693870063863,
+      "loss": 4.5738,
+      "step": 72900
+    },
+    {
+      "epoch": 0.628013085399449,
+      "grad_norm": 1.765625,
+      "learning_rate": 0.00019249664580038282,
+      "loss": 4.5589,
+      "step": 72950
+    },
+    {
+      "epoch": 0.6284435261707989,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00019248634611107227,
+      "loss": 4.9839,
+      "step": 73000
+    },
+    {
+      "epoch": 0.6288739669421488,
+      "grad_norm": 1.640625,
+      "learning_rate": 0.00019247603963346287,
+      "loss": 4.5437,
+      "step": 73050
+    },
+    {
+      "epoch": 0.6293044077134986,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.00019246572636831113,
+      "loss": 5.2066,
+      "step": 73100
+    },
+    {
+      "epoch": 0.6297348484848485,
+      "grad_norm": 3.8125,
+      "learning_rate": 0.00019245540631637403,
+      "loss": 4.955,
+      "step": 73150
+    },
+    {
+      "epoch": 0.6301652892561983,
+      "grad_norm": 1.578125,
+      "learning_rate": 0.00019244507947840903,
+      "loss": 4.8249,
+      "step": 73200
+    },
+    {
+      "epoch": 0.6305957300275482,
+      "grad_norm": 1.3203125,
+      "learning_rate": 0.0001924347458551741,
+      "loss": 5.0265,
+      "step": 73250
+    },
+    {
+      "epoch": 0.6310261707988981,
+      "grad_norm": 1.71875,
+      "learning_rate": 0.00019242440544742768,
+      "loss": 4.9371,
+      "step": 73300
+    },
+    {
+      "epoch": 0.6314566115702479,
+      "grad_norm": 2.265625,
+      "learning_rate": 0.00019241405825592877,
+      "loss": 4.8961,
+      "step": 73350
+    },
+    {
+      "epoch": 0.6318870523415978,
+      "grad_norm": 2.265625,
+      "learning_rate": 0.00019240370428143683,
+      "loss": 4.7428,
+      "step": 73400
+    },
+    {
+      "epoch": 0.6323174931129476,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00019239334352471178,
+      "loss": 4.5569,
+      "step": 73450
+    },
+    {
+      "epoch": 0.6327479338842975,
+      "grad_norm": 1.25,
+      "learning_rate": 0.0001923829759865141,
+      "loss": 4.952,
+      "step": 73500
+    },
+    {
+      "epoch": 0.6331783746556474,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.0001923726016676048,
+      "loss": 4.6965,
+      "step": 73550
+    },
+    {
+      "epoch": 0.6336088154269972,
+      "grad_norm": 2.265625,
+      "learning_rate": 0.00019236222056874525,
+      "loss": 4.6287,
+      "step": 73600
+    },
+    {
+      "epoch": 0.6340392561983471,
+      "grad_norm": 1.6796875,
+      "learning_rate": 0.00019235183269069746,
+      "loss": 5.0443,
+      "step": 73650
+    },
+    {
+      "epoch": 0.634469696969697,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.0001923414380342238,
+      "loss": 4.6662,
+      "step": 73700
+    },
+    {
+      "epoch": 0.6349001377410468,
+      "grad_norm": 3.0,
+      "learning_rate": 0.0001923310366000873,
+      "loss": 4.4377,
+      "step": 73750
+    },
+    {
+      "epoch": 0.6353305785123967,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00019232062838905135,
+      "loss": 4.8102,
+      "step": 73800
+    },
+    {
+      "epoch": 0.6357610192837465,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.00019231021340187994,
+      "loss": 4.8413,
+      "step": 73850
+    },
+    {
+      "epoch": 0.6361914600550964,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00019229979163933744,
+      "loss": 4.8162,
+      "step": 73900
+    },
+    {
+      "epoch": 0.6366219008264463,
+      "grad_norm": 1.3984375,
+      "learning_rate": 0.00019228936310218883,
+      "loss": 4.8704,
+      "step": 73950
+    },
+    {
+      "epoch": 0.6370523415977961,
+      "grad_norm": 1.8515625,
+      "learning_rate": 0.00019227892779119957,
+      "loss": 4.5698,
+      "step": 74000
+    },
+    {
+      "epoch": 0.637482782369146,
+      "grad_norm": 1.828125,
+      "learning_rate": 0.0001922684857071355,
+      "loss": 5.0687,
+      "step": 74050
+    },
+    {
+      "epoch": 0.6379132231404959,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.00019225803685076316,
+      "loss": 4.6932,
+      "step": 74100
+    },
+    {
+      "epoch": 0.6383436639118457,
+      "grad_norm": 1.515625,
+      "learning_rate": 0.00019224758122284938,
+      "loss": 4.6711,
+      "step": 74150
+    },
+    {
+      "epoch": 0.6387741046831956,
+      "grad_norm": 1.8359375,
+      "learning_rate": 0.0001922371188241616,
+      "loss": 4.8663,
+      "step": 74200
+    },
+    {
+      "epoch": 0.6392045454545454,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.00019222664965546775,
+      "loss": 4.992,
+      "step": 74250
+    },
+    {
+      "epoch": 0.6396349862258953,
+      "grad_norm": 1.390625,
+      "learning_rate": 0.00019221617371753625,
+      "loss": 4.7318,
+      "step": 74300
+    },
+    {
+      "epoch": 0.6400654269972452,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.00019220569101113607,
+      "loss": 5.1225,
+      "step": 74350
+    },
+    {
+      "epoch": 0.640495867768595,
+      "grad_norm": 4.625,
+      "learning_rate": 0.00019219520153703647,
+      "loss": 4.9056,
+      "step": 74400
+    },
+    {
+      "epoch": 0.6409263085399449,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.00019218470529600747,
+      "loss": 4.8433,
+      "step": 74450
+    },
+    {
+      "epoch": 0.6413567493112947,
+      "grad_norm": 2.125,
+      "learning_rate": 0.00019217420228881946,
+      "loss": 4.6725,
+      "step": 74500
+    },
+    {
+      "epoch": 0.6417871900826446,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.00019216369251624332,
+      "loss": 4.8702,
+      "step": 74550
+    },
+    {
+      "epoch": 0.6422176308539945,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.00019215317597905043,
+      "loss": 4.7003,
+      "step": 74600
+    },
+    {
+      "epoch": 0.6426480716253443,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.00019214265267801267,
+      "loss": 4.9039,
+      "step": 74650
+    },
+    {
+      "epoch": 0.6430785123966942,
+      "grad_norm": 3.96875,
+      "learning_rate": 0.0001921321226139025,
+      "loss": 4.6823,
+      "step": 74700
+    },
+    {
+      "epoch": 0.6435089531680441,
+      "grad_norm": 1.84375,
+      "learning_rate": 0.00019212158578749272,
+      "loss": 4.9247,
+      "step": 74750
+    },
+    {
+      "epoch": 0.6439393939393939,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.00019211104219955677,
+      "loss": 4.6949,
+      "step": 74800
+    },
+    {
+      "epoch": 0.6443698347107438,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00019210049185086853,
+      "loss": 4.7005,
+      "step": 74850
+    },
+    {
+      "epoch": 0.6448002754820936,
+      "grad_norm": 1.890625,
+      "learning_rate": 0.00019208993474220228,
+      "loss": 4.8432,
+      "step": 74900
+    },
+    {
+      "epoch": 0.6452307162534435,
+      "grad_norm": 3.171875,
+      "learning_rate": 0.000192079370874333,
+      "loss": 4.7056,
+      "step": 74950
+    },
+    {
+      "epoch": 0.6456611570247934,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00019206880024803601,
+      "loss": 4.8833,
+      "step": 75000
+    },
+    {
+      "epoch": 0.6456611570247934,
+      "eval_loss": 5.383493900299072,
+      "eval_runtime": 21.7227,
+      "eval_samples_per_second": 29.462,
+      "eval_steps_per_second": 14.731,
+      "eval_tts_loss": 6.751945592916089,
+      "step": 75000
+    },
+    {
+      "epoch": 0.6460915977961432,
+      "grad_norm": 1.21875,
+      "learning_rate": 0.00019205822286408718,
+      "loss": 4.8866,
+      "step": 75050
+    },
+    {
+      "epoch": 0.6465220385674931,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00019204763872326285,
+      "loss": 4.9559,
+      "step": 75100
+    },
+    {
+      "epoch": 0.6469524793388429,
+      "grad_norm": 1.6328125,
+      "learning_rate": 0.00019203704782633988,
+      "loss": 4.8879,
+      "step": 75150
+    },
+    {
+      "epoch": 0.6473829201101928,
+      "grad_norm": 1.78125,
+      "learning_rate": 0.00019202645017409563,
+      "loss": 4.1499,
+      "step": 75200
+    },
+    {
+      "epoch": 0.6478133608815427,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00019201584576730794,
+      "loss": 4.6582,
+      "step": 75250
+    },
+    {
+      "epoch": 0.6482438016528925,
+      "grad_norm": 1.6171875,
+      "learning_rate": 0.0001920052346067551,
+      "loss": 5.1874,
+      "step": 75300
+    },
+    {
+      "epoch": 0.6486742424242424,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.00019199461669321604,
+      "loss": 5.3677,
+      "step": 75350
+    },
+    {
+      "epoch": 0.6491046831955923,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.00019198399202747005,
+      "loss": 4.675,
+      "step": 75400
+    },
+    {
+      "epoch": 0.6495351239669421,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.00019197336061029695,
+      "loss": 4.7404,
+      "step": 75450
+    },
+    {
+      "epoch": 0.649965564738292,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.00019196272244247707,
+      "loss": 4.6372,
+      "step": 75500
+    },
+    {
+      "epoch": 0.6503960055096418,
+      "grad_norm": 1.921875,
+      "learning_rate": 0.0001919520775247912,
+      "loss": 4.7628,
+      "step": 75550
+    },
+    {
+      "epoch": 0.6508264462809917,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00019194142585802072,
+      "loss": 4.6766,
+      "step": 75600
+    },
+    {
+      "epoch": 0.6512568870523416,
+      "grad_norm": 2.296875,
+      "learning_rate": 0.00019193076744294737,
+      "loss": 4.5604,
+      "step": 75650
+    },
+    {
+      "epoch": 0.6516873278236914,
+      "grad_norm": 2.296875,
+      "learning_rate": 0.00019192010228035348,
+      "loss": 5.1918,
+      "step": 75700
+    },
+    {
+      "epoch": 0.6521177685950413,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.00019190943037102187,
+      "loss": 4.7353,
+      "step": 75750
+    },
+    {
+      "epoch": 0.6525482093663911,
+      "grad_norm": 1.921875,
+      "learning_rate": 0.00019189875171573583,
+      "loss": 4.7499,
+      "step": 75800
+    },
+    {
+      "epoch": 0.652978650137741,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.00019188806631527912,
+      "loss": 4.61,
+      "step": 75850
+    },
+    {
+      "epoch": 0.6534090909090909,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.00019187737417043603,
+      "loss": 4.6194,
+      "step": 75900
+    },
+    {
+      "epoch": 0.6538395316804407,
+      "grad_norm": 2.125,
+      "learning_rate": 0.00019186667528199136,
+      "loss": 4.4737,
+      "step": 75950
+    },
+    {
+      "epoch": 0.6542699724517906,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.0001918559696507304,
+      "loss": 4.7515,
+      "step": 76000
+    },
+    {
+      "epoch": 0.6547004132231405,
+      "grad_norm": 3.203125,
+      "learning_rate": 0.00019184525727743889,
+      "loss": 4.6163,
+      "step": 76050
+    },
+    {
+      "epoch": 0.6551308539944903,
+      "grad_norm": 4.375,
+      "learning_rate": 0.00019183453816290312,
+      "loss": 4.987,
+      "step": 76100
+    },
+    {
+      "epoch": 0.6555612947658402,
+      "grad_norm": 1.21875,
+      "learning_rate": 0.00019182381230790983,
+      "loss": 4.328,
+      "step": 76150
+    },
+    {
+      "epoch": 0.65599173553719,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.00019181307971324625,
+      "loss": 4.7604,
+      "step": 76200
+    },
+    {
+      "epoch": 0.6564221763085399,
+      "grad_norm": 3.609375,
+      "learning_rate": 0.0001918023403797002,
+      "loss": 4.6351,
+      "step": 76250
+    },
+    {
+      "epoch": 0.6568526170798898,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.00019179159430805987,
+      "loss": 4.8329,
+      "step": 76300
+    },
+    {
+      "epoch": 0.6572830578512396,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.00019178084149911403,
+      "loss": 4.6761,
+      "step": 76350
+    },
+    {
+      "epoch": 0.6577134986225895,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.00019177008195365186,
+      "loss": 4.7454,
+      "step": 76400
+    },
+    {
+      "epoch": 0.6581439393939394,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.00019175931567246314,
+      "loss": 4.5504,
+      "step": 76450
+    },
+    {
+      "epoch": 0.6585743801652892,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.00019174854265633807,
+      "loss": 4.7031,
+      "step": 76500
+    },
+    {
+      "epoch": 0.6590048209366391,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.00019173776290606738,
+      "loss": 4.6054,
+      "step": 76550
+    },
+    {
+      "epoch": 0.6594352617079889,
+      "grad_norm": 2.265625,
+      "learning_rate": 0.00019172697642244225,
+      "loss": 4.5167,
+      "step": 76600
+    },
+    {
+      "epoch": 0.6598657024793388,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.00019171618320625445,
+      "loss": 5.1212,
+      "step": 76650
+    },
+    {
+      "epoch": 0.6602961432506887,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.0001917053832582961,
+      "loss": 4.7065,
+      "step": 76700
+    },
+    {
+      "epoch": 0.6607265840220385,
+      "grad_norm": 1.71875,
+      "learning_rate": 0.00019169457657935995,
+      "loss": 4.8365,
+      "step": 76750
+    },
+    {
+      "epoch": 0.6611570247933884,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.00019168376317023916,
+      "loss": 4.9541,
+      "step": 76800
+    },
+    {
+      "epoch": 0.6615874655647382,
+      "grad_norm": 1.3984375,
+      "learning_rate": 0.0001916729430317274,
+      "loss": 4.766,
+      "step": 76850
+    },
+    {
+      "epoch": 0.6620179063360881,
+      "grad_norm": 1.25,
+      "learning_rate": 0.00019166211616461885,
+      "loss": 4.9038,
+      "step": 76900
+    },
+    {
+      "epoch": 0.662448347107438,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.00019165128256970823,
+      "loss": 5.0386,
+      "step": 76950
+    },
+    {
+      "epoch": 0.6628787878787878,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.00019164044224779064,
+      "loss": 4.7015,
+      "step": 77000
+    },
+    {
+      "epoch": 0.6633092286501377,
+      "grad_norm": 3.5625,
+      "learning_rate": 0.00019162959519966174,
+      "loss": 4.6228,
+      "step": 77050
+    },
+    {
+      "epoch": 0.6637396694214877,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00019161874142611772,
+      "loss": 4.5877,
+      "step": 77100
+    },
+    {
+      "epoch": 0.6641701101928374,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.0001916078809279552,
+      "loss": 4.9524,
+      "step": 77150
+    },
+    {
+      "epoch": 0.6646005509641874,
+      "grad_norm": 1.4140625,
+      "learning_rate": 0.0001915970137059713,
+      "loss": 4.42,
+      "step": 77200
+    },
+    {
+      "epoch": 0.6650309917355371,
+      "grad_norm": 1.9921875,
+      "learning_rate": 0.00019158613976096373,
+      "loss": 4.789,
+      "step": 77250
+    },
+    {
+      "epoch": 0.665461432506887,
+      "grad_norm": 1.78125,
+      "learning_rate": 0.0001915752590937305,
+      "loss": 4.9101,
+      "step": 77300
+    },
+    {
+      "epoch": 0.665891873278237,
+      "grad_norm": 3.421875,
+      "learning_rate": 0.0001915643717050703,
+      "loss": 4.7548,
+      "step": 77350
+    },
+    {
+      "epoch": 0.6663223140495868,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.00019155347759578222,
+      "loss": 4.7368,
+      "step": 77400
+    },
+    {
+      "epoch": 0.6667527548209367,
+      "grad_norm": 2.234375,
+      "learning_rate": 0.00019154257676666587,
+      "loss": 4.8501,
+      "step": 77450
+    },
+    {
+      "epoch": 0.6671831955922864,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.00019153166921852136,
+      "loss": 4.6698,
+      "step": 77500
+    },
+    {
+      "epoch": 0.6676136363636364,
+      "grad_norm": 3.4375,
+      "learning_rate": 0.0001915207549521492,
+      "loss": 4.8337,
+      "step": 77550
+    },
+    {
+      "epoch": 0.6680440771349863,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.0001915098339683506,
+      "loss": 4.7476,
+      "step": 77600
+    },
+    {
+      "epoch": 0.668474517906336,
+      "grad_norm": 2.625,
+      "learning_rate": 0.00019149890626792707,
+      "loss": 4.4777,
+      "step": 77650
+    },
+    {
+      "epoch": 0.668904958677686,
+      "grad_norm": 1.7109375,
+      "learning_rate": 0.00019148797185168065,
+      "loss": 4.7891,
+      "step": 77700
+    },
+    {
+      "epoch": 0.6693353994490359,
+      "grad_norm": 1.3203125,
+      "learning_rate": 0.000191477030720414,
+      "loss": 4.6878,
+      "step": 77750
+    },
+    {
+      "epoch": 0.6697658402203857,
+      "grad_norm": 1.71875,
+      "learning_rate": 0.00019146608287493004,
+      "loss": 4.7881,
+      "step": 77800
+    },
+    {
+      "epoch": 0.6701962809917356,
+      "grad_norm": 1.328125,
+      "learning_rate": 0.00019145512831603243,
+      "loss": 4.6572,
+      "step": 77850
+    },
+    {
+      "epoch": 0.6706267217630854,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.00019144416704452516,
+      "loss": 4.8722,
+      "step": 77900
+    },
+    {
+      "epoch": 0.6710571625344353,
+      "grad_norm": 1.4453125,
+      "learning_rate": 0.00019143319906121277,
+      "loss": 4.7615,
+      "step": 77950
+    },
+    {
+      "epoch": 0.6714876033057852,
+      "grad_norm": 1.59375,
+      "learning_rate": 0.00019142222436690034,
+      "loss": 4.8712,
+      "step": 78000
+    },
+    {
+      "epoch": 0.6714876033057852,
+      "eval_loss": 5.375321865081787,
+      "eval_runtime": 22.0388,
+      "eval_samples_per_second": 29.04,
+      "eval_steps_per_second": 14.52,
+      "eval_tts_loss": 6.785668295288904,
+      "step": 78000
+    },
+    {
+      "epoch": 0.671918044077135,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.0001914112429623933,
+      "loss": 4.644,
+      "step": 78050
+    },
+    {
+      "epoch": 0.6723484848484849,
+      "grad_norm": 3.4375,
+      "learning_rate": 0.0001914002548484977,
+      "loss": 4.6583,
+      "step": 78100
+    },
+    {
+      "epoch": 0.6727789256198347,
+      "grad_norm": 1.1953125,
+      "learning_rate": 0.00019138926002602004,
+      "loss": 4.907,
+      "step": 78150
+    },
+    {
+      "epoch": 0.6732093663911846,
+      "grad_norm": 1.9609375,
+      "learning_rate": 0.00019137825849576735,
+      "loss": 4.6658,
+      "step": 78200
+    },
+    {
+      "epoch": 0.6736398071625345,
+      "grad_norm": 1.203125,
+      "learning_rate": 0.00019136725025854708,
+      "loss": 4.7018,
+      "step": 78250
+    },
+    {
+      "epoch": 0.6740702479338843,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.00019135623531516723,
+      "loss": 4.685,
+      "step": 78300
+    },
+    {
+      "epoch": 0.6745006887052342,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00019134521366643626,
+      "loss": 4.8097,
+      "step": 78350
+    },
+    {
+      "epoch": 0.6749311294765841,
+      "grad_norm": 1.4609375,
+      "learning_rate": 0.00019133418531316315,
+      "loss": 4.8626,
+      "step": 78400
+    },
+    {
+      "epoch": 0.6753615702479339,
+      "grad_norm": 3.015625,
+      "learning_rate": 0.00019132315025615734,
+      "loss": 4.5555,
+      "step": 78450
+    },
+    {
+      "epoch": 0.6757920110192838,
+      "grad_norm": 6.40625,
+      "learning_rate": 0.00019131210849622878,
+      "loss": 4.5359,
+      "step": 78500
+    },
+    {
+      "epoch": 0.6762224517906336,
+      "grad_norm": 3.15625,
+      "learning_rate": 0.00019130106003418794,
+      "loss": 4.9026,
+      "step": 78550
+    },
+    {
+      "epoch": 0.6766528925619835,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.00019129000487084572,
+      "loss": 4.3526,
+      "step": 78600
+    },
+    {
+      "epoch": 0.6770833333333334,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.00019127894300701356,
+      "loss": 4.7109,
+      "step": 78650
+    },
+    {
+      "epoch": 0.6775137741046832,
+      "grad_norm": 2.984375,
+      "learning_rate": 0.00019126787444350338,
+      "loss": 4.8723,
+      "step": 78700
+    },
+    {
+      "epoch": 0.6779442148760331,
+      "grad_norm": 0.98828125,
+      "learning_rate": 0.00019125679918112755,
+      "loss": 4.4127,
+      "step": 78750
+    },
+    {
+      "epoch": 0.678374655647383,
+      "grad_norm": 2.375,
+      "learning_rate": 0.00019124571722069903,
+      "loss": 5.064,
+      "step": 78800
+    },
+    {
+      "epoch": 0.6788050964187328,
+      "grad_norm": 2.265625,
+      "learning_rate": 0.00019123462856303122,
+      "loss": 4.74,
+      "step": 78850
+    },
+    {
+      "epoch": 0.6792355371900827,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.0001912235332089379,
+      "loss": 4.6177,
+      "step": 78900
+    },
+    {
+      "epoch": 0.6796659779614325,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.00019121243115923357,
+      "loss": 4.7881,
+      "step": 78950
+    },
+    {
+      "epoch": 0.6800964187327824,
+      "grad_norm": 3.09375,
+      "learning_rate": 0.00019120132241473304,
+      "loss": 4.9935,
+      "step": 79000
+    },
+    {
+      "epoch": 0.6805268595041323,
+      "grad_norm": 1.5859375,
+      "learning_rate": 0.00019119020697625164,
+      "loss": 4.9343,
+      "step": 79050
+    },
+    {
+      "epoch": 0.6809573002754821,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.00019117908484460528,
+      "loss": 4.6888,
+      "step": 79100
+    },
+    {
+      "epoch": 0.681387741046832,
+      "grad_norm": 1.8046875,
+      "learning_rate": 0.00019116795602061023,
+      "loss": 4.8264,
+      "step": 79150
+    },
+    {
+      "epoch": 0.6818181818181818,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.0001911568205050834,
+      "loss": 4.7404,
+      "step": 79200
+    },
+    {
+      "epoch": 0.6822486225895317,
+      "grad_norm": 3.640625,
+      "learning_rate": 0.00019114567829884205,
+      "loss": 4.8228,
+      "step": 79250
+    },
+    {
+      "epoch": 0.6826790633608816,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.00019113452940270402,
+      "loss": 4.9754,
+      "step": 79300
+    },
+    {
+      "epoch": 0.6831095041322314,
+      "grad_norm": 2.875,
+      "learning_rate": 0.00019112337381748762,
+      "loss": 4.9272,
+      "step": 79350
+    },
+    {
+      "epoch": 0.6835399449035813,
+      "grad_norm": 0.78515625,
+      "learning_rate": 0.00019111221154401165,
+      "loss": 4.8657,
+      "step": 79400
+    },
+    {
+      "epoch": 0.6839703856749312,
+      "grad_norm": 3.046875,
+      "learning_rate": 0.00019110104258309536,
+      "loss": 4.6926,
+      "step": 79450
+    },
+    {
+      "epoch": 0.684400826446281,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.00019108986693555858,
+      "loss": 4.8111,
+      "step": 79500
+    },
+    {
+      "epoch": 0.6848312672176309,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.00019107868460222153,
+      "loss": 4.7785,
+      "step": 79550
+    },
+    {
+      "epoch": 0.6852617079889807,
+      "grad_norm": 1.2265625,
+      "learning_rate": 0.000191067495583905,
+      "loss": 4.86,
+      "step": 79600
+    },
+    {
+      "epoch": 0.6856921487603306,
+      "grad_norm": 2.25,
+      "learning_rate": 0.0001910562998814302,
+      "loss": 4.6237,
+      "step": 79650
+    },
+    {
+      "epoch": 0.6861225895316805,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.00019104509749561896,
+      "loss": 4.7586,
+      "step": 79700
+    },
+    {
+      "epoch": 0.6865530303030303,
+      "grad_norm": 3.328125,
+      "learning_rate": 0.00019103388842729342,
+      "loss": 5.1268,
+      "step": 79750
+    },
+    {
+      "epoch": 0.6869834710743802,
+      "grad_norm": 2.921875,
+      "learning_rate": 0.00019102267267727632,
+      "loss": 5.0966,
+      "step": 79800
+    },
+    {
+      "epoch": 0.68741391184573,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.00019101145024639092,
+      "loss": 4.7019,
+      "step": 79850
+    },
+    {
+      "epoch": 0.6878443526170799,
+      "grad_norm": 3.25,
+      "learning_rate": 0.00019100022113546086,
+      "loss": 4.4957,
+      "step": 79900
+    },
+    {
+      "epoch": 0.6882747933884298,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00019098898534531034,
+      "loss": 5.061,
+      "step": 79950
+    },
+    {
+      "epoch": 0.6887052341597796,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00019097774287676414,
+      "loss": 4.7823,
+      "step": 80000
+    },
+    {
+      "epoch": 0.6891356749311295,
+      "grad_norm": 2.984375,
+      "learning_rate": 0.0001909664937306473,
+      "loss": 4.8135,
+      "step": 80050
+    },
+    {
+      "epoch": 0.6895661157024794,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.00019095523790778554,
+      "loss": 4.794,
+      "step": 80100
+    },
+    {
+      "epoch": 0.6899965564738292,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.00019094397540900504,
+      "loss": 4.8347,
+      "step": 80150
+    },
+    {
+      "epoch": 0.6904269972451791,
+      "grad_norm": 3.265625,
+      "learning_rate": 0.0001909327062351324,
+      "loss": 4.5087,
+      "step": 80200
+    },
+    {
+      "epoch": 0.6908574380165289,
+      "grad_norm": 1.453125,
+      "learning_rate": 0.00019092143038699474,
+      "loss": 4.6743,
+      "step": 80250
+    },
+    {
+      "epoch": 0.6912878787878788,
+      "grad_norm": 0.8828125,
+      "learning_rate": 0.00019091014786541975,
+      "loss": 4.8636,
+      "step": 80300
+    },
+    {
+      "epoch": 0.6917183195592287,
+      "grad_norm": 2.921875,
+      "learning_rate": 0.00019089885867123552,
+      "loss": 4.6687,
+      "step": 80350
+    },
+    {
+      "epoch": 0.6921487603305785,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.0001908875628052706,
+      "loss": 4.7632,
+      "step": 80400
+    },
+    {
+      "epoch": 0.6925792011019284,
+      "grad_norm": 2.921875,
+      "learning_rate": 0.00019087626026835413,
+      "loss": 4.6997,
+      "step": 80450
+    },
+    {
+      "epoch": 0.6930096418732782,
+      "grad_norm": 3.4375,
+      "learning_rate": 0.00019086495106131567,
+      "loss": 4.7043,
+      "step": 80500
+    },
+    {
+      "epoch": 0.6934400826446281,
+      "grad_norm": 3.3125,
+      "learning_rate": 0.00019085363518498534,
+      "loss": 4.4696,
+      "step": 80550
+    },
+    {
+      "epoch": 0.693870523415978,
+      "grad_norm": 1.625,
+      "learning_rate": 0.00019084231264019364,
+      "loss": 5.0989,
+      "step": 80600
+    },
+    {
+      "epoch": 0.6943009641873278,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00019083098342777163,
+      "loss": 4.6036,
+      "step": 80650
+    },
+    {
+      "epoch": 0.6947314049586777,
+      "grad_norm": 1.625,
+      "learning_rate": 0.00019081964754855092,
+      "loss": 4.8058,
+      "step": 80700
+    },
+    {
+      "epoch": 0.6951618457300276,
+      "grad_norm": 1.8828125,
+      "learning_rate": 0.0001908083050033634,
+      "loss": 4.4008,
+      "step": 80750
+    },
+    {
+      "epoch": 0.6955922865013774,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.00019079695579304174,
+      "loss": 4.7096,
+      "step": 80800
+    },
+    {
+      "epoch": 0.6960227272727273,
+      "grad_norm": 1.2890625,
+      "learning_rate": 0.00019078559991841885,
+      "loss": 4.8174,
+      "step": 80850
+    },
+    {
+      "epoch": 0.6964531680440771,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00019077423738032826,
+      "loss": 4.9544,
+      "step": 80900
+    },
+    {
+      "epoch": 0.696883608815427,
+      "grad_norm": 3.515625,
+      "learning_rate": 0.00019076286817960395,
+      "loss": 4.8073,
+      "step": 80950
+    },
+    {
+      "epoch": 0.6973140495867769,
+      "grad_norm": 1.984375,
+      "learning_rate": 0.0001907514923170804,
+      "loss": 5.1697,
+      "step": 81000
+    },
+    {
+      "epoch": 0.6973140495867769,
+      "eval_loss": 5.368470191955566,
+      "eval_runtime": 21.8057,
+      "eval_samples_per_second": 29.35,
+      "eval_steps_per_second": 14.675,
+      "eval_tts_loss": 6.716216654162234,
+      "step": 81000
+    },
+    {
+      "epoch": 0.6977444903581267,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00019074010979359257,
+      "loss": 4.5545,
+      "step": 81050
+    },
+    {
+      "epoch": 0.6981749311294766,
+      "grad_norm": 1.203125,
+      "learning_rate": 0.00019072872060997594,
+      "loss": 4.9125,
+      "step": 81100
+    },
+    {
+      "epoch": 0.6986053719008265,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00019071732476706637,
+      "loss": 4.8031,
+      "step": 81150
+    },
+    {
+      "epoch": 0.6990358126721763,
+      "grad_norm": 4.78125,
+      "learning_rate": 0.00019070592226570038,
+      "loss": 4.9183,
+      "step": 81200
+    },
+    {
+      "epoch": 0.6994662534435262,
+      "grad_norm": 1.5703125,
+      "learning_rate": 0.00019069451310671487,
+      "loss": 4.7762,
+      "step": 81250
+    },
+    {
+      "epoch": 0.699896694214876,
+      "grad_norm": 2.375,
+      "learning_rate": 0.0001906830972909472,
+      "loss": 4.8935,
+      "step": 81300
+    },
+    {
+      "epoch": 0.7003271349862259,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.00019067167481923533,
+      "loss": 4.9102,
+      "step": 81350
+    },
+    {
+      "epoch": 0.7007575757575758,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00019066024569241758,
+      "loss": 4.6657,
+      "step": 81400
+    },
+    {
+      "epoch": 0.7011880165289256,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00019064880991133287,
+      "loss": 4.589,
+      "step": 81450
+    },
+    {
+      "epoch": 0.7016184573002755,
+      "grad_norm": 2.984375,
+      "learning_rate": 0.00019063736747682055,
+      "loss": 4.8951,
+      "step": 81500
+    },
+    {
+      "epoch": 0.7020488980716253,
+      "grad_norm": 3.046875,
+      "learning_rate": 0.00019062591838972046,
+      "loss": 4.8286,
+      "step": 81550
+    },
+    {
+      "epoch": 0.7024793388429752,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.00019061446265087296,
+      "loss": 5.2785,
+      "step": 81600
+    },
+    {
+      "epoch": 0.7029097796143251,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.00019060300026111886,
+      "loss": 4.6883,
+      "step": 81650
+    },
+    {
+      "epoch": 0.7033402203856749,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.00019059153122129947,
+      "loss": 4.4538,
+      "step": 81700
+    },
+    {
+      "epoch": 0.7037706611570248,
+      "grad_norm": 3.171875,
+      "learning_rate": 0.0001905800555322566,
+      "loss": 4.5534,
+      "step": 81750
+    },
+    {
+      "epoch": 0.7042011019283747,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.00019056857319483256,
+      "loss": 4.9203,
+      "step": 81800
+    },
+    {
+      "epoch": 0.7046315426997245,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00019055708420987011,
+      "loss": 4.862,
+      "step": 81850
+    },
+    {
+      "epoch": 0.7050619834710744,
+      "grad_norm": 2.125,
+      "learning_rate": 0.0001905455885782125,
+      "loss": 4.6505,
+      "step": 81900
+    },
+    {
+      "epoch": 0.7054924242424242,
+      "grad_norm": 3.90625,
+      "learning_rate": 0.0001905340863007035,
+      "loss": 4.6438,
+      "step": 81950
+    },
+    {
+      "epoch": 0.7059228650137741,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.00019052257737818738,
+      "loss": 4.8286,
+      "step": 82000
+    },
+    {
+      "epoch": 0.706353305785124,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00019051106181150883,
+      "loss": 4.4646,
+      "step": 82050
+    },
+    {
+      "epoch": 0.7067837465564738,
+      "grad_norm": 2.125,
+      "learning_rate": 0.00019049953960151308,
+      "loss": 4.5731,
+      "step": 82100
+    },
+    {
+      "epoch": 0.7072141873278237,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.00019048801074904583,
+      "loss": 4.9183,
+      "step": 82150
+    },
+    {
+      "epoch": 0.7076446280991735,
+      "grad_norm": 0.8984375,
+      "learning_rate": 0.00019047647525495328,
+      "loss": 4.328,
+      "step": 82200
+    },
+    {
+      "epoch": 0.7080750688705234,
+      "grad_norm": 0.91015625,
+      "learning_rate": 0.00019046493312008212,
+      "loss": 4.5565,
+      "step": 82250
+    },
+    {
+      "epoch": 0.7085055096418733,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.00019045338434527952,
+      "loss": 4.7026,
+      "step": 82300
+    },
+    {
+      "epoch": 0.7089359504132231,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.00019044182893139313,
+      "loss": 4.4458,
+      "step": 82350
+    },
+    {
+      "epoch": 0.709366391184573,
+      "grad_norm": 2.578125,
+      "learning_rate": 0.00019043026687927105,
+      "loss": 4.9716,
+      "step": 82400
+    },
+    {
+      "epoch": 0.7097968319559229,
+      "grad_norm": 1.703125,
+      "learning_rate": 0.00019041869818976198,
+      "loss": 4.6185,
+      "step": 82450
+    },
+    {
+      "epoch": 0.7102272727272727,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.000190407122863715,
+      "loss": 4.4424,
+      "step": 82500
+    },
+    {
+      "epoch": 0.7106577134986226,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.0001903955409019797,
+      "loss": 4.8352,
+      "step": 82550
+    },
+    {
+      "epoch": 0.7110881542699724,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.0001903839523054062,
+      "loss": 4.9915,
+      "step": 82600
+    },
+    {
+      "epoch": 0.7115185950413223,
+      "grad_norm": 1.2890625,
+      "learning_rate": 0.00019037235707484507,
+      "loss": 4.9429,
+      "step": 82650
+    },
+    {
+      "epoch": 0.7119490358126722,
+      "grad_norm": 3.25,
+      "learning_rate": 0.00019036075521114738,
+      "loss": 4.8566,
+      "step": 82700
+    },
+    {
+      "epoch": 0.712379476584022,
+      "grad_norm": 2.125,
+      "learning_rate": 0.00019034914671516465,
+      "loss": 5.0509,
+      "step": 82750
+    },
+    {
+      "epoch": 0.7128099173553719,
+      "grad_norm": 1.4765625,
+      "learning_rate": 0.00019033753158774898,
+      "loss": 4.3636,
+      "step": 82800
+    },
+    {
+      "epoch": 0.7132403581267218,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.00019032590982975283,
+      "loss": 4.8871,
+      "step": 82850
+    },
+    {
+      "epoch": 0.7136707988980716,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.00019031428144202926,
+      "loss": 4.775,
+      "step": 82900
+    },
+    {
+      "epoch": 0.7141012396694215,
+      "grad_norm": 3.515625,
+      "learning_rate": 0.0001903026464254317,
+      "loss": 4.8247,
+      "step": 82950
+    },
+    {
+      "epoch": 0.7145316804407713,
+      "grad_norm": 2.265625,
+      "learning_rate": 0.00019029100478081424,
+      "loss": 4.5673,
+      "step": 83000
+    },
+    {
+      "epoch": 0.7149621212121212,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.00019027935650903124,
+      "loss": 4.3138,
+      "step": 83050
+    },
+    {
+      "epoch": 0.7153925619834711,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00019026770161093776,
+      "loss": 4.7773,
+      "step": 83100
+    },
+    {
+      "epoch": 0.7158230027548209,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.00019025604008738917,
+      "loss": 4.9081,
+      "step": 83150
+    },
+    {
+      "epoch": 0.7162534435261708,
+      "grad_norm": 1.234375,
+      "learning_rate": 0.00019024437193924145,
+      "loss": 4.5025,
+      "step": 83200
+    },
+    {
+      "epoch": 0.7166838842975206,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00019023269716735098,
+      "loss": 4.8034,
+      "step": 83250
+    },
+    {
+      "epoch": 0.7171143250688705,
+      "grad_norm": 4.5,
+      "learning_rate": 0.0001902210157725747,
+      "loss": 5.0936,
+      "step": 83300
+    },
+    {
+      "epoch": 0.7175447658402204,
+      "grad_norm": 1.4296875,
+      "learning_rate": 0.00019020932775576995,
+      "loss": 4.7048,
+      "step": 83350
+    },
+    {
+      "epoch": 0.7179752066115702,
+      "grad_norm": 1.765625,
+      "learning_rate": 0.00019019763311779465,
+      "loss": 4.576,
+      "step": 83400
+    },
+    {
+      "epoch": 0.7184056473829201,
+      "grad_norm": 1.25,
+      "learning_rate": 0.00019018593185950712,
+      "loss": 5.0109,
+      "step": 83450
+    },
+    {
+      "epoch": 0.71883608815427,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00019017422398176625,
+      "loss": 4.8809,
+      "step": 83500
+    },
+    {
+      "epoch": 0.7192665289256198,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.00019016250948543134,
+      "loss": 4.7406,
+      "step": 83550
+    },
+    {
+      "epoch": 0.7196969696969697,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00019015078837136227,
+      "loss": 4.3789,
+      "step": 83600
+    },
+    {
+      "epoch": 0.7201274104683195,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.00019013906064041923,
+      "loss": 4.5741,
+      "step": 83650
+    },
+    {
+      "epoch": 0.7205578512396694,
+      "grad_norm": 2.25,
+      "learning_rate": 0.00019012732629346312,
+      "loss": 5.1825,
+      "step": 83700
+    },
+    {
+      "epoch": 0.7209882920110193,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.0001901155853313552,
+      "loss": 4.5367,
+      "step": 83750
+    },
+    {
+      "epoch": 0.7214187327823691,
+      "grad_norm": 3.84375,
+      "learning_rate": 0.00019010383775495715,
+      "loss": 4.3739,
+      "step": 83800
+    },
+    {
+      "epoch": 0.721849173553719,
+      "grad_norm": 3.59375,
+      "learning_rate": 0.00019009208356513132,
+      "loss": 4.621,
+      "step": 83850
+    },
+    {
+      "epoch": 0.7222796143250688,
+      "grad_norm": 1.4453125,
+      "learning_rate": 0.00019008032276274036,
+      "loss": 4.3862,
+      "step": 83900
+    },
+    {
+      "epoch": 0.7227100550964187,
+      "grad_norm": 3.28125,
+      "learning_rate": 0.00019006855534864752,
+      "loss": 4.5218,
+      "step": 83950
+    },
+    {
+      "epoch": 0.7231404958677686,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.00019005678132371653,
+      "loss": 5.0559,
+      "step": 84000
+    },
+    {
+      "epoch": 0.7231404958677686,
+      "eval_loss": 5.360166549682617,
+      "eval_runtime": 22.1767,
+      "eval_samples_per_second": 28.859,
+      "eval_steps_per_second": 14.43,
+      "eval_tts_loss": 6.821026823562931,
+      "step": 84000
+    },
+    {
+      "epoch": 0.7235709366391184,
+      "grad_norm": 1.8515625,
+      "learning_rate": 0.00019004500068881152,
+      "loss": 4.8812,
+      "step": 84050
+    },
+    {
+      "epoch": 0.7240013774104683,
+      "grad_norm": 3.140625,
+      "learning_rate": 0.00019003321344479723,
+      "loss": 4.5547,
+      "step": 84100
+    },
+    {
+      "epoch": 0.7244318181818182,
+      "grad_norm": 3.171875,
+      "learning_rate": 0.00019002141959253876,
+      "loss": 4.4966,
+      "step": 84150
+    },
+    {
+      "epoch": 0.724862258953168,
+      "grad_norm": 1.953125,
+      "learning_rate": 0.0001900096191329018,
+      "loss": 4.7886,
+      "step": 84200
+    },
+    {
+      "epoch": 0.7252926997245179,
+      "grad_norm": 3.546875,
+      "learning_rate": 0.00018999781206675246,
+      "loss": 4.5249,
+      "step": 84250
+    },
+    {
+      "epoch": 0.7257231404958677,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.00018998599839495733,
+      "loss": 4.7126,
+      "step": 84300
+    },
+    {
+      "epoch": 0.7261535812672176,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.00018997417811838352,
+      "loss": 4.8049,
+      "step": 84350
+    },
+    {
+      "epoch": 0.7265840220385675,
+      "grad_norm": 3.0625,
+      "learning_rate": 0.00018996235123789864,
+      "loss": 4.6832,
+      "step": 84400
+    },
+    {
+      "epoch": 0.7270144628099173,
+      "grad_norm": 3.28125,
+      "learning_rate": 0.00018995051775437074,
+      "loss": 4.3974,
+      "step": 84450
+    },
+    {
+      "epoch": 0.7274449035812672,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.00018993867766866834,
+      "loss": 4.8823,
+      "step": 84500
+    },
+    {
+      "epoch": 0.727875344352617,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.00018992683098166055,
+      "loss": 4.6495,
+      "step": 84550
+    },
+    {
+      "epoch": 0.7283057851239669,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.00018991497769421683,
+      "loss": 4.7448,
+      "step": 84600
+    },
+    {
+      "epoch": 0.7287362258953168,
+      "grad_norm": 3.28125,
+      "learning_rate": 0.0001899031178072072,
+      "loss": 4.7228,
+      "step": 84650
+    },
+    {
+      "epoch": 0.7291666666666666,
+      "grad_norm": 2.25,
+      "learning_rate": 0.00018989125132150213,
+      "loss": 4.7001,
+      "step": 84700
+    },
+    {
+      "epoch": 0.7295971074380165,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00018987937823797266,
+      "loss": 5.123,
+      "step": 84750
+    },
+    {
+      "epoch": 0.7300275482093664,
+      "grad_norm": 2.875,
+      "learning_rate": 0.0001898674985574902,
+      "loss": 4.9325,
+      "step": 84800
+    },
+    {
+      "epoch": 0.7304579889807162,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.0001898556122809267,
+      "loss": 5.0725,
+      "step": 84850
+    },
+    {
+      "epoch": 0.7308884297520661,
+      "grad_norm": 2.5,
+      "learning_rate": 0.00018984371940915457,
+      "loss": 4.6041,
+      "step": 84900
+    },
+    {
+      "epoch": 0.7313188705234159,
+      "grad_norm": 3.25,
+      "learning_rate": 0.00018983181994304676,
+      "loss": 4.5216,
+      "step": 84950
+    },
+    {
+      "epoch": 0.7317493112947658,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.00018981991388347663,
+      "loss": 4.3788,
+      "step": 85000
+    },
+    {
+      "epoch": 0.7321797520661157,
+      "grad_norm": 1.6796875,
+      "learning_rate": 0.00018980800123131808,
+      "loss": 4.8238,
+      "step": 85050
+    },
+    {
+      "epoch": 0.7326101928374655,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00018979608198744545,
+      "loss": 5.1704,
+      "step": 85100
+    },
+    {
+      "epoch": 0.7330406336088154,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.00018978415615273364,
+      "loss": 4.882,
+      "step": 85150
+    },
+    {
+      "epoch": 0.7334710743801653,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.00018977222372805792,
+      "loss": 4.5257,
+      "step": 85200
+    },
+    {
+      "epoch": 0.7339015151515151,
+      "grad_norm": 3.171875,
+      "learning_rate": 0.00018976028471429413,
+      "loss": 4.7512,
+      "step": 85250
+    },
+    {
+      "epoch": 0.734331955922865,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00018974833911231858,
+      "loss": 4.7346,
+      "step": 85300
+    },
+    {
+      "epoch": 0.7347623966942148,
+      "grad_norm": 2.9375,
+      "learning_rate": 0.00018973638692300798,
+      "loss": 4.6258,
+      "step": 85350
+    },
+    {
+      "epoch": 0.7351928374655647,
+      "grad_norm": 1.53125,
+      "learning_rate": 0.00018972442814723972,
+      "loss": 4.9166,
+      "step": 85400
+    },
+    {
+      "epoch": 0.7356232782369146,
+      "grad_norm": 4.09375,
+      "learning_rate": 0.00018971246278589145,
+      "loss": 4.5802,
+      "step": 85450
+    },
+    {
+      "epoch": 0.7360537190082644,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.00018970049083984146,
+      "loss": 4.6552,
+      "step": 85500
+    },
+    {
+      "epoch": 0.7364841597796143,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.0001896885123099684,
+      "loss": 4.9737,
+      "step": 85550
+    },
+    {
+      "epoch": 0.7369146005509641,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.00018967652719715155,
+      "loss": 4.7488,
+      "step": 85600
+    },
+    {
+      "epoch": 0.737345041322314,
+      "grad_norm": 2.9375,
+      "learning_rate": 0.0001896645355022705,
+      "loss": 5.1506,
+      "step": 85650
+    },
+    {
+      "epoch": 0.737775482093664,
+      "grad_norm": 3.015625,
+      "learning_rate": 0.0001896525372262055,
+      "loss": 4.7965,
+      "step": 85700
+    },
+    {
+      "epoch": 0.7382059228650137,
+      "grad_norm": 1.359375,
+      "learning_rate": 0.00018964053236983713,
+      "loss": 4.5038,
+      "step": 85750
+    },
+    {
+      "epoch": 0.7386363636363636,
+      "grad_norm": 1.984375,
+      "learning_rate": 0.00018962852093404656,
+      "loss": 4.8172,
+      "step": 85800
+    },
+    {
+      "epoch": 0.7390668044077136,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.00018961650291971543,
+      "loss": 4.5635,
+      "step": 85850
+    },
+    {
+      "epoch": 0.7394972451790633,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.00018960447832772577,
+      "loss": 4.7879,
+      "step": 85900
+    },
+    {
+      "epoch": 0.7399276859504132,
+      "grad_norm": 4.15625,
+      "learning_rate": 0.00018959244715896015,
+      "loss": 4.6641,
+      "step": 85950
+    },
+    {
+      "epoch": 0.740358126721763,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.00018958040941430176,
+      "loss": 4.4173,
+      "step": 86000
+    },
+    {
+      "epoch": 0.740788567493113,
+      "grad_norm": 1.5,
+      "learning_rate": 0.000189568365094634,
+      "loss": 4.8023,
+      "step": 86050
+    },
+    {
+      "epoch": 0.7412190082644629,
+      "grad_norm": 3.71875,
+      "learning_rate": 0.00018955631420084098,
+      "loss": 4.9974,
+      "step": 86100
+    },
+    {
+      "epoch": 0.7416494490358126,
+      "grad_norm": 3.0625,
+      "learning_rate": 0.00018954425673380713,
+      "loss": 4.6843,
+      "step": 86150
+    },
+    {
+      "epoch": 0.7420798898071626,
+      "grad_norm": 3.671875,
+      "learning_rate": 0.00018953219269441754,
+      "loss": 4.5927,
+      "step": 86200
+    },
+    {
+      "epoch": 0.7425103305785123,
+      "grad_norm": 3.140625,
+      "learning_rate": 0.00018952012208355763,
+      "loss": 5.0064,
+      "step": 86250
+    },
+    {
+      "epoch": 0.7429407713498623,
+      "grad_norm": 1.9609375,
+      "learning_rate": 0.00018950804490211335,
+      "loss": 4.7797,
+      "step": 86300
+    },
+    {
+      "epoch": 0.7433712121212122,
+      "grad_norm": 3.015625,
+      "learning_rate": 0.00018949596115097118,
+      "loss": 4.8961,
+      "step": 86350
+    },
+    {
+      "epoch": 0.743801652892562,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00018948387083101803,
+      "loss": 4.9575,
+      "step": 86400
+    },
+    {
+      "epoch": 0.7442320936639119,
+      "grad_norm": 1.796875,
+      "learning_rate": 0.00018947177394314127,
+      "loss": 4.7509,
+      "step": 86450
+    },
+    {
+      "epoch": 0.7446625344352618,
+      "grad_norm": 1.2734375,
+      "learning_rate": 0.0001894596704882288,
+      "loss": 4.542,
+      "step": 86500
+    },
+    {
+      "epoch": 0.7450929752066116,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00018944756046716902,
+      "loss": 4.4015,
+      "step": 86550
+    },
+    {
+      "epoch": 0.7455234159779615,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.0001894354438808507,
+      "loss": 4.4901,
+      "step": 86600
+    },
+    {
+      "epoch": 0.7459538567493113,
+      "grad_norm": 1.578125,
+      "learning_rate": 0.00018942332073016327,
+      "loss": 4.6675,
+      "step": 86650
+    },
+    {
+      "epoch": 0.7463842975206612,
+      "grad_norm": 3.765625,
+      "learning_rate": 0.00018941119101599646,
+      "loss": 4.9081,
+      "step": 86700
+    },
+    {
+      "epoch": 0.7468147382920111,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.00018939905473924065,
+      "loss": 4.8476,
+      "step": 86750
+    },
+    {
+      "epoch": 0.7472451790633609,
+      "grad_norm": 1.5703125,
+      "learning_rate": 0.00018938691190078654,
+      "loss": 4.4907,
+      "step": 86800
+    },
+    {
+      "epoch": 0.7476756198347108,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.00018937476250152543,
+      "loss": 3.9543,
+      "step": 86850
+    },
+    {
+      "epoch": 0.7481060606060606,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.00018936260654234903,
+      "loss": 4.5492,
+      "step": 86900
+    },
+    {
+      "epoch": 0.7485365013774105,
+      "grad_norm": 4.625,
+      "learning_rate": 0.00018935044402414958,
+      "loss": 4.6187,
+      "step": 86950
+    },
+    {
+      "epoch": 0.7489669421487604,
+      "grad_norm": 3.671875,
+      "learning_rate": 0.00018933827494781982,
+      "loss": 4.553,
+      "step": 87000
+    },
+    {
+      "epoch": 0.7489669421487604,
+      "eval_loss": 5.351483345031738,
+      "eval_runtime": 22.7065,
+      "eval_samples_per_second": 28.186,
+      "eval_steps_per_second": 14.093,
+      "eval_tts_loss": 6.789242874380564,
+      "step": 87000
+    },
+    {
+      "epoch": 0.7493973829201102,
+      "grad_norm": 1.90625,
+      "learning_rate": 0.00018932609931425284,
+      "loss": 4.4619,
+      "step": 87050
+    },
+    {
+      "epoch": 0.7498278236914601,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.00018931391712434238,
+      "loss": 4.9802,
+      "step": 87100
+    },
+    {
+      "epoch": 0.75025826446281,
+      "grad_norm": 3.453125,
+      "learning_rate": 0.0001893017283789826,
+      "loss": 4.4043,
+      "step": 87150
+    },
+    {
+      "epoch": 0.7506887052341598,
+      "grad_norm": 1.6953125,
+      "learning_rate": 0.00018928953307906804,
+      "loss": 4.5111,
+      "step": 87200
+    },
+    {
+      "epoch": 0.7511191460055097,
+      "grad_norm": 3.59375,
+      "learning_rate": 0.0001892773312254939,
+      "loss": 4.7247,
+      "step": 87250
+    },
+    {
+      "epoch": 0.7515495867768595,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.0001892651228191557,
+      "loss": 4.7363,
+      "step": 87300
+    },
+    {
+      "epoch": 0.7519800275482094,
+      "grad_norm": 1.9296875,
+      "learning_rate": 0.0001892529078609496,
+      "loss": 4.6117,
+      "step": 87350
+    },
+    {
+      "epoch": 0.7524104683195593,
+      "grad_norm": 2.921875,
+      "learning_rate": 0.00018924068635177205,
+      "loss": 4.7823,
+      "step": 87400
+    },
+    {
+      "epoch": 0.7528409090909091,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.00018922845829252017,
+      "loss": 4.9833,
+      "step": 87450
+    },
+    {
+      "epoch": 0.753271349862259,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.00018921622368409138,
+      "loss": 4.5446,
+      "step": 87500
+    },
+    {
+      "epoch": 0.7537017906336089,
+      "grad_norm": 1.8984375,
+      "learning_rate": 0.00018920398252738374,
+      "loss": 4.7755,
+      "step": 87550
+    },
+    {
+      "epoch": 0.7541322314049587,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.00018919173482329573,
+      "loss": 4.3477,
+      "step": 87600
+    },
+    {
+      "epoch": 0.7545626721763086,
+      "grad_norm": 1.3359375,
+      "learning_rate": 0.00018917948057272627,
+      "loss": 4.348,
+      "step": 87650
+    },
+    {
+      "epoch": 0.7549931129476584,
+      "grad_norm": 2.21875,
+      "learning_rate": 0.0001891672197765748,
+      "loss": 4.7291,
+      "step": 87700
+    },
+    {
+      "epoch": 0.7554235537190083,
+      "grad_norm": 2.0,
+      "learning_rate": 0.00018915495243574125,
+      "loss": 4.8584,
+      "step": 87750
+    },
+    {
+      "epoch": 0.7558539944903582,
+      "grad_norm": 1.40625,
+      "learning_rate": 0.00018914267855112604,
+      "loss": 4.7376,
+      "step": 87800
+    },
+    {
+      "epoch": 0.756284435261708,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.00018913039812363,
+      "loss": 4.7964,
+      "step": 87850
+    },
+    {
+      "epoch": 0.7567148760330579,
+      "grad_norm": 1.921875,
+      "learning_rate": 0.0001891181111541545,
+      "loss": 4.977,
+      "step": 87900
+    },
+    {
+      "epoch": 0.7571453168044077,
+      "grad_norm": 0.66015625,
+      "learning_rate": 0.0001891058176436014,
+      "loss": 4.6962,
+      "step": 87950
+    },
+    {
+      "epoch": 0.7575757575757576,
+      "grad_norm": 1.796875,
+      "learning_rate": 0.000189093517592873,
+      "loss": 4.6742,
+      "step": 88000
+    },
+    {
+      "epoch": 0.7580061983471075,
+      "grad_norm": 3.078125,
+      "learning_rate": 0.0001890812110028721,
+      "loss": 4.6078,
+      "step": 88050
+    },
+    {
+      "epoch": 0.7584366391184573,
+      "grad_norm": 3.0,
+      "learning_rate": 0.00018906889787450196,
+      "loss": 5.1462,
+      "step": 88100
+    },
+    {
+      "epoch": 0.7588670798898072,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00018905657820866636,
+      "loss": 4.8582,
+      "step": 88150
+    },
+    {
+      "epoch": 0.7592975206611571,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.00018904425200626953,
+      "loss": 4.7233,
+      "step": 88200
+    },
+    {
+      "epoch": 0.7597279614325069,
+      "grad_norm": 1.8515625,
+      "learning_rate": 0.0001890319192682162,
+      "loss": 4.5037,
+      "step": 88250
+    },
+    {
+      "epoch": 0.7601584022038568,
+      "grad_norm": 1.90625,
+      "learning_rate": 0.00018901957999541152,
+      "loss": 4.9128,
+      "step": 88300
+    },
+    {
+      "epoch": 0.7605888429752066,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.00018900723418876125,
+      "loss": 5.3711,
+      "step": 88350
+    },
+    {
+      "epoch": 0.7610192837465565,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.00018899488184917146,
+      "loss": 4.6903,
+      "step": 88400
+    },
+    {
+      "epoch": 0.7614497245179064,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.00018898252297754883,
+      "loss": 5.1068,
+      "step": 88450
+    },
+    {
+      "epoch": 0.7618801652892562,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.00018897015757480046,
+      "loss": 4.8245,
+      "step": 88500
+    },
+    {
+      "epoch": 0.7623106060606061,
+      "grad_norm": 4.125,
+      "learning_rate": 0.00018895778564183394,
+      "loss": 4.9601,
+      "step": 88550
+    },
+    {
+      "epoch": 0.7627410468319559,
+      "grad_norm": 1.6171875,
+      "learning_rate": 0.00018894540717955734,
+      "loss": 4.9908,
+      "step": 88600
+    },
+    {
+      "epoch": 0.7631714876033058,
+      "grad_norm": 1.5,
+      "learning_rate": 0.00018893302218887925,
+      "loss": 4.387,
+      "step": 88650
+    },
+    {
+      "epoch": 0.7636019283746557,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00018892063067070866,
+      "loss": 4.7879,
+      "step": 88700
+    },
+    {
+      "epoch": 0.7640323691460055,
+      "grad_norm": 1.4375,
+      "learning_rate": 0.00018890823262595512,
+      "loss": 4.1621,
+      "step": 88750
+    },
+    {
+      "epoch": 0.7644628099173554,
+      "grad_norm": 1.4375,
+      "learning_rate": 0.00018889582805552856,
+      "loss": 4.7262,
+      "step": 88800
+    },
+    {
+      "epoch": 0.7648932506887053,
+      "grad_norm": 2.875,
+      "learning_rate": 0.0001888834169603395,
+      "loss": 4.8107,
+      "step": 88850
+    },
+    {
+      "epoch": 0.7653236914600551,
+      "grad_norm": 2.140625,
+      "learning_rate": 0.0001888709993412989,
+      "loss": 4.5427,
+      "step": 88900
+    },
+    {
+      "epoch": 0.765754132231405,
+      "grad_norm": 3.09375,
+      "learning_rate": 0.0001888585751993181,
+      "loss": 4.7586,
+      "step": 88950
+    },
+    {
+      "epoch": 0.7661845730027548,
+      "grad_norm": 1.46875,
+      "learning_rate": 0.00018884614453530912,
+      "loss": 4.5743,
+      "step": 89000
+    },
+    {
+      "epoch": 0.7666150137741047,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.0001888337073501843,
+      "loss": 4.961,
+      "step": 89050
+    },
+    {
+      "epoch": 0.7670454545454546,
+      "grad_norm": 3.0,
+      "learning_rate": 0.00018882126364485645,
+      "loss": 4.7848,
+      "step": 89100
+    },
+    {
+      "epoch": 0.7674758953168044,
+      "grad_norm": 3.109375,
+      "learning_rate": 0.00018880881342023896,
+      "loss": 4.4636,
+      "step": 89150
+    },
+    {
+      "epoch": 0.7679063360881543,
+      "grad_norm": 3.671875,
+      "learning_rate": 0.00018879635667724563,
+      "loss": 4.5567,
+      "step": 89200
+    },
+    {
+      "epoch": 0.7683367768595041,
+      "grad_norm": 3.375,
+      "learning_rate": 0.00018878389341679078,
+      "loss": 4.6852,
+      "step": 89250
+    },
+    {
+      "epoch": 0.768767217630854,
+      "grad_norm": 1.5,
+      "learning_rate": 0.0001887714236397892,
+      "loss": 4.824,
+      "step": 89300
+    },
+    {
+      "epoch": 0.7691976584022039,
+      "grad_norm": 1.4296875,
+      "learning_rate": 0.0001887589473471561,
+      "loss": 4.7838,
+      "step": 89350
+    },
+    {
+      "epoch": 0.7696280991735537,
+      "grad_norm": 1.859375,
+      "learning_rate": 0.00018874646453980725,
+      "loss": 4.5263,
+      "step": 89400
+    },
+    {
+      "epoch": 0.7700585399449036,
+      "grad_norm": 3.125,
+      "learning_rate": 0.00018873397521865882,
+      "loss": 4.9256,
+      "step": 89450
+    },
+    {
+      "epoch": 0.7704889807162535,
+      "grad_norm": 1.3359375,
+      "learning_rate": 0.00018872147938462752,
+      "loss": 4.6703,
+      "step": 89500
+    },
+    {
+      "epoch": 0.7709194214876033,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.00018870897703863055,
+      "loss": 4.8892,
+      "step": 89550
+    },
+    {
+      "epoch": 0.7713498622589532,
+      "grad_norm": 1.859375,
+      "learning_rate": 0.00018869646818158554,
+      "loss": 5.0574,
+      "step": 89600
+    },
+    {
+      "epoch": 0.771780303030303,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.00018868395281441055,
+      "loss": 5.0371,
+      "step": 89650
+    },
+    {
+      "epoch": 0.7722107438016529,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.0001886714309380243,
+      "loss": 4.862,
+      "step": 89700
+    },
+    {
+      "epoch": 0.7726411845730028,
+      "grad_norm": 1.171875,
+      "learning_rate": 0.00018865890255334572,
+      "loss": 4.4832,
+      "step": 89750
+    },
+    {
+      "epoch": 0.7730716253443526,
+      "grad_norm": 1.3671875,
+      "learning_rate": 0.0001886463676612945,
+      "loss": 4.6223,
+      "step": 89800
+    },
+    {
+      "epoch": 0.7735020661157025,
+      "grad_norm": 1.4609375,
+      "learning_rate": 0.0001886338262627906,
+      "loss": 4.6485,
+      "step": 89850
+    },
+    {
+      "epoch": 0.7739325068870524,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00018862127835875457,
+      "loss": 4.7803,
+      "step": 89900
+    },
+    {
+      "epoch": 0.7743629476584022,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.00018860872395010738,
+      "loss": 4.8075,
+      "step": 89950
+    },
+    {
+      "epoch": 0.7747933884297521,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.00018859616303777048,
+      "loss": 4.6491,
+      "step": 90000
+    },
+    {
+      "epoch": 0.7747933884297521,
+      "eval_loss": 5.349773406982422,
+      "eval_runtime": 21.7904,
+      "eval_samples_per_second": 29.371,
+      "eval_steps_per_second": 14.685,
+      "eval_tts_loss": 6.816362142143376,
+      "step": 90000
+    },
+    {
+      "epoch": 0.7752238292011019,
+      "grad_norm": 3.125,
+      "learning_rate": 0.00018858359562266586,
+      "loss": 4.8141,
+      "step": 90050
+    },
+    {
+      "epoch": 0.7756542699724518,
+      "grad_norm": 1.125,
+      "learning_rate": 0.0001885710217057159,
+      "loss": 4.6386,
+      "step": 90100
+    },
+    {
+      "epoch": 0.7760847107438017,
+      "grad_norm": 3.21875,
+      "learning_rate": 0.00018855844128784353,
+      "loss": 4.6615,
+      "step": 90150
+    },
+    {
+      "epoch": 0.7765151515151515,
+      "grad_norm": 2.140625,
+      "learning_rate": 0.00018854585436997207,
+      "loss": 4.7206,
+      "step": 90200
+    },
+    {
+      "epoch": 0.7769455922865014,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00018853326095302544,
+      "loss": 4.9931,
+      "step": 90250
+    },
+    {
+      "epoch": 0.7773760330578512,
+      "grad_norm": 2.421875,
+      "learning_rate": 0.00018852066103792795,
+      "loss": 4.7669,
+      "step": 90300
+    },
+    {
+      "epoch": 0.7778064738292011,
+      "grad_norm": 1.7109375,
+      "learning_rate": 0.00018850805462560435,
+      "loss": 4.9046,
+      "step": 90350
+    },
+    {
+      "epoch": 0.778236914600551,
+      "grad_norm": 1.265625,
+      "learning_rate": 0.00018849544171698003,
+      "loss": 4.4504,
+      "step": 90400
+    },
+    {
+      "epoch": 0.7786673553719008,
+      "grad_norm": 1.4453125,
+      "learning_rate": 0.00018848282231298064,
+      "loss": 4.3823,
+      "step": 90450
+    },
+    {
+      "epoch": 0.7790977961432507,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.00018847019641453247,
+      "loss": 4.8484,
+      "step": 90500
+    },
+    {
+      "epoch": 0.7795282369146006,
+      "grad_norm": 2.625,
+      "learning_rate": 0.00018845756402256227,
+      "loss": 4.7266,
+      "step": 90550
+    },
+    {
+      "epoch": 0.7799586776859504,
+      "grad_norm": 2.9375,
+      "learning_rate": 0.00018844492513799718,
+      "loss": 4.9104,
+      "step": 90600
+    },
+    {
+      "epoch": 0.7803891184573003,
+      "grad_norm": 1.9453125,
+      "learning_rate": 0.00018843227976176485,
+      "loss": 4.8199,
+      "step": 90650
+    },
+    {
+      "epoch": 0.7808195592286501,
+      "grad_norm": 2.140625,
+      "learning_rate": 0.00018841962789479348,
+      "loss": 4.5509,
+      "step": 90700
+    },
+    {
+      "epoch": 0.78125,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.00018840696953801165,
+      "loss": 5.1658,
+      "step": 90750
+    },
+    {
+      "epoch": 0.7816804407713499,
+      "grad_norm": 2.21875,
+      "learning_rate": 0.00018839430469234846,
+      "loss": 4.6426,
+      "step": 90800
+    },
+    {
+      "epoch": 0.7821108815426997,
+      "grad_norm": 1.7109375,
+      "learning_rate": 0.0001883816333587335,
+      "loss": 4.529,
+      "step": 90850
+    },
+    {
+      "epoch": 0.7825413223140496,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00018836895553809685,
+      "loss": 4.9304,
+      "step": 90900
+    },
+    {
+      "epoch": 0.7829717630853994,
+      "grad_norm": 2.921875,
+      "learning_rate": 0.00018835627123136895,
+      "loss": 4.5608,
+      "step": 90950
+    },
+    {
+      "epoch": 0.7834022038567493,
+      "grad_norm": 3.15625,
+      "learning_rate": 0.00018834358043948087,
+      "loss": 4.2306,
+      "step": 91000
+    },
+    {
+      "epoch": 0.7838326446280992,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00018833088316336404,
+      "loss": 4.4561,
+      "step": 91050
+    },
+    {
+      "epoch": 0.784263085399449,
+      "grad_norm": 2.375,
+      "learning_rate": 0.0001883181794039504,
+      "loss": 4.9889,
+      "step": 91100
+    },
+    {
+      "epoch": 0.7846935261707989,
+      "grad_norm": 3.0,
+      "learning_rate": 0.00018830546916217246,
+      "loss": 5.0307,
+      "step": 91150
+    },
+    {
+      "epoch": 0.7851239669421488,
+      "grad_norm": 2.0,
+      "learning_rate": 0.00018829275243896308,
+      "loss": 4.514,
+      "step": 91200
+    },
+    {
+      "epoch": 0.7855544077134986,
+      "grad_norm": 2.375,
+      "learning_rate": 0.00018828002923525563,
+      "loss": 4.6886,
+      "step": 91250
+    },
+    {
+      "epoch": 0.7859848484848485,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00018826729955198396,
+      "loss": 5.1398,
+      "step": 91300
+    },
+    {
+      "epoch": 0.7864152892561983,
+      "grad_norm": 1.7578125,
+      "learning_rate": 0.00018825456339008242,
+      "loss": 4.6828,
+      "step": 91350
+    },
+    {
+      "epoch": 0.7868457300275482,
+      "grad_norm": 4.0,
+      "learning_rate": 0.00018824182075048582,
+      "loss": 4.8537,
+      "step": 91400
+    },
+    {
+      "epoch": 0.7872761707988981,
+      "grad_norm": 3.046875,
+      "learning_rate": 0.00018822907163412947,
+      "loss": 4.3437,
+      "step": 91450
+    },
+    {
+      "epoch": 0.7877066115702479,
+      "grad_norm": 1.78125,
+      "learning_rate": 0.00018821631604194904,
+      "loss": 4.4522,
+      "step": 91500
+    },
+    {
+      "epoch": 0.7881370523415978,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.00018820355397488082,
+      "loss": 4.9165,
+      "step": 91550
+    },
+    {
+      "epoch": 0.7885674931129476,
+      "grad_norm": 3.203125,
+      "learning_rate": 0.0001881907854338615,
+      "loss": 4.5211,
+      "step": 91600
+    },
+    {
+      "epoch": 0.7889979338842975,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.00018817801041982833,
+      "loss": 4.9461,
+      "step": 91650
+    },
+    {
+      "epoch": 0.7894283746556474,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.0001881652289337189,
+      "loss": 4.9684,
+      "step": 91700
+    },
+    {
+      "epoch": 0.7898588154269972,
+      "grad_norm": 2.21875,
+      "learning_rate": 0.00018815244097647132,
+      "loss": 4.8717,
+      "step": 91750
+    },
+    {
+      "epoch": 0.7902892561983471,
+      "grad_norm": 1.9375,
+      "learning_rate": 0.00018813964654902426,
+      "loss": 4.6948,
+      "step": 91800
+    },
+    {
+      "epoch": 0.790719696969697,
+      "grad_norm": 3.40625,
+      "learning_rate": 0.00018812684565231678,
+      "loss": 4.9747,
+      "step": 91850
+    },
+    {
+      "epoch": 0.7911501377410468,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00018811403828728843,
+      "loss": 4.1586,
+      "step": 91900
+    },
+    {
+      "epoch": 0.7915805785123967,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.00018810122445487925,
+      "loss": 4.6596,
+      "step": 91950
+    },
+    {
+      "epoch": 0.7920110192837465,
+      "grad_norm": 2.296875,
+      "learning_rate": 0.00018808840415602976,
+      "loss": 4.6828,
+      "step": 92000
+    },
+    {
+      "epoch": 0.7924414600550964,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.00018807557739168097,
+      "loss": 4.9194,
+      "step": 92050
+    },
+    {
+      "epoch": 0.7928719008264463,
+      "grad_norm": 1.5,
+      "learning_rate": 0.00018806274416277425,
+      "loss": 4.586,
+      "step": 92100
+    },
+    {
+      "epoch": 0.7933023415977961,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.0001880499044702516,
+      "loss": 4.4376,
+      "step": 92150
+    },
+    {
+      "epoch": 0.793732782369146,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.00018803705831505537,
+      "loss": 4.6227,
+      "step": 92200
+    },
+    {
+      "epoch": 0.7941632231404959,
+      "grad_norm": 1.875,
+      "learning_rate": 0.00018802420569812852,
+      "loss": 4.7545,
+      "step": 92250
+    },
+    {
+      "epoch": 0.7945936639118457,
+      "grad_norm": 1.7265625,
+      "learning_rate": 0.00018801134662041436,
+      "loss": 4.5774,
+      "step": 92300
+    },
+    {
+      "epoch": 0.7950241046831956,
+      "grad_norm": 3.5,
+      "learning_rate": 0.0001879984810828567,
+      "loss": 4.6669,
+      "step": 92350
+    },
+    {
+      "epoch": 0.7954545454545454,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.00018798560908639983,
+      "loss": 4.9228,
+      "step": 92400
+    },
+    {
+      "epoch": 0.7958849862258953,
+      "grad_norm": 1.71875,
+      "learning_rate": 0.00018797273063198862,
+      "loss": 4.5457,
+      "step": 92450
+    },
+    {
+      "epoch": 0.7963154269972452,
+      "grad_norm": 1.7421875,
+      "learning_rate": 0.00018795984572056825,
+      "loss": 5.0458,
+      "step": 92500
+    },
+    {
+      "epoch": 0.796745867768595,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.0001879469543530844,
+      "loss": 4.5707,
+      "step": 92550
+    },
+    {
+      "epoch": 0.7971763085399449,
+      "grad_norm": 4.25,
+      "learning_rate": 0.00018793405653048337,
+      "loss": 4.8589,
+      "step": 92600
+    },
+    {
+      "epoch": 0.7976067493112947,
+      "grad_norm": 3.328125,
+      "learning_rate": 0.00018792115225371177,
+      "loss": 4.4774,
+      "step": 92650
+    },
+    {
+      "epoch": 0.7980371900826446,
+      "grad_norm": 3.03125,
+      "learning_rate": 0.0001879082415237168,
+      "loss": 4.7471,
+      "step": 92700
+    },
+    {
+      "epoch": 0.7984676308539945,
+      "grad_norm": 1.984375,
+      "learning_rate": 0.00018789532434144602,
+      "loss": 4.5363,
+      "step": 92750
+    },
+    {
+      "epoch": 0.7988980716253443,
+      "grad_norm": 3.203125,
+      "learning_rate": 0.00018788240070784752,
+      "loss": 4.6892,
+      "step": 92800
+    },
+    {
+      "epoch": 0.7993285123966942,
+      "grad_norm": 4.5,
+      "learning_rate": 0.0001878694706238699,
+      "loss": 4.4629,
+      "step": 92850
+    },
+    {
+      "epoch": 0.7997589531680441,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.0001878565340904622,
+      "loss": 3.75,
+      "step": 92900
+    },
+    {
+      "epoch": 0.8001893939393939,
+      "grad_norm": 2.0,
+      "learning_rate": 0.00018784359110857394,
+      "loss": 4.4475,
+      "step": 92950
+    },
+    {
+      "epoch": 0.8006198347107438,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.0001878306416791551,
+      "loss": 4.9981,
+      "step": 93000
+    },
+    {
+      "epoch": 0.8006198347107438,
+      "eval_loss": 5.340872764587402,
+      "eval_runtime": 21.7484,
+      "eval_samples_per_second": 29.428,
+      "eval_steps_per_second": 14.714,
+      "eval_tts_loss": 6.779757897928123,
+      "step": 93000
+    },
+    {
+      "epoch": 0.8010502754820936,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.00018781768580315608,
+      "loss": 4.7714,
+      "step": 93050
+    },
+    {
+      "epoch": 0.8014807162534435,
+      "grad_norm": 1.875,
+      "learning_rate": 0.0001878047234815279,
+      "loss": 4.4299,
+      "step": 93100
+    },
+    {
+      "epoch": 0.8019111570247934,
+      "grad_norm": 3.421875,
+      "learning_rate": 0.00018779175471522194,
+      "loss": 4.516,
+      "step": 93150
+    },
+    {
+      "epoch": 0.8023415977961432,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.00018777877950519007,
+      "loss": 4.9042,
+      "step": 93200
+    },
+    {
+      "epoch": 0.8027720385674931,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.00018776579785238467,
+      "loss": 4.8926,
+      "step": 93250
+    },
+    {
+      "epoch": 0.8032024793388429,
+      "grad_norm": 1.78125,
+      "learning_rate": 0.0001877528097577585,
+      "loss": 4.9889,
+      "step": 93300
+    },
+    {
+      "epoch": 0.8036329201101928,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.0001877398152222649,
+      "loss": 4.6047,
+      "step": 93350
+    },
+    {
+      "epoch": 0.8040633608815427,
+      "grad_norm": 3.40625,
+      "learning_rate": 0.00018772681424685764,
+      "loss": 4.7777,
+      "step": 93400
+    },
+    {
+      "epoch": 0.8044938016528925,
+      "grad_norm": 0.671875,
+      "learning_rate": 0.00018771380683249098,
+      "loss": 4.6952,
+      "step": 93450
+    },
+    {
+      "epoch": 0.8049242424242424,
+      "grad_norm": 2.0,
+      "learning_rate": 0.00018770079298011962,
+      "loss": 4.8907,
+      "step": 93500
+    },
+    {
+      "epoch": 0.8053546831955923,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.00018768777269069875,
+      "loss": 4.865,
+      "step": 93550
+    },
+    {
+      "epoch": 0.8057851239669421,
+      "grad_norm": 1.28125,
+      "learning_rate": 0.00018767474596518403,
+      "loss": 4.768,
+      "step": 93600
+    },
+    {
+      "epoch": 0.806215564738292,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.00018766171280453156,
+      "loss": 4.7857,
+      "step": 93650
+    },
+    {
+      "epoch": 0.8066460055096418,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.000187648673209698,
+      "loss": 5.0161,
+      "step": 93700
+    },
+    {
+      "epoch": 0.8070764462809917,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.00018763562718164043,
+      "loss": 4.7486,
+      "step": 93750
+    },
+    {
+      "epoch": 0.8075068870523416,
+      "grad_norm": 1.6640625,
+      "learning_rate": 0.00018762257472131638,
+      "loss": 4.986,
+      "step": 93800
+    },
+    {
+      "epoch": 0.8079373278236914,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.00018760951582968383,
+      "loss": 5.0439,
+      "step": 93850
+    },
+    {
+      "epoch": 0.8083677685950413,
+      "grad_norm": 1.203125,
+      "learning_rate": 0.00018759645050770135,
+      "loss": 4.2721,
+      "step": 93900
+    },
+    {
+      "epoch": 0.8087982093663911,
+      "grad_norm": 3.296875,
+      "learning_rate": 0.00018758337875632782,
+      "loss": 4.8746,
+      "step": 93950
+    },
+    {
+      "epoch": 0.809228650137741,
+      "grad_norm": 3.59375,
+      "learning_rate": 0.00018757030057652279,
+      "loss": 4.8101,
+      "step": 94000
+    },
+    {
+      "epoch": 0.8096590909090909,
+      "grad_norm": 4.09375,
+      "learning_rate": 0.00018755721596924607,
+      "loss": 4.863,
+      "step": 94050
+    },
+    {
+      "epoch": 0.8100895316804407,
+      "grad_norm": 3.1875,
+      "learning_rate": 0.0001875441249354581,
+      "loss": 4.6973,
+      "step": 94100
+    },
+    {
+      "epoch": 0.8105199724517906,
+      "grad_norm": 4.53125,
+      "learning_rate": 0.00018753102747611972,
+      "loss": 4.6561,
+      "step": 94150
+    },
+    {
+      "epoch": 0.8109504132231405,
+      "grad_norm": 1.8203125,
+      "learning_rate": 0.00018751792359219226,
+      "loss": 4.9213,
+      "step": 94200
+    },
+    {
+      "epoch": 0.8113808539944903,
+      "grad_norm": 3.078125,
+      "learning_rate": 0.00018750481328463748,
+      "loss": 4.6864,
+      "step": 94250
+    },
+    {
+      "epoch": 0.8118112947658402,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.00018749169655441769,
+      "loss": 5.0092,
+      "step": 94300
+    },
+    {
+      "epoch": 0.81224173553719,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.0001874785734024956,
+      "loss": 4.976,
+      "step": 94350
+    },
+    {
+      "epoch": 0.8126721763085399,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.00018746544382983445,
+      "loss": 4.7885,
+      "step": 94400
+    },
+    {
+      "epoch": 0.8131026170798898,
+      "grad_norm": 3.578125,
+      "learning_rate": 0.00018745230783739794,
+      "loss": 4.9978,
+      "step": 94450
+    },
+    {
+      "epoch": 0.8135330578512396,
+      "grad_norm": 3.4375,
+      "learning_rate": 0.00018743916542615015,
+      "loss": 4.7951,
+      "step": 94500
+    },
+    {
+      "epoch": 0.8139634986225895,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00018742601659705577,
+      "loss": 4.7587,
+      "step": 94550
+    },
+    {
+      "epoch": 0.8143939393939394,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00018741286135107984,
+      "loss": 4.5513,
+      "step": 94600
+    },
+    {
+      "epoch": 0.8148243801652892,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.00018739969968918803,
+      "loss": 4.7726,
+      "step": 94650
+    },
+    {
+      "epoch": 0.8152548209366391,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.00018738653161234624,
+      "loss": 4.5777,
+      "step": 94700
+    },
+    {
+      "epoch": 0.8156852617079889,
+      "grad_norm": 1.609375,
+      "learning_rate": 0.00018737335712152109,
+      "loss": 4.7884,
+      "step": 94750
+    },
+    {
+      "epoch": 0.8161157024793388,
+      "grad_norm": 1.203125,
+      "learning_rate": 0.0001873601762176795,
+      "loss": 5.1044,
+      "step": 94800
+    },
+    {
+      "epoch": 0.8165461432506887,
+      "grad_norm": 2.578125,
+      "learning_rate": 0.00018734698890178893,
+      "loss": 5.3587,
+      "step": 94850
+    },
+    {
+      "epoch": 0.8169765840220385,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.0001873337951748173,
+      "loss": 4.839,
+      "step": 94900
+    },
+    {
+      "epoch": 0.8174070247933884,
+      "grad_norm": 1.296875,
+      "learning_rate": 0.00018732059503773302,
+      "loss": 4.2681,
+      "step": 94950
+    },
+    {
+      "epoch": 0.8178374655647382,
+      "grad_norm": 4.34375,
+      "learning_rate": 0.00018730738849150494,
+      "loss": 5.0213,
+      "step": 95000
+    },
+    {
+      "epoch": 0.8182679063360881,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.0001872941755371024,
+      "loss": 4.5184,
+      "step": 95050
+    },
+    {
+      "epoch": 0.818698347107438,
+      "grad_norm": 1.8046875,
+      "learning_rate": 0.00018728095617549518,
+      "loss": 4.903,
+      "step": 95100
+    },
+    {
+      "epoch": 0.8191287878787878,
+      "grad_norm": 1.203125,
+      "learning_rate": 0.00018726773040765354,
+      "loss": 4.7532,
+      "step": 95150
+    },
+    {
+      "epoch": 0.8195592286501377,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.00018725449823454827,
+      "loss": 4.7824,
+      "step": 95200
+    },
+    {
+      "epoch": 0.8199896694214877,
+      "grad_norm": 1.4765625,
+      "learning_rate": 0.0001872412596571506,
+      "loss": 4.8679,
+      "step": 95250
+    },
+    {
+      "epoch": 0.8204201101928374,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00018722801467643214,
+      "loss": 5.0606,
+      "step": 95300
+    },
+    {
+      "epoch": 0.8208505509641874,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.0001872147632933651,
+      "loss": 4.7142,
+      "step": 95350
+    },
+    {
+      "epoch": 0.8212809917355371,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.00018720150550892206,
+      "loss": 5.1643,
+      "step": 95400
+    },
+    {
+      "epoch": 0.821711432506887,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.00018718824132407612,
+      "loss": 4.6024,
+      "step": 95450
+    },
+    {
+      "epoch": 0.822141873278237,
+      "grad_norm": 3.546875,
+      "learning_rate": 0.0001871749707398009,
+      "loss": 4.7573,
+      "step": 95500
+    },
+    {
+      "epoch": 0.8225723140495868,
+      "grad_norm": 1.4921875,
+      "learning_rate": 0.00018716169375707038,
+      "loss": 4.4401,
+      "step": 95550
+    },
+    {
+      "epoch": 0.8230027548209367,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00018714841037685903,
+      "loss": 4.3197,
+      "step": 95600
+    },
+    {
+      "epoch": 0.8234331955922864,
+      "grad_norm": 3.203125,
+      "learning_rate": 0.0001871351206001419,
+      "loss": 4.851,
+      "step": 95650
+    },
+    {
+      "epoch": 0.8238636363636364,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.0001871218244278944,
+      "loss": 4.9826,
+      "step": 95700
+    },
+    {
+      "epoch": 0.8242940771349863,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.0001871085218610924,
+      "loss": 5.0037,
+      "step": 95750
+    },
+    {
+      "epoch": 0.824724517906336,
+      "grad_norm": 1.625,
+      "learning_rate": 0.00018709521290071236,
+      "loss": 4.2328,
+      "step": 95800
+    },
+    {
+      "epoch": 0.825154958677686,
+      "grad_norm": 1.9453125,
+      "learning_rate": 0.00018708189754773104,
+      "loss": 4.2897,
+      "step": 95850
+    },
+    {
+      "epoch": 0.8255853994490359,
+      "grad_norm": 2.0,
+      "learning_rate": 0.00018706857580312583,
+      "loss": 4.7004,
+      "step": 95900
+    },
+    {
+      "epoch": 0.8260158402203857,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00018705524766787448,
+      "loss": 4.8402,
+      "step": 95950
+    },
+    {
+      "epoch": 0.8264462809917356,
+      "grad_norm": 2.296875,
+      "learning_rate": 0.00018704191314295525,
+      "loss": 5.0452,
+      "step": 96000
+    },
+    {
+      "epoch": 0.8264462809917356,
+      "eval_loss": 5.334881782531738,
+      "eval_runtime": 22.0398,
+      "eval_samples_per_second": 29.038,
+      "eval_steps_per_second": 14.519,
+      "eval_tts_loss": 6.7898586608838105,
+      "step": 96000
+    },
+    {
+      "epoch": 0.8268767217630854,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.00018702857222934686,
+      "loss": 4.9609,
+      "step": 96050
+    },
+    {
+      "epoch": 0.8273071625344353,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.00018701522492802855,
+      "loss": 4.6768,
+      "step": 96100
+    },
+    {
+      "epoch": 0.8277376033057852,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.0001870018712399799,
+      "loss": 4.9732,
+      "step": 96150
+    },
+    {
+      "epoch": 0.828168044077135,
+      "grad_norm": 1.8359375,
+      "learning_rate": 0.00018698851116618115,
+      "loss": 5.2723,
+      "step": 96200
+    },
+    {
+      "epoch": 0.8285984848484849,
+      "grad_norm": 1.5546875,
+      "learning_rate": 0.0001869751447076128,
+      "loss": 4.7225,
+      "step": 96250
+    },
+    {
+      "epoch": 0.8290289256198347,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.00018696177186525596,
+      "loss": 4.5981,
+      "step": 96300
+    },
+    {
+      "epoch": 0.8294593663911846,
+      "grad_norm": 1.765625,
+      "learning_rate": 0.00018694839264009222,
+      "loss": 4.6406,
+      "step": 96350
+    },
+    {
+      "epoch": 0.8298898071625345,
+      "grad_norm": 3.34375,
+      "learning_rate": 0.00018693500703310348,
+      "loss": 4.1615,
+      "step": 96400
+    },
+    {
+      "epoch": 0.8303202479338843,
+      "grad_norm": 1.5703125,
+      "learning_rate": 0.00018692161504527225,
+      "loss": 4.7821,
+      "step": 96450
+    },
+    {
+      "epoch": 0.8307506887052342,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.00018690821667758154,
+      "loss": 4.5311,
+      "step": 96500
+    },
+    {
+      "epoch": 0.8311811294765841,
+      "grad_norm": 0.8984375,
+      "learning_rate": 0.0001868948119310147,
+      "loss": 4.3184,
+      "step": 96550
+    },
+    {
+      "epoch": 0.8316115702479339,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.00018688140080655563,
+      "loss": 4.4898,
+      "step": 96600
+    },
+    {
+      "epoch": 0.8320420110192838,
+      "grad_norm": 1.40625,
+      "learning_rate": 0.00018686798330518864,
+      "loss": 4.9426,
+      "step": 96650
+    },
+    {
+      "epoch": 0.8324724517906336,
+      "grad_norm": 1.9375,
+      "learning_rate": 0.00018685455942789861,
+      "loss": 4.8363,
+      "step": 96700
+    },
+    {
+      "epoch": 0.8329028925619835,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.0001868411291756708,
+      "loss": 4.52,
+      "step": 96750
+    },
+    {
+      "epoch": 0.8333333333333334,
+      "grad_norm": 3.046875,
+      "learning_rate": 0.0001868276925494909,
+      "loss": 4.7537,
+      "step": 96800
+    },
+    {
+      "epoch": 0.8337637741046832,
+      "grad_norm": 2.421875,
+      "learning_rate": 0.00018681424955034522,
+      "loss": 4.7974,
+      "step": 96850
+    },
+    {
+      "epoch": 0.8341942148760331,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.00018680080017922042,
+      "loss": 4.9911,
+      "step": 96900
+    },
+    {
+      "epoch": 0.834624655647383,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00018678734443710362,
+      "loss": 4.3628,
+      "step": 96950
+    },
+    {
+      "epoch": 0.8350550964187328,
+      "grad_norm": 0.6328125,
+      "learning_rate": 0.00018677388232498247,
+      "loss": 4.537,
+      "step": 97000
+    },
+    {
+      "epoch": 0.8354855371900827,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.00018676041384384505,
+      "loss": 4.675,
+      "step": 97050
+    },
+    {
+      "epoch": 0.8359159779614325,
+      "grad_norm": 3.0625,
+      "learning_rate": 0.00018674693899467995,
+      "loss": 4.6626,
+      "step": 97100
+    },
+    {
+      "epoch": 0.8363464187327824,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.00018673345777847614,
+      "loss": 4.7018,
+      "step": 97150
+    },
+    {
+      "epoch": 0.8367768595041323,
+      "grad_norm": 3.203125,
+      "learning_rate": 0.00018671997019622315,
+      "loss": 5.0525,
+      "step": 97200
+    },
+    {
+      "epoch": 0.8372073002754821,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.00018670647624891094,
+      "loss": 4.7487,
+      "step": 97250
+    },
+    {
+      "epoch": 0.837637741046832,
+      "grad_norm": 1.703125,
+      "learning_rate": 0.0001866929759375299,
+      "loss": 4.6557,
+      "step": 97300
+    },
+    {
+      "epoch": 0.8380681818181818,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.00018667946926307098,
+      "loss": 4.5398,
+      "step": 97350
+    },
+    {
+      "epoch": 0.8384986225895317,
+      "grad_norm": 3.09375,
+      "learning_rate": 0.0001866659562265255,
+      "loss": 4.7489,
+      "step": 97400
+    },
+    {
+      "epoch": 0.8389290633608816,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00018665243682888528,
+      "loss": 4.9353,
+      "step": 97450
+    },
+    {
+      "epoch": 0.8393595041322314,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.00018663891107114267,
+      "loss": 4.6758,
+      "step": 97500
+    },
+    {
+      "epoch": 0.8397899449035813,
+      "grad_norm": 2.5,
+      "learning_rate": 0.0001866253789542904,
+      "loss": 4.5842,
+      "step": 97550
+    },
+    {
+      "epoch": 0.8402203856749312,
+      "grad_norm": 1.4921875,
+      "learning_rate": 0.00018661184047932166,
+      "loss": 4.4563,
+      "step": 97600
+    },
+    {
+      "epoch": 0.840650826446281,
+      "grad_norm": 1.703125,
+      "learning_rate": 0.0001865982956472302,
+      "loss": 4.583,
+      "step": 97650
+    },
+    {
+      "epoch": 0.8410812672176309,
+      "grad_norm": 3.109375,
+      "learning_rate": 0.00018658474445901015,
+      "loss": 4.5257,
+      "step": 97700
+    },
+    {
+      "epoch": 0.8415117079889807,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00018657118691565617,
+      "loss": 4.6481,
+      "step": 97750
+    },
+    {
+      "epoch": 0.8419421487603306,
+      "grad_norm": 1.7265625,
+      "learning_rate": 0.0001865576230181633,
+      "loss": 4.539,
+      "step": 97800
+    },
+    {
+      "epoch": 0.8423725895316805,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.00018654405276752716,
+      "loss": 4.7829,
+      "step": 97850
+    },
+    {
+      "epoch": 0.8428030303030303,
+      "grad_norm": 1.625,
+      "learning_rate": 0.00018653047616474372,
+      "loss": 4.5292,
+      "step": 97900
+    },
+    {
+      "epoch": 0.8432334710743802,
+      "grad_norm": 1.4453125,
+      "learning_rate": 0.00018651689321080953,
+      "loss": 4.7529,
+      "step": 97950
+    },
+    {
+      "epoch": 0.84366391184573,
+      "grad_norm": 1.7890625,
+      "learning_rate": 0.00018650330390672154,
+      "loss": 4.9715,
+      "step": 98000
+    },
+    {
+      "epoch": 0.8440943526170799,
+      "grad_norm": 2.296875,
+      "learning_rate": 0.00018648970825347713,
+      "loss": 4.2531,
+      "step": 98050
+    },
+    {
+      "epoch": 0.8445247933884298,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.00018647610625207423,
+      "loss": 5.0176,
+      "step": 98100
+    },
+    {
+      "epoch": 0.8449552341597796,
+      "grad_norm": 1.4453125,
+      "learning_rate": 0.00018646249790351118,
+      "loss": 4.4414,
+      "step": 98150
+    },
+    {
+      "epoch": 0.8453856749311295,
+      "grad_norm": 2.375,
+      "learning_rate": 0.00018644888320878684,
+      "loss": 4.302,
+      "step": 98200
+    },
+    {
+      "epoch": 0.8458161157024794,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.00018643526216890045,
+      "loss": 4.5189,
+      "step": 98250
+    },
+    {
+      "epoch": 0.8462465564738292,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.0001864216347848518,
+      "loss": 4.624,
+      "step": 98300
+    },
+    {
+      "epoch": 0.8466769972451791,
+      "grad_norm": 3.703125,
+      "learning_rate": 0.00018640800105764108,
+      "loss": 4.6469,
+      "step": 98350
+    },
+    {
+      "epoch": 0.8471074380165289,
+      "grad_norm": 3.625,
+      "learning_rate": 0.000186394360988269,
+      "loss": 4.6857,
+      "step": 98400
+    },
+    {
+      "epoch": 0.8475378787878788,
+      "grad_norm": 2.625,
+      "learning_rate": 0.0001863807145777367,
+      "loss": 4.8966,
+      "step": 98450
+    },
+    {
+      "epoch": 0.8479683195592287,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.00018636706182704584,
+      "loss": 4.5156,
+      "step": 98500
+    },
+    {
+      "epoch": 0.8483987603305785,
+      "grad_norm": 1.6796875,
+      "learning_rate": 0.00018635340273719843,
+      "loss": 4.7085,
+      "step": 98550
+    },
+    {
+      "epoch": 0.8488292011019284,
+      "grad_norm": 4.40625,
+      "learning_rate": 0.00018633973730919706,
+      "loss": 4.7127,
+      "step": 98600
+    },
+    {
+      "epoch": 0.8492596418732782,
+      "grad_norm": 2.625,
+      "learning_rate": 0.00018632606554404475,
+      "loss": 4.4613,
+      "step": 98650
+    },
+    {
+      "epoch": 0.8496900826446281,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.00018631238744274496,
+      "loss": 5.0779,
+      "step": 98700
+    },
+    {
+      "epoch": 0.850120523415978,
+      "grad_norm": 3.53125,
+      "learning_rate": 0.0001862987030063016,
+      "loss": 4.7219,
+      "step": 98750
+    },
+    {
+      "epoch": 0.8505509641873278,
+      "grad_norm": 1.859375,
+      "learning_rate": 0.00018628501223571917,
+      "loss": 4.5248,
+      "step": 98800
+    },
+    {
+      "epoch": 0.8509814049586777,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.00018627131513200245,
+      "loss": 4.5576,
+      "step": 98850
+    },
+    {
+      "epoch": 0.8514118457300276,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.00018625761169615684,
+      "loss": 4.5564,
+      "step": 98900
+    },
+    {
+      "epoch": 0.8518422865013774,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.00018624390192918813,
+      "loss": 4.5104,
+      "step": 98950
+    },
+    {
+      "epoch": 0.8522727272727273,
+      "grad_norm": 1.3515625,
+      "learning_rate": 0.00018623018583210254,
+      "loss": 5.0488,
+      "step": 99000
+    },
+    {
+      "epoch": 0.8522727272727273,
+      "eval_loss": 5.325535774230957,
+      "eval_runtime": 21.8544,
+      "eval_samples_per_second": 29.285,
+      "eval_steps_per_second": 14.642,
+      "eval_tts_loss": 6.811281706750041,
+      "step": 99000
+    },
+    {
+      "epoch": 0.8527031680440771,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.00018621646340590684,
+      "loss": 5.008,
+      "step": 99050
+    },
+    {
+      "epoch": 0.853133608815427,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.00018620273465160824,
+      "loss": 4.7662,
+      "step": 99100
+    },
+    {
+      "epoch": 0.8535640495867769,
+      "grad_norm": 3.984375,
+      "learning_rate": 0.00018618899957021441,
+      "loss": 4.8999,
+      "step": 99150
+    },
+    {
+      "epoch": 0.8539944903581267,
+      "grad_norm": 1.6015625,
+      "learning_rate": 0.00018617525816273342,
+      "loss": 4.7202,
+      "step": 99200
+    },
+    {
+      "epoch": 0.8544249311294766,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.0001861615104301739,
+      "loss": 4.3764,
+      "step": 99250
+    },
+    {
+      "epoch": 0.8548553719008265,
+      "grad_norm": 4.25,
+      "learning_rate": 0.0001861477563735449,
+      "loss": 4.6332,
+      "step": 99300
+    },
+    {
+      "epoch": 0.8552858126721763,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00018613399599385594,
+      "loss": 4.6435,
+      "step": 99350
+    },
+    {
+      "epoch": 0.8557162534435262,
+      "grad_norm": 2.625,
+      "learning_rate": 0.000186120229292117,
+      "loss": 5.0863,
+      "step": 99400
+    },
+    {
+      "epoch": 0.856146694214876,
+      "grad_norm": 1.328125,
+      "learning_rate": 0.0001861064562693385,
+      "loss": 4.4357,
+      "step": 99450
+    },
+    {
+      "epoch": 0.8565771349862259,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.00018609267692653138,
+      "loss": 4.6186,
+      "step": 99500
+    },
+    {
+      "epoch": 0.8570075757575758,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.000186078891264707,
+      "loss": 4.4482,
+      "step": 99550
+    },
+    {
+      "epoch": 0.8574380165289256,
+      "grad_norm": 2.21875,
+      "learning_rate": 0.00018606509928487724,
+      "loss": 4.771,
+      "step": 99600
+    },
+    {
+      "epoch": 0.8578684573002755,
+      "grad_norm": 3.625,
+      "learning_rate": 0.00018605130098805432,
+      "loss": 4.4148,
+      "step": 99650
+    },
+    {
+      "epoch": 0.8582988980716253,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.0001860374963752511,
+      "loss": 4.5078,
+      "step": 99700
+    },
+    {
+      "epoch": 0.8587293388429752,
+      "grad_norm": 0.8203125,
+      "learning_rate": 0.00018602368544748073,
+      "loss": 4.7904,
+      "step": 99750
+    },
+    {
+      "epoch": 0.8591597796143251,
+      "grad_norm": 1.65625,
+      "learning_rate": 0.0001860098682057569,
+      "loss": 4.6289,
+      "step": 99800
+    },
+    {
+      "epoch": 0.8595902203856749,
+      "grad_norm": 1.9453125,
+      "learning_rate": 0.00018599604465109387,
+      "loss": 4.9563,
+      "step": 99850
+    },
+    {
+      "epoch": 0.8600206611570248,
+      "grad_norm": 3.046875,
+      "learning_rate": 0.00018598221478450614,
+      "loss": 4.7274,
+      "step": 99900
+    },
+    {
+      "epoch": 0.8604511019283747,
+      "grad_norm": 2.0,
+      "learning_rate": 0.00018596837860700887,
+      "loss": 4.5151,
+      "step": 99950
+    },
+    {
+      "epoch": 0.8608815426997245,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00018595453611961756,
+      "loss": 4.7865,
+      "step": 100000
+    },
+    {
+      "epoch": 0.8613119834710744,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.00018594068732334822,
+      "loss": 4.7826,
+      "step": 100050
+    },
+    {
+      "epoch": 0.8617424242424242,
+      "grad_norm": 1.765625,
+      "learning_rate": 0.00018592683221921736,
+      "loss": 4.6698,
+      "step": 100100
+    },
+    {
+      "epoch": 0.8621728650137741,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.0001859129708082419,
+      "loss": 4.7061,
+      "step": 100150
+    },
+    {
+      "epoch": 0.862603305785124,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.0001858991030914392,
+      "loss": 4.3987,
+      "step": 100200
+    },
+    {
+      "epoch": 0.8630337465564738,
+      "grad_norm": 1.65625,
+      "learning_rate": 0.00018588522906982715,
+      "loss": 4.8183,
+      "step": 100250
+    },
+    {
+      "epoch": 0.8634641873278237,
+      "grad_norm": 5.0,
+      "learning_rate": 0.0001858713487444241,
+      "loss": 4.6674,
+      "step": 100300
+    },
+    {
+      "epoch": 0.8638946280991735,
+      "grad_norm": 2.875,
+      "learning_rate": 0.00018585746211624877,
+      "loss": 4.9843,
+      "step": 100350
+    },
+    {
+      "epoch": 0.8643250688705234,
+      "grad_norm": 1.296875,
+      "learning_rate": 0.00018584356918632046,
+      "loss": 4.2949,
+      "step": 100400
+    },
+    {
+      "epoch": 0.8647555096418733,
+      "grad_norm": 3.109375,
+      "learning_rate": 0.0001858296699556589,
+      "loss": 4.3828,
+      "step": 100450
+    },
+    {
+      "epoch": 0.8651859504132231,
+      "grad_norm": 3.296875,
+      "learning_rate": 0.0001858157644252842,
+      "loss": 5.1659,
+      "step": 100500
+    },
+    {
+      "epoch": 0.865616391184573,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.00018580185259621702,
+      "loss": 4.993,
+      "step": 100550
+    },
+    {
+      "epoch": 0.8660468319559229,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00018578793446947852,
+      "loss": 4.7621,
+      "step": 100600
+    },
+    {
+      "epoch": 0.8664772727272727,
+      "grad_norm": 1.765625,
+      "learning_rate": 0.00018577401004609017,
+      "loss": 4.7898,
+      "step": 100650
+    },
+    {
+      "epoch": 0.8669077134986226,
+      "grad_norm": 3.53125,
+      "learning_rate": 0.00018576007932707405,
+      "loss": 4.0649,
+      "step": 100700
+    },
+    {
+      "epoch": 0.8673381542699724,
+      "grad_norm": 1.3046875,
+      "learning_rate": 0.0001857461423134526,
+      "loss": 4.8609,
+      "step": 100750
+    },
+    {
+      "epoch": 0.8677685950413223,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.0001857321990062488,
+      "loss": 5.0485,
+      "step": 100800
+    },
+    {
+      "epoch": 0.8681990358126722,
+      "grad_norm": 3.328125,
+      "learning_rate": 0.00018571824940648605,
+      "loss": 4.8174,
+      "step": 100850
+    },
+    {
+      "epoch": 0.868629476584022,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.00018570429351518822,
+      "loss": 4.5981,
+      "step": 100900
+    },
+    {
+      "epoch": 0.8690599173553719,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00018569033133337964,
+      "loss": 4.8208,
+      "step": 100950
+    },
+    {
+      "epoch": 0.8694903581267218,
+      "grad_norm": 3.078125,
+      "learning_rate": 0.00018567636286208515,
+      "loss": 4.4019,
+      "step": 101000
+    },
+    {
+      "epoch": 0.8699207988980716,
+      "grad_norm": 1.3359375,
+      "learning_rate": 0.0001856623881023299,
+      "loss": 4.6871,
+      "step": 101050
+    },
+    {
+      "epoch": 0.8703512396694215,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.0001856484070551397,
+      "loss": 4.8664,
+      "step": 101100
+    },
+    {
+      "epoch": 0.8707816804407713,
+      "grad_norm": 1.3046875,
+      "learning_rate": 0.00018563441972154073,
+      "loss": 4.7675,
+      "step": 101150
+    },
+    {
+      "epoch": 0.8712121212121212,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00018562042610255954,
+      "loss": 4.828,
+      "step": 101200
+    },
+    {
+      "epoch": 0.8716425619834711,
+      "grad_norm": 1.8984375,
+      "learning_rate": 0.00018560642619922332,
+      "loss": 4.5816,
+      "step": 101250
+    },
+    {
+      "epoch": 0.8720730027548209,
+      "grad_norm": 1.703125,
+      "learning_rate": 0.0001855924200125596,
+      "loss": 4.7028,
+      "step": 101300
+    },
+    {
+      "epoch": 0.8725034435261708,
+      "grad_norm": 1.5,
+      "learning_rate": 0.0001855784075435964,
+      "loss": 4.6827,
+      "step": 101350
+    },
+    {
+      "epoch": 0.8729338842975206,
+      "grad_norm": 3.21875,
+      "learning_rate": 0.00018556438879336224,
+      "loss": 4.8132,
+      "step": 101400
+    },
+    {
+      "epoch": 0.8733643250688705,
+      "grad_norm": 4.84375,
+      "learning_rate": 0.000185550363762886,
+      "loss": 4.915,
+      "step": 101450
+    },
+    {
+      "epoch": 0.8737947658402204,
+      "grad_norm": 1.6875,
+      "learning_rate": 0.00018553633245319717,
+      "loss": 4.6117,
+      "step": 101500
+    },
+    {
+      "epoch": 0.8742252066115702,
+      "grad_norm": 2.234375,
+      "learning_rate": 0.00018552229486532555,
+      "loss": 4.8032,
+      "step": 101550
+    },
+    {
+      "epoch": 0.8746556473829201,
+      "grad_norm": 2.140625,
+      "learning_rate": 0.0001855082510003015,
+      "loss": 4.4768,
+      "step": 101600
+    },
+    {
+      "epoch": 0.87508608815427,
+      "grad_norm": 1.78125,
+      "learning_rate": 0.0001854942008591558,
+      "loss": 4.6589,
+      "step": 101650
+    },
+    {
+      "epoch": 0.8755165289256198,
+      "grad_norm": 3.546875,
+      "learning_rate": 0.0001854801444429197,
+      "loss": 4.6911,
+      "step": 101700
+    },
+    {
+      "epoch": 0.8759469696969697,
+      "grad_norm": 2.25,
+      "learning_rate": 0.00018546608175262492,
+      "loss": 4.6856,
+      "step": 101750
+    },
+    {
+      "epoch": 0.8763774104683195,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00018545201278930362,
+      "loss": 4.6395,
+      "step": 101800
+    },
+    {
+      "epoch": 0.8768078512396694,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.00018543793755398843,
+      "loss": 4.8894,
+      "step": 101850
+    },
+    {
+      "epoch": 0.8772382920110193,
+      "grad_norm": 2.9375,
+      "learning_rate": 0.00018542385604771244,
+      "loss": 4.9716,
+      "step": 101900
+    },
+    {
+      "epoch": 0.8776687327823691,
+      "grad_norm": 2.125,
+      "learning_rate": 0.00018540976827150925,
+      "loss": 4.4848,
+      "step": 101950
+    },
+    {
+      "epoch": 0.878099173553719,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.0001853956742264128,
+      "loss": 4.6957,
+      "step": 102000
+    },
+    {
+      "epoch": 0.878099173553719,
+      "eval_loss": 5.325944423675537,
+      "eval_runtime": 21.7557,
+      "eval_samples_per_second": 29.418,
+      "eval_steps_per_second": 14.709,
+      "eval_tts_loss": 6.826836223579188,
+      "step": 102000
+    },
+    {
+      "epoch": 0.8785296143250688,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.0001853815739134576,
+      "loss": 4.9113,
+      "step": 102050
+    },
+    {
+      "epoch": 0.8789600550964187,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.00018536746733367857,
+      "loss": 4.8703,
+      "step": 102100
+    },
+    {
+      "epoch": 0.8793904958677686,
+      "grad_norm": 2.21875,
+      "learning_rate": 0.00018535335448811112,
+      "loss": 4.7105,
+      "step": 102150
+    },
+    {
+      "epoch": 0.8798209366391184,
+      "grad_norm": 1.3359375,
+      "learning_rate": 0.0001853392353777911,
+      "loss": 4.1233,
+      "step": 102200
+    },
+    {
+      "epoch": 0.8802513774104683,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.00018532511000375482,
+      "loss": 4.4242,
+      "step": 102250
+    },
+    {
+      "epoch": 0.8806818181818182,
+      "grad_norm": 1.125,
+      "learning_rate": 0.00018531097836703902,
+      "loss": 4.6817,
+      "step": 102300
+    },
+    {
+      "epoch": 0.881112258953168,
+      "grad_norm": 2.265625,
+      "learning_rate": 0.00018529684046868098,
+      "loss": 4.3535,
+      "step": 102350
+    },
+    {
+      "epoch": 0.8815426997245179,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00018528269630971837,
+      "loss": 4.6127,
+      "step": 102400
+    },
+    {
+      "epoch": 0.8819731404958677,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.0001852685458911893,
+      "loss": 4.9314,
+      "step": 102450
+    },
+    {
+      "epoch": 0.8824035812672176,
+      "grad_norm": 0.94140625,
+      "learning_rate": 0.0001852543892141325,
+      "loss": 4.7783,
+      "step": 102500
+    },
+    {
+      "epoch": 0.8828340220385675,
+      "grad_norm": 3.09375,
+      "learning_rate": 0.0001852402262795869,
+      "loss": 4.746,
+      "step": 102550
+    },
+    {
+      "epoch": 0.8832644628099173,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.0001852260570885921,
+      "loss": 4.5602,
+      "step": 102600
+    },
+    {
+      "epoch": 0.8836949035812672,
+      "grad_norm": 2.234375,
+      "learning_rate": 0.0001852118816421881,
+      "loss": 4.7384,
+      "step": 102650
+    },
+    {
+      "epoch": 0.884125344352617,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.0001851976999414153,
+      "loss": 4.5998,
+      "step": 102700
+    },
+    {
+      "epoch": 0.8845557851239669,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00018518351198731464,
+      "loss": 4.8648,
+      "step": 102750
+    },
+    {
+      "epoch": 0.8849862258953168,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.0001851693177809275,
+      "loss": 4.5639,
+      "step": 102800
+    },
+    {
+      "epoch": 0.8854166666666666,
+      "grad_norm": 1.265625,
+      "learning_rate": 0.00018515511732329566,
+      "loss": 4.6472,
+      "step": 102850
+    },
+    {
+      "epoch": 0.8858471074380165,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.00018514091061546143,
+      "loss": 4.4335,
+      "step": 102900
+    },
+    {
+      "epoch": 0.8862775482093664,
+      "grad_norm": 1.2265625,
+      "learning_rate": 0.00018512669765846754,
+      "loss": 4.8827,
+      "step": 102950
+    },
+    {
+      "epoch": 0.8867079889807162,
+      "grad_norm": 1.4921875,
+      "learning_rate": 0.0001851124784533572,
+      "loss": 4.7631,
+      "step": 103000
+    },
+    {
+      "epoch": 0.8871384297520661,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.0001850982530011741,
+      "loss": 4.5358,
+      "step": 103050
+    },
+    {
+      "epoch": 0.8875688705234159,
+      "grad_norm": 2.625,
+      "learning_rate": 0.00018508402130296226,
+      "loss": 5.1086,
+      "step": 103100
+    },
+    {
+      "epoch": 0.8879993112947658,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.00018506978335976638,
+      "loss": 4.6903,
+      "step": 103150
+    },
+    {
+      "epoch": 0.8884297520661157,
+      "grad_norm": 4.8125,
+      "learning_rate": 0.0001850555391726314,
+      "loss": 4.7302,
+      "step": 103200
+    },
+    {
+      "epoch": 0.8888601928374655,
+      "grad_norm": 2.421875,
+      "learning_rate": 0.00018504128874260282,
+      "loss": 4.8309,
+      "step": 103250
+    },
+    {
+      "epoch": 0.8892906336088154,
+      "grad_norm": 3.1875,
+      "learning_rate": 0.00018502703207072668,
+      "loss": 4.5214,
+      "step": 103300
+    },
+    {
+      "epoch": 0.8897210743801653,
+      "grad_norm": 3.0625,
+      "learning_rate": 0.00018501276915804926,
+      "loss": 4.7718,
+      "step": 103350
+    },
+    {
+      "epoch": 0.8901515151515151,
+      "grad_norm": 1.6953125,
+      "learning_rate": 0.00018499850000561752,
+      "loss": 4.9352,
+      "step": 103400
+    },
+    {
+      "epoch": 0.890581955922865,
+      "grad_norm": 0.65625,
+      "learning_rate": 0.0001849842246144788,
+      "loss": 4.6532,
+      "step": 103450
+    },
+    {
+      "epoch": 0.8910123966942148,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.00018496994298568078,
+      "loss": 4.7838,
+      "step": 103500
+    },
+    {
+      "epoch": 0.8914428374655647,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.0001849556551202718,
+      "loss": 4.6994,
+      "step": 103550
+    },
+    {
+      "epoch": 0.8918732782369146,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.00018494136101930052,
+      "loss": 4.7826,
+      "step": 103600
+    },
+    {
+      "epoch": 0.8923037190082644,
+      "grad_norm": 1.890625,
+      "learning_rate": 0.00018492706068381609,
+      "loss": 4.5258,
+      "step": 103650
+    },
+    {
+      "epoch": 0.8927341597796143,
+      "grad_norm": 2.9375,
+      "learning_rate": 0.00018491275411486812,
+      "loss": 5.2615,
+      "step": 103700
+    },
+    {
+      "epoch": 0.8931646005509641,
+      "grad_norm": 2.25,
+      "learning_rate": 0.0001848984413135067,
+      "loss": 4.8224,
+      "step": 103750
+    },
+    {
+      "epoch": 0.893595041322314,
+      "grad_norm": 1.953125,
+      "learning_rate": 0.00018488412228078236,
+      "loss": 4.4506,
+      "step": 103800
+    },
+    {
+      "epoch": 0.894025482093664,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.00018486979701774607,
+      "loss": 4.6331,
+      "step": 103850
+    },
+    {
+      "epoch": 0.8944559228650137,
+      "grad_norm": 4.03125,
+      "learning_rate": 0.0001848554655254493,
+      "loss": 4.556,
+      "step": 103900
+    },
+    {
+      "epoch": 0.8948863636363636,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.0001848411278049439,
+      "loss": 4.9729,
+      "step": 103950
+    },
+    {
+      "epoch": 0.8953168044077136,
+      "grad_norm": 1.796875,
+      "learning_rate": 0.00018482678385728227,
+      "loss": 4.4746,
+      "step": 104000
+    },
+    {
+      "epoch": 0.8957472451790633,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.0001848124336835172,
+      "loss": 4.6459,
+      "step": 104050
+    },
+    {
+      "epoch": 0.8961776859504132,
+      "grad_norm": 2.421875,
+      "learning_rate": 0.000184798077284702,
+      "loss": 5.2198,
+      "step": 104100
+    },
+    {
+      "epoch": 0.896608126721763,
+      "grad_norm": 2.421875,
+      "learning_rate": 0.00018478371466189034,
+      "loss": 4.6399,
+      "step": 104150
+    },
+    {
+      "epoch": 0.897038567493113,
+      "grad_norm": 2.140625,
+      "learning_rate": 0.0001847693458161365,
+      "loss": 4.7582,
+      "step": 104200
+    },
+    {
+      "epoch": 0.8974690082644629,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.00018475497074849502,
+      "loss": 4.8269,
+      "step": 104250
+    },
+    {
+      "epoch": 0.8978994490358126,
+      "grad_norm": 1.71875,
+      "learning_rate": 0.00018474058946002107,
+      "loss": 4.8294,
+      "step": 104300
+    },
+    {
+      "epoch": 0.8983298898071626,
+      "grad_norm": 1.7890625,
+      "learning_rate": 0.00018472620195177016,
+      "loss": 4.8011,
+      "step": 104350
+    },
+    {
+      "epoch": 0.8987603305785123,
+      "grad_norm": 3.203125,
+      "learning_rate": 0.0001847118082247983,
+      "loss": 4.9976,
+      "step": 104400
+    },
+    {
+      "epoch": 0.8991907713498623,
+      "grad_norm": 0.5859375,
+      "learning_rate": 0.000184697408280162,
+      "loss": 4.667,
+      "step": 104450
+    },
+    {
+      "epoch": 0.8996212121212122,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.00018468300211891818,
+      "loss": 4.4345,
+      "step": 104500
+    },
+    {
+      "epoch": 0.900051652892562,
+      "grad_norm": 3.609375,
+      "learning_rate": 0.00018466858974212417,
+      "loss": 4.7446,
+      "step": 104550
+    },
+    {
+      "epoch": 0.9004820936639119,
+      "grad_norm": 3.03125,
+      "learning_rate": 0.00018465417115083788,
+      "loss": 4.4933,
+      "step": 104600
+    },
+    {
+      "epoch": 0.9009125344352618,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00018463974634611754,
+      "loss": 4.6429,
+      "step": 104650
+    },
+    {
+      "epoch": 0.9013429752066116,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.00018462531532902193,
+      "loss": 4.6575,
+      "step": 104700
+    },
+    {
+      "epoch": 0.9017734159779615,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00018461087810061027,
+      "loss": 4.8047,
+      "step": 104750
+    },
+    {
+      "epoch": 0.9022038567493113,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00018459643466194216,
+      "loss": 4.6825,
+      "step": 104800
+    },
+    {
+      "epoch": 0.9026342975206612,
+      "grad_norm": 1.953125,
+      "learning_rate": 0.0001845819850140778,
+      "loss": 4.1799,
+      "step": 104850
+    },
+    {
+      "epoch": 0.9030647382920111,
+      "grad_norm": 2.625,
+      "learning_rate": 0.0001845675291580777,
+      "loss": 4.3134,
+      "step": 104900
+    },
+    {
+      "epoch": 0.9034951790633609,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.0001845530670950029,
+      "loss": 4.5509,
+      "step": 104950
+    },
+    {
+      "epoch": 0.9039256198347108,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00018453859882591492,
+      "loss": 4.6409,
+      "step": 105000
+    },
+    {
+      "epoch": 0.9039256198347108,
+      "eval_loss": 5.319069862365723,
+      "eval_runtime": 21.8425,
+      "eval_samples_per_second": 29.301,
+      "eval_steps_per_second": 14.65,
+      "eval_tts_loss": 6.835954681379342,
+      "step": 105000
+    },
+    {
+      "epoch": 0.9043560606060606,
+      "grad_norm": 1.96875,
+      "learning_rate": 0.0001845241243518757,
+      "loss": 4.8657,
+      "step": 105050
+    },
+    {
+      "epoch": 0.9047865013774105,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.00018450964367394758,
+      "loss": 4.9152,
+      "step": 105100
+    },
+    {
+      "epoch": 0.9052169421487604,
+      "grad_norm": 3.984375,
+      "learning_rate": 0.00018449515679319343,
+      "loss": 4.8435,
+      "step": 105150
+    },
+    {
+      "epoch": 0.9056473829201102,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.00018448066371067658,
+      "loss": 4.9524,
+      "step": 105200
+    },
+    {
+      "epoch": 0.9060778236914601,
+      "grad_norm": 2.921875,
+      "learning_rate": 0.0001844661644274608,
+      "loss": 5.0342,
+      "step": 105250
+    },
+    {
+      "epoch": 0.90650826446281,
+      "grad_norm": 3.625,
+      "learning_rate": 0.00018445165894461022,
+      "loss": 4.576,
+      "step": 105300
+    },
+    {
+      "epoch": 0.9069387052341598,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00018443714726318965,
+      "loss": 4.5225,
+      "step": 105350
+    },
+    {
+      "epoch": 0.9073691460055097,
+      "grad_norm": 2.234375,
+      "learning_rate": 0.00018442262938426408,
+      "loss": 4.7826,
+      "step": 105400
+    },
+    {
+      "epoch": 0.9077995867768595,
+      "grad_norm": 1.359375,
+      "learning_rate": 0.00018440810530889922,
+      "loss": 4.2423,
+      "step": 105450
+    },
+    {
+      "epoch": 0.9082300275482094,
+      "grad_norm": 1.734375,
+      "learning_rate": 0.000184393575038161,
+      "loss": 4.6326,
+      "step": 105500
+    },
+    {
+      "epoch": 0.9086604683195593,
+      "grad_norm": 1.953125,
+      "learning_rate": 0.00018437903857311592,
+      "loss": 4.9193,
+      "step": 105550
+    },
+    {
+      "epoch": 0.9090909090909091,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.000184364495914831,
+      "loss": 4.759,
+      "step": 105600
+    },
+    {
+      "epoch": 0.909521349862259,
+      "grad_norm": 1.9453125,
+      "learning_rate": 0.00018434994706437359,
+      "loss": 4.8403,
+      "step": 105650
+    },
+    {
+      "epoch": 0.9099517906336089,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00018433539202281153,
+      "loss": 4.8115,
+      "step": 105700
+    },
+    {
+      "epoch": 0.9103822314049587,
+      "grad_norm": 3.8125,
+      "learning_rate": 0.00018432083079121317,
+      "loss": 5.1318,
+      "step": 105750
+    },
+    {
+      "epoch": 0.9108126721763086,
+      "grad_norm": 1.953125,
+      "learning_rate": 0.00018430626337064725,
+      "loss": 4.4289,
+      "step": 105800
+    },
+    {
+      "epoch": 0.9112431129476584,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.00018429168976218297,
+      "loss": 4.851,
+      "step": 105850
+    },
+    {
+      "epoch": 0.9116735537190083,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00018427710996689003,
+      "loss": 4.0955,
+      "step": 105900
+    },
+    {
+      "epoch": 0.9121039944903582,
+      "grad_norm": 2.625,
+      "learning_rate": 0.00018426252398583853,
+      "loss": 4.7545,
+      "step": 105950
+    },
+    {
+      "epoch": 0.912534435261708,
+      "grad_norm": 1.390625,
+      "learning_rate": 0.0001842479318200991,
+      "loss": 4.7372,
+      "step": 106000
+    },
+    {
+      "epoch": 0.9129648760330579,
+      "grad_norm": 0.83984375,
+      "learning_rate": 0.00018423333347074273,
+      "loss": 4.5969,
+      "step": 106050
+    },
+    {
+      "epoch": 0.9133953168044077,
+      "grad_norm": 3.90625,
+      "learning_rate": 0.0001842187289388409,
+      "loss": 4.7007,
+      "step": 106100
+    },
+    {
+      "epoch": 0.9138257575757576,
+      "grad_norm": 3.34375,
+      "learning_rate": 0.00018420411822546554,
+      "loss": 4.5481,
+      "step": 106150
+    },
+    {
+      "epoch": 0.9142561983471075,
+      "grad_norm": 2.296875,
+      "learning_rate": 0.0001841895013316891,
+      "loss": 4.8891,
+      "step": 106200
+    },
+    {
+      "epoch": 0.9146866391184573,
+      "grad_norm": 3.578125,
+      "learning_rate": 0.0001841748782585844,
+      "loss": 4.6893,
+      "step": 106250
+    },
+    {
+      "epoch": 0.9151170798898072,
+      "grad_norm": 3.0,
+      "learning_rate": 0.00018416024900722478,
+      "loss": 4.3488,
+      "step": 106300
+    },
+    {
+      "epoch": 0.9155475206611571,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.0001841456135786839,
+      "loss": 4.8745,
+      "step": 106350
+    },
+    {
+      "epoch": 0.9159779614325069,
+      "grad_norm": 1.875,
+      "learning_rate": 0.00018413097197403605,
+      "loss": 4.272,
+      "step": 106400
+    },
+    {
+      "epoch": 0.9164084022038568,
+      "grad_norm": 3.578125,
+      "learning_rate": 0.00018411632419435586,
+      "loss": 5.0796,
+      "step": 106450
+    },
+    {
+      "epoch": 0.9168388429752066,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.00018410167024071847,
+      "loss": 4.5094,
+      "step": 106500
+    },
+    {
+      "epoch": 0.9172692837465565,
+      "grad_norm": 1.96875,
+      "learning_rate": 0.00018408701011419945,
+      "loss": 4.217,
+      "step": 106550
+    },
+    {
+      "epoch": 0.9176997245179064,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.00018407234381587477,
+      "loss": 4.8997,
+      "step": 106600
+    },
+    {
+      "epoch": 0.9181301652892562,
+      "grad_norm": 2.296875,
+      "learning_rate": 0.00018405767134682094,
+      "loss": 4.6582,
+      "step": 106650
+    },
+    {
+      "epoch": 0.9185606060606061,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.0001840429927081149,
+      "loss": 4.8529,
+      "step": 106700
+    },
+    {
+      "epoch": 0.9189910468319559,
+      "grad_norm": 2.25,
+      "learning_rate": 0.00018402830790083404,
+      "loss": 4.7468,
+      "step": 106750
+    },
+    {
+      "epoch": 0.9194214876033058,
+      "grad_norm": 0.9453125,
+      "learning_rate": 0.00018401361692605615,
+      "loss": 4.274,
+      "step": 106800
+    },
+    {
+      "epoch": 0.9198519283746557,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00018399891978485955,
+      "loss": 4.9797,
+      "step": 106850
+    },
+    {
+      "epoch": 0.9202823691460055,
+      "grad_norm": 3.578125,
+      "learning_rate": 0.00018398421647832294,
+      "loss": 4.5853,
+      "step": 106900
+    },
+    {
+      "epoch": 0.9207128099173554,
+      "grad_norm": 3.921875,
+      "learning_rate": 0.0001839695070075256,
+      "loss": 4.5254,
+      "step": 106950
+    },
+    {
+      "epoch": 0.9211432506887053,
+      "grad_norm": 2.375,
+      "learning_rate": 0.00018395479137354702,
+      "loss": 4.9453,
+      "step": 107000
+    },
+    {
+      "epoch": 0.9215736914600551,
+      "grad_norm": 1.609375,
+      "learning_rate": 0.00018394006957746744,
+      "loss": 4.334,
+      "step": 107050
+    },
+    {
+      "epoch": 0.922004132231405,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00018392534162036734,
+      "loss": 4.6963,
+      "step": 107100
+    },
+    {
+      "epoch": 0.9224345730027548,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.00018391060750332775,
+      "loss": 4.6619,
+      "step": 107150
+    },
+    {
+      "epoch": 0.9228650137741047,
+      "grad_norm": 3.390625,
+      "learning_rate": 0.00018389586722743011,
+      "loss": 4.6646,
+      "step": 107200
+    },
+    {
+      "epoch": 0.9232954545454546,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.0001838811207937563,
+      "loss": 4.9815,
+      "step": 107250
+    },
+    {
+      "epoch": 0.9237258953168044,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.0001838663682033887,
+      "loss": 4.5707,
+      "step": 107300
+    },
+    {
+      "epoch": 0.9241563360881543,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.00018385160945741014,
+      "loss": 5.164,
+      "step": 107350
+    },
+    {
+      "epoch": 0.9245867768595041,
+      "grad_norm": 1.984375,
+      "learning_rate": 0.00018383684455690384,
+      "loss": 3.9016,
+      "step": 107400
+    },
+    {
+      "epoch": 0.925017217630854,
+      "grad_norm": 3.578125,
+      "learning_rate": 0.00018382207350295352,
+      "loss": 4.6926,
+      "step": 107450
+    },
+    {
+      "epoch": 0.9254476584022039,
+      "grad_norm": 3.953125,
+      "learning_rate": 0.00018380729629664336,
+      "loss": 4.6408,
+      "step": 107500
+    },
+    {
+      "epoch": 0.9258780991735537,
+      "grad_norm": 3.984375,
+      "learning_rate": 0.000183792512939058,
+      "loss": 4.909,
+      "step": 107550
+    },
+    {
+      "epoch": 0.9263085399449036,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00018377772343128242,
+      "loss": 4.9433,
+      "step": 107600
+    },
+    {
+      "epoch": 0.9267389807162535,
+      "grad_norm": 3.015625,
+      "learning_rate": 0.00018376292777440222,
+      "loss": 4.8536,
+      "step": 107650
+    },
+    {
+      "epoch": 0.9271694214876033,
+      "grad_norm": 1.640625,
+      "learning_rate": 0.00018374812596950335,
+      "loss": 4.5243,
+      "step": 107700
+    },
+    {
+      "epoch": 0.9275998622589532,
+      "grad_norm": 1.671875,
+      "learning_rate": 0.00018373331801767223,
+      "loss": 4.8739,
+      "step": 107750
+    },
+    {
+      "epoch": 0.928030303030303,
+      "grad_norm": 0.796875,
+      "learning_rate": 0.0001837185039199957,
+      "loss": 4.587,
+      "step": 107800
+    },
+    {
+      "epoch": 0.9284607438016529,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.0001837036836775611,
+      "loss": 4.3264,
+      "step": 107850
+    },
+    {
+      "epoch": 0.9288911845730028,
+      "grad_norm": 1.8671875,
+      "learning_rate": 0.00018368885729145627,
+      "loss": 4.7946,
+      "step": 107900
+    },
+    {
+      "epoch": 0.9293216253443526,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.00018367402476276933,
+      "loss": 4.7081,
+      "step": 107950
+    },
+    {
+      "epoch": 0.9297520661157025,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.00018365918609258898,
+      "loss": 4.5729,
+      "step": 108000
+    },
+    {
+      "epoch": 0.9297520661157025,
+      "eval_loss": 5.310000419616699,
+      "eval_runtime": 22.3217,
+      "eval_samples_per_second": 28.672,
+      "eval_steps_per_second": 14.336,
+      "eval_tts_loss": 6.829048421016426,
+      "step": 108000
+    },
+    {
+      "epoch": 0.9301825068870524,
+      "grad_norm": 1.5078125,
+      "learning_rate": 0.0001836443412820044,
+      "loss": 4.8927,
+      "step": 108050
+    },
+    {
+      "epoch": 0.9306129476584022,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.00018362949033210515,
+      "loss": 4.6305,
+      "step": 108100
+    },
+    {
+      "epoch": 0.9310433884297521,
+      "grad_norm": 2.375,
+      "learning_rate": 0.0001836146332439812,
+      "loss": 5.0496,
+      "step": 108150
+    },
+    {
+      "epoch": 0.9314738292011019,
+      "grad_norm": 3.1875,
+      "learning_rate": 0.00018359977001872308,
+      "loss": 4.814,
+      "step": 108200
+    },
+    {
+      "epoch": 0.9319042699724518,
+      "grad_norm": 1.0390625,
+      "learning_rate": 0.00018358490065742173,
+      "loss": 4.9844,
+      "step": 108250
+    },
+    {
+      "epoch": 0.9323347107438017,
+      "grad_norm": 1.9375,
+      "learning_rate": 0.0001835700251611685,
+      "loss": 4.6898,
+      "step": 108300
+    },
+    {
+      "epoch": 0.9327651515151515,
+      "grad_norm": 2.421875,
+      "learning_rate": 0.00018355514353105525,
+      "loss": 4.841,
+      "step": 108350
+    },
+    {
+      "epoch": 0.9331955922865014,
+      "grad_norm": 2.5,
+      "learning_rate": 0.0001835402557681742,
+      "loss": 4.7741,
+      "step": 108400
+    },
+    {
+      "epoch": 0.9336260330578512,
+      "grad_norm": 1.9921875,
+      "learning_rate": 0.00018352536187361815,
+      "loss": 4.7507,
+      "step": 108450
+    },
+    {
+      "epoch": 0.9340564738292011,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00018351046184848023,
+      "loss": 4.6393,
+      "step": 108500
+    },
+    {
+      "epoch": 0.934486914600551,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.0001834955556938541,
+      "loss": 4.8567,
+      "step": 108550
+    },
+    {
+      "epoch": 0.9349173553719008,
+      "grad_norm": 4.625,
+      "learning_rate": 0.0001834806434108338,
+      "loss": 4.4327,
+      "step": 108600
+    },
+    {
+      "epoch": 0.9353477961432507,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.00018346572500051392,
+      "loss": 4.7422,
+      "step": 108650
+    },
+    {
+      "epoch": 0.9357782369146006,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.0001834508004639894,
+      "loss": 4.1282,
+      "step": 108700
+    },
+    {
+      "epoch": 0.9362086776859504,
+      "grad_norm": 1.8359375,
+      "learning_rate": 0.00018343586980235568,
+      "loss": 4.7681,
+      "step": 108750
+    },
+    {
+      "epoch": 0.9366391184573003,
+      "grad_norm": 1.3984375,
+      "learning_rate": 0.00018342093301670863,
+      "loss": 4.829,
+      "step": 108800
+    },
+    {
+      "epoch": 0.9370695592286501,
+      "grad_norm": 2.375,
+      "learning_rate": 0.00018340599010814457,
+      "loss": 5.1187,
+      "step": 108850
+    },
+    {
+      "epoch": 0.9375,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.00018339104107776029,
+      "loss": 4.7694,
+      "step": 108900
+    },
+    {
+      "epoch": 0.9379304407713499,
+      "grad_norm": 1.78125,
+      "learning_rate": 0.00018337608592665304,
+      "loss": 4.8252,
+      "step": 108950
+    },
+    {
+      "epoch": 0.9383608815426997,
+      "grad_norm": 1.4140625,
+      "learning_rate": 0.0001833611246559204,
+      "loss": 4.726,
+      "step": 109000
+    },
+    {
+      "epoch": 0.9387913223140496,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.00018334615726666063,
+      "loss": 4.6152,
+      "step": 109050
+    },
+    {
+      "epoch": 0.9392217630853994,
+      "grad_norm": 1.9296875,
+      "learning_rate": 0.00018333118375997223,
+      "loss": 4.592,
+      "step": 109100
+    },
+    {
+      "epoch": 0.9396522038567493,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.00018331620413695424,
+      "loss": 4.6736,
+      "step": 109150
+    },
+    {
+      "epoch": 0.9400826446280992,
+      "grad_norm": 3.125,
+      "learning_rate": 0.0001833012183987061,
+      "loss": 4.689,
+      "step": 109200
+    },
+    {
+      "epoch": 0.940513085399449,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00018328622654632777,
+      "loss": 4.704,
+      "step": 109250
+    },
+    {
+      "epoch": 0.9409435261707989,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.00018327122858091964,
+      "loss": 4.7188,
+      "step": 109300
+    },
+    {
+      "epoch": 0.9413739669421488,
+      "grad_norm": 1.84375,
+      "learning_rate": 0.00018325622450358244,
+      "loss": 4.9638,
+      "step": 109350
+    },
+    {
+      "epoch": 0.9418044077134986,
+      "grad_norm": 3.1875,
+      "learning_rate": 0.00018324121431541752,
+      "loss": 4.8302,
+      "step": 109400
+    },
+    {
+      "epoch": 0.9422348484848485,
+      "grad_norm": 3.125,
+      "learning_rate": 0.00018322619801752657,
+      "loss": 4.6308,
+      "step": 109450
+    },
+    {
+      "epoch": 0.9426652892561983,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.00018321117561101175,
+      "loss": 4.7341,
+      "step": 109500
+    },
+    {
+      "epoch": 0.9430957300275482,
+      "grad_norm": 1.6015625,
+      "learning_rate": 0.00018319614709697566,
+      "loss": 5.0833,
+      "step": 109550
+    },
+    {
+      "epoch": 0.9435261707988981,
+      "grad_norm": 1.4453125,
+      "learning_rate": 0.00018318111247652138,
+      "loss": 4.6262,
+      "step": 109600
+    },
+    {
+      "epoch": 0.9439566115702479,
+      "grad_norm": 1.4296875,
+      "learning_rate": 0.0001831660717507524,
+      "loss": 4.7921,
+      "step": 109650
+    },
+    {
+      "epoch": 0.9443870523415978,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00018315102492077272,
+      "loss": 4.4389,
+      "step": 109700
+    },
+    {
+      "epoch": 0.9448174931129476,
+      "grad_norm": 0.9609375,
+      "learning_rate": 0.0001831359719876867,
+      "loss": 4.8367,
+      "step": 109750
+    },
+    {
+      "epoch": 0.9452479338842975,
+      "grad_norm": 3.765625,
+      "learning_rate": 0.0001831209129525992,
+      "loss": 4.8597,
+      "step": 109800
+    },
+    {
+      "epoch": 0.9456783746556474,
+      "grad_norm": 3.234375,
+      "learning_rate": 0.00018310584781661553,
+      "loss": 4.7801,
+      "step": 109850
+    },
+    {
+      "epoch": 0.9461088154269972,
+      "grad_norm": 3.125,
+      "learning_rate": 0.00018309077658084145,
+      "loss": 4.9747,
+      "step": 109900
+    },
+    {
+      "epoch": 0.9465392561983471,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.00018307569924638312,
+      "loss": 4.5881,
+      "step": 109950
+    },
+    {
+      "epoch": 0.946969696969697,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00018306061581434724,
+      "loss": 5.0307,
+      "step": 110000
+    },
+    {
+      "epoch": 0.9474001377410468,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00018304552628584087,
+      "loss": 4.7241,
+      "step": 110050
+    },
+    {
+      "epoch": 0.9478305785123967,
+      "grad_norm": 3.1875,
+      "learning_rate": 0.0001830304306619715,
+      "loss": 4.8156,
+      "step": 110100
+    },
+    {
+      "epoch": 0.9482610192837465,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.0001830153289438472,
+      "loss": 4.7013,
+      "step": 110150
+    },
+    {
+      "epoch": 0.9486914600550964,
+      "grad_norm": 1.203125,
+      "learning_rate": 0.00018300022113257636,
+      "loss": 4.6494,
+      "step": 110200
+    },
+    {
+      "epoch": 0.9491219008264463,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.00018298510722926787,
+      "loss": 4.476,
+      "step": 110250
+    },
+    {
+      "epoch": 0.9495523415977961,
+      "grad_norm": 1.3125,
+      "learning_rate": 0.00018296998723503106,
+      "loss": 4.2874,
+      "step": 110300
+    },
+    {
+      "epoch": 0.949982782369146,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.0001829548611509757,
+      "loss": 4.5065,
+      "step": 110350
+    },
+    {
+      "epoch": 0.9504132231404959,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.00018293972897821204,
+      "loss": 4.6838,
+      "step": 110400
+    },
+    {
+      "epoch": 0.9508436639118457,
+      "grad_norm": 1.9453125,
+      "learning_rate": 0.0001829245907178507,
+      "loss": 4.4963,
+      "step": 110450
+    },
+    {
+      "epoch": 0.9512741046831956,
+      "grad_norm": 1.796875,
+      "learning_rate": 0.00018290944637100278,
+      "loss": 4.9266,
+      "step": 110500
+    },
+    {
+      "epoch": 0.9517045454545454,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00018289429593877993,
+      "loss": 4.4374,
+      "step": 110550
+    },
+    {
+      "epoch": 0.9521349862258953,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.0001828791394222941,
+      "loss": 5.0353,
+      "step": 110600
+    },
+    {
+      "epoch": 0.9525654269972452,
+      "grad_norm": 2.984375,
+      "learning_rate": 0.00018286397682265774,
+      "loss": 4.4734,
+      "step": 110650
+    },
+    {
+      "epoch": 0.952995867768595,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.00018284880814098376,
+      "loss": 4.7412,
+      "step": 110700
+    },
+    {
+      "epoch": 0.9534263085399449,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00018283363337838555,
+      "loss": 4.5173,
+      "step": 110750
+    },
+    {
+      "epoch": 0.9538567493112947,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.00018281845253597686,
+      "loss": 4.4637,
+      "step": 110800
+    },
+    {
+      "epoch": 0.9542871900826446,
+      "grad_norm": 3.03125,
+      "learning_rate": 0.00018280326561487194,
+      "loss": 4.6908,
+      "step": 110850
+    },
+    {
+      "epoch": 0.9547176308539945,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00018278807261618544,
+      "loss": 4.3467,
+      "step": 110900
+    },
+    {
+      "epoch": 0.9551480716253443,
+      "grad_norm": 1.46875,
+      "learning_rate": 0.00018277287354103255,
+      "loss": 5.0616,
+      "step": 110950
+    },
+    {
+      "epoch": 0.9555785123966942,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.00018275766839052888,
+      "loss": 4.7608,
+      "step": 111000
+    },
+    {
+      "epoch": 0.9555785123966942,
+      "eval_loss": 5.300333023071289,
+      "eval_runtime": 21.9596,
+      "eval_samples_per_second": 29.144,
+      "eval_steps_per_second": 14.572,
+      "eval_tts_loss": 6.812118898581518,
+      "step": 111000
+    },
+    {
+      "epoch": 0.9560089531680441,
+      "grad_norm": 3.140625,
+      "learning_rate": 0.0001827424571657904,
+      "loss": 4.9825,
+      "step": 111050
+    },
+    {
+      "epoch": 0.9564393939393939,
+      "grad_norm": 2.25,
+      "learning_rate": 0.00018272723986793355,
+      "loss": 4.8704,
+      "step": 111100
+    },
+    {
+      "epoch": 0.9568698347107438,
+      "grad_norm": 1.625,
+      "learning_rate": 0.0001827120164980753,
+      "loss": 4.8763,
+      "step": 111150
+    },
+    {
+      "epoch": 0.9573002754820936,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.000182696787057333,
+      "loss": 4.807,
+      "step": 111200
+    },
+    {
+      "epoch": 0.9577307162534435,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.00018268155154682446,
+      "loss": 4.9365,
+      "step": 111250
+    },
+    {
+      "epoch": 0.9581611570247934,
+      "grad_norm": 2.125,
+      "learning_rate": 0.00018266630996766794,
+      "loss": 4.5768,
+      "step": 111300
+    },
+    {
+      "epoch": 0.9585915977961432,
+      "grad_norm": 1.46875,
+      "learning_rate": 0.0001826510623209821,
+      "loss": 4.7168,
+      "step": 111350
+    },
+    {
+      "epoch": 0.9590220385674931,
+      "grad_norm": 3.515625,
+      "learning_rate": 0.00018263580860788614,
+      "loss": 4.6275,
+      "step": 111400
+    },
+    {
+      "epoch": 0.9594524793388429,
+      "grad_norm": 1.7734375,
+      "learning_rate": 0.00018262054882949962,
+      "loss": 4.9334,
+      "step": 111450
+    },
+    {
+      "epoch": 0.9598829201101928,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.0001826052829869426,
+      "loss": 4.7279,
+      "step": 111500
+    },
+    {
+      "epoch": 0.9603133608815427,
+      "grad_norm": 2.625,
+      "learning_rate": 0.0001825900110813355,
+      "loss": 4.8701,
+      "step": 111550
+    },
+    {
+      "epoch": 0.9607438016528925,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.00018257473311379929,
+      "loss": 4.9257,
+      "step": 111600
+    },
+    {
+      "epoch": 0.9611742424242424,
+      "grad_norm": 0.8125,
+      "learning_rate": 0.00018255944908545532,
+      "loss": 4.71,
+      "step": 111650
+    },
+    {
+      "epoch": 0.9616046831955923,
+      "grad_norm": 3.015625,
+      "learning_rate": 0.00018254415899742545,
+      "loss": 4.9018,
+      "step": 111700
+    },
+    {
+      "epoch": 0.9620351239669421,
+      "grad_norm": 1.703125,
+      "learning_rate": 0.0001825288628508319,
+      "loss": 4.6962,
+      "step": 111750
+    },
+    {
+      "epoch": 0.962465564738292,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00018251356064679738,
+      "loss": 4.4214,
+      "step": 111800
+    },
+    {
+      "epoch": 0.9628960055096418,
+      "grad_norm": 2.21875,
+      "learning_rate": 0.00018249825238644504,
+      "loss": 5.217,
+      "step": 111850
+    },
+    {
+      "epoch": 0.9633264462809917,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.00018248293807089847,
+      "loss": 4.8202,
+      "step": 111900
+    },
+    {
+      "epoch": 0.9637568870523416,
+      "grad_norm": 3.75,
+      "learning_rate": 0.0001824676177012817,
+      "loss": 4.3443,
+      "step": 111950
+    },
+    {
+      "epoch": 0.9641873278236914,
+      "grad_norm": 3.671875,
+      "learning_rate": 0.00018245229127871926,
+      "loss": 4.6238,
+      "step": 112000
+    },
+    {
+      "epoch": 0.9646177685950413,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00018243695880433602,
+      "loss": 4.5155,
+      "step": 112050
+    },
+    {
+      "epoch": 0.9650482093663911,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00018242162027925737,
+      "loss": 5.1373,
+      "step": 112100
+    },
+    {
+      "epoch": 0.965478650137741,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.00018240627570460913,
+      "loss": 4.6647,
+      "step": 112150
+    },
+    {
+      "epoch": 0.9659090909090909,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.0001823909250815176,
+      "loss": 4.6351,
+      "step": 112200
+    },
+    {
+      "epoch": 0.9663395316804407,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.00018237556841110936,
+      "loss": 4.6715,
+      "step": 112250
+    },
+    {
+      "epoch": 0.9667699724517906,
+      "grad_norm": 3.515625,
+      "learning_rate": 0.00018236020569451167,
+      "loss": 4.8067,
+      "step": 112300
+    },
+    {
+      "epoch": 0.9672004132231405,
+      "grad_norm": 3.40625,
+      "learning_rate": 0.00018234483693285207,
+      "loss": 4.8991,
+      "step": 112350
+    },
+    {
+      "epoch": 0.9676308539944903,
+      "grad_norm": 1.6875,
+      "learning_rate": 0.00018232946212725864,
+      "loss": 4.2828,
+      "step": 112400
+    },
+    {
+      "epoch": 0.9680612947658402,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.0001823140812788598,
+      "loss": 4.6571,
+      "step": 112450
+    },
+    {
+      "epoch": 0.96849173553719,
+      "grad_norm": 1.21875,
+      "learning_rate": 0.00018229869438878452,
+      "loss": 4.5357,
+      "step": 112500
+    },
+    {
+      "epoch": 0.9689221763085399,
+      "grad_norm": 2.21875,
+      "learning_rate": 0.0001822833014581621,
+      "loss": 4.9415,
+      "step": 112550
+    },
+    {
+      "epoch": 0.9693526170798898,
+      "grad_norm": 3.28125,
+      "learning_rate": 0.0001822679024881224,
+      "loss": 4.646,
+      "step": 112600
+    },
+    {
+      "epoch": 0.9697830578512396,
+      "grad_norm": 3.171875,
+      "learning_rate": 0.0001822524974797957,
+      "loss": 4.6626,
+      "step": 112650
+    },
+    {
+      "epoch": 0.9702134986225895,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.00018223708643431264,
+      "loss": 4.8407,
+      "step": 112700
+    },
+    {
+      "epoch": 0.9706439393939394,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.00018222166935280435,
+      "loss": 4.2004,
+      "step": 112750
+    },
+    {
+      "epoch": 0.9710743801652892,
+      "grad_norm": 2.375,
+      "learning_rate": 0.00018220624623640247,
+      "loss": 4.9094,
+      "step": 112800
+    },
+    {
+      "epoch": 0.9715048209366391,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.00018219081708623896,
+      "loss": 4.444,
+      "step": 112850
+    },
+    {
+      "epoch": 0.9719352617079889,
+      "grad_norm": 3.3125,
+      "learning_rate": 0.0001821753819034463,
+      "loss": 4.8196,
+      "step": 112900
+    },
+    {
+      "epoch": 0.9723657024793388,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00018215994068915744,
+      "loss": 4.358,
+      "step": 112950
+    },
+    {
+      "epoch": 0.9727961432506887,
+      "grad_norm": 1.9609375,
+      "learning_rate": 0.0001821444934445057,
+      "loss": 4.9661,
+      "step": 113000
+    },
+    {
+      "epoch": 0.9732265840220385,
+      "grad_norm": 3.125,
+      "learning_rate": 0.00018212904017062491,
+      "loss": 4.8289,
+      "step": 113050
+    },
+    {
+      "epoch": 0.9736570247933884,
+      "grad_norm": 2.578125,
+      "learning_rate": 0.00018211358086864924,
+      "loss": 4.964,
+      "step": 113100
+    },
+    {
+      "epoch": 0.9740874655647382,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.0001820981155397134,
+      "loss": 4.5162,
+      "step": 113150
+    },
+    {
+      "epoch": 0.9745179063360881,
+      "grad_norm": 3.8125,
+      "learning_rate": 0.00018208264418495253,
+      "loss": 4.5643,
+      "step": 113200
+    },
+    {
+      "epoch": 0.974948347107438,
+      "grad_norm": 1.765625,
+      "learning_rate": 0.0001820671668055022,
+      "loss": 4.6342,
+      "step": 113250
+    },
+    {
+      "epoch": 0.9753787878787878,
+      "grad_norm": 2.9375,
+      "learning_rate": 0.00018205168340249838,
+      "loss": 4.9864,
+      "step": 113300
+    },
+    {
+      "epoch": 0.9758092286501377,
+      "grad_norm": 3.90625,
+      "learning_rate": 0.00018203619397707751,
+      "loss": 4.7318,
+      "step": 113350
+    },
+    {
+      "epoch": 0.9762396694214877,
+      "grad_norm": 1.6171875,
+      "learning_rate": 0.00018202069853037655,
+      "loss": 4.8014,
+      "step": 113400
+    },
+    {
+      "epoch": 0.9766701101928374,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.00018200519706353276,
+      "loss": 5.043,
+      "step": 113450
+    },
+    {
+      "epoch": 0.9771005509641874,
+      "grad_norm": 3.171875,
+      "learning_rate": 0.00018198968957768398,
+      "loss": 5.0049,
+      "step": 113500
+    },
+    {
+      "epoch": 0.9775309917355371,
+      "grad_norm": 2.625,
+      "learning_rate": 0.0001819741760739684,
+      "loss": 4.7359,
+      "step": 113550
+    },
+    {
+      "epoch": 0.977961432506887,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.0001819586565535246,
+      "loss": 4.519,
+      "step": 113600
+    },
+    {
+      "epoch": 0.978391873278237,
+      "grad_norm": 1.6953125,
+      "learning_rate": 0.0001819431310174918,
+      "loss": 4.6648,
+      "step": 113650
+    },
+    {
+      "epoch": 0.9788223140495868,
+      "grad_norm": 1.6484375,
+      "learning_rate": 0.00018192759946700945,
+      "loss": 4.6725,
+      "step": 113700
+    },
+    {
+      "epoch": 0.9792527548209367,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.0001819120619032176,
+      "loss": 4.8179,
+      "step": 113750
+    },
+    {
+      "epoch": 0.9796831955922864,
+      "grad_norm": 0.76953125,
+      "learning_rate": 0.00018189651832725665,
+      "loss": 4.6351,
+      "step": 113800
+    },
+    {
+      "epoch": 0.9801136363636364,
+      "grad_norm": 1.578125,
+      "learning_rate": 0.00018188096874026746,
+      "loss": 4.5547,
+      "step": 113850
+    },
+    {
+      "epoch": 0.9805440771349863,
+      "grad_norm": 3.4375,
+      "learning_rate": 0.00018186541314339132,
+      "loss": 4.9056,
+      "step": 113900
+    },
+    {
+      "epoch": 0.980974517906336,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.00018184985153777,
+      "loss": 4.4375,
+      "step": 113950
+    },
+    {
+      "epoch": 0.981404958677686,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.00018183428392454568,
+      "loss": 4.8116,
+      "step": 114000
+    },
+    {
+      "epoch": 0.981404958677686,
+      "eval_loss": 5.294703006744385,
+      "eval_runtime": 21.9006,
+      "eval_samples_per_second": 29.223,
+      "eval_steps_per_second": 14.611,
+      "eval_tts_loss": 6.886583222421584,
+      "step": 114000
+    },
+    {
+      "epoch": 0.9818353994490359,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.000181818710304861,
+      "loss": 4.935,
+      "step": 114050
+    },
+    {
+      "epoch": 0.9822658402203857,
+      "grad_norm": 3.265625,
+      "learning_rate": 0.00018180313067985902,
+      "loss": 4.5252,
+      "step": 114100
+    },
+    {
+      "epoch": 0.9826962809917356,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.00018178754505068323,
+      "loss": 4.5541,
+      "step": 114150
+    },
+    {
+      "epoch": 0.9831267217630854,
+      "grad_norm": 1.9765625,
+      "learning_rate": 0.0001817719534184776,
+      "loss": 4.732,
+      "step": 114200
+    },
+    {
+      "epoch": 0.9835571625344353,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.00018175635578438656,
+      "loss": 4.6602,
+      "step": 114250
+    },
+    {
+      "epoch": 0.9839876033057852,
+      "grad_norm": 1.890625,
+      "learning_rate": 0.00018174075214955488,
+      "loss": 4.8676,
+      "step": 114300
+    },
+    {
+      "epoch": 0.984418044077135,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.00018172514251512789,
+      "loss": 4.43,
+      "step": 114350
+    },
+    {
+      "epoch": 0.9848484848484849,
+      "grad_norm": 1.8984375,
+      "learning_rate": 0.00018170952688225124,
+      "loss": 4.8476,
+      "step": 114400
+    },
+    {
+      "epoch": 0.9852789256198347,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00018169390525207116,
+      "loss": 4.5011,
+      "step": 114450
+    },
+    {
+      "epoch": 0.9857093663911846,
+      "grad_norm": 1.9609375,
+      "learning_rate": 0.0001816782776257342,
+      "loss": 4.4709,
+      "step": 114500
+    },
+    {
+      "epoch": 0.9861398071625345,
+      "grad_norm": 2.625,
+      "learning_rate": 0.00018166264400438738,
+      "loss": 5.0821,
+      "step": 114550
+    },
+    {
+      "epoch": 0.9865702479338843,
+      "grad_norm": 0.57421875,
+      "learning_rate": 0.00018164700438917815,
+      "loss": 5.1266,
+      "step": 114600
+    },
+    {
+      "epoch": 0.9870006887052342,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.00018163135878125454,
+      "loss": 4.5975,
+      "step": 114650
+    },
+    {
+      "epoch": 0.9874311294765841,
+      "grad_norm": 3.453125,
+      "learning_rate": 0.0001816157071817648,
+      "loss": 4.861,
+      "step": 114700
+    },
+    {
+      "epoch": 0.9878615702479339,
+      "grad_norm": 1.921875,
+      "learning_rate": 0.00018160004959185777,
+      "loss": 4.4884,
+      "step": 114750
+    },
+    {
+      "epoch": 0.9882920110192838,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.00018158438601268264,
+      "loss": 4.5202,
+      "step": 114800
+    },
+    {
+      "epoch": 0.9887224517906336,
+      "grad_norm": 1.796875,
+      "learning_rate": 0.0001815687164453891,
+      "loss": 4.0955,
+      "step": 114850
+    },
+    {
+      "epoch": 0.9891528925619835,
+      "grad_norm": 3.828125,
+      "learning_rate": 0.0001815530408911273,
+      "loss": 4.4795,
+      "step": 114900
+    },
+    {
+      "epoch": 0.9895833333333334,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00018153735935104776,
+      "loss": 5.0941,
+      "step": 114950
+    },
+    {
+      "epoch": 0.9900137741046832,
+      "grad_norm": 1.515625,
+      "learning_rate": 0.00018152167182630146,
+      "loss": 4.6825,
+      "step": 115000
+    },
+    {
+      "epoch": 0.9904442148760331,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.00018150597831803985,
+      "loss": 4.8007,
+      "step": 115050
+    },
+    {
+      "epoch": 0.990874655647383,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.00018149027882741482,
+      "loss": 4.787,
+      "step": 115100
+    },
+    {
+      "epoch": 0.9913050964187328,
+      "grad_norm": 1.3359375,
+      "learning_rate": 0.00018147457335557864,
+      "loss": 4.6287,
+      "step": 115150
+    },
+    {
+      "epoch": 0.9917355371900827,
+      "grad_norm": 3.75,
+      "learning_rate": 0.00018145886190368407,
+      "loss": 4.7001,
+      "step": 115200
+    },
+    {
+      "epoch": 0.9921659779614325,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.0001814431444728843,
+      "loss": 4.709,
+      "step": 115250
+    },
+    {
+      "epoch": 0.9925964187327824,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.00018142742106433296,
+      "loss": 4.9991,
+      "step": 115300
+    },
+    {
+      "epoch": 0.9930268595041323,
+      "grad_norm": 2.921875,
+      "learning_rate": 0.0001814116916791841,
+      "loss": 4.2271,
+      "step": 115350
+    },
+    {
+      "epoch": 0.9934573002754821,
+      "grad_norm": 1.3359375,
+      "learning_rate": 0.00018139595631859226,
+      "loss": 4.7682,
+      "step": 115400
+    },
+    {
+      "epoch": 0.993887741046832,
+      "grad_norm": 3.5,
+      "learning_rate": 0.0001813802149837123,
+      "loss": 4.4572,
+      "step": 115450
+    },
+    {
+      "epoch": 0.9943181818181818,
+      "grad_norm": 2.375,
+      "learning_rate": 0.00018136446767569973,
+      "loss": 4.6668,
+      "step": 115500
+    },
+    {
+      "epoch": 0.9947486225895317,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.00018134871439571024,
+      "loss": 4.6613,
+      "step": 115550
+    },
+    {
+      "epoch": 0.9951790633608816,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00018133295514490013,
+      "loss": 4.4614,
+      "step": 115600
+    },
+    {
+      "epoch": 0.9956095041322314,
+      "grad_norm": 3.4375,
+      "learning_rate": 0.00018131718992442613,
+      "loss": 4.9357,
+      "step": 115650
+    },
+    {
+      "epoch": 0.9960399449035813,
+      "grad_norm": 3.5,
+      "learning_rate": 0.00018130141873544536,
+      "loss": 4.2927,
+      "step": 115700
+    },
+    {
+      "epoch": 0.9964703856749312,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.00018128564157911535,
+      "loss": 4.3704,
+      "step": 115750
+    },
+    {
+      "epoch": 0.996900826446281,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00018126985845659416,
+      "loss": 4.3264,
+      "step": 115800
+    },
+    {
+      "epoch": 0.9973312672176309,
+      "grad_norm": 1.921875,
+      "learning_rate": 0.00018125406936904018,
+      "loss": 5.1619,
+      "step": 115850
+    },
+    {
+      "epoch": 0.9977617079889807,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.00018123827431761237,
+      "loss": 4.6115,
+      "step": 115900
+    },
+    {
+      "epoch": 0.9981921487603306,
+      "grad_norm": 4.0625,
+      "learning_rate": 0.00018122247330347002,
+      "loss": 4.5484,
+      "step": 115950
+    },
+    {
+      "epoch": 0.9986225895316805,
+      "grad_norm": 1.4609375,
+      "learning_rate": 0.00018120666632777285,
+      "loss": 4.9301,
+      "step": 116000
+    },
+    {
+      "epoch": 0.9990530303030303,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.00018119085339168112,
+      "loss": 4.8622,
+      "step": 116050
+    },
+    {
+      "epoch": 0.9994834710743802,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.00018117503449635542,
+      "loss": 5.0018,
+      "step": 116100
+    },
+    {
+      "epoch": 0.99991391184573,
+      "grad_norm": 1.2578125,
+      "learning_rate": 0.00018115920964295685,
+      "loss": 4.6004,
+      "step": 116150
+    },
+    {
+      "epoch": 1.0003443526170799,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.0001811433788326469,
+      "loss": 4.6662,
+      "step": 116200
+    },
+    {
+      "epoch": 1.0007747933884297,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.00018112754206658753,
+      "loss": 4.9016,
+      "step": 116250
+    },
+    {
+      "epoch": 1.0012052341597797,
+      "grad_norm": 2.25,
+      "learning_rate": 0.00018111169934594115,
+      "loss": 4.6042,
+      "step": 116300
+    },
+    {
+      "epoch": 1.0016356749311295,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.00018109585067187052,
+      "loss": 4.9825,
+      "step": 116350
+    },
+    {
+      "epoch": 1.0020661157024793,
+      "grad_norm": 4.34375,
+      "learning_rate": 0.00018107999604553894,
+      "loss": 4.514,
+      "step": 116400
+    },
+    {
+      "epoch": 1.0024965564738293,
+      "grad_norm": 0.875,
+      "learning_rate": 0.0001810641354681101,
+      "loss": 4.6825,
+      "step": 116450
+    },
+    {
+      "epoch": 1.002926997245179,
+      "grad_norm": 2.921875,
+      "learning_rate": 0.00018104826894074812,
+      "loss": 4.6185,
+      "step": 116500
+    },
+    {
+      "epoch": 1.0033574380165289,
+      "grad_norm": 3.171875,
+      "learning_rate": 0.00018103239646461762,
+      "loss": 4.6123,
+      "step": 116550
+    },
+    {
+      "epoch": 1.003787878787879,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.0001810165180408835,
+      "loss": 4.8008,
+      "step": 116600
+    },
+    {
+      "epoch": 1.0042183195592287,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.00018100063367071134,
+      "loss": 4.9084,
+      "step": 116650
+    },
+    {
+      "epoch": 1.0046487603305785,
+      "grad_norm": 2.140625,
+      "learning_rate": 0.00018098474335526688,
+      "loss": 4.4487,
+      "step": 116700
+    },
+    {
+      "epoch": 1.0050792011019283,
+      "grad_norm": 3.828125,
+      "learning_rate": 0.0001809688470957166,
+      "loss": 4.6371,
+      "step": 116750
+    },
+    {
+      "epoch": 1.0055096418732783,
+      "grad_norm": 3.390625,
+      "learning_rate": 0.00018095294489322704,
+      "loss": 4.7429,
+      "step": 116800
+    },
+    {
+      "epoch": 1.005940082644628,
+      "grad_norm": 1.8359375,
+      "learning_rate": 0.00018093703674896557,
+      "loss": 4.2184,
+      "step": 116850
+    },
+    {
+      "epoch": 1.0063705234159779,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.0001809211226640997,
+      "loss": 4.8171,
+      "step": 116900
+    },
+    {
+      "epoch": 1.006800964187328,
+      "grad_norm": 4.78125,
+      "learning_rate": 0.00018090520263979752,
+      "loss": 4.2607,
+      "step": 116950
+    },
+    {
+      "epoch": 1.0072314049586777,
+      "grad_norm": 3.71875,
+      "learning_rate": 0.0001808892766772276,
+      "loss": 4.5191,
+      "step": 117000
+    },
+    {
+      "epoch": 1.0072314049586777,
+      "eval_loss": 5.293093204498291,
+      "eval_runtime": 21.7119,
+      "eval_samples_per_second": 29.477,
+      "eval_steps_per_second": 14.738,
+      "eval_tts_loss": 6.988506331182041,
+      "step": 117000
+    },
+    {
+      "epoch": 1.0076618457300275,
+      "grad_norm": 3.5,
+      "learning_rate": 0.00018087334477755879,
+      "loss": 4.5956,
+      "step": 117050
+    },
+    {
+      "epoch": 1.0080922865013775,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00018085740694196044,
+      "loss": 4.5534,
+      "step": 117100
+    },
+    {
+      "epoch": 1.0085227272727273,
+      "grad_norm": 3.6875,
+      "learning_rate": 0.00018084146317160242,
+      "loss": 4.6093,
+      "step": 117150
+    },
+    {
+      "epoch": 1.008953168044077,
+      "grad_norm": 1.828125,
+      "learning_rate": 0.00018082551346765494,
+      "loss": 4.579,
+      "step": 117200
+    },
+    {
+      "epoch": 1.009383608815427,
+      "grad_norm": 1.453125,
+      "learning_rate": 0.00018080955783128866,
+      "loss": 4.7456,
+      "step": 117250
+    },
+    {
+      "epoch": 1.009814049586777,
+      "grad_norm": 5.59375,
+      "learning_rate": 0.0001807935962636747,
+      "loss": 4.3437,
+      "step": 117300
+    },
+    {
+      "epoch": 1.0102444903581267,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00018077762876598464,
+      "loss": 4.7591,
+      "step": 117350
+    },
+    {
+      "epoch": 1.0106749311294765,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00018076165533939038,
+      "loss": 4.666,
+      "step": 117400
+    },
+    {
+      "epoch": 1.0111053719008265,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.0001807456759850644,
+      "loss": 4.5435,
+      "step": 117450
+    },
+    {
+      "epoch": 1.0115358126721763,
+      "grad_norm": 1.203125,
+      "learning_rate": 0.00018072969070417953,
+      "loss": 4.6361,
+      "step": 117500
+    },
+    {
+      "epoch": 1.011966253443526,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.00018071369949790907,
+      "loss": 4.9383,
+      "step": 117550
+    },
+    {
+      "epoch": 1.012396694214876,
+      "grad_norm": 3.859375,
+      "learning_rate": 0.00018069770236742668,
+      "loss": 4.5216,
+      "step": 117600
+    },
+    {
+      "epoch": 1.012827134986226,
+      "grad_norm": 1.6015625,
+      "learning_rate": 0.00018068169931390661,
+      "loss": 4.8076,
+      "step": 117650
+    },
+    {
+      "epoch": 1.0132575757575757,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.0001806656903385234,
+      "loss": 4.3783,
+      "step": 117700
+    },
+    {
+      "epoch": 1.0136880165289257,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00018064967544245206,
+      "loss": 4.2773,
+      "step": 117750
+    },
+    {
+      "epoch": 1.0141184573002755,
+      "grad_norm": 1.359375,
+      "learning_rate": 0.00018063365462686804,
+      "loss": 4.6129,
+      "step": 117800
+    },
+    {
+      "epoch": 1.0145488980716253,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00018061762789294732,
+      "loss": 4.8391,
+      "step": 117850
+    },
+    {
+      "epoch": 1.0149793388429753,
+      "grad_norm": 1.2734375,
+      "learning_rate": 0.00018060159524186612,
+      "loss": 4.5374,
+      "step": 117900
+    },
+    {
+      "epoch": 1.015409779614325,
+      "grad_norm": 2.375,
+      "learning_rate": 0.0001805855566748013,
+      "loss": 4.0976,
+      "step": 117950
+    },
+    {
+      "epoch": 1.015840220385675,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.0001805695121929299,
+      "loss": 4.9281,
+      "step": 118000
+    },
+    {
+      "epoch": 1.0162706611570247,
+      "grad_norm": 1.53125,
+      "learning_rate": 0.00018055346179742977,
+      "loss": 4.6582,
+      "step": 118050
+    },
+    {
+      "epoch": 1.0167011019283747,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00018053740548947878,
+      "loss": 4.776,
+      "step": 118100
+    },
+    {
+      "epoch": 1.0171315426997245,
+      "grad_norm": 1.6328125,
+      "learning_rate": 0.00018052134327025554,
+      "loss": 4.6826,
+      "step": 118150
+    },
+    {
+      "epoch": 1.0175619834710743,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.00018050527514093894,
+      "loss": 4.5394,
+      "step": 118200
+    },
+    {
+      "epoch": 1.0179924242424243,
+      "grad_norm": 0.9921875,
+      "learning_rate": 0.0001804892011027084,
+      "loss": 4.7749,
+      "step": 118250
+    },
+    {
+      "epoch": 1.018422865013774,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.00018047312115674363,
+      "loss": 4.6088,
+      "step": 118300
+    },
+    {
+      "epoch": 1.018853305785124,
+      "grad_norm": 1.171875,
+      "learning_rate": 0.0001804570353042249,
+      "loss": 5.0518,
+      "step": 118350
+    },
+    {
+      "epoch": 1.019283746556474,
+      "grad_norm": 3.96875,
+      "learning_rate": 0.00018044094354633292,
+      "loss": 4.9322,
+      "step": 118400
+    },
+    {
+      "epoch": 1.0197141873278237,
+      "grad_norm": 1.84375,
+      "learning_rate": 0.00018042484588424876,
+      "loss": 4.6429,
+      "step": 118450
+    },
+    {
+      "epoch": 1.0201446280991735,
+      "grad_norm": 2.296875,
+      "learning_rate": 0.0001804087423191539,
+      "loss": 4.4183,
+      "step": 118500
+    },
+    {
+      "epoch": 1.0205750688705235,
+      "grad_norm": 3.6875,
+      "learning_rate": 0.0001803926328522304,
+      "loss": 4.5771,
+      "step": 118550
+    },
+    {
+      "epoch": 1.0210055096418733,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.00018037651748466061,
+      "loss": 4.3772,
+      "step": 118600
+    },
+    {
+      "epoch": 1.021435950413223,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.00018036039621762738,
+      "loss": 4.8166,
+      "step": 118650
+    },
+    {
+      "epoch": 1.021866391184573,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.00018034426905231396,
+      "loss": 4.7993,
+      "step": 118700
+    },
+    {
+      "epoch": 1.022296831955923,
+      "grad_norm": 1.671875,
+      "learning_rate": 0.00018032813598990407,
+      "loss": 4.9895,
+      "step": 118750
+    },
+    {
+      "epoch": 1.0227272727272727,
+      "grad_norm": 5.375,
+      "learning_rate": 0.00018031199703158178,
+      "loss": 4.2557,
+      "step": 118800
+    },
+    {
+      "epoch": 1.0231577134986225,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00018029585217853175,
+      "loss": 4.3125,
+      "step": 118850
+    },
+    {
+      "epoch": 1.0235881542699725,
+      "grad_norm": 3.609375,
+      "learning_rate": 0.0001802797014319389,
+      "loss": 4.3647,
+      "step": 118900
+    },
+    {
+      "epoch": 1.0240185950413223,
+      "grad_norm": 4.09375,
+      "learning_rate": 0.0001802635447929887,
+      "loss": 4.7287,
+      "step": 118950
+    },
+    {
+      "epoch": 1.024449035812672,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.00018024738226286703,
+      "loss": 4.3629,
+      "step": 119000
+    },
+    {
+      "epoch": 1.0248794765840221,
+      "grad_norm": 1.484375,
+      "learning_rate": 0.00018023121384276014,
+      "loss": 5.0144,
+      "step": 119050
+    },
+    {
+      "epoch": 1.025309917355372,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.0001802150395338548,
+      "loss": 4.2512,
+      "step": 119100
+    },
+    {
+      "epoch": 1.0257403581267217,
+      "grad_norm": 1.8828125,
+      "learning_rate": 0.0001801988593373381,
+      "loss": 4.4129,
+      "step": 119150
+    },
+    {
+      "epoch": 1.0261707988980717,
+      "grad_norm": 0.796875,
+      "learning_rate": 0.0001801826732543977,
+      "loss": 4.3623,
+      "step": 119200
+    },
+    {
+      "epoch": 1.0266012396694215,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.0001801664812862216,
+      "loss": 4.8789,
+      "step": 119250
+    },
+    {
+      "epoch": 1.0270316804407713,
+      "grad_norm": 3.796875,
+      "learning_rate": 0.00018015028343399826,
+      "loss": 4.4223,
+      "step": 119300
+    },
+    {
+      "epoch": 1.027462121212121,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00018013407969891655,
+      "loss": 4.5086,
+      "step": 119350
+    },
+    {
+      "epoch": 1.0278925619834711,
+      "grad_norm": 3.90625,
+      "learning_rate": 0.00018011787008216583,
+      "loss": 4.6112,
+      "step": 119400
+    },
+    {
+      "epoch": 1.028323002754821,
+      "grad_norm": 1.28125,
+      "learning_rate": 0.0001801016545849358,
+      "loss": 4.4403,
+      "step": 119450
+    },
+    {
+      "epoch": 1.0287534435261707,
+      "grad_norm": 3.125,
+      "learning_rate": 0.00018008543320841667,
+      "loss": 4.5919,
+      "step": 119500
+    },
+    {
+      "epoch": 1.0291838842975207,
+      "grad_norm": 2.25,
+      "learning_rate": 0.00018006920595379903,
+      "loss": 4.5554,
+      "step": 119550
+    },
+    {
+      "epoch": 1.0296143250688705,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.000180052972822274,
+      "loss": 4.7866,
+      "step": 119600
+    },
+    {
+      "epoch": 1.0300447658402203,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.00018003673381503297,
+      "loss": 4.6108,
+      "step": 119650
+    },
+    {
+      "epoch": 1.0304752066115703,
+      "grad_norm": 2.984375,
+      "learning_rate": 0.00018002048893326787,
+      "loss": 4.3531,
+      "step": 119700
+    },
+    {
+      "epoch": 1.0309056473829201,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.00018000423817817108,
+      "loss": 4.5188,
+      "step": 119750
+    },
+    {
+      "epoch": 1.03133608815427,
+      "grad_norm": 4.09375,
+      "learning_rate": 0.00017998798155093533,
+      "loss": 4.8692,
+      "step": 119800
+    },
+    {
+      "epoch": 1.03176652892562,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.00017997171905275382,
+      "loss": 4.794,
+      "step": 119850
+    },
+    {
+      "epoch": 1.0321969696969697,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.0001799554506848202,
+      "loss": 4.1908,
+      "step": 119900
+    },
+    {
+      "epoch": 1.0326274104683195,
+      "grad_norm": 1.265625,
+      "learning_rate": 0.00017993917644832855,
+      "loss": 4.6614,
+      "step": 119950
+    },
+    {
+      "epoch": 1.0330578512396693,
+      "grad_norm": 2.421875,
+      "learning_rate": 0.0001799228963444733,
+      "loss": 4.9469,
+      "step": 120000
+    },
+    {
+      "epoch": 1.0330578512396693,
+      "eval_loss": 5.277187824249268,
+      "eval_runtime": 21.8386,
+      "eval_samples_per_second": 29.306,
+      "eval_steps_per_second": 14.653,
+      "eval_tts_loss": 6.978889931368823,
+      "step": 120000
+    },
+    {
+      "epoch": 1.0334882920110193,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.00017990661037444946,
+      "loss": 4.5529,
+      "step": 120050
+    },
+    {
+      "epoch": 1.0339187327823691,
+      "grad_norm": 1.578125,
+      "learning_rate": 0.00017989031853945232,
+      "loss": 4.6823,
+      "step": 120100
+    },
+    {
+      "epoch": 1.034349173553719,
+      "grad_norm": 3.21875,
+      "learning_rate": 0.0001798740208406777,
+      "loss": 4.581,
+      "step": 120150
+    },
+    {
+      "epoch": 1.034779614325069,
+      "grad_norm": 2.5,
+      "learning_rate": 0.0001798577172793218,
+      "loss": 4.252,
+      "step": 120200
+    },
+    {
+      "epoch": 1.0352100550964187,
+      "grad_norm": 1.8125,
+      "learning_rate": 0.00017984140785658125,
+      "loss": 4.7233,
+      "step": 120250
+    },
+    {
+      "epoch": 1.0356404958677685,
+      "grad_norm": 2.375,
+      "learning_rate": 0.00017982509257365316,
+      "loss": 4.4834,
+      "step": 120300
+    },
+    {
+      "epoch": 1.0360709366391185,
+      "grad_norm": 1.4609375,
+      "learning_rate": 0.00017980877143173506,
+      "loss": 4.3067,
+      "step": 120350
+    },
+    {
+      "epoch": 1.0365013774104683,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.00017979244443202478,
+      "loss": 4.4205,
+      "step": 120400
+    },
+    {
+      "epoch": 1.0369318181818181,
+      "grad_norm": 1.7109375,
+      "learning_rate": 0.00017977611157572078,
+      "loss": 4.7334,
+      "step": 120450
+    },
+    {
+      "epoch": 1.0373622589531681,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00017975977286402187,
+      "loss": 4.6972,
+      "step": 120500
+    },
+    {
+      "epoch": 1.037792699724518,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.00017974342829812722,
+      "loss": 4.6294,
+      "step": 120550
+    },
+    {
+      "epoch": 1.0382231404958677,
+      "grad_norm": 2.234375,
+      "learning_rate": 0.0001797270778792365,
+      "loss": 4.2679,
+      "step": 120600
+    },
+    {
+      "epoch": 1.0386535812672175,
+      "grad_norm": 1.9921875,
+      "learning_rate": 0.0001797107216085498,
+      "loss": 4.788,
+      "step": 120650
+    },
+    {
+      "epoch": 1.0390840220385675,
+      "grad_norm": 1.3671875,
+      "learning_rate": 0.00017969435948726764,
+      "loss": 4.6148,
+      "step": 120700
+    },
+    {
+      "epoch": 1.0395144628099173,
+      "grad_norm": 3.15625,
+      "learning_rate": 0.00017967799151659094,
+      "loss": 4.5744,
+      "step": 120750
+    },
+    {
+      "epoch": 1.0399449035812671,
+      "grad_norm": 2.5,
+      "learning_rate": 0.0001796616176977211,
+      "loss": 4.8067,
+      "step": 120800
+    },
+    {
+      "epoch": 1.0403753443526171,
+      "grad_norm": 3.46875,
+      "learning_rate": 0.00017964523803185993,
+      "loss": 4.392,
+      "step": 120850
+    },
+    {
+      "epoch": 1.040805785123967,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.00017962885252020965,
+      "loss": 4.6037,
+      "step": 120900
+    },
+    {
+      "epoch": 1.0412362258953167,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00017961246116397295,
+      "loss": 4.5907,
+      "step": 120950
+    },
+    {
+      "epoch": 1.0416666666666667,
+      "grad_norm": 3.90625,
+      "learning_rate": 0.00017959606396435285,
+      "loss": 4.7166,
+      "step": 121000
+    },
+    {
+      "epoch": 1.0420971074380165,
+      "grad_norm": 4.59375,
+      "learning_rate": 0.0001795796609225529,
+      "loss": 4.5639,
+      "step": 121050
+    },
+    {
+      "epoch": 1.0425275482093663,
+      "grad_norm": 1.375,
+      "learning_rate": 0.0001795632520397771,
+      "loss": 4.7798,
+      "step": 121100
+    },
+    {
+      "epoch": 1.0429579889807163,
+      "grad_norm": 1.296875,
+      "learning_rate": 0.00017954683731722976,
+      "loss": 4.6755,
+      "step": 121150
+    },
+    {
+      "epoch": 1.0433884297520661,
+      "grad_norm": 1.5859375,
+      "learning_rate": 0.00017953041675611574,
+      "loss": 4.4991,
+      "step": 121200
+    },
+    {
+      "epoch": 1.043818870523416,
+      "grad_norm": 2.5,
+      "learning_rate": 0.0001795139903576402,
+      "loss": 5.0583,
+      "step": 121250
+    },
+    {
+      "epoch": 1.044249311294766,
+      "grad_norm": 3.453125,
+      "learning_rate": 0.00017949755812300886,
+      "loss": 4.678,
+      "step": 121300
+    },
+    {
+      "epoch": 1.0446797520661157,
+      "grad_norm": 3.125,
+      "learning_rate": 0.00017948112005342778,
+      "loss": 4.6192,
+      "step": 121350
+    },
+    {
+      "epoch": 1.0451101928374655,
+      "grad_norm": 3.03125,
+      "learning_rate": 0.00017946467615010355,
+      "loss": 4.9013,
+      "step": 121400
+    },
+    {
+      "epoch": 1.0455406336088153,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.000179448226414243,
+      "loss": 4.86,
+      "step": 121450
+    },
+    {
+      "epoch": 1.0459710743801653,
+      "grad_norm": 1.7578125,
+      "learning_rate": 0.0001794317708470536,
+      "loss": 4.3527,
+      "step": 121500
+    },
+    {
+      "epoch": 1.0464015151515151,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.00017941530944974314,
+      "loss": 4.6214,
+      "step": 121550
+    },
+    {
+      "epoch": 1.046831955922865,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.0001793988422235198,
+      "loss": 4.7467,
+      "step": 121600
+    },
+    {
+      "epoch": 1.047262396694215,
+      "grad_norm": 1.7578125,
+      "learning_rate": 0.00017938236916959231,
+      "loss": 4.5602,
+      "step": 121650
+    },
+    {
+      "epoch": 1.0476928374655647,
+      "grad_norm": 1.6796875,
+      "learning_rate": 0.00017936589028916968,
+      "loss": 4.2249,
+      "step": 121700
+    },
+    {
+      "epoch": 1.0481232782369145,
+      "grad_norm": 3.46875,
+      "learning_rate": 0.00017934940558346146,
+      "loss": 4.1235,
+      "step": 121750
+    },
+    {
+      "epoch": 1.0485537190082646,
+      "grad_norm": 4.46875,
+      "learning_rate": 0.00017933291505367764,
+      "loss": 4.6966,
+      "step": 121800
+    },
+    {
+      "epoch": 1.0489841597796143,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00017931641870102852,
+      "loss": 4.5515,
+      "step": 121850
+    },
+    {
+      "epoch": 1.0494146005509641,
+      "grad_norm": 1.40625,
+      "learning_rate": 0.00017929991652672494,
+      "loss": 4.4808,
+      "step": 121900
+    },
+    {
+      "epoch": 1.0498450413223142,
+      "grad_norm": 3.8125,
+      "learning_rate": 0.00017928340853197807,
+      "loss": 4.7231,
+      "step": 121950
+    },
+    {
+      "epoch": 1.050275482093664,
+      "grad_norm": 5.375,
+      "learning_rate": 0.0001792668947179996,
+      "loss": 4.7616,
+      "step": 122000
+    },
+    {
+      "epoch": 1.0507059228650137,
+      "grad_norm": 4.03125,
+      "learning_rate": 0.00017925037508600159,
+      "loss": 4.3451,
+      "step": 122050
+    },
+    {
+      "epoch": 1.0511363636363635,
+      "grad_norm": 1.46875,
+      "learning_rate": 0.00017923384963719661,
+      "loss": 4.5132,
+      "step": 122100
+    },
+    {
+      "epoch": 1.0515668044077136,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.00017921731837279751,
+      "loss": 4.5463,
+      "step": 122150
+    },
+    {
+      "epoch": 1.0519972451790633,
+      "grad_norm": 5.375,
+      "learning_rate": 0.0001792007812940177,
+      "loss": 5.1975,
+      "step": 122200
+    },
+    {
+      "epoch": 1.0524276859504131,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00017918423840207094,
+      "loss": 4.6246,
+      "step": 122250
+    },
+    {
+      "epoch": 1.0528581267217632,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.00017916768969817145,
+      "loss": 4.584,
+      "step": 122300
+    },
+    {
+      "epoch": 1.053288567493113,
+      "grad_norm": 3.609375,
+      "learning_rate": 0.00017915113518353386,
+      "loss": 4.6767,
+      "step": 122350
+    },
+    {
+      "epoch": 1.0537190082644627,
+      "grad_norm": 0.6484375,
+      "learning_rate": 0.00017913457485937325,
+      "loss": 4.7067,
+      "step": 122400
+    },
+    {
+      "epoch": 1.0541494490358128,
+      "grad_norm": 2.9375,
+      "learning_rate": 0.0001791180087269051,
+      "loss": 4.7701,
+      "step": 122450
+    },
+    {
+      "epoch": 1.0545798898071626,
+      "grad_norm": 1.484375,
+      "learning_rate": 0.00017910143678734537,
+      "loss": 4.8197,
+      "step": 122500
+    },
+    {
+      "epoch": 1.0550103305785123,
+      "grad_norm": 2.25,
+      "learning_rate": 0.00017908485904191034,
+      "loss": 4.3686,
+      "step": 122550
+    },
+    {
+      "epoch": 1.0554407713498624,
+      "grad_norm": 3.828125,
+      "learning_rate": 0.0001790682754918168,
+      "loss": 4.9852,
+      "step": 122600
+    },
+    {
+      "epoch": 1.0558712121212122,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.000179051686138282,
+      "loss": 4.6628,
+      "step": 122650
+    },
+    {
+      "epoch": 1.056301652892562,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.00017903509098252349,
+      "loss": 4.3822,
+      "step": 122700
+    },
+    {
+      "epoch": 1.0567320936639117,
+      "grad_norm": 1.46875,
+      "learning_rate": 0.00017901849002575936,
+      "loss": 4.7333,
+      "step": 122750
+    },
+    {
+      "epoch": 1.0571625344352618,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.00017900188326920808,
+      "loss": 4.8762,
+      "step": 122800
+    },
+    {
+      "epoch": 1.0575929752066116,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.00017898527071408854,
+      "loss": 5.1889,
+      "step": 122850
+    },
+    {
+      "epoch": 1.0580234159779613,
+      "grad_norm": 1.625,
+      "learning_rate": 0.00017896865236162007,
+      "loss": 4.7234,
+      "step": 122900
+    },
+    {
+      "epoch": 1.0584538567493114,
+      "grad_norm": 1.4296875,
+      "learning_rate": 0.0001789520282130224,
+      "loss": 4.2801,
+      "step": 122950
+    },
+    {
+      "epoch": 1.0588842975206612,
+      "grad_norm": 3.0625,
+      "learning_rate": 0.00017893539826951578,
+      "loss": 4.4105,
+      "step": 123000
+    },
+    {
+      "epoch": 1.0588842975206612,
+      "eval_loss": 5.272480010986328,
+      "eval_runtime": 21.8098,
+      "eval_samples_per_second": 29.345,
+      "eval_steps_per_second": 14.672,
+      "eval_tts_loss": 6.962516161429058,
+      "step": 123000
+    },
+    {
+      "epoch": 1.059314738292011,
+      "grad_norm": 4.8125,
+      "learning_rate": 0.0001789187625323207,
+      "loss": 5.0393,
+      "step": 123050
+    },
+    {
+      "epoch": 1.059745179063361,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.0001789021210026583,
+      "loss": 4.7434,
+      "step": 123100
+    },
+    {
+      "epoch": 1.0601756198347108,
+      "grad_norm": 3.703125,
+      "learning_rate": 0.00017888547368174996,
+      "loss": 4.9379,
+      "step": 123150
+    },
+    {
+      "epoch": 1.0606060606060606,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.0001788688205708176,
+      "loss": 4.8474,
+      "step": 123200
+    },
+    {
+      "epoch": 1.0610365013774106,
+      "grad_norm": 1.96875,
+      "learning_rate": 0.00017885216167108346,
+      "loss": 4.621,
+      "step": 123250
+    },
+    {
+      "epoch": 1.0614669421487604,
+      "grad_norm": 1.828125,
+      "learning_rate": 0.00017883549698377034,
+      "loss": 4.2299,
+      "step": 123300
+    },
+    {
+      "epoch": 1.0618973829201102,
+      "grad_norm": 3.015625,
+      "learning_rate": 0.00017881882651010135,
+      "loss": 4.7655,
+      "step": 123350
+    },
+    {
+      "epoch": 1.06232782369146,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.00017880215025130008,
+      "loss": 5.0066,
+      "step": 123400
+    },
+    {
+      "epoch": 1.06275826446281,
+      "grad_norm": 3.203125,
+      "learning_rate": 0.0001787854682085905,
+      "loss": 4.9436,
+      "step": 123450
+    },
+    {
+      "epoch": 1.0631887052341598,
+      "grad_norm": 1.3046875,
+      "learning_rate": 0.0001787687803831971,
+      "loss": 4.3164,
+      "step": 123500
+    },
+    {
+      "epoch": 1.0636191460055096,
+      "grad_norm": 2.75,
+      "learning_rate": 0.0001787520867763447,
+      "loss": 4.3095,
+      "step": 123550
+    },
+    {
+      "epoch": 1.0640495867768596,
+      "grad_norm": 2.21875,
+      "learning_rate": 0.00017873538738925855,
+      "loss": 3.9903,
+      "step": 123600
+    },
+    {
+      "epoch": 1.0644800275482094,
+      "grad_norm": 3.453125,
+      "learning_rate": 0.00017871868222316437,
+      "loss": 4.5933,
+      "step": 123650
+    },
+    {
+      "epoch": 1.0649104683195592,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.0001787019712792883,
+      "loss": 4.5854,
+      "step": 123700
+    },
+    {
+      "epoch": 1.0653409090909092,
+      "grad_norm": 4.9375,
+      "learning_rate": 0.00017868525455885686,
+      "loss": 5.0299,
+      "step": 123750
+    },
+    {
+      "epoch": 1.065771349862259,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.00017866853206309702,
+      "loss": 4.4883,
+      "step": 123800
+    },
+    {
+      "epoch": 1.0662017906336088,
+      "grad_norm": 1.5546875,
+      "learning_rate": 0.00017865180379323617,
+      "loss": 4.3379,
+      "step": 123850
+    },
+    {
+      "epoch": 1.0666322314049588,
+      "grad_norm": 2.234375,
+      "learning_rate": 0.00017863506975050218,
+      "loss": 4.3761,
+      "step": 123900
+    },
+    {
+      "epoch": 1.0670626721763086,
+      "grad_norm": 0.9765625,
+      "learning_rate": 0.00017861832993612325,
+      "loss": 4.4775,
+      "step": 123950
+    },
+    {
+      "epoch": 1.0674931129476584,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00017860158435132806,
+      "loss": 4.8059,
+      "step": 124000
+    },
+    {
+      "epoch": 1.0679235537190084,
+      "grad_norm": 2.875,
+      "learning_rate": 0.00017858483299734569,
+      "loss": 4.6028,
+      "step": 124050
+    },
+    {
+      "epoch": 1.0683539944903582,
+      "grad_norm": 3.390625,
+      "learning_rate": 0.00017856807587540564,
+      "loss": 4.6802,
+      "step": 124100
+    },
+    {
+      "epoch": 1.068784435261708,
+      "grad_norm": 1.859375,
+      "learning_rate": 0.0001785513129867379,
+      "loss": 4.3151,
+      "step": 124150
+    },
+    {
+      "epoch": 1.0692148760330578,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.00017853454433257277,
+      "loss": 4.4292,
+      "step": 124200
+    },
+    {
+      "epoch": 1.0696453168044078,
+      "grad_norm": 1.625,
+      "learning_rate": 0.00017851776991414104,
+      "loss": 4.2712,
+      "step": 124250
+    },
+    {
+      "epoch": 1.0700757575757576,
+      "grad_norm": 4.71875,
+      "learning_rate": 0.00017850098973267397,
+      "loss": 4.4969,
+      "step": 124300
+    },
+    {
+      "epoch": 1.0705061983471074,
+      "grad_norm": 1.6171875,
+      "learning_rate": 0.00017848420378940313,
+      "loss": 4.9214,
+      "step": 124350
+    },
+    {
+      "epoch": 1.0709366391184574,
+      "grad_norm": 5.40625,
+      "learning_rate": 0.0001784674120855606,
+      "loss": 4.6779,
+      "step": 124400
+    },
+    {
+      "epoch": 1.0713670798898072,
+      "grad_norm": 1.7890625,
+      "learning_rate": 0.00017845061462237887,
+      "loss": 4.3168,
+      "step": 124450
+    },
+    {
+      "epoch": 1.071797520661157,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.0001784338114010908,
+      "loss": 4.7684,
+      "step": 124500
+    },
+    {
+      "epoch": 1.072227961432507,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.0001784170024229297,
+      "loss": 4.7823,
+      "step": 124550
+    },
+    {
+      "epoch": 1.0726584022038568,
+      "grad_norm": 3.3125,
+      "learning_rate": 0.00017840018768912937,
+      "loss": 4.8714,
+      "step": 124600
+    },
+    {
+      "epoch": 1.0730888429752066,
+      "grad_norm": 1.875,
+      "learning_rate": 0.00017838336720092394,
+      "loss": 4.8894,
+      "step": 124650
+    },
+    {
+      "epoch": 1.0735192837465566,
+      "grad_norm": 1.78125,
+      "learning_rate": 0.00017836654095954802,
+      "loss": 4.6314,
+      "step": 124700
+    },
+    {
+      "epoch": 1.0739497245179064,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.00017834970896623658,
+      "loss": 5.0466,
+      "step": 124750
+    },
+    {
+      "epoch": 1.0743801652892562,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.0001783328712222251,
+      "loss": 4.6268,
+      "step": 124800
+    },
+    {
+      "epoch": 1.074810606060606,
+      "grad_norm": 1.6328125,
+      "learning_rate": 0.00017831602772874937,
+      "loss": 4.2365,
+      "step": 124850
+    },
+    {
+      "epoch": 1.075241046831956,
+      "grad_norm": 1.859375,
+      "learning_rate": 0.00017829917848704574,
+      "loss": 4.7067,
+      "step": 124900
+    },
+    {
+      "epoch": 1.0756714876033058,
+      "grad_norm": 4.21875,
+      "learning_rate": 0.00017828232349835087,
+      "loss": 4.6713,
+      "step": 124950
+    },
+    {
+      "epoch": 1.0761019283746556,
+      "grad_norm": 2.140625,
+      "learning_rate": 0.0001782654627639019,
+      "loss": 4.8225,
+      "step": 125000
+    },
+    {
+      "epoch": 1.0765323691460056,
+      "grad_norm": 3.125,
+      "learning_rate": 0.00017824859628493632,
+      "loss": 4.6114,
+      "step": 125050
+    },
+    {
+      "epoch": 1.0769628099173554,
+      "grad_norm": 3.75,
+      "learning_rate": 0.00017823172406269215,
+      "loss": 4.5789,
+      "step": 125100
+    },
+    {
+      "epoch": 1.0773932506887052,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.00017821484609840775,
+      "loss": 4.5315,
+      "step": 125150
+    },
+    {
+      "epoch": 1.0778236914600552,
+      "grad_norm": 1.65625,
+      "learning_rate": 0.0001781979623933219,
+      "loss": 4.6303,
+      "step": 125200
+    },
+    {
+      "epoch": 1.078254132231405,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.0001781810729486739,
+      "loss": 4.2937,
+      "step": 125250
+    },
+    {
+      "epoch": 1.0786845730027548,
+      "grad_norm": 2.921875,
+      "learning_rate": 0.00017816417776570336,
+      "loss": 4.7388,
+      "step": 125300
+    },
+    {
+      "epoch": 1.0791150137741048,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.0001781472768456503,
+      "loss": 4.4114,
+      "step": 125350
+    },
+    {
+      "epoch": 1.0795454545454546,
+      "grad_norm": 2.9375,
+      "learning_rate": 0.0001781303701897553,
+      "loss": 4.5467,
+      "step": 125400
+    },
+    {
+      "epoch": 1.0799758953168044,
+      "grad_norm": 3.453125,
+      "learning_rate": 0.0001781134577992592,
+      "loss": 4.4847,
+      "step": 125450
+    },
+    {
+      "epoch": 1.0804063360881542,
+      "grad_norm": 3.875,
+      "learning_rate": 0.0001780965396754034,
+      "loss": 4.4528,
+      "step": 125500
+    },
+    {
+      "epoch": 1.0808367768595042,
+      "grad_norm": 2.875,
+      "learning_rate": 0.0001780796158194296,
+      "loss": 4.5023,
+      "step": 125550
+    },
+    {
+      "epoch": 1.081267217630854,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.00017806268623257997,
+      "loss": 4.565,
+      "step": 125600
+    },
+    {
+      "epoch": 1.0816976584022038,
+      "grad_norm": 3.609375,
+      "learning_rate": 0.00017804575091609715,
+      "loss": 4.2734,
+      "step": 125650
+    },
+    {
+      "epoch": 1.0821280991735538,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.00017802880987122412,
+      "loss": 4.6331,
+      "step": 125700
+    },
+    {
+      "epoch": 1.0825585399449036,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.00017801186309920436,
+      "loss": 4.8683,
+      "step": 125750
+    },
+    {
+      "epoch": 1.0829889807162534,
+      "grad_norm": 3.171875,
+      "learning_rate": 0.00017799491060128167,
+      "loss": 4.5946,
+      "step": 125800
+    },
+    {
+      "epoch": 1.0834194214876034,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.00017797795237870035,
+      "loss": 4.7209,
+      "step": 125850
+    },
+    {
+      "epoch": 1.0838498622589532,
+      "grad_norm": 1.8125,
+      "learning_rate": 0.0001779609884327051,
+      "loss": 5.0445,
+      "step": 125900
+    },
+    {
+      "epoch": 1.084280303030303,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.00017794401876454103,
+      "loss": 4.6825,
+      "step": 125950
+    },
+    {
+      "epoch": 1.084710743801653,
+      "grad_norm": 1.9609375,
+      "learning_rate": 0.0001779270433754537,
+      "loss": 4.7229,
+      "step": 126000
+    },
+    {
+      "epoch": 1.084710743801653,
+      "eval_loss": 5.262632846832275,
+      "eval_runtime": 21.9137,
+      "eval_samples_per_second": 29.205,
+      "eval_steps_per_second": 14.603,
+      "eval_tts_loss": 6.942831247074225,
+      "step": 126000
+    },
+    {
+      "epoch": 1.0851411845730028,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00017791006226668902,
+      "loss": 4.9246,
+      "step": 126050
+    },
+    {
+      "epoch": 1.0855716253443526,
+      "grad_norm": 2.625,
+      "learning_rate": 0.00017789307543949345,
+      "loss": 4.3588,
+      "step": 126100
+    },
+    {
+      "epoch": 1.0860020661157024,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.0001778760828951137,
+      "loss": 4.8105,
+      "step": 126150
+    },
+    {
+      "epoch": 1.0864325068870524,
+      "grad_norm": 3.65625,
+      "learning_rate": 0.00017785908463479702,
+      "loss": 4.7491,
+      "step": 126200
+    },
+    {
+      "epoch": 1.0868629476584022,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00017784208065979104,
+      "loss": 4.7102,
+      "step": 126250
+    },
+    {
+      "epoch": 1.087293388429752,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00017782507097134382,
+      "loss": 4.8335,
+      "step": 126300
+    },
+    {
+      "epoch": 1.087723829201102,
+      "grad_norm": 6.40625,
+      "learning_rate": 0.00017780805557070388,
+      "loss": 5.1412,
+      "step": 126350
+    },
+    {
+      "epoch": 1.0881542699724518,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00017779103445912003,
+      "loss": 4.5067,
+      "step": 126400
+    },
+    {
+      "epoch": 1.0885847107438016,
+      "grad_norm": 3.734375,
+      "learning_rate": 0.00017777400763784162,
+      "loss": 4.3604,
+      "step": 126450
+    },
+    {
+      "epoch": 1.0890151515151516,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.0001777569751081184,
+      "loss": 4.4491,
+      "step": 126500
+    },
+    {
+      "epoch": 1.0894455922865014,
+      "grad_norm": 2.625,
+      "learning_rate": 0.0001777399368712005,
+      "loss": 4.5945,
+      "step": 126550
+    },
+    {
+      "epoch": 1.0898760330578512,
+      "grad_norm": 1.75,
+      "learning_rate": 0.0001777228929283385,
+      "loss": 4.5258,
+      "step": 126600
+    },
+    {
+      "epoch": 1.0903064738292012,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.00017770584328078337,
+      "loss": 4.9029,
+      "step": 126650
+    },
+    {
+      "epoch": 1.090736914600551,
+      "grad_norm": 1.9921875,
+      "learning_rate": 0.00017768878792978653,
+      "loss": 5.0176,
+      "step": 126700
+    },
+    {
+      "epoch": 1.0911673553719008,
+      "grad_norm": 4.6875,
+      "learning_rate": 0.0001776717268765998,
+      "loss": 4.3295,
+      "step": 126750
+    },
+    {
+      "epoch": 1.0915977961432506,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00017765466012247547,
+      "loss": 4.6499,
+      "step": 126800
+    },
+    {
+      "epoch": 1.0920282369146006,
+      "grad_norm": 1.875,
+      "learning_rate": 0.00017763758766866614,
+      "loss": 4.8432,
+      "step": 126850
+    },
+    {
+      "epoch": 1.0924586776859504,
+      "grad_norm": 2.625,
+      "learning_rate": 0.00017762050951642495,
+      "loss": 4.4676,
+      "step": 126900
+    },
+    {
+      "epoch": 1.0928891184573002,
+      "grad_norm": 2.25,
+      "learning_rate": 0.0001776034256670053,
+      "loss": 4.7084,
+      "step": 126950
+    },
+    {
+      "epoch": 1.0933195592286502,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.00017758633612166124,
+      "loss": 4.8007,
+      "step": 127000
+    },
+    {
+      "epoch": 1.09375,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.00017756924088164702,
+      "loss": 4.1653,
+      "step": 127050
+    },
+    {
+      "epoch": 1.0941804407713498,
+      "grad_norm": 1.8984375,
+      "learning_rate": 0.0001775521399482174,
+      "loss": 4.2092,
+      "step": 127100
+    },
+    {
+      "epoch": 1.0946108815426998,
+      "grad_norm": 1.9375,
+      "learning_rate": 0.00017753503332262755,
+      "loss": 4.2951,
+      "step": 127150
+    },
+    {
+      "epoch": 1.0950413223140496,
+      "grad_norm": 2.578125,
+      "learning_rate": 0.00017751792100613312,
+      "loss": 4.758,
+      "step": 127200
+    },
+    {
+      "epoch": 1.0954717630853994,
+      "grad_norm": 3.5,
+      "learning_rate": 0.00017750080299999005,
+      "loss": 4.7231,
+      "step": 127250
+    },
+    {
+      "epoch": 1.0959022038567494,
+      "grad_norm": 1.796875,
+      "learning_rate": 0.0001774836793054548,
+      "loss": 4.3676,
+      "step": 127300
+    },
+    {
+      "epoch": 1.0963326446280992,
+      "grad_norm": 1.4765625,
+      "learning_rate": 0.00017746654992378415,
+      "loss": 4.0514,
+      "step": 127350
+    },
+    {
+      "epoch": 1.096763085399449,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.00017744941485623548,
+      "loss": 4.7052,
+      "step": 127400
+    },
+    {
+      "epoch": 1.0971935261707988,
+      "grad_norm": 6.28125,
+      "learning_rate": 0.00017743227410406634,
+      "loss": 4.974,
+      "step": 127450
+    },
+    {
+      "epoch": 1.0976239669421488,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00017741512766853492,
+      "loss": 4.4805,
+      "step": 127500
+    },
+    {
+      "epoch": 1.0980544077134986,
+      "grad_norm": 1.8359375,
+      "learning_rate": 0.00017739797555089965,
+      "loss": 4.9846,
+      "step": 127550
+    },
+    {
+      "epoch": 1.0984848484848484,
+      "grad_norm": 0.93359375,
+      "learning_rate": 0.00017738081775241953,
+      "loss": 4.2951,
+      "step": 127600
+    },
+    {
+      "epoch": 1.0989152892561984,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.0001773636542743539,
+      "loss": 5.0558,
+      "step": 127650
+    },
+    {
+      "epoch": 1.0993457300275482,
+      "grad_norm": 3.390625,
+      "learning_rate": 0.00017734648511796242,
+      "loss": 4.5819,
+      "step": 127700
+    },
+    {
+      "epoch": 1.099776170798898,
+      "grad_norm": 3.8125,
+      "learning_rate": 0.00017732931028450541,
+      "loss": 4.9886,
+      "step": 127750
+    },
+    {
+      "epoch": 1.100206611570248,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00017731212977524342,
+      "loss": 4.7401,
+      "step": 127800
+    },
+    {
+      "epoch": 1.1006370523415978,
+      "grad_norm": 3.28125,
+      "learning_rate": 0.00017729494359143742,
+      "loss": 4.6309,
+      "step": 127850
+    },
+    {
+      "epoch": 1.1010674931129476,
+      "grad_norm": 3.765625,
+      "learning_rate": 0.00017727775173434883,
+      "loss": 4.3819,
+      "step": 127900
+    },
+    {
+      "epoch": 1.1014979338842976,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00017726055420523958,
+      "loss": 4.9363,
+      "step": 127950
+    },
+    {
+      "epoch": 1.1019283746556474,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.00017724335100537187,
+      "loss": 4.8072,
+      "step": 128000
+    },
+    {
+      "epoch": 1.1023588154269972,
+      "grad_norm": 3.890625,
+      "learning_rate": 0.00017722614213600837,
+      "loss": 4.7794,
+      "step": 128050
+    },
+    {
+      "epoch": 1.102789256198347,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.0001772089275984122,
+      "loss": 4.6653,
+      "step": 128100
+    },
+    {
+      "epoch": 1.103219696969697,
+      "grad_norm": 2.234375,
+      "learning_rate": 0.00017719170739384686,
+      "loss": 4.5847,
+      "step": 128150
+    },
+    {
+      "epoch": 1.1036501377410468,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.0001771744815235763,
+      "loss": 4.84,
+      "step": 128200
+    },
+    {
+      "epoch": 1.1040805785123966,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00017715724998886484,
+      "loss": 4.6217,
+      "step": 128250
+    },
+    {
+      "epoch": 1.1045110192837466,
+      "grad_norm": 3.109375,
+      "learning_rate": 0.00017714001279097724,
+      "loss": 4.6454,
+      "step": 128300
+    },
+    {
+      "epoch": 1.1049414600550964,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.00017712276993117864,
+      "loss": 4.3933,
+      "step": 128350
+    },
+    {
+      "epoch": 1.1053719008264462,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.00017710552141073468,
+      "loss": 4.8818,
+      "step": 128400
+    },
+    {
+      "epoch": 1.1058023415977962,
+      "grad_norm": 2.625,
+      "learning_rate": 0.00017708826723091136,
+      "loss": 4.6885,
+      "step": 128450
+    },
+    {
+      "epoch": 1.106232782369146,
+      "grad_norm": 3.140625,
+      "learning_rate": 0.00017707100739297507,
+      "loss": 4.5041,
+      "step": 128500
+    },
+    {
+      "epoch": 1.1066632231404958,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00017705374189819266,
+      "loss": 4.8692,
+      "step": 128550
+    },
+    {
+      "epoch": 1.1070936639118458,
+      "grad_norm": 1.59375,
+      "learning_rate": 0.00017703647074783139,
+      "loss": 4.4058,
+      "step": 128600
+    },
+    {
+      "epoch": 1.1075241046831956,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.00017701919394315892,
+      "loss": 5.008,
+      "step": 128650
+    },
+    {
+      "epoch": 1.1079545454545454,
+      "grad_norm": 1.3515625,
+      "learning_rate": 0.00017700191148544328,
+      "loss": 4.3127,
+      "step": 128700
+    },
+    {
+      "epoch": 1.1083849862258952,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00017698462337595308,
+      "loss": 4.3472,
+      "step": 128750
+    },
+    {
+      "epoch": 1.1088154269972452,
+      "grad_norm": 3.40625,
+      "learning_rate": 0.0001769673296159571,
+      "loss": 4.6588,
+      "step": 128800
+    },
+    {
+      "epoch": 1.109245867768595,
+      "grad_norm": 5.53125,
+      "learning_rate": 0.00017695003020672475,
+      "loss": 4.8604,
+      "step": 128850
+    },
+    {
+      "epoch": 1.1096763085399448,
+      "grad_norm": 2.5,
+      "learning_rate": 0.00017693272514952575,
+      "loss": 4.7601,
+      "step": 128900
+    },
+    {
+      "epoch": 1.1101067493112948,
+      "grad_norm": 3.859375,
+      "learning_rate": 0.00017691541444563025,
+      "loss": 4.7327,
+      "step": 128950
+    },
+    {
+      "epoch": 1.1105371900826446,
+      "grad_norm": 1.9296875,
+      "learning_rate": 0.0001768980980963088,
+      "loss": 4.8305,
+      "step": 129000
+    },
+    {
+      "epoch": 1.1105371900826446,
+      "eval_loss": 5.248120307922363,
+      "eval_runtime": 21.7745,
+      "eval_samples_per_second": 29.392,
+      "eval_steps_per_second": 14.696,
+      "eval_tts_loss": 6.979110696898764,
+      "step": 129000
+    },
+    {
+      "epoch": 1.1109676308539944,
+      "grad_norm": 4.21875,
+      "learning_rate": 0.00017688077610283242,
+      "loss": 4.9047,
+      "step": 129050
+    },
+    {
+      "epoch": 1.1113980716253444,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.00017686344846647249,
+      "loss": 4.9494,
+      "step": 129100
+    },
+    {
+      "epoch": 1.1118285123966942,
+      "grad_norm": 3.375,
+      "learning_rate": 0.0001768461151885008,
+      "loss": 4.5499,
+      "step": 129150
+    },
+    {
+      "epoch": 1.112258953168044,
+      "grad_norm": 1.8125,
+      "learning_rate": 0.0001768287762701896,
+      "loss": 4.6235,
+      "step": 129200
+    },
+    {
+      "epoch": 1.112689393939394,
+      "grad_norm": 3.265625,
+      "learning_rate": 0.00017681143171281152,
+      "loss": 4.6963,
+      "step": 129250
+    },
+    {
+      "epoch": 1.1131198347107438,
+      "grad_norm": 1.8203125,
+      "learning_rate": 0.00017679408151763962,
+      "loss": 4.2662,
+      "step": 129300
+    },
+    {
+      "epoch": 1.1135502754820936,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.00017677672568594739,
+      "loss": 4.3946,
+      "step": 129350
+    },
+    {
+      "epoch": 1.1139807162534434,
+      "grad_norm": 1.8359375,
+      "learning_rate": 0.00017675936421900862,
+      "loss": 4.6193,
+      "step": 129400
+    },
+    {
+      "epoch": 1.1144111570247934,
+      "grad_norm": 1.546875,
+      "learning_rate": 0.00017674199711809773,
+      "loss": 4.847,
+      "step": 129450
+    },
+    {
+      "epoch": 1.1148415977961432,
+      "grad_norm": 2.984375,
+      "learning_rate": 0.00017672462438448933,
+      "loss": 4.8154,
+      "step": 129500
+    },
+    {
+      "epoch": 1.115272038567493,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00017670724601945862,
+      "loss": 4.6874,
+      "step": 129550
+    },
+    {
+      "epoch": 1.115702479338843,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.00017668986202428106,
+      "loss": 4.685,
+      "step": 129600
+    },
+    {
+      "epoch": 1.1161329201101928,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.00017667247240023263,
+      "loss": 4.1753,
+      "step": 129650
+    },
+    {
+      "epoch": 1.1165633608815426,
+      "grad_norm": 3.375,
+      "learning_rate": 0.0001766550771485897,
+      "loss": 4.4835,
+      "step": 129700
+    },
+    {
+      "epoch": 1.1169938016528926,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.000176637676270629,
+      "loss": 4.5487,
+      "step": 129750
+    },
+    {
+      "epoch": 1.1174242424242424,
+      "grad_norm": 3.328125,
+      "learning_rate": 0.0001766202697676278,
+      "loss": 4.5791,
+      "step": 129800
+    },
+    {
+      "epoch": 1.1178546831955922,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00017660285764086364,
+      "loss": 4.5032,
+      "step": 129850
+    },
+    {
+      "epoch": 1.1182851239669422,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00017658543989161453,
+      "loss": 4.3983,
+      "step": 129900
+    },
+    {
+      "epoch": 1.118715564738292,
+      "grad_norm": 3.140625,
+      "learning_rate": 0.00017656801652115892,
+      "loss": 4.5268,
+      "step": 129950
+    },
+    {
+      "epoch": 1.1191460055096418,
+      "grad_norm": 2.5,
+      "learning_rate": 0.0001765505875307756,
+      "loss": 4.8744,
+      "step": 130000
+    },
+    {
+      "epoch": 1.1195764462809916,
+      "grad_norm": 1.7109375,
+      "learning_rate": 0.00017653315292174394,
+      "loss": 4.5245,
+      "step": 130050
+    },
+    {
+      "epoch": 1.1200068870523416,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00017651571269534347,
+      "loss": 4.7077,
+      "step": 130100
+    },
+    {
+      "epoch": 1.1204373278236914,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.0001764982668528543,
+      "loss": 4.8312,
+      "step": 130150
+    },
+    {
+      "epoch": 1.1208677685950412,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.00017648081539555697,
+      "loss": 4.7503,
+      "step": 130200
+    },
+    {
+      "epoch": 1.1212982093663912,
+      "grad_norm": 1.53125,
+      "learning_rate": 0.00017646335832473232,
+      "loss": 4.768,
+      "step": 130250
+    },
+    {
+      "epoch": 1.121728650137741,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.0001764458956416617,
+      "loss": 4.7932,
+      "step": 130300
+    },
+    {
+      "epoch": 1.1221590909090908,
+      "grad_norm": 3.8125,
+      "learning_rate": 0.00017642842734762678,
+      "loss": 4.5183,
+      "step": 130350
+    },
+    {
+      "epoch": 1.1225895316804408,
+      "grad_norm": 2.265625,
+      "learning_rate": 0.0001764109534439098,
+      "loss": 4.7769,
+      "step": 130400
+    },
+    {
+      "epoch": 1.1230199724517906,
+      "grad_norm": 0.349609375,
+      "learning_rate": 0.00017639347393179317,
+      "loss": 4.6036,
+      "step": 130450
+    },
+    {
+      "epoch": 1.1234504132231404,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.00017637598881255995,
+      "loss": 4.1616,
+      "step": 130500
+    },
+    {
+      "epoch": 1.1238808539944904,
+      "grad_norm": 1.7421875,
+      "learning_rate": 0.0001763584980874935,
+      "loss": 4.3915,
+      "step": 130550
+    },
+    {
+      "epoch": 1.1243112947658402,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00017634100175787753,
+      "loss": 4.3013,
+      "step": 130600
+    },
+    {
+      "epoch": 1.12474173553719,
+      "grad_norm": 2.875,
+      "learning_rate": 0.00017632349982499632,
+      "loss": 4.6347,
+      "step": 130650
+    },
+    {
+      "epoch": 1.1251721763085398,
+      "grad_norm": 3.1875,
+      "learning_rate": 0.00017630599229013442,
+      "loss": 4.547,
+      "step": 130700
+    },
+    {
+      "epoch": 1.1256026170798898,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.00017628847915457686,
+      "loss": 4.2865,
+      "step": 130750
+    },
+    {
+      "epoch": 1.1260330578512396,
+      "grad_norm": 1.7421875,
+      "learning_rate": 0.0001762709604196091,
+      "loss": 4.6399,
+      "step": 130800
+    },
+    {
+      "epoch": 1.1264634986225897,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.0001762534360865169,
+      "loss": 4.5703,
+      "step": 130850
+    },
+    {
+      "epoch": 1.1268939393939394,
+      "grad_norm": 4.09375,
+      "learning_rate": 0.00017623590615658657,
+      "loss": 4.4677,
+      "step": 130900
+    },
+    {
+      "epoch": 1.1273243801652892,
+      "grad_norm": 4.625,
+      "learning_rate": 0.00017621837063110479,
+      "loss": 4.5387,
+      "step": 130950
+    },
+    {
+      "epoch": 1.127754820936639,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.00017620082951135856,
+      "loss": 4.4496,
+      "step": 131000
+    },
+    {
+      "epoch": 1.128185261707989,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.0001761832827986354,
+      "loss": 4.598,
+      "step": 131050
+    },
+    {
+      "epoch": 1.1286157024793388,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.0001761657304942232,
+      "loss": 4.7248,
+      "step": 131100
+    },
+    {
+      "epoch": 1.1290461432506886,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00017614817259941026,
+      "loss": 4.7915,
+      "step": 131150
+    },
+    {
+      "epoch": 1.1294765840220387,
+      "grad_norm": 1.859375,
+      "learning_rate": 0.00017613060911548527,
+      "loss": 4.6542,
+      "step": 131200
+    },
+    {
+      "epoch": 1.1299070247933884,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00017611304004373735,
+      "loss": 4.7307,
+      "step": 131250
+    },
+    {
+      "epoch": 1.1303374655647382,
+      "grad_norm": 1.90625,
+      "learning_rate": 0.0001760954653854561,
+      "loss": 4.5429,
+      "step": 131300
+    },
+    {
+      "epoch": 1.130767906336088,
+      "grad_norm": 3.578125,
+      "learning_rate": 0.0001760778851419314,
+      "loss": 4.9145,
+      "step": 131350
+    },
+    {
+      "epoch": 1.131198347107438,
+      "grad_norm": 1.640625,
+      "learning_rate": 0.0001760602993144536,
+      "loss": 4.5879,
+      "step": 131400
+    },
+    {
+      "epoch": 1.1316287878787878,
+      "grad_norm": 1.8515625,
+      "learning_rate": 0.00017604270790431352,
+      "loss": 4.6708,
+      "step": 131450
+    },
+    {
+      "epoch": 1.1320592286501379,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.00017602511091280225,
+      "loss": 4.6827,
+      "step": 131500
+    },
+    {
+      "epoch": 1.1324896694214877,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.00017600750834121143,
+      "loss": 4.3763,
+      "step": 131550
+    },
+    {
+      "epoch": 1.1329201101928374,
+      "grad_norm": 3.84375,
+      "learning_rate": 0.00017598990019083305,
+      "loss": 4.5403,
+      "step": 131600
+    },
+    {
+      "epoch": 1.1333505509641872,
+      "grad_norm": 3.890625,
+      "learning_rate": 0.00017597228646295943,
+      "loss": 4.704,
+      "step": 131650
+    },
+    {
+      "epoch": 1.1337809917355373,
+      "grad_norm": 3.625,
+      "learning_rate": 0.00017595466715888348,
+      "loss": 4.7661,
+      "step": 131700
+    },
+    {
+      "epoch": 1.134211432506887,
+      "grad_norm": 2.25,
+      "learning_rate": 0.0001759370422798984,
+      "loss": 4.697,
+      "step": 131750
+    },
+    {
+      "epoch": 1.1346418732782368,
+      "grad_norm": 3.578125,
+      "learning_rate": 0.00017591941182729776,
+      "loss": 4.6169,
+      "step": 131800
+    },
+    {
+      "epoch": 1.1350723140495869,
+      "grad_norm": 1.40625,
+      "learning_rate": 0.0001759017758023757,
+      "loss": 4.504,
+      "step": 131850
+    },
+    {
+      "epoch": 1.1355027548209367,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.00017588413420642653,
+      "loss": 4.6758,
+      "step": 131900
+    },
+    {
+      "epoch": 1.1359331955922864,
+      "grad_norm": 3.453125,
+      "learning_rate": 0.0001758664870407452,
+      "loss": 5.0743,
+      "step": 131950
+    },
+    {
+      "epoch": 1.1363636363636362,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00017584883430662697,
+      "loss": 4.6312,
+      "step": 132000
+    },
+    {
+      "epoch": 1.1363636363636362,
+      "eval_loss": 5.238390922546387,
+      "eval_runtime": 21.7258,
+      "eval_samples_per_second": 29.458,
+      "eval_steps_per_second": 14.729,
+      "eval_tts_loss": 6.946057393402399,
+      "step": 132000
+    },
+    {
+      "epoch": 1.1367940771349863,
+      "grad_norm": 3.109375,
+      "learning_rate": 0.00017583117600536747,
+      "loss": 4.5312,
+      "step": 132050
+    },
+    {
+      "epoch": 1.137224517906336,
+      "grad_norm": 1.734375,
+      "learning_rate": 0.00017581351213826282,
+      "loss": 4.5015,
+      "step": 132100
+    },
+    {
+      "epoch": 1.137654958677686,
+      "grad_norm": 3.515625,
+      "learning_rate": 0.00017579584270660948,
+      "loss": 4.7982,
+      "step": 132150
+    },
+    {
+      "epoch": 1.1380853994490359,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.0001757781677117044,
+      "loss": 4.7285,
+      "step": 132200
+    },
+    {
+      "epoch": 1.1385158402203857,
+      "grad_norm": 3.5625,
+      "learning_rate": 0.00017576048715484482,
+      "loss": 4.53,
+      "step": 132250
+    },
+    {
+      "epoch": 1.1389462809917354,
+      "grad_norm": 0.9375,
+      "learning_rate": 0.00017574280103732854,
+      "loss": 4.7411,
+      "step": 132300
+    },
+    {
+      "epoch": 1.1393767217630855,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.00017572510936045358,
+      "loss": 4.7267,
+      "step": 132350
+    },
+    {
+      "epoch": 1.1398071625344353,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00017570741212551854,
+      "loss": 4.8826,
+      "step": 132400
+    },
+    {
+      "epoch": 1.140237603305785,
+      "grad_norm": 1.5390625,
+      "learning_rate": 0.0001756897093338223,
+      "loss": 4.0393,
+      "step": 132450
+    },
+    {
+      "epoch": 1.140668044077135,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.00017567200098666426,
+      "loss": 4.3133,
+      "step": 132500
+    },
+    {
+      "epoch": 1.1410984848484849,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00017565428708534418,
+      "loss": 4.5982,
+      "step": 132550
+    },
+    {
+      "epoch": 1.1415289256198347,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00017563656763116222,
+      "loss": 4.4531,
+      "step": 132600
+    },
+    {
+      "epoch": 1.1419593663911844,
+      "grad_norm": 0.62890625,
+      "learning_rate": 0.00017561884262541887,
+      "loss": 4.7224,
+      "step": 132650
+    },
+    {
+      "epoch": 1.1423898071625345,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.00017560111206941522,
+      "loss": 4.2587,
+      "step": 132700
+    },
+    {
+      "epoch": 1.1428202479338843,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.0001755833759644526,
+      "loss": 4.9991,
+      "step": 132750
+    },
+    {
+      "epoch": 1.1432506887052343,
+      "grad_norm": 2.9375,
+      "learning_rate": 0.00017556563431183277,
+      "loss": 4.6288,
+      "step": 132800
+    },
+    {
+      "epoch": 1.143681129476584,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.00017554788711285802,
+      "loss": 5.0109,
+      "step": 132850
+    },
+    {
+      "epoch": 1.1441115702479339,
+      "grad_norm": 4.9375,
+      "learning_rate": 0.00017553013436883086,
+      "loss": 4.6113,
+      "step": 132900
+    },
+    {
+      "epoch": 1.1445420110192837,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.00017551237608105438,
+      "loss": 4.7877,
+      "step": 132950
+    },
+    {
+      "epoch": 1.1449724517906337,
+      "grad_norm": 1.8125,
+      "learning_rate": 0.0001754946122508319,
+      "loss": 4.4095,
+      "step": 133000
+    },
+    {
+      "epoch": 1.1454028925619835,
+      "grad_norm": 1.6953125,
+      "learning_rate": 0.00017547684287946738,
+      "loss": 4.3929,
+      "step": 133050
+    },
+    {
+      "epoch": 1.1458333333333333,
+      "grad_norm": 3.734375,
+      "learning_rate": 0.00017545906796826496,
+      "loss": 4.9235,
+      "step": 133100
+    },
+    {
+      "epoch": 1.1462637741046833,
+      "grad_norm": 1.5546875,
+      "learning_rate": 0.0001754412875185293,
+      "loss": 4.7911,
+      "step": 133150
+    },
+    {
+      "epoch": 1.146694214876033,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.00017542350153156545,
+      "loss": 4.792,
+      "step": 133200
+    },
+    {
+      "epoch": 1.1471246556473829,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00017540571000867886,
+      "loss": 4.572,
+      "step": 133250
+    },
+    {
+      "epoch": 1.1475550964187329,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.00017538791295117543,
+      "loss": 4.9435,
+      "step": 133300
+    },
+    {
+      "epoch": 1.1479855371900827,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.00017537011036036136,
+      "loss": 4.7983,
+      "step": 133350
+    },
+    {
+      "epoch": 1.1484159779614325,
+      "grad_norm": 3.640625,
+      "learning_rate": 0.00017535230223754335,
+      "loss": 4.3793,
+      "step": 133400
+    },
+    {
+      "epoch": 1.1488464187327825,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.0001753344885840285,
+      "loss": 4.278,
+      "step": 133450
+    },
+    {
+      "epoch": 1.1492768595041323,
+      "grad_norm": 1.390625,
+      "learning_rate": 0.00017531666940112425,
+      "loss": 4.6926,
+      "step": 133500
+    },
+    {
+      "epoch": 1.149707300275482,
+      "grad_norm": 2.578125,
+      "learning_rate": 0.00017529884469013853,
+      "loss": 4.5613,
+      "step": 133550
+    },
+    {
+      "epoch": 1.1501377410468319,
+      "grad_norm": 0.70703125,
+      "learning_rate": 0.00017528101445237965,
+      "loss": 4.2171,
+      "step": 133600
+    },
+    {
+      "epoch": 1.1505681818181819,
+      "grad_norm": 1.7421875,
+      "learning_rate": 0.00017526317868915624,
+      "loss": 4.646,
+      "step": 133650
+    },
+    {
+      "epoch": 1.1509986225895317,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.00017524533740177742,
+      "loss": 4.8612,
+      "step": 133700
+    },
+    {
+      "epoch": 1.1514290633608815,
+      "grad_norm": 3.265625,
+      "learning_rate": 0.00017522749059155278,
+      "loss": 4.2613,
+      "step": 133750
+    },
+    {
+      "epoch": 1.1518595041322315,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.00017520963825979217,
+      "loss": 4.434,
+      "step": 133800
+    },
+    {
+      "epoch": 1.1522899449035813,
+      "grad_norm": 4.21875,
+      "learning_rate": 0.00017519178040780592,
+      "loss": 4.3308,
+      "step": 133850
+    },
+    {
+      "epoch": 1.152720385674931,
+      "grad_norm": 2.375,
+      "learning_rate": 0.00017517391703690478,
+      "loss": 4.4181,
+      "step": 133900
+    },
+    {
+      "epoch": 1.153150826446281,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.00017515604814839986,
+      "loss": 4.7665,
+      "step": 133950
+    },
+    {
+      "epoch": 1.1535812672176309,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.00017513817374360273,
+      "loss": 4.6749,
+      "step": 134000
+    },
+    {
+      "epoch": 1.1540117079889807,
+      "grad_norm": 3.8125,
+      "learning_rate": 0.00017512029382382531,
+      "loss": 4.1754,
+      "step": 134050
+    },
+    {
+      "epoch": 1.1544421487603307,
+      "grad_norm": 3.546875,
+      "learning_rate": 0.00017510240839037992,
+      "loss": 4.4877,
+      "step": 134100
+    },
+    {
+      "epoch": 1.1548725895316805,
+      "grad_norm": 4.46875,
+      "learning_rate": 0.00017508451744457934,
+      "loss": 4.5622,
+      "step": 134150
+    },
+    {
+      "epoch": 1.1553030303030303,
+      "grad_norm": 3.671875,
+      "learning_rate": 0.00017506662098773677,
+      "loss": 4.7159,
+      "step": 134200
+    },
+    {
+      "epoch": 1.15573347107438,
+      "grad_norm": 1.921875,
+      "learning_rate": 0.0001750487190211657,
+      "loss": 4.1376,
+      "step": 134250
+    },
+    {
+      "epoch": 1.15616391184573,
+      "grad_norm": 3.03125,
+      "learning_rate": 0.00017503081154618012,
+      "loss": 4.8649,
+      "step": 134300
+    },
+    {
+      "epoch": 1.1565943526170799,
+      "grad_norm": 3.234375,
+      "learning_rate": 0.0001750128985640944,
+      "loss": 4.6983,
+      "step": 134350
+    },
+    {
+      "epoch": 1.1570247933884297,
+      "grad_norm": 3.109375,
+      "learning_rate": 0.00017499498007622337,
+      "loss": 4.9357,
+      "step": 134400
+    },
+    {
+      "epoch": 1.1574552341597797,
+      "grad_norm": 3.265625,
+      "learning_rate": 0.0001749770560838821,
+      "loss": 4.5523,
+      "step": 134450
+    },
+    {
+      "epoch": 1.1578856749311295,
+      "grad_norm": 3.890625,
+      "learning_rate": 0.00017495912658838626,
+      "loss": 4.8399,
+      "step": 134500
+    },
+    {
+      "epoch": 1.1583161157024793,
+      "grad_norm": 3.875,
+      "learning_rate": 0.0001749411915910518,
+      "loss": 4.736,
+      "step": 134550
+    },
+    {
+      "epoch": 1.1587465564738293,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.00017492325109319512,
+      "loss": 4.4008,
+      "step": 134600
+    },
+    {
+      "epoch": 1.159176997245179,
+      "grad_norm": 0.9921875,
+      "learning_rate": 0.00017490530509613303,
+      "loss": 4.2166,
+      "step": 134650
+    },
+    {
+      "epoch": 1.1596074380165289,
+      "grad_norm": 1.625,
+      "learning_rate": 0.00017488735360118266,
+      "loss": 5.058,
+      "step": 134700
+    },
+    {
+      "epoch": 1.160037878787879,
+      "grad_norm": 4.1875,
+      "learning_rate": 0.00017486939660966172,
+      "loss": 4.625,
+      "step": 134750
+    },
+    {
+      "epoch": 1.1604683195592287,
+      "grad_norm": 1.140625,
+      "learning_rate": 0.00017485143412288812,
+      "loss": 4.2593,
+      "step": 134800
+    },
+    {
+      "epoch": 1.1608987603305785,
+      "grad_norm": 2.265625,
+      "learning_rate": 0.0001748334661421803,
+      "loss": 4.8134,
+      "step": 134850
+    },
+    {
+      "epoch": 1.1613292011019283,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.00017481549266885709,
+      "loss": 4.8117,
+      "step": 134900
+    },
+    {
+      "epoch": 1.1617596418732783,
+      "grad_norm": 1.296875,
+      "learning_rate": 0.00017479751370423768,
+      "loss": 4.5121,
+      "step": 134950
+    },
+    {
+      "epoch": 1.162190082644628,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.00017477952924964169,
+      "loss": 4.7107,
+      "step": 135000
+    },
+    {
+      "epoch": 1.162190082644628,
+      "eval_loss": 5.229957580566406,
+      "eval_runtime": 21.9034,
+      "eval_samples_per_second": 29.219,
+      "eval_steps_per_second": 14.61,
+      "eval_tts_loss": 7.000764441175345,
+      "step": 135000
+    },
+    {
+      "epoch": 1.1626205234159779,
+      "grad_norm": 1.203125,
+      "learning_rate": 0.00017476153930638915,
+      "loss": 4.3452,
+      "step": 135050
+    },
+    {
+      "epoch": 1.163050964187328,
+      "grad_norm": 3.359375,
+      "learning_rate": 0.00017474354387580045,
+      "loss": 4.7001,
+      "step": 135100
+    },
+    {
+      "epoch": 1.1634814049586777,
+      "grad_norm": 5.71875,
+      "learning_rate": 0.0001747255429591965,
+      "loss": 4.404,
+      "step": 135150
+    },
+    {
+      "epoch": 1.1639118457300275,
+      "grad_norm": 3.34375,
+      "learning_rate": 0.00017470753655789842,
+      "loss": 4.6584,
+      "step": 135200
+    },
+    {
+      "epoch": 1.1643422865013775,
+      "grad_norm": 2.265625,
+      "learning_rate": 0.00017468952467322794,
+      "loss": 4.3545,
+      "step": 135250
+    },
+    {
+      "epoch": 1.1647727272727273,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00017467150730650702,
+      "loss": 4.4773,
+      "step": 135300
+    },
+    {
+      "epoch": 1.165203168044077,
+      "grad_norm": 3.3125,
+      "learning_rate": 0.0001746534844590581,
+      "loss": 4.7949,
+      "step": 135350
+    },
+    {
+      "epoch": 1.165633608815427,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00017463545613220408,
+      "loss": 4.1822,
+      "step": 135400
+    },
+    {
+      "epoch": 1.166064049586777,
+      "grad_norm": 3.8125,
+      "learning_rate": 0.00017461742232726817,
+      "loss": 4.853,
+      "step": 135450
+    },
+    {
+      "epoch": 1.1664944903581267,
+      "grad_norm": 2.375,
+      "learning_rate": 0.00017459938304557396,
+      "loss": 4.443,
+      "step": 135500
+    },
+    {
+      "epoch": 1.1669249311294765,
+      "grad_norm": 3.5,
+      "learning_rate": 0.00017458133828844558,
+      "loss": 4.301,
+      "step": 135550
+    },
+    {
+      "epoch": 1.1673553719008265,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.00017456328805720746,
+      "loss": 4.9344,
+      "step": 135600
+    },
+    {
+      "epoch": 1.1677858126721763,
+      "grad_norm": 1.875,
+      "learning_rate": 0.00017454523235318437,
+      "loss": 4.4283,
+      "step": 135650
+    },
+    {
+      "epoch": 1.168216253443526,
+      "grad_norm": 1.8203125,
+      "learning_rate": 0.00017452717117770163,
+      "loss": 4.5411,
+      "step": 135700
+    },
+    {
+      "epoch": 1.168646694214876,
+      "grad_norm": 3.40625,
+      "learning_rate": 0.00017450910453208483,
+      "loss": 4.6708,
+      "step": 135750
+    },
+    {
+      "epoch": 1.169077134986226,
+      "grad_norm": 2.234375,
+      "learning_rate": 0.00017449103241766017,
+      "loss": 5.1022,
+      "step": 135800
+    },
+    {
+      "epoch": 1.1695075757575757,
+      "grad_norm": 3.0625,
+      "learning_rate": 0.00017447295483575393,
+      "loss": 4.0141,
+      "step": 135850
+    },
+    {
+      "epoch": 1.1699380165289257,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.00017445487178769305,
+      "loss": 4.846,
+      "step": 135900
+    },
+    {
+      "epoch": 1.1703684573002755,
+      "grad_norm": 3.671875,
+      "learning_rate": 0.00017443678327480478,
+      "loss": 4.5868,
+      "step": 135950
+    },
+    {
+      "epoch": 1.1707988980716253,
+      "grad_norm": 1.2890625,
+      "learning_rate": 0.00017441868929841676,
+      "loss": 4.2212,
+      "step": 136000
+    },
+    {
+      "epoch": 1.1712293388429753,
+      "grad_norm": 3.234375,
+      "learning_rate": 0.00017440058985985707,
+      "loss": 5.0071,
+      "step": 136050
+    },
+    {
+      "epoch": 1.171659779614325,
+      "grad_norm": 3.203125,
+      "learning_rate": 0.00017438248496045414,
+      "loss": 4.5556,
+      "step": 136100
+    },
+    {
+      "epoch": 1.172090220385675,
+      "grad_norm": 1.25,
+      "learning_rate": 0.0001743643746015369,
+      "loss": 4.8886,
+      "step": 136150
+    },
+    {
+      "epoch": 1.1725206611570247,
+      "grad_norm": 3.0,
+      "learning_rate": 0.00017434625878443453,
+      "loss": 4.6341,
+      "step": 136200
+    },
+    {
+      "epoch": 1.1729511019283747,
+      "grad_norm": 5.65625,
+      "learning_rate": 0.00017432813751047674,
+      "loss": 4.278,
+      "step": 136250
+    },
+    {
+      "epoch": 1.1733815426997245,
+      "grad_norm": 3.59375,
+      "learning_rate": 0.0001743100107809936,
+      "loss": 4.5555,
+      "step": 136300
+    },
+    {
+      "epoch": 1.1738119834710743,
+      "grad_norm": 5.15625,
+      "learning_rate": 0.0001742918785973155,
+      "loss": 4.6431,
+      "step": 136350
+    },
+    {
+      "epoch": 1.1742424242424243,
+      "grad_norm": 4.34375,
+      "learning_rate": 0.00017427374096077336,
+      "loss": 4.5562,
+      "step": 136400
+    },
+    {
+      "epoch": 1.174672865013774,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.00017425559787269848,
+      "loss": 4.4798,
+      "step": 136450
+    },
+    {
+      "epoch": 1.175103305785124,
+      "grad_norm": 3.8125,
+      "learning_rate": 0.00017423744933442242,
+      "loss": 4.4478,
+      "step": 136500
+    },
+    {
+      "epoch": 1.175533746556474,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.00017421929534727735,
+      "loss": 4.6886,
+      "step": 136550
+    },
+    {
+      "epoch": 1.1759641873278237,
+      "grad_norm": 1.9609375,
+      "learning_rate": 0.00017420113591259568,
+      "loss": 4.5264,
+      "step": 136600
+    },
+    {
+      "epoch": 1.1763946280991735,
+      "grad_norm": 1.8046875,
+      "learning_rate": 0.00017418297103171027,
+      "loss": 4.8383,
+      "step": 136650
+    },
+    {
+      "epoch": 1.1768250688705235,
+      "grad_norm": 3.171875,
+      "learning_rate": 0.00017416480070595443,
+      "loss": 4.7409,
+      "step": 136700
+    },
+    {
+      "epoch": 1.1772555096418733,
+      "grad_norm": 1.8359375,
+      "learning_rate": 0.00017414662493666174,
+      "loss": 4.5214,
+      "step": 136750
+    },
+    {
+      "epoch": 1.177685950413223,
+      "grad_norm": 3.5625,
+      "learning_rate": 0.00017412844372516632,
+      "loss": 4.6205,
+      "step": 136800
+    },
+    {
+      "epoch": 1.178116391184573,
+      "grad_norm": 3.109375,
+      "learning_rate": 0.00017411025707280263,
+      "loss": 4.6916,
+      "step": 136850
+    },
+    {
+      "epoch": 1.178546831955923,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.00017409206498090554,
+      "loss": 4.3895,
+      "step": 136900
+    },
+    {
+      "epoch": 1.1789772727272727,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00017407386745081028,
+      "loss": 4.3466,
+      "step": 136950
+    },
+    {
+      "epoch": 1.1794077134986225,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.00017405566448385254,
+      "loss": 4.6823,
+      "step": 137000
+    },
+    {
+      "epoch": 1.1798381542699725,
+      "grad_norm": 4.28125,
+      "learning_rate": 0.00017403745608136835,
+      "loss": 4.1412,
+      "step": 137050
+    },
+    {
+      "epoch": 1.1802685950413223,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.00017401924224469414,
+      "loss": 4.5917,
+      "step": 137100
+    },
+    {
+      "epoch": 1.180699035812672,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00017400102297516687,
+      "loss": 4.5987,
+      "step": 137150
+    },
+    {
+      "epoch": 1.1811294765840221,
+      "grad_norm": 3.546875,
+      "learning_rate": 0.00017398279827412369,
+      "loss": 4.1835,
+      "step": 137200
+    },
+    {
+      "epoch": 1.181559917355372,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00017396456814290233,
+      "loss": 4.5742,
+      "step": 137250
+    },
+    {
+      "epoch": 1.1819903581267217,
+      "grad_norm": 3.671875,
+      "learning_rate": 0.00017394633258284076,
+      "loss": 4.5937,
+      "step": 137300
+    },
+    {
+      "epoch": 1.1824207988980717,
+      "grad_norm": 3.890625,
+      "learning_rate": 0.0001739280915952775,
+      "loss": 4.6926,
+      "step": 137350
+    },
+    {
+      "epoch": 1.1828512396694215,
+      "grad_norm": 3.46875,
+      "learning_rate": 0.00017390984518155138,
+      "loss": 4.3917,
+      "step": 137400
+    },
+    {
+      "epoch": 1.1832816804407713,
+      "grad_norm": 1.296875,
+      "learning_rate": 0.00017389159334300165,
+      "loss": 4.4966,
+      "step": 137450
+    },
+    {
+      "epoch": 1.183712121212121,
+      "grad_norm": 2.9375,
+      "learning_rate": 0.00017387333608096794,
+      "loss": 4.4538,
+      "step": 137500
+    },
+    {
+      "epoch": 1.1841425619834711,
+      "grad_norm": 4.09375,
+      "learning_rate": 0.0001738550733967903,
+      "loss": 4.8217,
+      "step": 137550
+    },
+    {
+      "epoch": 1.184573002754821,
+      "grad_norm": 1.5234375,
+      "learning_rate": 0.00017383680529180922,
+      "loss": 4.5733,
+      "step": 137600
+    },
+    {
+      "epoch": 1.1850034435261707,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00017381853176736545,
+      "loss": 4.3104,
+      "step": 137650
+    },
+    {
+      "epoch": 1.1854338842975207,
+      "grad_norm": 1.6953125,
+      "learning_rate": 0.0001738002528248003,
+      "loss": 4.766,
+      "step": 137700
+    },
+    {
+      "epoch": 1.1858643250688705,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00017378196846545534,
+      "loss": 4.4348,
+      "step": 137750
+    },
+    {
+      "epoch": 1.1862947658402203,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.00017376367869067268,
+      "loss": 4.7157,
+      "step": 137800
+    },
+    {
+      "epoch": 1.1867252066115703,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.0001737453835017947,
+      "loss": 4.4045,
+      "step": 137850
+    },
+    {
+      "epoch": 1.1871556473829201,
+      "grad_norm": 3.15625,
+      "learning_rate": 0.00017372708290016423,
+      "loss": 4.7365,
+      "step": 137900
+    },
+    {
+      "epoch": 1.18758608815427,
+      "grad_norm": 1.4140625,
+      "learning_rate": 0.00017370877688712447,
+      "loss": 4.3317,
+      "step": 137950
+    },
+    {
+      "epoch": 1.18801652892562,
+      "grad_norm": 3.125,
+      "learning_rate": 0.0001736904654640191,
+      "loss": 4.591,
+      "step": 138000
+    },
+    {
+      "epoch": 1.18801652892562,
+      "eval_loss": 5.221219062805176,
+      "eval_runtime": 21.6774,
+      "eval_samples_per_second": 29.524,
+      "eval_steps_per_second": 14.762,
+      "eval_tts_loss": 7.074069197268985,
+      "step": 138000
+    },
+    {
+      "epoch": 1.1884469696969697,
+      "grad_norm": 1.2421875,
+      "learning_rate": 0.00017367214863219215,
+      "loss": 4.481,
+      "step": 138050
+    },
+    {
+      "epoch": 1.1888774104683195,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.00017365382639298796,
+      "loss": 4.6663,
+      "step": 138100
+    },
+    {
+      "epoch": 1.1893078512396693,
+      "grad_norm": 1.1875,
+      "learning_rate": 0.0001736354987477514,
+      "loss": 4.6344,
+      "step": 138150
+    },
+    {
+      "epoch": 1.1897382920110193,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.00017361716569782765,
+      "loss": 4.5625,
+      "step": 138200
+    },
+    {
+      "epoch": 1.1901687327823691,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00017359882724456233,
+      "loss": 4.4223,
+      "step": 138250
+    },
+    {
+      "epoch": 1.190599173553719,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.00017358048338930146,
+      "loss": 4.7565,
+      "step": 138300
+    },
+    {
+      "epoch": 1.191029614325069,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00017356213413339142,
+      "loss": 4.3504,
+      "step": 138350
+    },
+    {
+      "epoch": 1.1914600550964187,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.00017354377947817899,
+      "loss": 4.7485,
+      "step": 138400
+    },
+    {
+      "epoch": 1.1918904958677685,
+      "grad_norm": 2.25,
+      "learning_rate": 0.0001735254194250114,
+      "loss": 4.3816,
+      "step": 138450
+    },
+    {
+      "epoch": 1.1923209366391185,
+      "grad_norm": 3.28125,
+      "learning_rate": 0.00017350705397523622,
+      "loss": 4.7629,
+      "step": 138500
+    },
+    {
+      "epoch": 1.1927513774104683,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.0001734886831302014,
+      "loss": 4.496,
+      "step": 138550
+    },
+    {
+      "epoch": 1.1931818181818181,
+      "grad_norm": 4.625,
+      "learning_rate": 0.0001734703068912554,
+      "loss": 4.7535,
+      "step": 138600
+    },
+    {
+      "epoch": 1.1936122589531681,
+      "grad_norm": 1.7265625,
+      "learning_rate": 0.0001734519252597469,
+      "loss": 4.6305,
+      "step": 138650
+    },
+    {
+      "epoch": 1.194042699724518,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.00017343353823702514,
+      "loss": 4.7616,
+      "step": 138700
+    },
+    {
+      "epoch": 1.1944731404958677,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.0001734151458244397,
+      "loss": 4.5813,
+      "step": 138750
+    },
+    {
+      "epoch": 1.1949035812672175,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.00017339674802334048,
+      "loss": 4.5628,
+      "step": 138800
+    },
+    {
+      "epoch": 1.1953340220385675,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00017337834483507786,
+      "loss": 4.5174,
+      "step": 138850
+    },
+    {
+      "epoch": 1.1957644628099173,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.00017335993626100264,
+      "loss": 4.704,
+      "step": 138900
+    },
+    {
+      "epoch": 1.1961949035812671,
+      "grad_norm": 3.28125,
+      "learning_rate": 0.00017334152230246593,
+      "loss": 4.5928,
+      "step": 138950
+    },
+    {
+      "epoch": 1.1966253443526171,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00017332310296081928,
+      "loss": 4.8787,
+      "step": 139000
+    },
+    {
+      "epoch": 1.197055785123967,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.00017330467823741463,
+      "loss": 4.4581,
+      "step": 139050
+    },
+    {
+      "epoch": 1.1974862258953167,
+      "grad_norm": 3.234375,
+      "learning_rate": 0.0001732862481336043,
+      "loss": 4.8128,
+      "step": 139100
+    },
+    {
+      "epoch": 1.1979166666666667,
+      "grad_norm": 3.296875,
+      "learning_rate": 0.00017326781265074103,
+      "loss": 4.6923,
+      "step": 139150
+    },
+    {
+      "epoch": 1.1983471074380165,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.00017324937179017798,
+      "loss": 4.732,
+      "step": 139200
+    },
+    {
+      "epoch": 1.1987775482093663,
+      "grad_norm": 1.6484375,
+      "learning_rate": 0.00017323092555326865,
+      "loss": 4.7558,
+      "step": 139250
+    },
+    {
+      "epoch": 1.1992079889807163,
+      "grad_norm": 2.625,
+      "learning_rate": 0.00017321247394136693,
+      "loss": 4.7072,
+      "step": 139300
+    },
+    {
+      "epoch": 1.1996384297520661,
+      "grad_norm": 3.90625,
+      "learning_rate": 0.00017319401695582712,
+      "loss": 4.7033,
+      "step": 139350
+    },
+    {
+      "epoch": 1.200068870523416,
+      "grad_norm": 3.1875,
+      "learning_rate": 0.00017317555459800396,
+      "loss": 4.3473,
+      "step": 139400
+    },
+    {
+      "epoch": 1.2004993112947657,
+      "grad_norm": 4.25,
+      "learning_rate": 0.00017315708686925255,
+      "loss": 4.7289,
+      "step": 139450
+    },
+    {
+      "epoch": 1.2009297520661157,
+      "grad_norm": 1.1875,
+      "learning_rate": 0.00017313861377092838,
+      "loss": 4.594,
+      "step": 139500
+    },
+    {
+      "epoch": 1.2013601928374655,
+      "grad_norm": 3.078125,
+      "learning_rate": 0.0001731201353043873,
+      "loss": 4.549,
+      "step": 139550
+    },
+    {
+      "epoch": 1.2017906336088156,
+      "grad_norm": 3.140625,
+      "learning_rate": 0.00017310165147098558,
+      "loss": 4.599,
+      "step": 139600
+    },
+    {
+      "epoch": 1.2022210743801653,
+      "grad_norm": 4.15625,
+      "learning_rate": 0.00017308316227207998,
+      "loss": 4.7083,
+      "step": 139650
+    },
+    {
+      "epoch": 1.2026515151515151,
+      "grad_norm": 2.984375,
+      "learning_rate": 0.00017306466770902745,
+      "loss": 3.994,
+      "step": 139700
+    },
+    {
+      "epoch": 1.203081955922865,
+      "grad_norm": 2.625,
+      "learning_rate": 0.00017304616778318556,
+      "loss": 4.6629,
+      "step": 139750
+    },
+    {
+      "epoch": 1.203512396694215,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.0001730276624959121,
+      "loss": 5.0117,
+      "step": 139800
+    },
+    {
+      "epoch": 1.2039428374655647,
+      "grad_norm": 3.109375,
+      "learning_rate": 0.00017300915184856533,
+      "loss": 4.5509,
+      "step": 139850
+    },
+    {
+      "epoch": 1.2043732782369145,
+      "grad_norm": 1.6796875,
+      "learning_rate": 0.00017299063584250393,
+      "loss": 4.5971,
+      "step": 139900
+    },
+    {
+      "epoch": 1.2048037190082646,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.00017297211447908685,
+      "loss": 4.5729,
+      "step": 139950
+    },
+    {
+      "epoch": 1.2052341597796143,
+      "grad_norm": 3.25,
+      "learning_rate": 0.0001729535877596736,
+      "loss": 4.608,
+      "step": 140000
+    },
+    {
+      "epoch": 1.2056646005509641,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.00017293505568562398,
+      "loss": 4.6049,
+      "step": 140050
+    },
+    {
+      "epoch": 1.206095041322314,
+      "grad_norm": 4.75,
+      "learning_rate": 0.00017291651825829818,
+      "loss": 4.6043,
+      "step": 140100
+    },
+    {
+      "epoch": 1.206525482093664,
+      "grad_norm": 3.5,
+      "learning_rate": 0.00017289797547905686,
+      "loss": 4.3925,
+      "step": 140150
+    },
+    {
+      "epoch": 1.2069559228650137,
+      "grad_norm": 3.078125,
+      "learning_rate": 0.00017287942734926093,
+      "loss": 4.3748,
+      "step": 140200
+    },
+    {
+      "epoch": 1.2073863636363638,
+      "grad_norm": 3.609375,
+      "learning_rate": 0.00017286087387027185,
+      "loss": 4.7009,
+      "step": 140250
+    },
+    {
+      "epoch": 1.2078168044077136,
+      "grad_norm": 1.4609375,
+      "learning_rate": 0.0001728423150434514,
+      "loss": 4.4022,
+      "step": 140300
+    },
+    {
+      "epoch": 1.2082472451790633,
+      "grad_norm": 3.1875,
+      "learning_rate": 0.0001728237508701618,
+      "loss": 4.6236,
+      "step": 140350
+    },
+    {
+      "epoch": 1.2086776859504131,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.0001728051813517655,
+      "loss": 4.4696,
+      "step": 140400
+    },
+    {
+      "epoch": 1.2091081267217632,
+      "grad_norm": 3.921875,
+      "learning_rate": 0.00017278660648962558,
+      "loss": 4.4388,
+      "step": 140450
+    },
+    {
+      "epoch": 1.209538567493113,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00017276802628510531,
+      "loss": 4.1941,
+      "step": 140500
+    },
+    {
+      "epoch": 1.2099690082644627,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.0001727494407395685,
+      "loss": 4.5451,
+      "step": 140550
+    },
+    {
+      "epoch": 1.2103994490358128,
+      "grad_norm": 3.40625,
+      "learning_rate": 0.00017273084985437923,
+      "loss": 4.5245,
+      "step": 140600
+    },
+    {
+      "epoch": 1.2108298898071626,
+      "grad_norm": 4.40625,
+      "learning_rate": 0.0001727122536309021,
+      "loss": 4.3755,
+      "step": 140650
+    },
+    {
+      "epoch": 1.2112603305785123,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.000172693652070502,
+      "loss": 4.4781,
+      "step": 140700
+    },
+    {
+      "epoch": 1.2116907713498621,
+      "grad_norm": 3.328125,
+      "learning_rate": 0.00017267504517454424,
+      "loss": 4.1627,
+      "step": 140750
+    },
+    {
+      "epoch": 1.2121212121212122,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.0001726564329443945,
+      "loss": 4.6936,
+      "step": 140800
+    },
+    {
+      "epoch": 1.212551652892562,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.00017263781538141894,
+      "loss": 4.4489,
+      "step": 140850
+    },
+    {
+      "epoch": 1.212982093663912,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.000172619192486984,
+      "loss": 4.6993,
+      "step": 140900
+    },
+    {
+      "epoch": 1.2134125344352618,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.00017260056426245658,
+      "loss": 4.441,
+      "step": 140950
+    },
+    {
+      "epoch": 1.2138429752066116,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.00017258193070920395,
+      "loss": 4.4734,
+      "step": 141000
+    },
+    {
+      "epoch": 1.2138429752066116,
+      "eval_loss": 5.210572719573975,
+      "eval_runtime": 22.0704,
+      "eval_samples_per_second": 28.998,
+      "eval_steps_per_second": 14.499,
+      "eval_tts_loss": 7.0870365647599645,
+      "step": 141000
+    },
+    {
+      "epoch": 1.2142734159779613,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.00017256329182859377,
+      "loss": 4.6075,
+      "step": 141050
+    },
+    {
+      "epoch": 1.2147038567493114,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.00017254464762199407,
+      "loss": 4.1875,
+      "step": 141100
+    },
+    {
+      "epoch": 1.2151342975206612,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.0001725259980907733,
+      "loss": 4.7463,
+      "step": 141150
+    },
+    {
+      "epoch": 1.215564738292011,
+      "grad_norm": 0.81640625,
+      "learning_rate": 0.00017250734323630038,
+      "loss": 4.467,
+      "step": 141200
+    },
+    {
+      "epoch": 1.215995179063361,
+      "grad_norm": 2.125,
+      "learning_rate": 0.00017248868305994445,
+      "loss": 4.8383,
+      "step": 141250
+    },
+    {
+      "epoch": 1.2164256198347108,
+      "grad_norm": 3.078125,
+      "learning_rate": 0.00017247001756307516,
+      "loss": 4.849,
+      "step": 141300
+    },
+    {
+      "epoch": 1.2168560606060606,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.00017245134674706244,
+      "loss": 4.5783,
+      "step": 141350
+    },
+    {
+      "epoch": 1.2172865013774103,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.0001724326706132768,
+      "loss": 4.8766,
+      "step": 141400
+    },
+    {
+      "epoch": 1.2177169421487604,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.000172413989163089,
+      "loss": 4.3604,
+      "step": 141450
+    },
+    {
+      "epoch": 1.2181473829201102,
+      "grad_norm": 2.21875,
+      "learning_rate": 0.00017239530239787016,
+      "loss": 4.611,
+      "step": 141500
+    },
+    {
+      "epoch": 1.2185778236914602,
+      "grad_norm": 2.375,
+      "learning_rate": 0.0001723766103189919,
+      "loss": 4.4779,
+      "step": 141550
+    },
+    {
+      "epoch": 1.21900826446281,
+      "grad_norm": 3.546875,
+      "learning_rate": 0.00017235791292782615,
+      "loss": 4.6852,
+      "step": 141600
+    },
+    {
+      "epoch": 1.2194387052341598,
+      "grad_norm": 2.5,
+      "learning_rate": 0.0001723392102257453,
+      "loss": 4.564,
+      "step": 141650
+    },
+    {
+      "epoch": 1.2198691460055096,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00017232050221412207,
+      "loss": 4.9894,
+      "step": 141700
+    },
+    {
+      "epoch": 1.2202995867768596,
+      "grad_norm": 3.890625,
+      "learning_rate": 0.00017230178889432953,
+      "loss": 5.0421,
+      "step": 141750
+    },
+    {
+      "epoch": 1.2207300275482094,
+      "grad_norm": 1.546875,
+      "learning_rate": 0.0001722830702677413,
+      "loss": 4.5726,
+      "step": 141800
+    },
+    {
+      "epoch": 1.2211604683195592,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.0001722643463357312,
+      "loss": 4.3865,
+      "step": 141850
+    },
+    {
+      "epoch": 1.2215909090909092,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.00017224561709967354,
+      "loss": 4.5016,
+      "step": 141900
+    },
+    {
+      "epoch": 1.222021349862259,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00017222688256094307,
+      "loss": 4.5605,
+      "step": 141950
+    },
+    {
+      "epoch": 1.2224517906336088,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.0001722081427209148,
+      "loss": 4.6273,
+      "step": 142000
+    },
+    {
+      "epoch": 1.2228822314049586,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00017218939758096422,
+      "loss": 4.294,
+      "step": 142050
+    },
+    {
+      "epoch": 1.2233126721763086,
+      "grad_norm": 3.8125,
+      "learning_rate": 0.00017217064714246716,
+      "loss": 4.6043,
+      "step": 142100
+    },
+    {
+      "epoch": 1.2237431129476584,
+      "grad_norm": 2.125,
+      "learning_rate": 0.00017215189140679988,
+      "loss": 4.548,
+      "step": 142150
+    },
+    {
+      "epoch": 1.2241735537190084,
+      "grad_norm": 2.0,
+      "learning_rate": 0.00017213313037533904,
+      "loss": 4.503,
+      "step": 142200
+    },
+    {
+      "epoch": 1.2246039944903582,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.0001721143640494616,
+      "loss": 4.1412,
+      "step": 142250
+    },
+    {
+      "epoch": 1.225034435261708,
+      "grad_norm": 2.578125,
+      "learning_rate": 0.000172095592430545,
+      "loss": 4.5492,
+      "step": 142300
+    },
+    {
+      "epoch": 1.2254648760330578,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.00017207681551996705,
+      "loss": 4.9064,
+      "step": 142350
+    },
+    {
+      "epoch": 1.2258953168044078,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.0001720580333191059,
+      "loss": 4.6188,
+      "step": 142400
+    },
+    {
+      "epoch": 1.2263257575757576,
+      "grad_norm": 3.28125,
+      "learning_rate": 0.00017203924582934014,
+      "loss": 4.3496,
+      "step": 142450
+    },
+    {
+      "epoch": 1.2267561983471074,
+      "grad_norm": 2.984375,
+      "learning_rate": 0.00017202045305204875,
+      "loss": 5.1003,
+      "step": 142500
+    },
+    {
+      "epoch": 1.2271866391184574,
+      "grad_norm": 3.859375,
+      "learning_rate": 0.00017200165498861106,
+      "loss": 4.4041,
+      "step": 142550
+    },
+    {
+      "epoch": 1.2276170798898072,
+      "grad_norm": 3.6875,
+      "learning_rate": 0.00017198285164040682,
+      "loss": 4.5644,
+      "step": 142600
+    },
+    {
+      "epoch": 1.228047520661157,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.00017196404300881614,
+      "loss": 4.4952,
+      "step": 142650
+    },
+    {
+      "epoch": 1.228477961432507,
+      "grad_norm": 3.46875,
+      "learning_rate": 0.00017194522909521947,
+      "loss": 4.7108,
+      "step": 142700
+    },
+    {
+      "epoch": 1.2289084022038568,
+      "grad_norm": 3.84375,
+      "learning_rate": 0.00017192640990099786,
+      "loss": 4.4127,
+      "step": 142750
+    },
+    {
+      "epoch": 1.2293388429752066,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.0001719075854275325,
+      "loss": 4.2333,
+      "step": 142800
+    },
+    {
+      "epoch": 1.2297692837465566,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00017188875567620507,
+      "loss": 4.5475,
+      "step": 142850
+    },
+    {
+      "epoch": 1.2301997245179064,
+      "grad_norm": 3.15625,
+      "learning_rate": 0.00017186992064839766,
+      "loss": 4.4541,
+      "step": 142900
+    },
+    {
+      "epoch": 1.2306301652892562,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.00017185108034549267,
+      "loss": 4.2658,
+      "step": 142950
+    },
+    {
+      "epoch": 1.231060606060606,
+      "grad_norm": 0.9765625,
+      "learning_rate": 0.000171832234768873,
+      "loss": 4.6635,
+      "step": 143000
+    },
+    {
+      "epoch": 1.231491046831956,
+      "grad_norm": 1.7265625,
+      "learning_rate": 0.00017181338391992186,
+      "loss": 4.5951,
+      "step": 143050
+    },
+    {
+      "epoch": 1.2319214876033058,
+      "grad_norm": 3.53125,
+      "learning_rate": 0.00017179452780002286,
+      "loss": 4.8548,
+      "step": 143100
+    },
+    {
+      "epoch": 1.2323519283746556,
+      "grad_norm": 2.984375,
+      "learning_rate": 0.00017177566641055994,
+      "loss": 5.0237,
+      "step": 143150
+    },
+    {
+      "epoch": 1.2327823691460056,
+      "grad_norm": 2.5,
+      "learning_rate": 0.00017175679975291755,
+      "loss": 4.4822,
+      "step": 143200
+    },
+    {
+      "epoch": 1.2332128099173554,
+      "grad_norm": 3.734375,
+      "learning_rate": 0.00017173792782848048,
+      "loss": 4.8718,
+      "step": 143250
+    },
+    {
+      "epoch": 1.2336432506887052,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.00017171905063863386,
+      "loss": 4.4284,
+      "step": 143300
+    },
+    {
+      "epoch": 1.2340736914600552,
+      "grad_norm": 3.265625,
+      "learning_rate": 0.0001717001681847632,
+      "loss": 4.2758,
+      "step": 143350
+    },
+    {
+      "epoch": 1.234504132231405,
+      "grad_norm": 3.359375,
+      "learning_rate": 0.00017168128046825447,
+      "loss": 4.6864,
+      "step": 143400
+    },
+    {
+      "epoch": 1.2349345730027548,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.00017166238749049398,
+      "loss": 4.1353,
+      "step": 143450
+    },
+    {
+      "epoch": 1.2353650137741048,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00017164348925286844,
+      "loss": 4.3424,
+      "step": 143500
+    },
+    {
+      "epoch": 1.2357954545454546,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.00017162458575676492,
+      "loss": 4.51,
+      "step": 143550
+    },
+    {
+      "epoch": 1.2362258953168044,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.0001716056770035709,
+      "loss": 4.2697,
+      "step": 143600
+    },
+    {
+      "epoch": 1.2366563360881542,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.00017158676299467427,
+      "loss": 4.4364,
+      "step": 143650
+    },
+    {
+      "epoch": 1.2370867768595042,
+      "grad_norm": 4.25,
+      "learning_rate": 0.00017156784373146327,
+      "loss": 4.6218,
+      "step": 143700
+    },
+    {
+      "epoch": 1.237517217630854,
+      "grad_norm": 3.140625,
+      "learning_rate": 0.00017154891921532651,
+      "loss": 4.691,
+      "step": 143750
+    },
+    {
+      "epoch": 1.2379476584022038,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.000171529989447653,
+      "loss": 4.5416,
+      "step": 143800
+    },
+    {
+      "epoch": 1.2383780991735538,
+      "grad_norm": 3.109375,
+      "learning_rate": 0.00017151105442983223,
+      "loss": 4.651,
+      "step": 143850
+    },
+    {
+      "epoch": 1.2388085399449036,
+      "grad_norm": 1.4453125,
+      "learning_rate": 0.00017149211416325386,
+      "loss": 4.6083,
+      "step": 143900
+    },
+    {
+      "epoch": 1.2392389807162534,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00017147316864930814,
+      "loss": 4.3779,
+      "step": 143950
+    },
+    {
+      "epoch": 1.2396694214876034,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.00017145421788938565,
+      "loss": 4.4656,
+      "step": 144000
+    },
+    {
+      "epoch": 1.2396694214876034,
+      "eval_loss": 5.19620418548584,
+      "eval_runtime": 21.8281,
+      "eval_samples_per_second": 29.32,
+      "eval_steps_per_second": 14.66,
+      "eval_tts_loss": 7.050007847587131,
+      "step": 144000
+    },
+    {
+      "epoch": 1.2400998622589532,
+      "grad_norm": 0.984375,
+      "learning_rate": 0.00017143526188487726,
+      "loss": 4.7263,
+      "step": 144050
+    },
+    {
+      "epoch": 1.240530303030303,
+      "grad_norm": 4.0625,
+      "learning_rate": 0.0001714163006371744,
+      "loss": 4.5929,
+      "step": 144100
+    },
+    {
+      "epoch": 1.240960743801653,
+      "grad_norm": 4.34375,
+      "learning_rate": 0.00017139733414766868,
+      "loss": 4.6241,
+      "step": 144150
+    },
+    {
+      "epoch": 1.2413911845730028,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.0001713783624177523,
+      "loss": 4.5753,
+      "step": 144200
+    },
+    {
+      "epoch": 1.2418216253443526,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.0001713593854488176,
+      "loss": 4.59,
+      "step": 144250
+    },
+    {
+      "epoch": 1.2422520661157024,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00017134040324225764,
+      "loss": 4.6879,
+      "step": 144300
+    },
+    {
+      "epoch": 1.2426825068870524,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00017132141579946555,
+      "loss": 4.5068,
+      "step": 144350
+    },
+    {
+      "epoch": 1.2431129476584022,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00017130242312183498,
+      "loss": 4.7414,
+      "step": 144400
+    },
+    {
+      "epoch": 1.243543388429752,
+      "grad_norm": 1.734375,
+      "learning_rate": 0.00017128342521075996,
+      "loss": 4.7805,
+      "step": 144450
+    },
+    {
+      "epoch": 1.243973829201102,
+      "grad_norm": 4.53125,
+      "learning_rate": 0.00017126442206763488,
+      "loss": 4.8175,
+      "step": 144500
+    },
+    {
+      "epoch": 1.2444042699724518,
+      "grad_norm": 1.65625,
+      "learning_rate": 0.00017124541369385458,
+      "loss": 4.2131,
+      "step": 144550
+    },
+    {
+      "epoch": 1.2448347107438016,
+      "grad_norm": 1.8984375,
+      "learning_rate": 0.00017122640009081417,
+      "loss": 4.4931,
+      "step": 144600
+    },
+    {
+      "epoch": 1.2452651515151516,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.00017120738125990932,
+      "loss": 4.7294,
+      "step": 144650
+    },
+    {
+      "epoch": 1.2456955922865014,
+      "grad_norm": 3.5625,
+      "learning_rate": 0.00017118835720253582,
+      "loss": 4.7757,
+      "step": 144700
+    },
+    {
+      "epoch": 1.2461260330578512,
+      "grad_norm": 1.796875,
+      "learning_rate": 0.00017116932792009008,
+      "loss": 4.3835,
+      "step": 144750
+    },
+    {
+      "epoch": 1.2465564738292012,
+      "grad_norm": 4.5,
+      "learning_rate": 0.0001711502934139688,
+      "loss": 4.8918,
+      "step": 144800
+    },
+    {
+      "epoch": 1.246986914600551,
+      "grad_norm": 3.46875,
+      "learning_rate": 0.0001711312536855691,
+      "loss": 4.6512,
+      "step": 144850
+    },
+    {
+      "epoch": 1.2474173553719008,
+      "grad_norm": 1.6796875,
+      "learning_rate": 0.00017111220873628838,
+      "loss": 4.7372,
+      "step": 144900
+    },
+    {
+      "epoch": 1.2478477961432506,
+      "grad_norm": 3.75,
+      "learning_rate": 0.00017109315856752455,
+      "loss": 4.6873,
+      "step": 144950
+    },
+    {
+      "epoch": 1.2482782369146006,
+      "grad_norm": 4.65625,
+      "learning_rate": 0.00017107410318067582,
+      "loss": 4.3637,
+      "step": 145000
+    },
+    {
+      "epoch": 1.2487086776859504,
+      "grad_norm": 3.265625,
+      "learning_rate": 0.00017105504257714088,
+      "loss": 4.6875,
+      "step": 145050
+    },
+    {
+      "epoch": 1.2491391184573002,
+      "grad_norm": 3.578125,
+      "learning_rate": 0.0001710359767583187,
+      "loss": 4.4667,
+      "step": 145100
+    },
+    {
+      "epoch": 1.2495695592286502,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00017101690572560863,
+      "loss": 4.5576,
+      "step": 145150
+    },
+    {
+      "epoch": 1.25,
+      "grad_norm": 4.0,
+      "learning_rate": 0.00017099782948041048,
+      "loss": 4.7459,
+      "step": 145200
+    },
+    {
+      "epoch": 1.2504304407713498,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.0001709787480241244,
+      "loss": 4.8452,
+      "step": 145250
+    },
+    {
+      "epoch": 1.2508608815426996,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.00017095966135815093,
+      "loss": 5.022,
+      "step": 145300
+    },
+    {
+      "epoch": 1.2512913223140496,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.000170940569483891,
+      "loss": 4.8232,
+      "step": 145350
+    },
+    {
+      "epoch": 1.2517217630853994,
+      "grad_norm": 3.390625,
+      "learning_rate": 0.0001709214724027459,
+      "loss": 4.9928,
+      "step": 145400
+    },
+    {
+      "epoch": 1.2521522038567494,
+      "grad_norm": 0.828125,
+      "learning_rate": 0.0001709023701161173,
+      "loss": 4.7853,
+      "step": 145450
+    },
+    {
+      "epoch": 1.2525826446280992,
+      "grad_norm": 1.859375,
+      "learning_rate": 0.00017088326262540733,
+      "loss": 4.1903,
+      "step": 145500
+    },
+    {
+      "epoch": 1.253013085399449,
+      "grad_norm": 2.0,
+      "learning_rate": 0.00017086414993201836,
+      "loss": 4.553,
+      "step": 145550
+    },
+    {
+      "epoch": 1.2534435261707988,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00017084503203735322,
+      "loss": 4.5055,
+      "step": 145600
+    },
+    {
+      "epoch": 1.2538739669421488,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.0001708259089428152,
+      "loss": 4.3498,
+      "step": 145650
+    },
+    {
+      "epoch": 1.2543044077134986,
+      "grad_norm": 4.625,
+      "learning_rate": 0.00017080678064980784,
+      "loss": 4.765,
+      "step": 145700
+    },
+    {
+      "epoch": 1.2547348484848486,
+      "grad_norm": 4.40625,
+      "learning_rate": 0.00017078764715973513,
+      "loss": 4.595,
+      "step": 145750
+    },
+    {
+      "epoch": 1.2551652892561984,
+      "grad_norm": 3.515625,
+      "learning_rate": 0.00017076850847400144,
+      "loss": 4.4891,
+      "step": 145800
+    },
+    {
+      "epoch": 1.2555957300275482,
+      "grad_norm": 1.46875,
+      "learning_rate": 0.00017074936459401146,
+      "loss": 4.5071,
+      "step": 145850
+    },
+    {
+      "epoch": 1.256026170798898,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.00017073021552117034,
+      "loss": 4.5827,
+      "step": 145900
+    },
+    {
+      "epoch": 1.256456611570248,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.00017071106125688358,
+      "loss": 4.9186,
+      "step": 145950
+    },
+    {
+      "epoch": 1.2568870523415978,
+      "grad_norm": 2.5,
+      "learning_rate": 0.0001706919018025571,
+      "loss": 4.7654,
+      "step": 146000
+    },
+    {
+      "epoch": 1.2573174931129476,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.00017067273715959709,
+      "loss": 4.5248,
+      "step": 146050
+    },
+    {
+      "epoch": 1.2577479338842976,
+      "grad_norm": 1.515625,
+      "learning_rate": 0.00017065356732941022,
+      "loss": 4.8473,
+      "step": 146100
+    },
+    {
+      "epoch": 1.2581783746556474,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00017063439231340354,
+      "loss": 3.9998,
+      "step": 146150
+    },
+    {
+      "epoch": 1.2586088154269972,
+      "grad_norm": 1.953125,
+      "learning_rate": 0.00017061521211298442,
+      "loss": 4.1946,
+      "step": 146200
+    },
+    {
+      "epoch": 1.259039256198347,
+      "grad_norm": 3.265625,
+      "learning_rate": 0.0001705960267295607,
+      "loss": 4.3768,
+      "step": 146250
+    },
+    {
+      "epoch": 1.259469696969697,
+      "grad_norm": 1.9921875,
+      "learning_rate": 0.00017057683616454047,
+      "loss": 4.7711,
+      "step": 146300
+    },
+    {
+      "epoch": 1.2599001377410468,
+      "grad_norm": 3.3125,
+      "learning_rate": 0.00017055764041933237,
+      "loss": 4.4223,
+      "step": 146350
+    },
+    {
+      "epoch": 1.2603305785123968,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.00017053843949534523,
+      "loss": 4.1182,
+      "step": 146400
+    },
+    {
+      "epoch": 1.2607610192837466,
+      "grad_norm": 1.546875,
+      "learning_rate": 0.00017051923339398838,
+      "loss": 4.5495,
+      "step": 146450
+    },
+    {
+      "epoch": 1.2611914600550964,
+      "grad_norm": 3.78125,
+      "learning_rate": 0.00017050002211667158,
+      "loss": 4.3128,
+      "step": 146500
+    },
+    {
+      "epoch": 1.2616219008264462,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.0001704808056648048,
+      "loss": 4.417,
+      "step": 146550
+    },
+    {
+      "epoch": 1.2620523415977962,
+      "grad_norm": 2.21875,
+      "learning_rate": 0.00017046158403979855,
+      "loss": 4.7308,
+      "step": 146600
+    },
+    {
+      "epoch": 1.262482782369146,
+      "grad_norm": 1.8828125,
+      "learning_rate": 0.00017044235724306361,
+      "loss": 4.5017,
+      "step": 146650
+    },
+    {
+      "epoch": 1.2629132231404958,
+      "grad_norm": 2.125,
+      "learning_rate": 0.00017042312527601122,
+      "loss": 4.491,
+      "step": 146700
+    },
+    {
+      "epoch": 1.2633436639118458,
+      "grad_norm": 1.2578125,
+      "learning_rate": 0.00017040388814005294,
+      "loss": 4.6107,
+      "step": 146750
+    },
+    {
+      "epoch": 1.2637741046831956,
+      "grad_norm": 1.3359375,
+      "learning_rate": 0.00017038464583660077,
+      "loss": 4.4281,
+      "step": 146800
+    },
+    {
+      "epoch": 1.2642045454545454,
+      "grad_norm": 3.625,
+      "learning_rate": 0.00017036539836706705,
+      "loss": 4.4565,
+      "step": 146850
+    },
+    {
+      "epoch": 1.2646349862258952,
+      "grad_norm": 3.640625,
+      "learning_rate": 0.00017034614573286443,
+      "loss": 4.3151,
+      "step": 146900
+    },
+    {
+      "epoch": 1.2650654269972452,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.0001703268879354061,
+      "loss": 4.4134,
+      "step": 146950
+    },
+    {
+      "epoch": 1.265495867768595,
+      "grad_norm": 1.8125,
+      "learning_rate": 0.00017030762497610553,
+      "loss": 4.6642,
+      "step": 147000
+    },
+    {
+      "epoch": 1.265495867768595,
+      "eval_loss": 5.188155651092529,
+      "eval_runtime": 21.8393,
+      "eval_samples_per_second": 29.305,
+      "eval_steps_per_second": 14.653,
+      "eval_tts_loss": 7.0855276917543994,
+      "step": 147000
+    },
+    {
+      "epoch": 1.265926308539945,
+      "grad_norm": 3.015625,
+      "learning_rate": 0.00017028835685637653,
+      "loss": 4.7554,
+      "step": 147050
+    },
+    {
+      "epoch": 1.2663567493112948,
+      "grad_norm": 1.265625,
+      "learning_rate": 0.00017026908357763336,
+      "loss": 4.7875,
+      "step": 147100
+    },
+    {
+      "epoch": 1.2667871900826446,
+      "grad_norm": 1.984375,
+      "learning_rate": 0.00017024980514129064,
+      "loss": 4.5771,
+      "step": 147150
+    },
+    {
+      "epoch": 1.2672176308539944,
+      "grad_norm": 1.84375,
+      "learning_rate": 0.0001702305215487634,
+      "loss": 4.3038,
+      "step": 147200
+    },
+    {
+      "epoch": 1.2676480716253444,
+      "grad_norm": 4.96875,
+      "learning_rate": 0.00017021123280146696,
+      "loss": 4.801,
+      "step": 147250
+    },
+    {
+      "epoch": 1.2680785123966942,
+      "grad_norm": 3.65625,
+      "learning_rate": 0.00017019193890081708,
+      "loss": 4.3845,
+      "step": 147300
+    },
+    {
+      "epoch": 1.268508953168044,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00017017263984822994,
+      "loss": 4.2734,
+      "step": 147350
+    },
+    {
+      "epoch": 1.268939393939394,
+      "grad_norm": 1.5625,
+      "learning_rate": 0.000170153335645122,
+      "loss": 4.4577,
+      "step": 147400
+    },
+    {
+      "epoch": 1.2693698347107438,
+      "grad_norm": 2.125,
+      "learning_rate": 0.00017013402629291019,
+      "loss": 3.9716,
+      "step": 147450
+    },
+    {
+      "epoch": 1.2698002754820936,
+      "grad_norm": 3.78125,
+      "learning_rate": 0.0001701147117930117,
+      "loss": 4.5053,
+      "step": 147500
+    },
+    {
+      "epoch": 1.2702307162534434,
+      "grad_norm": 1.5703125,
+      "learning_rate": 0.00017009539214684423,
+      "loss": 4.2726,
+      "step": 147550
+    },
+    {
+      "epoch": 1.2706611570247934,
+      "grad_norm": 4.125,
+      "learning_rate": 0.0001700760673558258,
+      "loss": 4.6494,
+      "step": 147600
+    },
+    {
+      "epoch": 1.2710915977961432,
+      "grad_norm": 5.65625,
+      "learning_rate": 0.0001700567374213748,
+      "loss": 4.7959,
+      "step": 147650
+    },
+    {
+      "epoch": 1.2715220385674932,
+      "grad_norm": 2.234375,
+      "learning_rate": 0.00017003740234491002,
+      "loss": 4.5812,
+      "step": 147700
+    },
+    {
+      "epoch": 1.271952479338843,
+      "grad_norm": 3.625,
+      "learning_rate": 0.00017001806212785058,
+      "loss": 4.5615,
+      "step": 147750
+    },
+    {
+      "epoch": 1.2723829201101928,
+      "grad_norm": 2.140625,
+      "learning_rate": 0.00016999871677161603,
+      "loss": 4.6125,
+      "step": 147800
+    },
+    {
+      "epoch": 1.2728133608815426,
+      "grad_norm": 3.5625,
+      "learning_rate": 0.00016997936627762626,
+      "loss": 4.3393,
+      "step": 147850
+    },
+    {
+      "epoch": 1.2732438016528926,
+      "grad_norm": 3.296875,
+      "learning_rate": 0.0001699600106473016,
+      "loss": 4.7559,
+      "step": 147900
+    },
+    {
+      "epoch": 1.2736742424242424,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.00016994064988206268,
+      "loss": 4.6604,
+      "step": 147950
+    },
+    {
+      "epoch": 1.2741046831955922,
+      "grad_norm": 1.703125,
+      "learning_rate": 0.0001699212839833305,
+      "loss": 4.7125,
+      "step": 148000
+    },
+    {
+      "epoch": 1.2745351239669422,
+      "grad_norm": 3.25,
+      "learning_rate": 0.00016990191295252656,
+      "loss": 4.5814,
+      "step": 148050
+    },
+    {
+      "epoch": 1.274965564738292,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00016988253679107258,
+      "loss": 4.4965,
+      "step": 148100
+    },
+    {
+      "epoch": 1.2753960055096418,
+      "grad_norm": 3.046875,
+      "learning_rate": 0.00016986315550039074,
+      "loss": 4.9009,
+      "step": 148150
+    },
+    {
+      "epoch": 1.2758264462809916,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.00016984376908190364,
+      "loss": 4.6258,
+      "step": 148200
+    },
+    {
+      "epoch": 1.2762568870523416,
+      "grad_norm": 3.671875,
+      "learning_rate": 0.00016982437753703414,
+      "loss": 4.5483,
+      "step": 148250
+    },
+    {
+      "epoch": 1.2766873278236914,
+      "grad_norm": 3.875,
+      "learning_rate": 0.0001698049808672056,
+      "loss": 4.4415,
+      "step": 148300
+    },
+    {
+      "epoch": 1.2771177685950414,
+      "grad_norm": 2.421875,
+      "learning_rate": 0.00016978557907384157,
+      "loss": 4.4254,
+      "step": 148350
+    },
+    {
+      "epoch": 1.2775482093663912,
+      "grad_norm": 3.203125,
+      "learning_rate": 0.00016976617215836627,
+      "loss": 4.495,
+      "step": 148400
+    },
+    {
+      "epoch": 1.277978650137741,
+      "grad_norm": 3.234375,
+      "learning_rate": 0.00016974676012220397,
+      "loss": 4.2674,
+      "step": 148450
+    },
+    {
+      "epoch": 1.2784090909090908,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.00016972734296677962,
+      "loss": 4.5779,
+      "step": 148500
+    },
+    {
+      "epoch": 1.2788395316804408,
+      "grad_norm": 1.8984375,
+      "learning_rate": 0.00016970792069351823,
+      "loss": 4.622,
+      "step": 148550
+    },
+    {
+      "epoch": 1.2792699724517906,
+      "grad_norm": 3.953125,
+      "learning_rate": 0.0001696884933038455,
+      "loss": 4.37,
+      "step": 148600
+    },
+    {
+      "epoch": 1.2797004132231404,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00016966906079918733,
+      "loss": 3.8526,
+      "step": 148650
+    },
+    {
+      "epoch": 1.2801308539944904,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.0001696496231809699,
+      "loss": 4.3262,
+      "step": 148700
+    },
+    {
+      "epoch": 1.2805612947658402,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.00016963018045062004,
+      "loss": 4.5249,
+      "step": 148750
+    },
+    {
+      "epoch": 1.28099173553719,
+      "grad_norm": 3.4375,
+      "learning_rate": 0.00016961073260956475,
+      "loss": 4.7711,
+      "step": 148800
+    },
+    {
+      "epoch": 1.2814221763085398,
+      "grad_norm": 3.3125,
+      "learning_rate": 0.00016959127965923142,
+      "loss": 4.559,
+      "step": 148850
+    },
+    {
+      "epoch": 1.2818526170798898,
+      "grad_norm": 1.6796875,
+      "learning_rate": 0.00016957182160104794,
+      "loss": 4.3972,
+      "step": 148900
+    },
+    {
+      "epoch": 1.2822830578512396,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.00016955235843644242,
+      "loss": 4.5368,
+      "step": 148950
+    },
+    {
+      "epoch": 1.2827134986225897,
+      "grad_norm": 4.25,
+      "learning_rate": 0.00016953289016684344,
+      "loss": 4.6124,
+      "step": 149000
+    },
+    {
+      "epoch": 1.2831439393939394,
+      "grad_norm": 4.375,
+      "learning_rate": 0.00016951341679367993,
+      "loss": 4.8719,
+      "step": 149050
+    },
+    {
+      "epoch": 1.2835743801652892,
+      "grad_norm": 1.6328125,
+      "learning_rate": 0.0001694939383183812,
+      "loss": 4.663,
+      "step": 149100
+    },
+    {
+      "epoch": 1.284004820936639,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.0001694744547423769,
+      "loss": 4.923,
+      "step": 149150
+    },
+    {
+      "epoch": 1.284435261707989,
+      "grad_norm": 6.5,
+      "learning_rate": 0.00016945496606709707,
+      "loss": 4.7873,
+      "step": 149200
+    },
+    {
+      "epoch": 1.2848657024793388,
+      "grad_norm": 3.234375,
+      "learning_rate": 0.00016943547229397222,
+      "loss": 4.4368,
+      "step": 149250
+    },
+    {
+      "epoch": 1.2852961432506886,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.00016941597342443306,
+      "loss": 4.3052,
+      "step": 149300
+    },
+    {
+      "epoch": 1.2857265840220387,
+      "grad_norm": 5.53125,
+      "learning_rate": 0.00016939646945991082,
+      "loss": 4.6331,
+      "step": 149350
+    },
+    {
+      "epoch": 1.2861570247933884,
+      "grad_norm": 3.453125,
+      "learning_rate": 0.00016937696040183705,
+      "loss": 4.2599,
+      "step": 149400
+    },
+    {
+      "epoch": 1.2865874655647382,
+      "grad_norm": 3.078125,
+      "learning_rate": 0.00016935744625164363,
+      "loss": 4.6169,
+      "step": 149450
+    },
+    {
+      "epoch": 1.287017906336088,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.0001693379270107629,
+      "loss": 4.684,
+      "step": 149500
+    },
+    {
+      "epoch": 1.287448347107438,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.0001693184026806275,
+      "loss": 4.5671,
+      "step": 149550
+    },
+    {
+      "epoch": 1.2878787878787878,
+      "grad_norm": 3.40625,
+      "learning_rate": 0.0001692988732626705,
+      "loss": 4.5103,
+      "step": 149600
+    },
+    {
+      "epoch": 1.2883092286501379,
+      "grad_norm": 3.671875,
+      "learning_rate": 0.00016927933875832525,
+      "loss": 4.4864,
+      "step": 149650
+    },
+    {
+      "epoch": 1.2887396694214877,
+      "grad_norm": 1.5546875,
+      "learning_rate": 0.00016925979916902563,
+      "loss": 4.4726,
+      "step": 149700
+    },
+    {
+      "epoch": 1.2891701101928374,
+      "grad_norm": 4.46875,
+      "learning_rate": 0.00016924025449620577,
+      "loss": 4.6283,
+      "step": 149750
+    },
+    {
+      "epoch": 1.2896005509641872,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.0001692207047413002,
+      "loss": 4.6297,
+      "step": 149800
+    },
+    {
+      "epoch": 1.2900309917355373,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.00016920114990574386,
+      "loss": 4.6217,
+      "step": 149850
+    },
+    {
+      "epoch": 1.290461432506887,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00016918158999097198,
+      "loss": 4.4836,
+      "step": 149900
+    },
+    {
+      "epoch": 1.2908918732782368,
+      "grad_norm": 4.3125,
+      "learning_rate": 0.00016916202499842026,
+      "loss": 4.7021,
+      "step": 149950
+    },
+    {
+      "epoch": 1.2913223140495869,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.00016914245492952466,
+      "loss": 4.7028,
+      "step": 150000
+    },
+    {
+      "epoch": 1.2913223140495869,
+      "eval_loss": 5.180142402648926,
+      "eval_runtime": 21.7269,
+      "eval_samples_per_second": 29.457,
+      "eval_steps_per_second": 14.728,
+      "eval_tts_loss": 7.046059679570362,
+      "step": 150000
+    },
+    {
+      "epoch": 1.2917527548209367,
+      "grad_norm": 3.546875,
+      "learning_rate": 0.0001691228797857217,
+      "loss": 4.5666,
+      "step": 150050
+    },
+    {
+      "epoch": 1.2921831955922864,
+      "grad_norm": 3.765625,
+      "learning_rate": 0.00016910329956844805,
+      "loss": 4.4699,
+      "step": 150100
+    },
+    {
+      "epoch": 1.2926136363636362,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.0001690837142791409,
+      "loss": 4.5514,
+      "step": 150150
+    },
+    {
+      "epoch": 1.2930440771349863,
+      "grad_norm": 1.984375,
+      "learning_rate": 0.00016906412391923777,
+      "loss": 4.2519,
+      "step": 150200
+    },
+    {
+      "epoch": 1.293474517906336,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.00016904452849017648,
+      "loss": 4.67,
+      "step": 150250
+    },
+    {
+      "epoch": 1.293904958677686,
+      "grad_norm": 3.0,
+      "learning_rate": 0.0001690249279933954,
+      "loss": 4.4038,
+      "step": 150300
+    },
+    {
+      "epoch": 1.2943353994490359,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.00016900532243033312,
+      "loss": 4.9448,
+      "step": 150350
+    },
+    {
+      "epoch": 1.2947658402203857,
+      "grad_norm": 3.46875,
+      "learning_rate": 0.00016898571180242863,
+      "loss": 4.4284,
+      "step": 150400
+    },
+    {
+      "epoch": 1.2951962809917354,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00016896609611112136,
+      "loss": 4.2791,
+      "step": 150450
+    },
+    {
+      "epoch": 1.2956267217630855,
+      "grad_norm": 3.71875,
+      "learning_rate": 0.000168946475357851,
+      "loss": 4.3141,
+      "step": 150500
+    },
+    {
+      "epoch": 1.2960571625344353,
+      "grad_norm": 3.609375,
+      "learning_rate": 0.00016892684954405765,
+      "loss": 4.5364,
+      "step": 150550
+    },
+    {
+      "epoch": 1.296487603305785,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.0001689072186711819,
+      "loss": 4.5861,
+      "step": 150600
+    },
+    {
+      "epoch": 1.296918044077135,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.0001688875827406645,
+      "loss": 4.4741,
+      "step": 150650
+    },
+    {
+      "epoch": 1.2973484848484849,
+      "grad_norm": 5.0625,
+      "learning_rate": 0.00016886794175394675,
+      "loss": 4.5311,
+      "step": 150700
+    },
+    {
+      "epoch": 1.2977789256198347,
+      "grad_norm": 3.421875,
+      "learning_rate": 0.0001688482957124703,
+      "loss": 4.314,
+      "step": 150750
+    },
+    {
+      "epoch": 1.2982093663911844,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.00016882864461767704,
+      "loss": 4.67,
+      "step": 150800
+    },
+    {
+      "epoch": 1.2986398071625345,
+      "grad_norm": 3.171875,
+      "learning_rate": 0.00016880898847100934,
+      "loss": 4.6059,
+      "step": 150850
+    },
+    {
+      "epoch": 1.2990702479338843,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.00016878932727390996,
+      "loss": 4.5343,
+      "step": 150900
+    },
+    {
+      "epoch": 1.2995006887052343,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.00016876966102782195,
+      "loss": 4.6155,
+      "step": 150950
+    },
+    {
+      "epoch": 1.299931129476584,
+      "grad_norm": 3.0625,
+      "learning_rate": 0.00016874998973418878,
+      "loss": 4.4232,
+      "step": 151000
+    },
+    {
+      "epoch": 1.3003615702479339,
+      "grad_norm": 2.875,
+      "learning_rate": 0.00016873031339445431,
+      "loss": 4.6204,
+      "step": 151050
+    },
+    {
+      "epoch": 1.3007920110192837,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.0001687106320100627,
+      "loss": 4.7969,
+      "step": 151100
+    },
+    {
+      "epoch": 1.3012224517906337,
+      "grad_norm": 3.25,
+      "learning_rate": 0.00016869094558245854,
+      "loss": 4.5595,
+      "step": 151150
+    },
+    {
+      "epoch": 1.3016528925619835,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.00016867125411308673,
+      "loss": 4.7172,
+      "step": 151200
+    },
+    {
+      "epoch": 1.3020833333333333,
+      "grad_norm": 1.78125,
+      "learning_rate": 0.00016865155760339265,
+      "loss": 4.6102,
+      "step": 151250
+    },
+    {
+      "epoch": 1.3025137741046833,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00016863185605482195,
+      "loss": 4.7415,
+      "step": 151300
+    },
+    {
+      "epoch": 1.302944214876033,
+      "grad_norm": 1.625,
+      "learning_rate": 0.00016861214946882065,
+      "loss": 4.3747,
+      "step": 151350
+    },
+    {
+      "epoch": 1.3033746556473829,
+      "grad_norm": 0.85546875,
+      "learning_rate": 0.00016859243784683522,
+      "loss": 4.0872,
+      "step": 151400
+    },
+    {
+      "epoch": 1.3038050964187327,
+      "grad_norm": 1.2265625,
+      "learning_rate": 0.00016857272119031244,
+      "loss": 4.4674,
+      "step": 151450
+    },
+    {
+      "epoch": 1.3042355371900827,
+      "grad_norm": 2.625,
+      "learning_rate": 0.00016855299950069947,
+      "loss": 4.5789,
+      "step": 151500
+    },
+    {
+      "epoch": 1.3046659779614325,
+      "grad_norm": 3.8125,
+      "learning_rate": 0.00016853327277944382,
+      "loss": 4.633,
+      "step": 151550
+    },
+    {
+      "epoch": 1.3050964187327825,
+      "grad_norm": 1.4140625,
+      "learning_rate": 0.0001685135410279934,
+      "loss": 4.5294,
+      "step": 151600
+    },
+    {
+      "epoch": 1.3055268595041323,
+      "grad_norm": 1.578125,
+      "learning_rate": 0.00016849380424779648,
+      "loss": 4.2456,
+      "step": 151650
+    },
+    {
+      "epoch": 1.305957300275482,
+      "grad_norm": 3.0625,
+      "learning_rate": 0.0001684740624403017,
+      "loss": 4.777,
+      "step": 151700
+    },
+    {
+      "epoch": 1.3063877410468319,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.00016845431560695806,
+      "loss": 4.3336,
+      "step": 151750
+    },
+    {
+      "epoch": 1.3068181818181819,
+      "grad_norm": 1.96875,
+      "learning_rate": 0.00016843456374921495,
+      "loss": 4.4481,
+      "step": 151800
+    },
+    {
+      "epoch": 1.3072486225895317,
+      "grad_norm": 1.65625,
+      "learning_rate": 0.00016841480686852206,
+      "loss": 4.4452,
+      "step": 151850
+    },
+    {
+      "epoch": 1.3076790633608815,
+      "grad_norm": 3.453125,
+      "learning_rate": 0.00016839504496632957,
+      "loss": 4.8353,
+      "step": 151900
+    },
+    {
+      "epoch": 1.3081095041322315,
+      "grad_norm": 2.875,
+      "learning_rate": 0.00016837527804408796,
+      "loss": 4.546,
+      "step": 151950
+    },
+    {
+      "epoch": 1.3085399449035813,
+      "grad_norm": 4.0625,
+      "learning_rate": 0.00016835550610324807,
+      "loss": 4.1746,
+      "step": 152000
+    },
+    {
+      "epoch": 1.308970385674931,
+      "grad_norm": 4.125,
+      "learning_rate": 0.00016833572914526102,
+      "loss": 4.6171,
+      "step": 152050
+    },
+    {
+      "epoch": 1.3094008264462809,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.00016831594717157855,
+      "loss": 4.8406,
+      "step": 152100
+    },
+    {
+      "epoch": 1.3098312672176309,
+      "grad_norm": 3.015625,
+      "learning_rate": 0.00016829616018365252,
+      "loss": 4.6286,
+      "step": 152150
+    },
+    {
+      "epoch": 1.3102617079889807,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.00016827636818293526,
+      "loss": 4.79,
+      "step": 152200
+    },
+    {
+      "epoch": 1.3106921487603307,
+      "grad_norm": 3.265625,
+      "learning_rate": 0.00016825657117087948,
+      "loss": 4.7157,
+      "step": 152250
+    },
+    {
+      "epoch": 1.3111225895316805,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.00016823676914893825,
+      "loss": 4.2032,
+      "step": 152300
+    },
+    {
+      "epoch": 1.3115530303030303,
+      "grad_norm": 3.40625,
+      "learning_rate": 0.00016821696211856497,
+      "loss": 4.2972,
+      "step": 152350
+    },
+    {
+      "epoch": 1.31198347107438,
+      "grad_norm": 1.8046875,
+      "learning_rate": 0.00016819715008121346,
+      "loss": 4.2005,
+      "step": 152400
+    },
+    {
+      "epoch": 1.31241391184573,
+      "grad_norm": 3.21875,
+      "learning_rate": 0.00016817733303833786,
+      "loss": 4.6816,
+      "step": 152450
+    },
+    {
+      "epoch": 1.3128443526170799,
+      "grad_norm": 3.078125,
+      "learning_rate": 0.00016815751099139267,
+      "loss": 4.6942,
+      "step": 152500
+    },
+    {
+      "epoch": 1.3132747933884297,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00016813768394183284,
+      "loss": 4.3114,
+      "step": 152550
+    },
+    {
+      "epoch": 1.3137052341597797,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00016811785189111363,
+      "loss": 4.4654,
+      "step": 152600
+    },
+    {
+      "epoch": 1.3141356749311295,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.00016809801484069062,
+      "loss": 4.6403,
+      "step": 152650
+    },
+    {
+      "epoch": 1.3145661157024793,
+      "grad_norm": 3.171875,
+      "learning_rate": 0.00016807817279201984,
+      "loss": 4.6006,
+      "step": 152700
+    },
+    {
+      "epoch": 1.314996556473829,
+      "grad_norm": 4.3125,
+      "learning_rate": 0.0001680583257465577,
+      "loss": 4.5661,
+      "step": 152750
+    },
+    {
+      "epoch": 1.315426997245179,
+      "grad_norm": 1.3671875,
+      "learning_rate": 0.00016803847370576084,
+      "loss": 4.1854,
+      "step": 152800
+    },
+    {
+      "epoch": 1.3158574380165289,
+      "grad_norm": 3.0625,
+      "learning_rate": 0.00016801861667108638,
+      "loss": 4.5186,
+      "step": 152850
+    },
+    {
+      "epoch": 1.316287878787879,
+      "grad_norm": 3.796875,
+      "learning_rate": 0.00016799875464399184,
+      "loss": 5.1423,
+      "step": 152900
+    },
+    {
+      "epoch": 1.3167183195592287,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.000167978887625935,
+      "loss": 4.4366,
+      "step": 152950
+    },
+    {
+      "epoch": 1.3171487603305785,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.00016795901561837405,
+      "loss": 4.3613,
+      "step": 153000
+    },
+    {
+      "epoch": 1.3171487603305785,
+      "eval_loss": 5.173136234283447,
+      "eval_runtime": 21.8333,
+      "eval_samples_per_second": 29.313,
+      "eval_steps_per_second": 14.657,
+      "eval_tts_loss": 7.084487522601173,
+      "step": 153000
+    },
+    {
+      "epoch": 1.3175792011019283,
+      "grad_norm": 1.953125,
+      "learning_rate": 0.00016793913862276757,
+      "loss": 4.6094,
+      "step": 153050
+    },
+    {
+      "epoch": 1.3180096418732783,
+      "grad_norm": 3.328125,
+      "learning_rate": 0.0001679192566405745,
+      "loss": 4.3604,
+      "step": 153100
+    },
+    {
+      "epoch": 1.318440082644628,
+      "grad_norm": 5.09375,
+      "learning_rate": 0.00016789936967325414,
+      "loss": 4.5244,
+      "step": 153150
+    },
+    {
+      "epoch": 1.3188705234159779,
+      "grad_norm": 4.40625,
+      "learning_rate": 0.00016787947772226613,
+      "loss": 4.3624,
+      "step": 153200
+    },
+    {
+      "epoch": 1.319300964187328,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.0001678595807890705,
+      "loss": 4.4661,
+      "step": 153250
+    },
+    {
+      "epoch": 1.3197314049586777,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.00016783967887512764,
+      "loss": 4.1379,
+      "step": 153300
+    },
+    {
+      "epoch": 1.3201618457300275,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.00016781977198189832,
+      "loss": 4.6748,
+      "step": 153350
+    },
+    {
+      "epoch": 1.3205922865013773,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.0001677998601108436,
+      "loss": 4.9102,
+      "step": 153400
+    },
+    {
+      "epoch": 1.3210227272727273,
+      "grad_norm": 3.234375,
+      "learning_rate": 0.0001677799432634251,
+      "loss": 4.4248,
+      "step": 153450
+    },
+    {
+      "epoch": 1.321453168044077,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.00016776002144110457,
+      "loss": 4.1825,
+      "step": 153500
+    },
+    {
+      "epoch": 1.321883608815427,
+      "grad_norm": 2.25,
+      "learning_rate": 0.00016774009464534426,
+      "loss": 4.1231,
+      "step": 153550
+    },
+    {
+      "epoch": 1.322314049586777,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.00016772016287760672,
+      "loss": 4.5703,
+      "step": 153600
+    },
+    {
+      "epoch": 1.3227444903581267,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00016770022613935495,
+      "loss": 4.2335,
+      "step": 153650
+    },
+    {
+      "epoch": 1.3231749311294765,
+      "grad_norm": 2.296875,
+      "learning_rate": 0.00016768028443205226,
+      "loss": 4.7468,
+      "step": 153700
+    },
+    {
+      "epoch": 1.3236053719008265,
+      "grad_norm": 1.515625,
+      "learning_rate": 0.00016766033775716228,
+      "loss": 4.4976,
+      "step": 153750
+    },
+    {
+      "epoch": 1.3240358126721763,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.0001676403861161491,
+      "loss": 4.8121,
+      "step": 153800
+    },
+    {
+      "epoch": 1.3244662534435263,
+      "grad_norm": 3.03125,
+      "learning_rate": 0.00016762042951047712,
+      "loss": 4.4706,
+      "step": 153850
+    },
+    {
+      "epoch": 1.324896694214876,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.00016760046794161109,
+      "loss": 4.5775,
+      "step": 153900
+    },
+    {
+      "epoch": 1.325327134986226,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.00016758050141101616,
+      "loss": 5.0554,
+      "step": 153950
+    },
+    {
+      "epoch": 1.3257575757575757,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.00016756052992015783,
+      "loss": 4.445,
+      "step": 154000
+    },
+    {
+      "epoch": 1.3261880165289255,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.00016754055347050194,
+      "loss": 4.3357,
+      "step": 154050
+    },
+    {
+      "epoch": 1.3266184573002755,
+      "grad_norm": 3.921875,
+      "learning_rate": 0.00016752057206351477,
+      "loss": 4.5034,
+      "step": 154100
+    },
+    {
+      "epoch": 1.3270488980716253,
+      "grad_norm": 3.171875,
+      "learning_rate": 0.0001675005857006629,
+      "loss": 4.7499,
+      "step": 154150
+    },
+    {
+      "epoch": 1.3274793388429753,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.00016748059438341325,
+      "loss": 4.1199,
+      "step": 154200
+    },
+    {
+      "epoch": 1.327909779614325,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.00016746059811323315,
+      "loss": 4.5631,
+      "step": 154250
+    },
+    {
+      "epoch": 1.328340220385675,
+      "grad_norm": 4.15625,
+      "learning_rate": 0.00016744059689159028,
+      "loss": 4.1595,
+      "step": 154300
+    },
+    {
+      "epoch": 1.3287706611570247,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00016742059071995276,
+      "loss": 4.6359,
+      "step": 154350
+    },
+    {
+      "epoch": 1.3292011019283747,
+      "grad_norm": 3.1875,
+      "learning_rate": 0.0001674005795997889,
+      "loss": 4.7987,
+      "step": 154400
+    },
+    {
+      "epoch": 1.3296315426997245,
+      "grad_norm": 5.0,
+      "learning_rate": 0.0001673805635325675,
+      "loss": 4.8978,
+      "step": 154450
+    },
+    {
+      "epoch": 1.3300619834710745,
+      "grad_norm": 4.5,
+      "learning_rate": 0.00016736054251975774,
+      "loss": 4.4571,
+      "step": 154500
+    },
+    {
+      "epoch": 1.3304924242424243,
+      "grad_norm": 5.53125,
+      "learning_rate": 0.00016734051656282909,
+      "loss": 4.4233,
+      "step": 154550
+    },
+    {
+      "epoch": 1.330922865013774,
+      "grad_norm": 2.921875,
+      "learning_rate": 0.0001673204856632514,
+      "loss": 4.5482,
+      "step": 154600
+    },
+    {
+      "epoch": 1.331353305785124,
+      "grad_norm": 3.0,
+      "learning_rate": 0.0001673004498224949,
+      "loss": 4.7337,
+      "step": 154650
+    },
+    {
+      "epoch": 1.3317837465564737,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00016728040904203018,
+      "loss": 4.5273,
+      "step": 154700
+    },
+    {
+      "epoch": 1.3322141873278237,
+      "grad_norm": 3.265625,
+      "learning_rate": 0.0001672603633233282,
+      "loss": 4.7702,
+      "step": 154750
+    },
+    {
+      "epoch": 1.3326446280991735,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.00016724031266786028,
+      "loss": 4.463,
+      "step": 154800
+    },
+    {
+      "epoch": 1.3330750688705235,
+      "grad_norm": 4.5625,
+      "learning_rate": 0.00016722025707709807,
+      "loss": 4.1421,
+      "step": 154850
+    },
+    {
+      "epoch": 1.3335055096418733,
+      "grad_norm": 3.140625,
+      "learning_rate": 0.00016720019655251363,
+      "loss": 4.437,
+      "step": 154900
+    },
+    {
+      "epoch": 1.333935950413223,
+      "grad_norm": 3.203125,
+      "learning_rate": 0.00016718013109557933,
+      "loss": 4.3212,
+      "step": 154950
+    },
+    {
+      "epoch": 1.334366391184573,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.00016716006070776796,
+      "loss": 4.8673,
+      "step": 155000
+    },
+    {
+      "epoch": 1.334796831955923,
+      "grad_norm": 3.3125,
+      "learning_rate": 0.0001671399853905526,
+      "loss": 4.5034,
+      "step": 155050
+    },
+    {
+      "epoch": 1.3352272727272727,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.00016711990514540684,
+      "loss": 4.3151,
+      "step": 155100
+    },
+    {
+      "epoch": 1.3356577134986227,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00016709981997380438,
+      "loss": 4.4263,
+      "step": 155150
+    },
+    {
+      "epoch": 1.3360881542699725,
+      "grad_norm": 3.03125,
+      "learning_rate": 0.00016707972987721953,
+      "loss": 4.665,
+      "step": 155200
+    },
+    {
+      "epoch": 1.3365185950413223,
+      "grad_norm": 4.28125,
+      "learning_rate": 0.00016705963485712686,
+      "loss": 4.498,
+      "step": 155250
+    },
+    {
+      "epoch": 1.336949035812672,
+      "grad_norm": 3.140625,
+      "learning_rate": 0.00016703953491500125,
+      "loss": 4.5109,
+      "step": 155300
+    },
+    {
+      "epoch": 1.3373794765840221,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.00016701943005231802,
+      "loss": 4.4373,
+      "step": 155350
+    },
+    {
+      "epoch": 1.337809917355372,
+      "grad_norm": 1.671875,
+      "learning_rate": 0.0001669993202705528,
+      "loss": 4.6884,
+      "step": 155400
+    },
+    {
+      "epoch": 1.3382403581267217,
+      "grad_norm": 3.5,
+      "learning_rate": 0.00016697920557118165,
+      "loss": 4.3216,
+      "step": 155450
+    },
+    {
+      "epoch": 1.3386707988980717,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.00016695908595568094,
+      "loss": 4.6211,
+      "step": 155500
+    },
+    {
+      "epoch": 1.3391012396694215,
+      "grad_norm": 1.953125,
+      "learning_rate": 0.00016693896142552735,
+      "loss": 4.4004,
+      "step": 155550
+    },
+    {
+      "epoch": 1.3395316804407713,
+      "grad_norm": 1.625,
+      "learning_rate": 0.000166918831982198,
+      "loss": 4.7398,
+      "step": 155600
+    },
+    {
+      "epoch": 1.339962121212121,
+      "grad_norm": 2.9375,
+      "learning_rate": 0.00016689869762717037,
+      "loss": 4.7281,
+      "step": 155650
+    },
+    {
+      "epoch": 1.3403925619834711,
+      "grad_norm": 4.0625,
+      "learning_rate": 0.0001668785583619223,
+      "loss": 4.6139,
+      "step": 155700
+    },
+    {
+      "epoch": 1.340823002754821,
+      "grad_norm": 3.0,
+      "learning_rate": 0.00016685841418793195,
+      "loss": 4.6512,
+      "step": 155750
+    },
+    {
+      "epoch": 1.341253443526171,
+      "grad_norm": 1.375,
+      "learning_rate": 0.0001668382651066778,
+      "loss": 4.4863,
+      "step": 155800
+    },
+    {
+      "epoch": 1.3416838842975207,
+      "grad_norm": 1.4375,
+      "learning_rate": 0.00016681811111963883,
+      "loss": 4.6242,
+      "step": 155850
+    },
+    {
+      "epoch": 1.3421143250688705,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.00016679795222829426,
+      "loss": 4.7162,
+      "step": 155900
+    },
+    {
+      "epoch": 1.3425447658402203,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.00016677778843412374,
+      "loss": 4.5976,
+      "step": 155950
+    },
+    {
+      "epoch": 1.3429752066115703,
+      "grad_norm": 2.125,
+      "learning_rate": 0.00016675761973860718,
+      "loss": 4.4855,
+      "step": 156000
+    },
+    {
+      "epoch": 1.3429752066115703,
+      "eval_loss": 5.165329456329346,
+      "eval_runtime": 21.8131,
+      "eval_samples_per_second": 29.34,
+      "eval_steps_per_second": 14.67,
+      "eval_tts_loss": 7.059406952023572,
+      "step": 156000
+    },
+    {
+      "epoch": 1.3434056473829201,
+      "grad_norm": 3.03125,
+      "learning_rate": 0.00016673744614322498,
+      "loss": 4.548,
+      "step": 156050
+    },
+    {
+      "epoch": 1.34383608815427,
+      "grad_norm": 1.4609375,
+      "learning_rate": 0.00016671726764945784,
+      "loss": 4.543,
+      "step": 156100
+    },
+    {
+      "epoch": 1.34426652892562,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.0001666970842587868,
+      "loss": 4.6498,
+      "step": 156150
+    },
+    {
+      "epoch": 1.3446969696969697,
+      "grad_norm": 4.4375,
+      "learning_rate": 0.00016667689597269324,
+      "loss": 4.3392,
+      "step": 156200
+    },
+    {
+      "epoch": 1.3451274104683195,
+      "grad_norm": 1.9765625,
+      "learning_rate": 0.000166656702792659,
+      "loss": 4.6564,
+      "step": 156250
+    },
+    {
+      "epoch": 1.3455578512396693,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.0001666365047201662,
+      "loss": 4.4692,
+      "step": 156300
+    },
+    {
+      "epoch": 1.3459882920110193,
+      "grad_norm": 1.9921875,
+      "learning_rate": 0.0001666163017566973,
+      "loss": 4.6336,
+      "step": 156350
+    },
+    {
+      "epoch": 1.3464187327823691,
+      "grad_norm": 3.484375,
+      "learning_rate": 0.0001665960939037352,
+      "loss": 4.5028,
+      "step": 156400
+    },
+    {
+      "epoch": 1.3468491735537191,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.0001665758811627631,
+      "loss": 4.7692,
+      "step": 156450
+    },
+    {
+      "epoch": 1.347279614325069,
+      "grad_norm": 1.546875,
+      "learning_rate": 0.00016655566353526456,
+      "loss": 4.4678,
+      "step": 156500
+    },
+    {
+      "epoch": 1.3477100550964187,
+      "grad_norm": 5.21875,
+      "learning_rate": 0.0001665354410227235,
+      "loss": 4.7771,
+      "step": 156550
+    },
+    {
+      "epoch": 1.3481404958677685,
+      "grad_norm": 0.96875,
+      "learning_rate": 0.00016651521362662423,
+      "loss": 4.8281,
+      "step": 156600
+    },
+    {
+      "epoch": 1.3485709366391185,
+      "grad_norm": 1.125,
+      "learning_rate": 0.00016649498134845142,
+      "loss": 4.6899,
+      "step": 156650
+    },
+    {
+      "epoch": 1.3490013774104683,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.00016647474418968997,
+      "loss": 4.3369,
+      "step": 156700
+    },
+    {
+      "epoch": 1.3494318181818181,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.0001664545021518254,
+      "loss": 4.3659,
+      "step": 156750
+    },
+    {
+      "epoch": 1.3498622589531681,
+      "grad_norm": 3.046875,
+      "learning_rate": 0.00016643425523634332,
+      "loss": 4.5338,
+      "step": 156800
+    },
+    {
+      "epoch": 1.350292699724518,
+      "grad_norm": 1.390625,
+      "learning_rate": 0.0001664140034447298,
+      "loss": 4.1846,
+      "step": 156850
+    },
+    {
+      "epoch": 1.3507231404958677,
+      "grad_norm": 1.671875,
+      "learning_rate": 0.0001663937467784714,
+      "loss": 4.5744,
+      "step": 156900
+    },
+    {
+      "epoch": 1.3511535812672175,
+      "grad_norm": 3.015625,
+      "learning_rate": 0.00016637348523905482,
+      "loss": 4.6909,
+      "step": 156950
+    },
+    {
+      "epoch": 1.3515840220385675,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.0001663532188279672,
+      "loss": 4.5887,
+      "step": 157000
+    },
+    {
+      "epoch": 1.3520144628099173,
+      "grad_norm": 3.28125,
+      "learning_rate": 0.00016633294754669607,
+      "loss": 4.2306,
+      "step": 157050
+    },
+    {
+      "epoch": 1.3524449035812673,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.00016631267139672934,
+      "loss": 4.4633,
+      "step": 157100
+    },
+    {
+      "epoch": 1.3528753443526171,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.0001662923903795552,
+      "loss": 4.6043,
+      "step": 157150
+    },
+    {
+      "epoch": 1.353305785123967,
+      "grad_norm": 2.921875,
+      "learning_rate": 0.00016627210449666224,
+      "loss": 4.4723,
+      "step": 157200
+    },
+    {
+      "epoch": 1.3537362258953167,
+      "grad_norm": 1.4140625,
+      "learning_rate": 0.00016625181374953937,
+      "loss": 4.8493,
+      "step": 157250
+    },
+    {
+      "epoch": 1.3541666666666667,
+      "grad_norm": 3.203125,
+      "learning_rate": 0.00016623151813967594,
+      "loss": 4.3919,
+      "step": 157300
+    },
+    {
+      "epoch": 1.3545971074380165,
+      "grad_norm": 3.265625,
+      "learning_rate": 0.00016621121766856157,
+      "loss": 4.9115,
+      "step": 157350
+    },
+    {
+      "epoch": 1.3550275482093663,
+      "grad_norm": 3.3125,
+      "learning_rate": 0.0001661909123376863,
+      "loss": 4.4,
+      "step": 157400
+    },
+    {
+      "epoch": 1.3554579889807163,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.00016617060214854046,
+      "loss": 4.796,
+      "step": 157450
+    },
+    {
+      "epoch": 1.3558884297520661,
+      "grad_norm": 6.125,
+      "learning_rate": 0.00016615028710261477,
+      "loss": 4.6368,
+      "step": 157500
+    },
+    {
+      "epoch": 1.356318870523416,
+      "grad_norm": 2.5,
+      "learning_rate": 0.00016612996720140037,
+      "loss": 4.7879,
+      "step": 157550
+    },
+    {
+      "epoch": 1.3567493112947657,
+      "grad_norm": 3.5,
+      "learning_rate": 0.00016610964244638865,
+      "loss": 4.2282,
+      "step": 157600
+    },
+    {
+      "epoch": 1.3571797520661157,
+      "grad_norm": 3.765625,
+      "learning_rate": 0.00016608931283907142,
+      "loss": 4.4556,
+      "step": 157650
+    },
+    {
+      "epoch": 1.3576101928374655,
+      "grad_norm": 3.34375,
+      "learning_rate": 0.0001660689783809408,
+      "loss": 4.8659,
+      "step": 157700
+    },
+    {
+      "epoch": 1.3580406336088156,
+      "grad_norm": 3.0,
+      "learning_rate": 0.00016604863907348935,
+      "loss": 4.2695,
+      "step": 157750
+    },
+    {
+      "epoch": 1.3584710743801653,
+      "grad_norm": 3.03125,
+      "learning_rate": 0.0001660282949182099,
+      "loss": 4.4837,
+      "step": 157800
+    },
+    {
+      "epoch": 1.3589015151515151,
+      "grad_norm": 1.734375,
+      "learning_rate": 0.00016600794591659562,
+      "loss": 4.2264,
+      "step": 157850
+    },
+    {
+      "epoch": 1.359331955922865,
+      "grad_norm": 1.8515625,
+      "learning_rate": 0.00016598759207014018,
+      "loss": 4.5096,
+      "step": 157900
+    },
+    {
+      "epoch": 1.359762396694215,
+      "grad_norm": 1.7890625,
+      "learning_rate": 0.00016596723338033743,
+      "loss": 4.637,
+      "step": 157950
+    },
+    {
+      "epoch": 1.3601928374655647,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00016594686984868168,
+      "loss": 4.5702,
+      "step": 158000
+    },
+    {
+      "epoch": 1.3606232782369145,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.0001659265014766676,
+      "loss": 4.687,
+      "step": 158050
+    },
+    {
+      "epoch": 1.3610537190082646,
+      "grad_norm": 3.203125,
+      "learning_rate": 0.0001659061282657901,
+      "loss": 4.3978,
+      "step": 158100
+    },
+    {
+      "epoch": 1.3614841597796143,
+      "grad_norm": 3.46875,
+      "learning_rate": 0.00016588575021754464,
+      "loss": 4.4582,
+      "step": 158150
+    },
+    {
+      "epoch": 1.3619146005509641,
+      "grad_norm": 4.15625,
+      "learning_rate": 0.00016586536733342684,
+      "loss": 4.3739,
+      "step": 158200
+    },
+    {
+      "epoch": 1.362345041322314,
+      "grad_norm": 3.265625,
+      "learning_rate": 0.0001658449796149328,
+      "loss": 4.4914,
+      "step": 158250
+    },
+    {
+      "epoch": 1.362775482093664,
+      "grad_norm": 1.8046875,
+      "learning_rate": 0.00016582458706355894,
+      "loss": 4.7275,
+      "step": 158300
+    },
+    {
+      "epoch": 1.3632059228650137,
+      "grad_norm": 2.9375,
+      "learning_rate": 0.000165804189680802,
+      "loss": 4.3543,
+      "step": 158350
+    },
+    {
+      "epoch": 1.3636363636363638,
+      "grad_norm": 4.15625,
+      "learning_rate": 0.0001657837874681591,
+      "loss": 4.5953,
+      "step": 158400
+    },
+    {
+      "epoch": 1.3640668044077136,
+      "grad_norm": 1.203125,
+      "learning_rate": 0.00016576338042712777,
+      "loss": 4.5376,
+      "step": 158450
+    },
+    {
+      "epoch": 1.3644972451790633,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.0001657429685592058,
+      "loss": 4.7171,
+      "step": 158500
+    },
+    {
+      "epoch": 1.3649276859504131,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.0001657225518658914,
+      "loss": 4.9503,
+      "step": 158550
+    },
+    {
+      "epoch": 1.3653581267217632,
+      "grad_norm": 1.9296875,
+      "learning_rate": 0.00016570213034868307,
+      "loss": 4.6662,
+      "step": 158600
+    },
+    {
+      "epoch": 1.365788567493113,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.00016568170400907975,
+      "loss": 4.7016,
+      "step": 158650
+    },
+    {
+      "epoch": 1.3662190082644627,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.0001656612728485807,
+      "loss": 4.5563,
+      "step": 158700
+    },
+    {
+      "epoch": 1.3666494490358128,
+      "grad_norm": 3.15625,
+      "learning_rate": 0.00016564083686868543,
+      "loss": 4.8444,
+      "step": 158750
+    },
+    {
+      "epoch": 1.3670798898071626,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.00016562039607089402,
+      "loss": 4.3974,
+      "step": 158800
+    },
+    {
+      "epoch": 1.3675103305785123,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.00016559995045670672,
+      "loss": 5.0154,
+      "step": 158850
+    },
+    {
+      "epoch": 1.3679407713498621,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00016557950002762416,
+      "loss": 4.4174,
+      "step": 158900
+    },
+    {
+      "epoch": 1.3683712121212122,
+      "grad_norm": 3.609375,
+      "learning_rate": 0.00016555904478514742,
+      "loss": 4.7853,
+      "step": 158950
+    },
+    {
+      "epoch": 1.368801652892562,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.00016553858473077786,
+      "loss": 3.9965,
+      "step": 159000
+    },
+    {
+      "epoch": 1.368801652892562,
+      "eval_loss": 5.159815311431885,
+      "eval_runtime": 21.7663,
+      "eval_samples_per_second": 29.403,
+      "eval_steps_per_second": 14.702,
+      "eval_tts_loss": 7.143070609614872,
+      "step": 159000
+    },
+    {
+      "epoch": 1.369232093663912,
+      "grad_norm": 3.71875,
+      "learning_rate": 0.00016551811986601717,
+      "loss": 4.6505,
+      "step": 159050
+    },
+    {
+      "epoch": 1.3696625344352618,
+      "grad_norm": 3.734375,
+      "learning_rate": 0.00016549765019236747,
+      "loss": 4.5226,
+      "step": 159100
+    },
+    {
+      "epoch": 1.3700929752066116,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00016547717571133115,
+      "loss": 4.4056,
+      "step": 159150
+    },
+    {
+      "epoch": 1.3705234159779613,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00016545669642441103,
+      "loss": 4.6952,
+      "step": 159200
+    },
+    {
+      "epoch": 1.3709538567493114,
+      "grad_norm": 3.65625,
+      "learning_rate": 0.0001654362123331102,
+      "loss": 4.3651,
+      "step": 159250
+    },
+    {
+      "epoch": 1.3713842975206612,
+      "grad_norm": 3.296875,
+      "learning_rate": 0.0001654157234389322,
+      "loss": 4.5984,
+      "step": 159300
+    },
+    {
+      "epoch": 1.371814738292011,
+      "grad_norm": 1.9921875,
+      "learning_rate": 0.00016539522974338083,
+      "loss": 4.7625,
+      "step": 159350
+    },
+    {
+      "epoch": 1.372245179063361,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00016537473124796027,
+      "loss": 4.357,
+      "step": 159400
+    },
+    {
+      "epoch": 1.3726756198347108,
+      "grad_norm": 1.4296875,
+      "learning_rate": 0.00016535422795417514,
+      "loss": 4.0961,
+      "step": 159450
+    },
+    {
+      "epoch": 1.3731060606060606,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.0001653337198635303,
+      "loss": 4.3652,
+      "step": 159500
+    },
+    {
+      "epoch": 1.3735365013774103,
+      "grad_norm": 1.4609375,
+      "learning_rate": 0.00016531320697753095,
+      "loss": 4.4528,
+      "step": 159550
+    },
+    {
+      "epoch": 1.3739669421487604,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.00016529268929768272,
+      "loss": 4.686,
+      "step": 159600
+    },
+    {
+      "epoch": 1.3743973829201102,
+      "grad_norm": 1.6484375,
+      "learning_rate": 0.00016527216682549163,
+      "loss": 4.4016,
+      "step": 159650
+    },
+    {
+      "epoch": 1.3748278236914602,
+      "grad_norm": 3.84375,
+      "learning_rate": 0.0001652516395624639,
+      "loss": 4.4997,
+      "step": 159700
+    },
+    {
+      "epoch": 1.37525826446281,
+      "grad_norm": 3.0,
+      "learning_rate": 0.0001652311075101062,
+      "loss": 4.4726,
+      "step": 159750
+    },
+    {
+      "epoch": 1.3756887052341598,
+      "grad_norm": 3.65625,
+      "learning_rate": 0.0001652105706699256,
+      "loss": 4.888,
+      "step": 159800
+    },
+    {
+      "epoch": 1.3761191460055096,
+      "grad_norm": 4.59375,
+      "learning_rate": 0.0001651900290434294,
+      "loss": 4.5492,
+      "step": 159850
+    },
+    {
+      "epoch": 1.3765495867768596,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.00016516948263212532,
+      "loss": 4.7991,
+      "step": 159900
+    },
+    {
+      "epoch": 1.3769800275482094,
+      "grad_norm": 3.84375,
+      "learning_rate": 0.00016514893143752146,
+      "loss": 4.5327,
+      "step": 159950
+    },
+    {
+      "epoch": 1.3774104683195592,
+      "grad_norm": 1.8671875,
+      "learning_rate": 0.00016512837546112617,
+      "loss": 4.3377,
+      "step": 160000
+    },
+    {
+      "epoch": 1.3778409090909092,
+      "grad_norm": 3.796875,
+      "learning_rate": 0.00016510781470444825,
+      "loss": 4.4729,
+      "step": 160050
+    },
+    {
+      "epoch": 1.378271349862259,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.00016508724916899684,
+      "loss": 4.6799,
+      "step": 160100
+    },
+    {
+      "epoch": 1.3787017906336088,
+      "grad_norm": 2.984375,
+      "learning_rate": 0.00016506667885628136,
+      "loss": 4.7163,
+      "step": 160150
+    },
+    {
+      "epoch": 1.3791322314049586,
+      "grad_norm": 3.6875,
+      "learning_rate": 0.00016504610376781166,
+      "loss": 4.5325,
+      "step": 160200
+    },
+    {
+      "epoch": 1.3795626721763086,
+      "grad_norm": 2.25,
+      "learning_rate": 0.00016502552390509787,
+      "loss": 4.7286,
+      "step": 160250
+    },
+    {
+      "epoch": 1.3799931129476584,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.00016500493926965054,
+      "loss": 4.6325,
+      "step": 160300
+    },
+    {
+      "epoch": 1.3804235537190084,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00016498434986298056,
+      "loss": 4.7649,
+      "step": 160350
+    },
+    {
+      "epoch": 1.3808539944903582,
+      "grad_norm": 1.328125,
+      "learning_rate": 0.00016496375568659908,
+      "loss": 4.7286,
+      "step": 160400
+    },
+    {
+      "epoch": 1.381284435261708,
+      "grad_norm": 3.28125,
+      "learning_rate": 0.0001649431567420177,
+      "loss": 4.6121,
+      "step": 160450
+    },
+    {
+      "epoch": 1.3817148760330578,
+      "grad_norm": 1.5078125,
+      "learning_rate": 0.00016492255303074838,
+      "loss": 4.1601,
+      "step": 160500
+    },
+    {
+      "epoch": 1.3821453168044078,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.00016490194455430333,
+      "loss": 4.2581,
+      "step": 160550
+    },
+    {
+      "epoch": 1.3825757575757576,
+      "grad_norm": 2.578125,
+      "learning_rate": 0.0001648813313141952,
+      "loss": 4.8742,
+      "step": 160600
+    },
+    {
+      "epoch": 1.3830061983471074,
+      "grad_norm": 1.7578125,
+      "learning_rate": 0.00016486071331193696,
+      "loss": 4.6559,
+      "step": 160650
+    },
+    {
+      "epoch": 1.3834366391184574,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.00016484009054904187,
+      "loss": 4.6533,
+      "step": 160700
+    },
+    {
+      "epoch": 1.3838670798898072,
+      "grad_norm": 5.65625,
+      "learning_rate": 0.00016481946302702367,
+      "loss": 4.5375,
+      "step": 160750
+    },
+    {
+      "epoch": 1.384297520661157,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.00016479883074739635,
+      "loss": 4.5037,
+      "step": 160800
+    },
+    {
+      "epoch": 1.3847279614325068,
+      "grad_norm": 3.640625,
+      "learning_rate": 0.00016477819371167428,
+      "loss": 4.5175,
+      "step": 160850
+    },
+    {
+      "epoch": 1.3851584022038568,
+      "grad_norm": 1.546875,
+      "learning_rate": 0.00016475755192137213,
+      "loss": 4.6932,
+      "step": 160900
+    },
+    {
+      "epoch": 1.3855888429752066,
+      "grad_norm": 2.875,
+      "learning_rate": 0.000164736905378005,
+      "loss": 4.2476,
+      "step": 160950
+    },
+    {
+      "epoch": 1.3860192837465566,
+      "grad_norm": 1.796875,
+      "learning_rate": 0.00016471625408308836,
+      "loss": 4.5071,
+      "step": 161000
+    },
+    {
+      "epoch": 1.3864497245179064,
+      "grad_norm": 3.421875,
+      "learning_rate": 0.00016469559803813786,
+      "loss": 4.4198,
+      "step": 161050
+    },
+    {
+      "epoch": 1.3868801652892562,
+      "grad_norm": 3.375,
+      "learning_rate": 0.00016467493724466964,
+      "loss": 4.6224,
+      "step": 161100
+    },
+    {
+      "epoch": 1.387310606060606,
+      "grad_norm": 3.9375,
+      "learning_rate": 0.00016465427170420024,
+      "loss": 5.0826,
+      "step": 161150
+    },
+    {
+      "epoch": 1.387741046831956,
+      "grad_norm": 3.921875,
+      "learning_rate": 0.00016463360141824636,
+      "loss": 4.6965,
+      "step": 161200
+    },
+    {
+      "epoch": 1.3881714876033058,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.0001646129263883252,
+      "loss": 5.0396,
+      "step": 161250
+    },
+    {
+      "epoch": 1.3886019283746556,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00016459224661595427,
+      "loss": 4.8041,
+      "step": 161300
+    },
+    {
+      "epoch": 1.3890323691460056,
+      "grad_norm": 1.8828125,
+      "learning_rate": 0.00016457156210265142,
+      "loss": 4.7975,
+      "step": 161350
+    },
+    {
+      "epoch": 1.3894628099173554,
+      "grad_norm": 2.578125,
+      "learning_rate": 0.00016455087284993484,
+      "loss": 4.6357,
+      "step": 161400
+    },
+    {
+      "epoch": 1.3898932506887052,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.00016453017885932308,
+      "loss": 4.7784,
+      "step": 161450
+    },
+    {
+      "epoch": 1.390323691460055,
+      "grad_norm": 1.828125,
+      "learning_rate": 0.00016450948013233503,
+      "loss": 4.7899,
+      "step": 161500
+    },
+    {
+      "epoch": 1.390754132231405,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.0001644887766704899,
+      "loss": 4.9918,
+      "step": 161550
+    },
+    {
+      "epoch": 1.3911845730027548,
+      "grad_norm": 1.5078125,
+      "learning_rate": 0.00016446806847530735,
+      "loss": 4.4822,
+      "step": 161600
+    },
+    {
+      "epoch": 1.3916150137741048,
+      "grad_norm": 3.5,
+      "learning_rate": 0.0001644473555483073,
+      "loss": 4.664,
+      "step": 161650
+    },
+    {
+      "epoch": 1.3920454545454546,
+      "grad_norm": 3.65625,
+      "learning_rate": 0.00016442663789100997,
+      "loss": 4.8967,
+      "step": 161700
+    },
+    {
+      "epoch": 1.3924758953168044,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.00016440591550493606,
+      "loss": 4.2254,
+      "step": 161750
+    },
+    {
+      "epoch": 1.3929063360881542,
+      "grad_norm": 2.75,
+      "learning_rate": 0.0001643851883916065,
+      "loss": 4.8994,
+      "step": 161800
+    },
+    {
+      "epoch": 1.3933367768595042,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.00016436445655254267,
+      "loss": 4.4175,
+      "step": 161850
+    },
+    {
+      "epoch": 1.393767217630854,
+      "grad_norm": 3.578125,
+      "learning_rate": 0.0001643437199892662,
+      "loss": 4.4351,
+      "step": 161900
+    },
+    {
+      "epoch": 1.3941976584022038,
+      "grad_norm": 4.8125,
+      "learning_rate": 0.00016432297870329913,
+      "loss": 4.7114,
+      "step": 161950
+    },
+    {
+      "epoch": 1.3946280991735538,
+      "grad_norm": 3.5625,
+      "learning_rate": 0.0001643022326961638,
+      "loss": 4.4985,
+      "step": 162000
+    },
+    {
+      "epoch": 1.3946280991735538,
+      "eval_loss": 5.1462507247924805,
+      "eval_runtime": 22.1152,
+      "eval_samples_per_second": 28.939,
+      "eval_steps_per_second": 14.47,
+      "eval_tts_loss": 7.109295402450957,
+      "step": 162000
+    },
+    {
+      "epoch": 1.3950585399449036,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.00016428148196938295,
+      "loss": 3.7045,
+      "step": 162050
+    },
+    {
+      "epoch": 1.3954889807162534,
+      "grad_norm": 3.34375,
+      "learning_rate": 0.00016426072652447963,
+      "loss": 4.4363,
+      "step": 162100
+    },
+    {
+      "epoch": 1.3959194214876032,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.00016423996636297724,
+      "loss": 4.4125,
+      "step": 162150
+    },
+    {
+      "epoch": 1.3963498622589532,
+      "grad_norm": 3.046875,
+      "learning_rate": 0.00016421920148639957,
+      "loss": 4.6629,
+      "step": 162200
+    },
+    {
+      "epoch": 1.396780303030303,
+      "grad_norm": 1.8515625,
+      "learning_rate": 0.0001641984318962706,
+      "loss": 4.8781,
+      "step": 162250
+    },
+    {
+      "epoch": 1.397210743801653,
+      "grad_norm": 3.828125,
+      "learning_rate": 0.00016417765759411493,
+      "loss": 4.6714,
+      "step": 162300
+    },
+    {
+      "epoch": 1.3976411845730028,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.00016415687858145725,
+      "loss": 4.8539,
+      "step": 162350
+    },
+    {
+      "epoch": 1.3980716253443526,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.00016413609485982274,
+      "loss": 4.546,
+      "step": 162400
+    },
+    {
+      "epoch": 1.3985020661157024,
+      "grad_norm": 1.8984375,
+      "learning_rate": 0.0001641153064307368,
+      "loss": 4.3027,
+      "step": 162450
+    },
+    {
+      "epoch": 1.3989325068870524,
+      "grad_norm": 3.953125,
+      "learning_rate": 0.00016409451329572536,
+      "loss": 4.5593,
+      "step": 162500
+    },
+    {
+      "epoch": 1.3993629476584022,
+      "grad_norm": 1.8203125,
+      "learning_rate": 0.00016407371545631457,
+      "loss": 4.276,
+      "step": 162550
+    },
+    {
+      "epoch": 1.399793388429752,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00016405291291403086,
+      "loss": 4.5486,
+      "step": 162600
+    },
+    {
+      "epoch": 1.400223829201102,
+      "grad_norm": 2.875,
+      "learning_rate": 0.00016403210567040119,
+      "loss": 4.7962,
+      "step": 162650
+    },
+    {
+      "epoch": 1.4006542699724518,
+      "grad_norm": 4.34375,
+      "learning_rate": 0.00016401129372695277,
+      "loss": 4.9889,
+      "step": 162700
+    },
+    {
+      "epoch": 1.4010847107438016,
+      "grad_norm": 3.125,
+      "learning_rate": 0.00016399047708521306,
+      "loss": 4.482,
+      "step": 162750
+    },
+    {
+      "epoch": 1.4015151515151514,
+      "grad_norm": 3.4375,
+      "learning_rate": 0.00016396965574671003,
+      "loss": 4.6839,
+      "step": 162800
+    },
+    {
+      "epoch": 1.4019455922865014,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.0001639488297129719,
+      "loss": 4.3398,
+      "step": 162850
+    },
+    {
+      "epoch": 1.4023760330578512,
+      "grad_norm": 1.1953125,
+      "learning_rate": 0.00016392799898552727,
+      "loss": 4.6088,
+      "step": 162900
+    },
+    {
+      "epoch": 1.4028064738292012,
+      "grad_norm": 1.671875,
+      "learning_rate": 0.00016390716356590504,
+      "loss": 4.1766,
+      "step": 162950
+    },
+    {
+      "epoch": 1.403236914600551,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.0001638863234556345,
+      "loss": 4.2907,
+      "step": 163000
+    },
+    {
+      "epoch": 1.4036673553719008,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00016386547865624527,
+      "loss": 4.5716,
+      "step": 163050
+    },
+    {
+      "epoch": 1.4040977961432506,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00016384462916926734,
+      "loss": 4.4531,
+      "step": 163100
+    },
+    {
+      "epoch": 1.4045282369146006,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.000163823774996231,
+      "loss": 4.5312,
+      "step": 163150
+    },
+    {
+      "epoch": 1.4049586776859504,
+      "grad_norm": 4.09375,
+      "learning_rate": 0.00016380291613866688,
+      "loss": 4.747,
+      "step": 163200
+    },
+    {
+      "epoch": 1.4053891184573004,
+      "grad_norm": 3.203125,
+      "learning_rate": 0.000163782052598106,
+      "loss": 4.3799,
+      "step": 163250
+    },
+    {
+      "epoch": 1.4058195592286502,
+      "grad_norm": 1.609375,
+      "learning_rate": 0.00016376118437607968,
+      "loss": 4.6006,
+      "step": 163300
+    },
+    {
+      "epoch": 1.40625,
+      "grad_norm": 3.40625,
+      "learning_rate": 0.0001637403114741196,
+      "loss": 4.7187,
+      "step": 163350
+    },
+    {
+      "epoch": 1.4066804407713498,
+      "grad_norm": 1.6796875,
+      "learning_rate": 0.0001637194338937578,
+      "loss": 4.2554,
+      "step": 163400
+    },
+    {
+      "epoch": 1.4071108815426996,
+      "grad_norm": 2.875,
+      "learning_rate": 0.0001636985516365267,
+      "loss": 4.7399,
+      "step": 163450
+    },
+    {
+      "epoch": 1.4075413223140496,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.00016367766470395893,
+      "loss": 4.4203,
+      "step": 163500
+    },
+    {
+      "epoch": 1.4079717630853994,
+      "grad_norm": 1.9453125,
+      "learning_rate": 0.00016365677309758755,
+      "loss": 4.2266,
+      "step": 163550
+    },
+    {
+      "epoch": 1.4084022038567494,
+      "grad_norm": 6.75,
+      "learning_rate": 0.000163635876818946,
+      "loss": 4.1866,
+      "step": 163600
+    },
+    {
+      "epoch": 1.4088326446280992,
+      "grad_norm": 3.15625,
+      "learning_rate": 0.000163614975869568,
+      "loss": 4.3326,
+      "step": 163650
+    },
+    {
+      "epoch": 1.409263085399449,
+      "grad_norm": 1.859375,
+      "learning_rate": 0.00016359407025098768,
+      "loss": 4.5099,
+      "step": 163700
+    },
+    {
+      "epoch": 1.4096935261707988,
+      "grad_norm": 1.625,
+      "learning_rate": 0.0001635731599647394,
+      "loss": 4.6093,
+      "step": 163750
+    },
+    {
+      "epoch": 1.4101239669421488,
+      "grad_norm": 1.75,
+      "learning_rate": 0.00016355224501235797,
+      "loss": 4.1367,
+      "step": 163800
+    },
+    {
+      "epoch": 1.4105544077134986,
+      "grad_norm": 4.3125,
+      "learning_rate": 0.0001635313253953785,
+      "loss": 4.2216,
+      "step": 163850
+    },
+    {
+      "epoch": 1.4109848484848486,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00016351040111533643,
+      "loss": 4.6569,
+      "step": 163900
+    },
+    {
+      "epoch": 1.4114152892561984,
+      "grad_norm": 1.3359375,
+      "learning_rate": 0.00016348947217376757,
+      "loss": 4.4413,
+      "step": 163950
+    },
+    {
+      "epoch": 1.4118457300275482,
+      "grad_norm": 1.7734375,
+      "learning_rate": 0.00016346853857220802,
+      "loss": 4.5604,
+      "step": 164000
+    },
+    {
+      "epoch": 1.412276170798898,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.0001634476003121943,
+      "loss": 3.991,
+      "step": 164050
+    },
+    {
+      "epoch": 1.412706611570248,
+      "grad_norm": 1.796875,
+      "learning_rate": 0.00016342665739526327,
+      "loss": 4.5542,
+      "step": 164100
+    },
+    {
+      "epoch": 1.4131370523415978,
+      "grad_norm": 3.65625,
+      "learning_rate": 0.000163405709822952,
+      "loss": 4.3613,
+      "step": 164150
+    },
+    {
+      "epoch": 1.4135674931129476,
+      "grad_norm": 3.125,
+      "learning_rate": 0.00016338475759679813,
+      "loss": 4.5651,
+      "step": 164200
+    },
+    {
+      "epoch": 1.4139979338842976,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.00016336380071833934,
+      "loss": 4.1581,
+      "step": 164250
+    },
+    {
+      "epoch": 1.4144283746556474,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.00016334283918911395,
+      "loss": 4.3491,
+      "step": 164300
+    },
+    {
+      "epoch": 1.4148588154269972,
+      "grad_norm": 1.78125,
+      "learning_rate": 0.00016332187301066044,
+      "loss": 4.4377,
+      "step": 164350
+    },
+    {
+      "epoch": 1.415289256198347,
+      "grad_norm": 1.9609375,
+      "learning_rate": 0.00016330090218451767,
+      "loss": 4.2134,
+      "step": 164400
+    },
+    {
+      "epoch": 1.415719696969697,
+      "grad_norm": 6.40625,
+      "learning_rate": 0.00016327992671222492,
+      "loss": 4.9712,
+      "step": 164450
+    },
+    {
+      "epoch": 1.4161501377410468,
+      "grad_norm": 3.65625,
+      "learning_rate": 0.0001632589465953217,
+      "loss": 4.3055,
+      "step": 164500
+    },
+    {
+      "epoch": 1.4165805785123968,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.00016323796183534788,
+      "loss": 4.5056,
+      "step": 164550
+    },
+    {
+      "epoch": 1.4170110192837466,
+      "grad_norm": 3.484375,
+      "learning_rate": 0.00016321697243384373,
+      "loss": 4.2459,
+      "step": 164600
+    },
+    {
+      "epoch": 1.4174414600550964,
+      "grad_norm": 1.4375,
+      "learning_rate": 0.0001631959783923498,
+      "loss": 4.7942,
+      "step": 164650
+    },
+    {
+      "epoch": 1.4178719008264462,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00016317497971240705,
+      "loss": 4.9029,
+      "step": 164700
+    },
+    {
+      "epoch": 1.4183023415977962,
+      "grad_norm": 3.453125,
+      "learning_rate": 0.0001631539763955567,
+      "loss": 4.7152,
+      "step": 164750
+    },
+    {
+      "epoch": 1.418732782369146,
+      "grad_norm": 3.640625,
+      "learning_rate": 0.00016313296844334036,
+      "loss": 4.8096,
+      "step": 164800
+    },
+    {
+      "epoch": 1.4191632231404958,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.00016311195585729999,
+      "loss": 4.4885,
+      "step": 164850
+    },
+    {
+      "epoch": 1.4195936639118458,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.00016309093863897783,
+      "loss": 4.3724,
+      "step": 164900
+    },
+    {
+      "epoch": 1.4200241046831956,
+      "grad_norm": 1.296875,
+      "learning_rate": 0.00016306991678991653,
+      "loss": 4.7366,
+      "step": 164950
+    },
+    {
+      "epoch": 1.4204545454545454,
+      "grad_norm": 3.96875,
+      "learning_rate": 0.000163048890311659,
+      "loss": 4.6269,
+      "step": 165000
+    },
+    {
+      "epoch": 1.4204545454545454,
+      "eval_loss": 5.140933036804199,
+      "eval_runtime": 21.9888,
+      "eval_samples_per_second": 29.106,
+      "eval_steps_per_second": 14.553,
+      "eval_tts_loss": 7.1610827883010035,
+      "step": 165000
+    },
+    {
+      "epoch": 1.4208849862258952,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.00016302785920574865,
+      "loss": 4.0005,
+      "step": 165050
+    },
+    {
+      "epoch": 1.4213154269972452,
+      "grad_norm": 2.21875,
+      "learning_rate": 0.00016300682347372898,
+      "loss": 5.0,
+      "step": 165100
+    },
+    {
+      "epoch": 1.421745867768595,
+      "grad_norm": 1.3984375,
+      "learning_rate": 0.00016298578311714403,
+      "loss": 4.4025,
+      "step": 165150
+    },
+    {
+      "epoch": 1.422176308539945,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.00016296473813753813,
+      "loss": 4.3078,
+      "step": 165200
+    },
+    {
+      "epoch": 1.4226067493112948,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.0001629436885364559,
+      "loss": 4.7616,
+      "step": 165250
+    },
+    {
+      "epoch": 1.4230371900826446,
+      "grad_norm": 4.09375,
+      "learning_rate": 0.00016292263431544243,
+      "loss": 4.6323,
+      "step": 165300
+    },
+    {
+      "epoch": 1.4234676308539944,
+      "grad_norm": 6.03125,
+      "learning_rate": 0.00016290157547604292,
+      "loss": 4.5392,
+      "step": 165350
+    },
+    {
+      "epoch": 1.4238980716253444,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00016288051201980311,
+      "loss": 4.6293,
+      "step": 165400
+    },
+    {
+      "epoch": 1.4243285123966942,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.00016285944394826905,
+      "loss": 4.4712,
+      "step": 165450
+    },
+    {
+      "epoch": 1.424758953168044,
+      "grad_norm": 3.640625,
+      "learning_rate": 0.000162838371262987,
+      "loss": 4.5395,
+      "step": 165500
+    },
+    {
+      "epoch": 1.425189393939394,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.0001628172939655037,
+      "loss": 4.8306,
+      "step": 165550
+    },
+    {
+      "epoch": 1.4256198347107438,
+      "grad_norm": 3.609375,
+      "learning_rate": 0.0001627962120573662,
+      "loss": 4.7463,
+      "step": 165600
+    },
+    {
+      "epoch": 1.4260502754820936,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.00016277512554012188,
+      "loss": 4.1847,
+      "step": 165650
+    },
+    {
+      "epoch": 1.4264807162534434,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.00016275403441531837,
+      "loss": 4.5358,
+      "step": 165700
+    },
+    {
+      "epoch": 1.4269111570247934,
+      "grad_norm": 2.5,
+      "learning_rate": 0.00016273293868450376,
+      "loss": 4.8443,
+      "step": 165750
+    },
+    {
+      "epoch": 1.4273415977961432,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.00016271183834922642,
+      "loss": 4.5064,
+      "step": 165800
+    },
+    {
+      "epoch": 1.4277720385674932,
+      "grad_norm": 3.359375,
+      "learning_rate": 0.00016269073341103503,
+      "loss": 3.9674,
+      "step": 165850
+    },
+    {
+      "epoch": 1.428202479338843,
+      "grad_norm": 1.984375,
+      "learning_rate": 0.00016266962387147875,
+      "loss": 4.3991,
+      "step": 165900
+    },
+    {
+      "epoch": 1.4286329201101928,
+      "grad_norm": 3.296875,
+      "learning_rate": 0.00016264850973210688,
+      "loss": 4.7838,
+      "step": 165950
+    },
+    {
+      "epoch": 1.4290633608815426,
+      "grad_norm": 3.203125,
+      "learning_rate": 0.0001626273909944692,
+      "loss": 4.3483,
+      "step": 166000
+    },
+    {
+      "epoch": 1.4294938016528926,
+      "grad_norm": 3.09375,
+      "learning_rate": 0.00016260626766011579,
+      "loss": 4.3884,
+      "step": 166050
+    },
+    {
+      "epoch": 1.4299242424242424,
+      "grad_norm": 3.625,
+      "learning_rate": 0.000162585139730597,
+      "loss": 4.5349,
+      "step": 166100
+    },
+    {
+      "epoch": 1.4303546831955922,
+      "grad_norm": 3.75,
+      "learning_rate": 0.00016256400720746363,
+      "loss": 4.3026,
+      "step": 166150
+    },
+    {
+      "epoch": 1.4307851239669422,
+      "grad_norm": 0.96875,
+      "learning_rate": 0.00016254287009226674,
+      "loss": 4.2588,
+      "step": 166200
+    },
+    {
+      "epoch": 1.431215564738292,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.00016252172838655775,
+      "loss": 4.3752,
+      "step": 166250
+    },
+    {
+      "epoch": 1.4316460055096418,
+      "grad_norm": 4.4375,
+      "learning_rate": 0.0001625005820918884,
+      "loss": 4.5317,
+      "step": 166300
+    },
+    {
+      "epoch": 1.4320764462809916,
+      "grad_norm": 1.84375,
+      "learning_rate": 0.00016247943120981082,
+      "loss": 4.6431,
+      "step": 166350
+    },
+    {
+      "epoch": 1.4325068870523416,
+      "grad_norm": 1.2578125,
+      "learning_rate": 0.00016245827574187742,
+      "loss": 4.848,
+      "step": 166400
+    },
+    {
+      "epoch": 1.4329373278236914,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00016243711568964092,
+      "loss": 4.6549,
+      "step": 166450
+    },
+    {
+      "epoch": 1.4333677685950414,
+      "grad_norm": 1.4609375,
+      "learning_rate": 0.0001624159510546545,
+      "loss": 4.4376,
+      "step": 166500
+    },
+    {
+      "epoch": 1.4337982093663912,
+      "grad_norm": 3.0,
+      "learning_rate": 0.00016239478183847157,
+      "loss": 4.783,
+      "step": 166550
+    },
+    {
+      "epoch": 1.434228650137741,
+      "grad_norm": 4.03125,
+      "learning_rate": 0.0001623736080426459,
+      "loss": 4.6719,
+      "step": 166600
+    },
+    {
+      "epoch": 1.4346590909090908,
+      "grad_norm": 3.90625,
+      "learning_rate": 0.00016235242966873158,
+      "loss": 4.398,
+      "step": 166650
+    },
+    {
+      "epoch": 1.4350895316804408,
+      "grad_norm": 6.15625,
+      "learning_rate": 0.00016233124671828308,
+      "loss": 4.7081,
+      "step": 166700
+    },
+    {
+      "epoch": 1.4355199724517906,
+      "grad_norm": 4.46875,
+      "learning_rate": 0.0001623100591928552,
+      "loss": 4.3986,
+      "step": 166750
+    },
+    {
+      "epoch": 1.4359504132231404,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.000162288867094003,
+      "loss": 4.6497,
+      "step": 166800
+    },
+    {
+      "epoch": 1.4363808539944904,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.000162267670423282,
+      "loss": 4.7188,
+      "step": 166850
+    },
+    {
+      "epoch": 1.4368112947658402,
+      "grad_norm": 3.0,
+      "learning_rate": 0.00016224646918224796,
+      "loss": 4.4252,
+      "step": 166900
+    },
+    {
+      "epoch": 1.43724173553719,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.00016222526337245704,
+      "loss": 4.4618,
+      "step": 166950
+    },
+    {
+      "epoch": 1.4376721763085398,
+      "grad_norm": 3.296875,
+      "learning_rate": 0.0001622040529954656,
+      "loss": 4.4996,
+      "step": 167000
+    },
+    {
+      "epoch": 1.4381026170798898,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.00016218283805283056,
+      "loss": 4.1393,
+      "step": 167050
+    },
+    {
+      "epoch": 1.4385330578512396,
+      "grad_norm": 4.125,
+      "learning_rate": 0.00016216161854610897,
+      "loss": 4.6059,
+      "step": 167100
+    },
+    {
+      "epoch": 1.4389634986225897,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.00016214039447685834,
+      "loss": 4.4537,
+      "step": 167150
+    },
+    {
+      "epoch": 1.4393939393939394,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.00016211916584663646,
+      "loss": 4.6736,
+      "step": 167200
+    },
+    {
+      "epoch": 1.4398243801652892,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.00016209793265700145,
+      "loss": 4.5261,
+      "step": 167250
+    },
+    {
+      "epoch": 1.440254820936639,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00016207669490951178,
+      "loss": 4.4991,
+      "step": 167300
+    },
+    {
+      "epoch": 1.440685261707989,
+      "grad_norm": 3.671875,
+      "learning_rate": 0.00016205545260572626,
+      "loss": 4.4003,
+      "step": 167350
+    },
+    {
+      "epoch": 1.4411157024793388,
+      "grad_norm": 1.9765625,
+      "learning_rate": 0.00016203420574720405,
+      "loss": 4.3696,
+      "step": 167400
+    },
+    {
+      "epoch": 1.4415461432506886,
+      "grad_norm": 1.9140625,
+      "learning_rate": 0.00016201295433550462,
+      "loss": 5.0686,
+      "step": 167450
+    },
+    {
+      "epoch": 1.4419765840220387,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.00016199169837218772,
+      "loss": 4.3716,
+      "step": 167500
+    },
+    {
+      "epoch": 1.4424070247933884,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.00016197043785881357,
+      "loss": 4.4272,
+      "step": 167550
+    },
+    {
+      "epoch": 1.4428374655647382,
+      "grad_norm": 3.0,
+      "learning_rate": 0.00016194917279694263,
+      "loss": 4.4247,
+      "step": 167600
+    },
+    {
+      "epoch": 1.443267906336088,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00016192790318813566,
+      "loss": 4.3105,
+      "step": 167650
+    },
+    {
+      "epoch": 1.443698347107438,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00016190662903395388,
+      "loss": 4.7667,
+      "step": 167700
+    },
+    {
+      "epoch": 1.4441287878787878,
+      "grad_norm": 3.8125,
+      "learning_rate": 0.00016188535033595866,
+      "loss": 4.5131,
+      "step": 167750
+    },
+    {
+      "epoch": 1.4445592286501379,
+      "grad_norm": 3.296875,
+      "learning_rate": 0.00016186406709571194,
+      "loss": 4.3132,
+      "step": 167800
+    },
+    {
+      "epoch": 1.4449896694214877,
+      "grad_norm": 2.921875,
+      "learning_rate": 0.00016184277931477578,
+      "loss": 4.5048,
+      "step": 167850
+    },
+    {
+      "epoch": 1.4454201101928374,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.00016182148699471266,
+      "loss": 4.4808,
+      "step": 167900
+    },
+    {
+      "epoch": 1.4458505509641872,
+      "grad_norm": 3.265625,
+      "learning_rate": 0.00016180019013708541,
+      "loss": 4.6591,
+      "step": 167950
+    },
+    {
+      "epoch": 1.4462809917355373,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.00016177888874345717,
+      "loss": 4.8858,
+      "step": 168000
+    },
+    {
+      "epoch": 1.4462809917355373,
+      "eval_loss": 5.131041526794434,
+      "eval_runtime": 21.9964,
+      "eval_samples_per_second": 29.096,
+      "eval_steps_per_second": 14.548,
+      "eval_tts_loss": 7.119232905988722,
+      "step": 168000
+    },
+    {
+      "epoch": 1.446711432506887,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.00016175758281539144,
+      "loss": 4.8953,
+      "step": 168050
+    },
+    {
+      "epoch": 1.4471418732782368,
+      "grad_norm": 3.890625,
+      "learning_rate": 0.00016173627235445198,
+      "loss": 4.8951,
+      "step": 168100
+    },
+    {
+      "epoch": 1.4475723140495869,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.00016171495736220297,
+      "loss": 4.7312,
+      "step": 168150
+    },
+    {
+      "epoch": 1.4480027548209367,
+      "grad_norm": 3.984375,
+      "learning_rate": 0.00016169363784020887,
+      "loss": 4.6748,
+      "step": 168200
+    },
+    {
+      "epoch": 1.4484331955922864,
+      "grad_norm": 4.25,
+      "learning_rate": 0.0001616723137900345,
+      "loss": 4.646,
+      "step": 168250
+    },
+    {
+      "epoch": 1.4488636363636362,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.00016165098521324497,
+      "loss": 4.3641,
+      "step": 168300
+    },
+    {
+      "epoch": 1.4492940771349863,
+      "grad_norm": 1.625,
+      "learning_rate": 0.0001616296521114058,
+      "loss": 4.0898,
+      "step": 168350
+    },
+    {
+      "epoch": 1.449724517906336,
+      "grad_norm": 3.0,
+      "learning_rate": 0.00016160831448608273,
+      "loss": 4.9034,
+      "step": 168400
+    },
+    {
+      "epoch": 1.450154958677686,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00016158697233884195,
+      "loss": 4.6063,
+      "step": 168450
+    },
+    {
+      "epoch": 1.4505853994490359,
+      "grad_norm": 1.4921875,
+      "learning_rate": 0.00016156562567124988,
+      "loss": 4.2414,
+      "step": 168500
+    },
+    {
+      "epoch": 1.4510158402203857,
+      "grad_norm": 1.34375,
+      "learning_rate": 0.00016154427448487337,
+      "loss": 4.4975,
+      "step": 168550
+    },
+    {
+      "epoch": 1.4514462809917354,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00016152291878127954,
+      "loss": 4.7147,
+      "step": 168600
+    },
+    {
+      "epoch": 1.4518767217630855,
+      "grad_norm": 3.703125,
+      "learning_rate": 0.00016150155856203583,
+      "loss": 4.6297,
+      "step": 168650
+    },
+    {
+      "epoch": 1.4523071625344353,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.00016148019382871006,
+      "loss": 4.464,
+      "step": 168700
+    },
+    {
+      "epoch": 1.452737603305785,
+      "grad_norm": 3.453125,
+      "learning_rate": 0.0001614588245828703,
+      "loss": 4.4556,
+      "step": 168750
+    },
+    {
+      "epoch": 1.453168044077135,
+      "grad_norm": 4.0625,
+      "learning_rate": 0.0001614374508260851,
+      "loss": 4.5968,
+      "step": 168800
+    },
+    {
+      "epoch": 1.4535984848484849,
+      "grad_norm": 2.625,
+      "learning_rate": 0.00016141607255992319,
+      "loss": 4.3659,
+      "step": 168850
+    },
+    {
+      "epoch": 1.4540289256198347,
+      "grad_norm": 4.71875,
+      "learning_rate": 0.00016139468978595365,
+      "loss": 4.696,
+      "step": 168900
+    },
+    {
+      "epoch": 1.4544593663911844,
+      "grad_norm": 3.015625,
+      "learning_rate": 0.000161373302505746,
+      "loss": 4.6857,
+      "step": 168950
+    },
+    {
+      "epoch": 1.4548898071625345,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.00016135191072086998,
+      "loss": 4.5399,
+      "step": 169000
+    },
+    {
+      "epoch": 1.4553202479338843,
+      "grad_norm": 1.7109375,
+      "learning_rate": 0.00016133051443289575,
+      "loss": 4.0464,
+      "step": 169050
+    },
+    {
+      "epoch": 1.4557506887052343,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00016130911364339368,
+      "loss": 4.4852,
+      "step": 169100
+    },
+    {
+      "epoch": 1.456181129476584,
+      "grad_norm": 3.65625,
+      "learning_rate": 0.00016128770835393462,
+      "loss": 4.5114,
+      "step": 169150
+    },
+    {
+      "epoch": 1.4566115702479339,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.0001612662985660896,
+      "loss": 4.591,
+      "step": 169200
+    },
+    {
+      "epoch": 1.4570420110192837,
+      "grad_norm": 3.21875,
+      "learning_rate": 0.00016124488428143007,
+      "loss": 4.5943,
+      "step": 169250
+    },
+    {
+      "epoch": 1.4574724517906337,
+      "grad_norm": 3.015625,
+      "learning_rate": 0.00016122346550152784,
+      "loss": 4.6022,
+      "step": 169300
+    },
+    {
+      "epoch": 1.4579028925619835,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.00016120204222795493,
+      "loss": 4.3769,
+      "step": 169350
+    },
+    {
+      "epoch": 1.4583333333333333,
+      "grad_norm": 3.578125,
+      "learning_rate": 0.00016118061446228384,
+      "loss": 4.8728,
+      "step": 169400
+    },
+    {
+      "epoch": 1.4587637741046833,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.00016115918220608726,
+      "loss": 4.464,
+      "step": 169450
+    },
+    {
+      "epoch": 1.459194214876033,
+      "grad_norm": 3.875,
+      "learning_rate": 0.0001611377454609383,
+      "loss": 4.2808,
+      "step": 169500
+    },
+    {
+      "epoch": 1.4596246556473829,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00016111630422841036,
+      "loss": 4.5325,
+      "step": 169550
+    },
+    {
+      "epoch": 1.4600550964187327,
+      "grad_norm": 3.640625,
+      "learning_rate": 0.00016109485851007718,
+      "loss": 4.4474,
+      "step": 169600
+    },
+    {
+      "epoch": 1.4604855371900827,
+      "grad_norm": 1.671875,
+      "learning_rate": 0.00016107340830751285,
+      "loss": 4.1831,
+      "step": 169650
+    },
+    {
+      "epoch": 1.4609159779614325,
+      "grad_norm": 3.21875,
+      "learning_rate": 0.00016105195362229173,
+      "loss": 4.6718,
+      "step": 169700
+    },
+    {
+      "epoch": 1.4613464187327825,
+      "grad_norm": 1.515625,
+      "learning_rate": 0.0001610304944559886,
+      "loss": 4.6422,
+      "step": 169750
+    },
+    {
+      "epoch": 1.4617768595041323,
+      "grad_norm": 4.125,
+      "learning_rate": 0.00016100903081017845,
+      "loss": 4.7274,
+      "step": 169800
+    },
+    {
+      "epoch": 1.462207300275482,
+      "grad_norm": 3.296875,
+      "learning_rate": 0.00016098756268643672,
+      "loss": 4.6843,
+      "step": 169850
+    },
+    {
+      "epoch": 1.4626377410468319,
+      "grad_norm": 2.25,
+      "learning_rate": 0.00016096609008633915,
+      "loss": 4.7582,
+      "step": 169900
+    },
+    {
+      "epoch": 1.4630681818181819,
+      "grad_norm": 1.375,
+      "learning_rate": 0.00016094461301146166,
+      "loss": 4.4084,
+      "step": 169950
+    },
+    {
+      "epoch": 1.4634986225895317,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00016092313146338075,
+      "loss": 4.3323,
+      "step": 170000
+    },
+    {
+      "epoch": 1.4639290633608815,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.00016090164544367306,
+      "loss": 4.8541,
+      "step": 170050
+    },
+    {
+      "epoch": 1.4643595041322315,
+      "grad_norm": 1.1484375,
+      "learning_rate": 0.00016088015495391563,
+      "loss": 4.8426,
+      "step": 170100
+    },
+    {
+      "epoch": 1.4647899449035813,
+      "grad_norm": 3.609375,
+      "learning_rate": 0.00016085865999568584,
+      "loss": 4.6635,
+      "step": 170150
+    },
+    {
+      "epoch": 1.465220385674931,
+      "grad_norm": 4.34375,
+      "learning_rate": 0.0001608371605705613,
+      "loss": 4.4282,
+      "step": 170200
+    },
+    {
+      "epoch": 1.4656508264462809,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.00016081565668012008,
+      "loss": 4.2122,
+      "step": 170250
+    },
+    {
+      "epoch": 1.4660812672176309,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.00016079414832594053,
+      "loss": 4.5892,
+      "step": 170300
+    },
+    {
+      "epoch": 1.4665117079889807,
+      "grad_norm": 3.59375,
+      "learning_rate": 0.00016077263550960127,
+      "loss": 4.7288,
+      "step": 170350
+    },
+    {
+      "epoch": 1.4669421487603307,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.00016075111823268134,
+      "loss": 4.1907,
+      "step": 170400
+    },
+    {
+      "epoch": 1.4673725895316805,
+      "grad_norm": 2.421875,
+      "learning_rate": 0.00016072959649676003,
+      "loss": 4.8959,
+      "step": 170450
+    },
+    {
+      "epoch": 1.4678030303030303,
+      "grad_norm": 2.984375,
+      "learning_rate": 0.000160708070303417,
+      "loss": 4.906,
+      "step": 170500
+    },
+    {
+      "epoch": 1.46823347107438,
+      "grad_norm": 3.75,
+      "learning_rate": 0.00016068653965423222,
+      "loss": 4.8136,
+      "step": 170550
+    },
+    {
+      "epoch": 1.46866391184573,
+      "grad_norm": 1.46875,
+      "learning_rate": 0.000160665004550786,
+      "loss": 4.5021,
+      "step": 170600
+    },
+    {
+      "epoch": 1.4690943526170799,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.00016064346499465902,
+      "loss": 4.3826,
+      "step": 170650
+    },
+    {
+      "epoch": 1.4695247933884297,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.00016062192098743213,
+      "loss": 4.61,
+      "step": 170700
+    },
+    {
+      "epoch": 1.4699552341597797,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.0001606003725306867,
+      "loss": 4.7178,
+      "step": 170750
+    },
+    {
+      "epoch": 1.4703856749311295,
+      "grad_norm": 2.296875,
+      "learning_rate": 0.00016057881962600428,
+      "loss": 4.34,
+      "step": 170800
+    },
+    {
+      "epoch": 1.4708161157024793,
+      "grad_norm": 1.4453125,
+      "learning_rate": 0.00016055726227496683,
+      "loss": 4.5995,
+      "step": 170850
+    },
+    {
+      "epoch": 1.471246556473829,
+      "grad_norm": 1.7734375,
+      "learning_rate": 0.00016053570047915667,
+      "loss": 4.5458,
+      "step": 170900
+    },
+    {
+      "epoch": 1.471676997245179,
+      "grad_norm": 4.59375,
+      "learning_rate": 0.00016051413424015632,
+      "loss": 4.8168,
+      "step": 170950
+    },
+    {
+      "epoch": 1.4721074380165289,
+      "grad_norm": 2.921875,
+      "learning_rate": 0.0001604925635595487,
+      "loss": 4.6998,
+      "step": 171000
+    },
+    {
+      "epoch": 1.4721074380165289,
+      "eval_loss": 5.125603675842285,
+      "eval_runtime": 21.8112,
+      "eval_samples_per_second": 29.343,
+      "eval_steps_per_second": 14.671,
+      "eval_tts_loss": 7.1247039040745905,
+      "step": 171000
+    },
+    {
+      "epoch": 1.472537878787879,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.0001604709884389171,
+      "loss": 4.6223,
+      "step": 171050
+    },
+    {
+      "epoch": 1.4729683195592287,
+      "grad_norm": 1.453125,
+      "learning_rate": 0.00016044940887984503,
+      "loss": 3.9704,
+      "step": 171100
+    },
+    {
+      "epoch": 1.4733987603305785,
+      "grad_norm": 3.65625,
+      "learning_rate": 0.00016042782488391642,
+      "loss": 4.463,
+      "step": 171150
+    },
+    {
+      "epoch": 1.4738292011019283,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00016040623645271546,
+      "loss": 4.5133,
+      "step": 171200
+    },
+    {
+      "epoch": 1.4742596418732783,
+      "grad_norm": 3.234375,
+      "learning_rate": 0.0001603846435878267,
+      "loss": 4.949,
+      "step": 171250
+    },
+    {
+      "epoch": 1.474690082644628,
+      "grad_norm": 1.8984375,
+      "learning_rate": 0.00016036304629083508,
+      "loss": 4.2462,
+      "step": 171300
+    },
+    {
+      "epoch": 1.4751205234159779,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.0001603414445633257,
+      "loss": 4.3681,
+      "step": 171350
+    },
+    {
+      "epoch": 1.475550964187328,
+      "grad_norm": 1.609375,
+      "learning_rate": 0.00016031983840688411,
+      "loss": 4.4016,
+      "step": 171400
+    },
+    {
+      "epoch": 1.4759814049586777,
+      "grad_norm": 4.125,
+      "learning_rate": 0.00016029822782309618,
+      "loss": 4.4909,
+      "step": 171450
+    },
+    {
+      "epoch": 1.4764118457300275,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.00016027661281354806,
+      "loss": 4.7323,
+      "step": 171500
+    },
+    {
+      "epoch": 1.4768422865013773,
+      "grad_norm": 3.453125,
+      "learning_rate": 0.00016025499337982626,
+      "loss": 4.4793,
+      "step": 171550
+    },
+    {
+      "epoch": 1.4772727272727273,
+      "grad_norm": 3.171875,
+      "learning_rate": 0.00016023336952351755,
+      "loss": 4.8037,
+      "step": 171600
+    },
+    {
+      "epoch": 1.477703168044077,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.00016021174124620914,
+      "loss": 4.2385,
+      "step": 171650
+    },
+    {
+      "epoch": 1.478133608815427,
+      "grad_norm": 3.234375,
+      "learning_rate": 0.00016019010854948846,
+      "loss": 4.4692,
+      "step": 171700
+    },
+    {
+      "epoch": 1.478564049586777,
+      "grad_norm": 2.140625,
+      "learning_rate": 0.0001601684714349433,
+      "loss": 4.3944,
+      "step": 171750
+    },
+    {
+      "epoch": 1.4789944903581267,
+      "grad_norm": 1.921875,
+      "learning_rate": 0.00016014682990416184,
+      "loss": 4.5051,
+      "step": 171800
+    },
+    {
+      "epoch": 1.4794249311294765,
+      "grad_norm": 0.89453125,
+      "learning_rate": 0.00016012518395873245,
+      "loss": 4.4617,
+      "step": 171850
+    },
+    {
+      "epoch": 1.4798553719008265,
+      "grad_norm": 1.9140625,
+      "learning_rate": 0.0001601035336002439,
+      "loss": 3.916,
+      "step": 171900
+    },
+    {
+      "epoch": 1.4802858126721763,
+      "grad_norm": 1.828125,
+      "learning_rate": 0.0001600818788302853,
+      "loss": 4.1627,
+      "step": 171950
+    },
+    {
+      "epoch": 1.4807162534435263,
+      "grad_norm": 3.640625,
+      "learning_rate": 0.00016006021965044612,
+      "loss": 4.5495,
+      "step": 172000
+    },
+    {
+      "epoch": 1.481146694214876,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.00016003855606231597,
+      "loss": 4.5397,
+      "step": 172050
+    },
+    {
+      "epoch": 1.481577134986226,
+      "grad_norm": 2.125,
+      "learning_rate": 0.000160016888067485,
+      "loss": 4.1693,
+      "step": 172100
+    },
+    {
+      "epoch": 1.4820075757575757,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.0001599952156675436,
+      "loss": 4.7106,
+      "step": 172150
+    },
+    {
+      "epoch": 1.4824380165289255,
+      "grad_norm": 4.84375,
+      "learning_rate": 0.00015997353886408245,
+      "loss": 4.5839,
+      "step": 172200
+    },
+    {
+      "epoch": 1.4828684573002755,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.00015995185765869257,
+      "loss": 4.1468,
+      "step": 172250
+    },
+    {
+      "epoch": 1.4832988980716253,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.00015993017205296535,
+      "loss": 4.5094,
+      "step": 172300
+    },
+    {
+      "epoch": 1.4837293388429753,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.0001599084820484924,
+      "loss": 4.4477,
+      "step": 172350
+    },
+    {
+      "epoch": 1.484159779614325,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.0001598867876468658,
+      "loss": 4.5997,
+      "step": 172400
+    },
+    {
+      "epoch": 1.484590220385675,
+      "grad_norm": 0.94921875,
+      "learning_rate": 0.0001598650888496778,
+      "loss": 4.4454,
+      "step": 172450
+    },
+    {
+      "epoch": 1.4850206611570247,
+      "grad_norm": 4.1875,
+      "learning_rate": 0.00015984338565852113,
+      "loss": 4.3022,
+      "step": 172500
+    },
+    {
+      "epoch": 1.4854511019283747,
+      "grad_norm": 3.25,
+      "learning_rate": 0.0001598216780749887,
+      "loss": 4.7081,
+      "step": 172550
+    },
+    {
+      "epoch": 1.4858815426997245,
+      "grad_norm": 1.8359375,
+      "learning_rate": 0.0001597999661006738,
+      "loss": 4.5399,
+      "step": 172600
+    },
+    {
+      "epoch": 1.4863119834710745,
+      "grad_norm": 4.15625,
+      "learning_rate": 0.00015977824973717003,
+      "loss": 4.6684,
+      "step": 172650
+    },
+    {
+      "epoch": 1.4867424242424243,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.00015975652898607136,
+      "loss": 4.6484,
+      "step": 172700
+    },
+    {
+      "epoch": 1.487172865013774,
+      "grad_norm": 2.984375,
+      "learning_rate": 0.00015973480384897206,
+      "loss": 4.4801,
+      "step": 172750
+    },
+    {
+      "epoch": 1.487603305785124,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.00015971307432746662,
+      "loss": 4.717,
+      "step": 172800
+    },
+    {
+      "epoch": 1.4880337465564737,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.00015969134042315006,
+      "loss": 4.7288,
+      "step": 172850
+    },
+    {
+      "epoch": 1.4884641873278237,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00015966960213761752,
+      "loss": 4.8935,
+      "step": 172900
+    },
+    {
+      "epoch": 1.4888946280991735,
+      "grad_norm": 1.4453125,
+      "learning_rate": 0.00015964785947246454,
+      "loss": 4.7739,
+      "step": 172950
+    },
+    {
+      "epoch": 1.4893250688705235,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00015962611242928705,
+      "loss": 4.4058,
+      "step": 173000
+    },
+    {
+      "epoch": 1.4897555096418733,
+      "grad_norm": 1.6640625,
+      "learning_rate": 0.00015960436100968117,
+      "loss": 4.6265,
+      "step": 173050
+    },
+    {
+      "epoch": 1.490185950413223,
+      "grad_norm": 2.921875,
+      "learning_rate": 0.00015958260521524347,
+      "loss": 4.5567,
+      "step": 173100
+    },
+    {
+      "epoch": 1.490616391184573,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.0001595608450475707,
+      "loss": 4.5831,
+      "step": 173150
+    },
+    {
+      "epoch": 1.491046831955923,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.00015953908050826007,
+      "loss": 4.1565,
+      "step": 173200
+    },
+    {
+      "epoch": 1.4914772727272727,
+      "grad_norm": 3.0,
+      "learning_rate": 0.00015951731159890903,
+      "loss": 4.1816,
+      "step": 173250
+    },
+    {
+      "epoch": 1.4919077134986227,
+      "grad_norm": 3.296875,
+      "learning_rate": 0.00015949553832111536,
+      "loss": 4.4599,
+      "step": 173300
+    },
+    {
+      "epoch": 1.4923381542699725,
+      "grad_norm": 4.3125,
+      "learning_rate": 0.00015947376067647722,
+      "loss": 4.6852,
+      "step": 173350
+    },
+    {
+      "epoch": 1.4927685950413223,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00015945197866659299,
+      "loss": 4.3893,
+      "step": 173400
+    },
+    {
+      "epoch": 1.493199035812672,
+      "grad_norm": 4.125,
+      "learning_rate": 0.00015943019229306145,
+      "loss": 4.891,
+      "step": 173450
+    },
+    {
+      "epoch": 1.4936294765840221,
+      "grad_norm": 4.46875,
+      "learning_rate": 0.00015940840155748162,
+      "loss": 4.1655,
+      "step": 173500
+    },
+    {
+      "epoch": 1.494059917355372,
+      "grad_norm": 1.8671875,
+      "learning_rate": 0.00015938660646145296,
+      "loss": 4.6828,
+      "step": 173550
+    },
+    {
+      "epoch": 1.4944903581267217,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.00015936480700657517,
+      "loss": 4.6513,
+      "step": 173600
+    },
+    {
+      "epoch": 1.4949207988980717,
+      "grad_norm": 3.53125,
+      "learning_rate": 0.00015934300319444828,
+      "loss": 4.2477,
+      "step": 173650
+    },
+    {
+      "epoch": 1.4953512396694215,
+      "grad_norm": 3.0,
+      "learning_rate": 0.0001593211950266726,
+      "loss": 4.7214,
+      "step": 173700
+    },
+    {
+      "epoch": 1.4957816804407713,
+      "grad_norm": 1.9453125,
+      "learning_rate": 0.00015929938250484885,
+      "loss": 4.2496,
+      "step": 173750
+    },
+    {
+      "epoch": 1.496212121212121,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.00015927756563057806,
+      "loss": 4.233,
+      "step": 173800
+    },
+    {
+      "epoch": 1.4966425619834711,
+      "grad_norm": 4.59375,
+      "learning_rate": 0.00015925574440546142,
+      "loss": 4.3277,
+      "step": 173850
+    },
+    {
+      "epoch": 1.497073002754821,
+      "grad_norm": 4.15625,
+      "learning_rate": 0.0001592339188311007,
+      "loss": 4.2822,
+      "step": 173900
+    },
+    {
+      "epoch": 1.497503443526171,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.00015921208890909774,
+      "loss": 4.3395,
+      "step": 173950
+    },
+    {
+      "epoch": 1.4979338842975207,
+      "grad_norm": 3.171875,
+      "learning_rate": 0.00015919025464105485,
+      "loss": 4.2793,
+      "step": 174000
+    },
+    {
+      "epoch": 1.4979338842975207,
+      "eval_loss": 5.116052150726318,
+      "eval_runtime": 21.9341,
+      "eval_samples_per_second": 29.178,
+      "eval_steps_per_second": 14.589,
+      "eval_tts_loss": 7.171269260336755,
+      "step": 174000
+    },
+    {
+      "epoch": 1.4983643250688705,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00015916841602857463,
+      "loss": 4.4916,
+      "step": 174050
+    },
+    {
+      "epoch": 1.4987947658402203,
+      "grad_norm": 2.375,
+      "learning_rate": 0.00015914657307325997,
+      "loss": 4.2553,
+      "step": 174100
+    },
+    {
+      "epoch": 1.4992252066115703,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.0001591247257767141,
+      "loss": 4.3339,
+      "step": 174150
+    },
+    {
+      "epoch": 1.4996556473829201,
+      "grad_norm": 1.828125,
+      "learning_rate": 0.0001591028741405406,
+      "loss": 4.4354,
+      "step": 174200
+    },
+    {
+      "epoch": 1.5000860881542701,
+      "grad_norm": 1.6640625,
+      "learning_rate": 0.00015908101816634328,
+      "loss": 4.3378,
+      "step": 174250
+    },
+    {
+      "epoch": 1.50051652892562,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00015905915785572634,
+      "loss": 4.7259,
+      "step": 174300
+    },
+    {
+      "epoch": 1.5009469696969697,
+      "grad_norm": 1.828125,
+      "learning_rate": 0.0001590372932102943,
+      "loss": 4.6156,
+      "step": 174350
+    },
+    {
+      "epoch": 1.5013774104683195,
+      "grad_norm": 3.625,
+      "learning_rate": 0.00015901542423165193,
+      "loss": 4.5779,
+      "step": 174400
+    },
+    {
+      "epoch": 1.5018078512396693,
+      "grad_norm": 3.109375,
+      "learning_rate": 0.00015899355092140443,
+      "loss": 4.6976,
+      "step": 174450
+    },
+    {
+      "epoch": 1.5022382920110193,
+      "grad_norm": 3.015625,
+      "learning_rate": 0.00015897167328115722,
+      "loss": 4.0949,
+      "step": 174500
+    },
+    {
+      "epoch": 1.5026687327823691,
+      "grad_norm": 2.5,
+      "learning_rate": 0.00015894979131251603,
+      "loss": 4.2544,
+      "step": 174550
+    },
+    {
+      "epoch": 1.5030991735537191,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00015892790501708702,
+      "loss": 4.5139,
+      "step": 174600
+    },
+    {
+      "epoch": 1.503529614325069,
+      "grad_norm": 6.53125,
+      "learning_rate": 0.00015890601439647658,
+      "loss": 4.9411,
+      "step": 174650
+    },
+    {
+      "epoch": 1.5039600550964187,
+      "grad_norm": 3.5,
+      "learning_rate": 0.00015888411945229138,
+      "loss": 4.7355,
+      "step": 174700
+    },
+    {
+      "epoch": 1.5043904958677685,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00015886222018613854,
+      "loss": 4.4482,
+      "step": 174750
+    },
+    {
+      "epoch": 1.5048209366391183,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.00015884031659962536,
+      "loss": 4.0418,
+      "step": 174800
+    },
+    {
+      "epoch": 1.5052513774104683,
+      "grad_norm": 1.609375,
+      "learning_rate": 0.00015881840869435955,
+      "loss": 4.3027,
+      "step": 174850
+    },
+    {
+      "epoch": 1.5056818181818183,
+      "grad_norm": 3.125,
+      "learning_rate": 0.00015879649647194914,
+      "loss": 4.4568,
+      "step": 174900
+    },
+    {
+      "epoch": 1.5061122589531681,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.00015877457993400233,
+      "loss": 4.4248,
+      "step": 174950
+    },
+    {
+      "epoch": 1.506542699724518,
+      "grad_norm": 1.96875,
+      "learning_rate": 0.00015875265908212784,
+      "loss": 4.1974,
+      "step": 175000
+    },
+    {
+      "epoch": 1.5069731404958677,
+      "grad_norm": 3.53125,
+      "learning_rate": 0.00015873073391793456,
+      "loss": 4.5189,
+      "step": 175050
+    },
+    {
+      "epoch": 1.5074035812672175,
+      "grad_norm": 3.375,
+      "learning_rate": 0.00015870880444303177,
+      "loss": 4.2854,
+      "step": 175100
+    },
+    {
+      "epoch": 1.5078340220385675,
+      "grad_norm": 4.5625,
+      "learning_rate": 0.00015868687065902912,
+      "loss": 4.2583,
+      "step": 175150
+    },
+    {
+      "epoch": 1.5082644628099173,
+      "grad_norm": 2.375,
+      "learning_rate": 0.00015866493256753636,
+      "loss": 4.3532,
+      "step": 175200
+    },
+    {
+      "epoch": 1.5086949035812673,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.0001586429901701638,
+      "loss": 4.573,
+      "step": 175250
+    },
+    {
+      "epoch": 1.5091253443526171,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.0001586210434685219,
+      "loss": 4.4158,
+      "step": 175300
+    },
+    {
+      "epoch": 1.509555785123967,
+      "grad_norm": 3.4375,
+      "learning_rate": 0.00015859909246422155,
+      "loss": 4.5303,
+      "step": 175350
+    },
+    {
+      "epoch": 1.5099862258953167,
+      "grad_norm": 3.515625,
+      "learning_rate": 0.00015857713715887394,
+      "loss": 4.5698,
+      "step": 175400
+    },
+    {
+      "epoch": 1.5104166666666665,
+      "grad_norm": 1.59375,
+      "learning_rate": 0.0001585551775540904,
+      "loss": 4.4585,
+      "step": 175450
+    },
+    {
+      "epoch": 1.5108471074380165,
+      "grad_norm": 3.25,
+      "learning_rate": 0.0001585332136514829,
+      "loss": 4.4231,
+      "step": 175500
+    },
+    {
+      "epoch": 1.5112775482093666,
+      "grad_norm": 4.84375,
+      "learning_rate": 0.00015851124545266342,
+      "loss": 4.2434,
+      "step": 175550
+    },
+    {
+      "epoch": 1.5117079889807163,
+      "grad_norm": 1.890625,
+      "learning_rate": 0.00015848927295924442,
+      "loss": 4.5051,
+      "step": 175600
+    },
+    {
+      "epoch": 1.5121384297520661,
+      "grad_norm": 3.28125,
+      "learning_rate": 0.0001584672961728386,
+      "loss": 4.2596,
+      "step": 175650
+    },
+    {
+      "epoch": 1.512568870523416,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.0001584453150950591,
+      "loss": 4.4772,
+      "step": 175700
+    },
+    {
+      "epoch": 1.5129993112947657,
+      "grad_norm": 1.875,
+      "learning_rate": 0.0001584233297275192,
+      "loss": 4.2572,
+      "step": 175750
+    },
+    {
+      "epoch": 1.5134297520661157,
+      "grad_norm": 1.7734375,
+      "learning_rate": 0.00015840134007183255,
+      "loss": 4.6716,
+      "step": 175800
+    },
+    {
+      "epoch": 1.5138601928374655,
+      "grad_norm": 3.5,
+      "learning_rate": 0.00015837934612961322,
+      "loss": 4.8343,
+      "step": 175850
+    },
+    {
+      "epoch": 1.5142906336088156,
+      "grad_norm": 4.125,
+      "learning_rate": 0.00015835734790247552,
+      "loss": 4.4672,
+      "step": 175900
+    },
+    {
+      "epoch": 1.5147210743801653,
+      "grad_norm": 4.4375,
+      "learning_rate": 0.00015833534539203402,
+      "loss": 4.4801,
+      "step": 175950
+    },
+    {
+      "epoch": 1.5151515151515151,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.00015831333859990364,
+      "loss": 4.3717,
+      "step": 176000
+    },
+    {
+      "epoch": 1.515581955922865,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00015829132752769972,
+      "loss": 4.2226,
+      "step": 176050
+    },
+    {
+      "epoch": 1.5160123966942147,
+      "grad_norm": 3.140625,
+      "learning_rate": 0.00015826931217703777,
+      "loss": 4.31,
+      "step": 176100
+    },
+    {
+      "epoch": 1.5164428374655647,
+      "grad_norm": 1.734375,
+      "learning_rate": 0.0001582472925495337,
+      "loss": 4.3418,
+      "step": 176150
+    },
+    {
+      "epoch": 1.5168732782369148,
+      "grad_norm": 4.0,
+      "learning_rate": 0.00015822526864680363,
+      "loss": 4.5728,
+      "step": 176200
+    },
+    {
+      "epoch": 1.5173037190082646,
+      "grad_norm": 6.0625,
+      "learning_rate": 0.00015820324047046414,
+      "loss": 4.3989,
+      "step": 176250
+    },
+    {
+      "epoch": 1.5177341597796143,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.000158181208022132,
+      "loss": 4.9379,
+      "step": 176300
+    },
+    {
+      "epoch": 1.5181646005509641,
+      "grad_norm": 2.265625,
+      "learning_rate": 0.0001581591713034244,
+      "loss": 4.6618,
+      "step": 176350
+    },
+    {
+      "epoch": 1.518595041322314,
+      "grad_norm": 1.859375,
+      "learning_rate": 0.00015813713031595877,
+      "loss": 4.2438,
+      "step": 176400
+    },
+    {
+      "epoch": 1.519025482093664,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.00015811508506135283,
+      "loss": 4.3404,
+      "step": 176450
+    },
+    {
+      "epoch": 1.5194559228650137,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.0001580930355412247,
+      "loss": 4.5756,
+      "step": 176500
+    },
+    {
+      "epoch": 1.5198863636363638,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00015807098175719278,
+      "loss": 4.6958,
+      "step": 176550
+    },
+    {
+      "epoch": 1.5203168044077136,
+      "grad_norm": 5.625,
+      "learning_rate": 0.0001580489237108757,
+      "loss": 4.6835,
+      "step": 176600
+    },
+    {
+      "epoch": 1.5207472451790633,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00015802686140389252,
+      "loss": 4.9358,
+      "step": 176650
+    },
+    {
+      "epoch": 1.5211776859504131,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.00015800479483786257,
+      "loss": 4.7549,
+      "step": 176700
+    },
+    {
+      "epoch": 1.521608126721763,
+      "grad_norm": 3.609375,
+      "learning_rate": 0.00015798272401440548,
+      "loss": 4.8873,
+      "step": 176750
+    },
+    {
+      "epoch": 1.522038567493113,
+      "grad_norm": 3.15625,
+      "learning_rate": 0.0001579606489351412,
+      "loss": 4.6265,
+      "step": 176800
+    },
+    {
+      "epoch": 1.522469008264463,
+      "grad_norm": 4.71875,
+      "learning_rate": 0.00015793856960168997,
+      "loss": 4.6135,
+      "step": 176850
+    },
+    {
+      "epoch": 1.5228994490358128,
+      "grad_norm": 3.4375,
+      "learning_rate": 0.00015791648601567242,
+      "loss": 4.9678,
+      "step": 176900
+    },
+    {
+      "epoch": 1.5233298898071626,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.0001578943981787094,
+      "loss": 4.5739,
+      "step": 176950
+    },
+    {
+      "epoch": 1.5237603305785123,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.0001578723060924221,
+      "loss": 4.6492,
+      "step": 177000
+    },
+    {
+      "epoch": 1.5237603305785123,
+      "eval_loss": 5.11269474029541,
+      "eval_runtime": 21.9227,
+      "eval_samples_per_second": 29.194,
+      "eval_steps_per_second": 14.597,
+      "eval_tts_loss": 7.1675677132981805,
+      "step": 177000
+    },
+    {
+      "epoch": 1.5241907713498621,
+      "grad_norm": 3.96875,
+      "learning_rate": 0.0001578502097584321,
+      "loss": 4.6664,
+      "step": 177050
+    },
+    {
+      "epoch": 1.5246212121212122,
+      "grad_norm": 4.875,
+      "learning_rate": 0.00015782810917836112,
+      "loss": 4.1409,
+      "step": 177100
+    },
+    {
+      "epoch": 1.525051652892562,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.00015780600435383134,
+      "loss": 4.3525,
+      "step": 177150
+    },
+    {
+      "epoch": 1.525482093663912,
+      "grad_norm": 3.28125,
+      "learning_rate": 0.00015778389528646523,
+      "loss": 4.3918,
+      "step": 177200
+    },
+    {
+      "epoch": 1.5259125344352618,
+      "grad_norm": 1.125,
+      "learning_rate": 0.00015776178197788557,
+      "loss": 4.5024,
+      "step": 177250
+    },
+    {
+      "epoch": 1.5263429752066116,
+      "grad_norm": 3.4375,
+      "learning_rate": 0.00015773966442971532,
+      "loss": 4.3126,
+      "step": 177300
+    },
+    {
+      "epoch": 1.5267734159779613,
+      "grad_norm": 3.75,
+      "learning_rate": 0.00015771754264357798,
+      "loss": 4.6656,
+      "step": 177350
+    },
+    {
+      "epoch": 1.5272038567493111,
+      "grad_norm": 2.375,
+      "learning_rate": 0.0001576954166210972,
+      "loss": 4.4017,
+      "step": 177400
+    },
+    {
+      "epoch": 1.5276342975206612,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00015767328636389694,
+      "loss": 4.3906,
+      "step": 177450
+    },
+    {
+      "epoch": 1.5280647382920112,
+      "grad_norm": 1.5,
+      "learning_rate": 0.00015765115187360156,
+      "loss": 4.4006,
+      "step": 177500
+    },
+    {
+      "epoch": 1.528495179063361,
+      "grad_norm": 3.203125,
+      "learning_rate": 0.00015762901315183566,
+      "loss": 4.5229,
+      "step": 177550
+    },
+    {
+      "epoch": 1.5289256198347108,
+      "grad_norm": 1.578125,
+      "learning_rate": 0.0001576068702002242,
+      "loss": 4.4675,
+      "step": 177600
+    },
+    {
+      "epoch": 1.5293560606060606,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.00015758472302039245,
+      "loss": 4.5464,
+      "step": 177650
+    },
+    {
+      "epoch": 1.5297865013774103,
+      "grad_norm": 3.46875,
+      "learning_rate": 0.00015756257161396588,
+      "loss": 4.4539,
+      "step": 177700
+    },
+    {
+      "epoch": 1.5302169421487604,
+      "grad_norm": 1.109375,
+      "learning_rate": 0.00015754041598257042,
+      "loss": 4.192,
+      "step": 177750
+    },
+    {
+      "epoch": 1.5306473829201102,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.0001575182561278322,
+      "loss": 4.6458,
+      "step": 177800
+    },
+    {
+      "epoch": 1.5310778236914602,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.00015749609205137778,
+      "loss": 4.2214,
+      "step": 177850
+    },
+    {
+      "epoch": 1.53150826446281,
+      "grad_norm": 4.0,
+      "learning_rate": 0.00015747392375483384,
+      "loss": 4.2799,
+      "step": 177900
+    },
+    {
+      "epoch": 1.5319387052341598,
+      "grad_norm": 4.3125,
+      "learning_rate": 0.00015745175123982764,
+      "loss": 4.77,
+      "step": 177950
+    },
+    {
+      "epoch": 1.5323691460055096,
+      "grad_norm": 3.65625,
+      "learning_rate": 0.00015742957450798646,
+      "loss": 4.6492,
+      "step": 178000
+    },
+    {
+      "epoch": 1.5327995867768593,
+      "grad_norm": 3.8125,
+      "learning_rate": 0.00015740739356093808,
+      "loss": 4.7453,
+      "step": 178050
+    },
+    {
+      "epoch": 1.5332300275482094,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00015738520840031056,
+      "loss": 4.8577,
+      "step": 178100
+    },
+    {
+      "epoch": 1.5336604683195594,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.00015736301902773215,
+      "loss": 4.4629,
+      "step": 178150
+    },
+    {
+      "epoch": 1.5340909090909092,
+      "grad_norm": 3.703125,
+      "learning_rate": 0.0001573408254448316,
+      "loss": 4.6416,
+      "step": 178200
+    },
+    {
+      "epoch": 1.534521349862259,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.00015731862765323783,
+      "loss": 4.3571,
+      "step": 178250
+    },
+    {
+      "epoch": 1.5349517906336088,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.0001572964256545801,
+      "loss": 4.4044,
+      "step": 178300
+    },
+    {
+      "epoch": 1.5353822314049586,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00015727421945048803,
+      "loss": 4.3758,
+      "step": 178350
+    },
+    {
+      "epoch": 1.5358126721763086,
+      "grad_norm": 1.5,
+      "learning_rate": 0.00015725200904259147,
+      "loss": 4.239,
+      "step": 178400
+    },
+    {
+      "epoch": 1.5362431129476584,
+      "grad_norm": 5.46875,
+      "learning_rate": 0.0001572297944325206,
+      "loss": 4.2754,
+      "step": 178450
+    },
+    {
+      "epoch": 1.5366735537190084,
+      "grad_norm": 1.921875,
+      "learning_rate": 0.000157207575621906,
+      "loss": 4.4921,
+      "step": 178500
+    },
+    {
+      "epoch": 1.5371039944903582,
+      "grad_norm": 1.203125,
+      "learning_rate": 0.00015718535261237838,
+      "loss": 4.2033,
+      "step": 178550
+    },
+    {
+      "epoch": 1.537534435261708,
+      "grad_norm": 1.5625,
+      "learning_rate": 0.000157163125405569,
+      "loss": 4.2648,
+      "step": 178600
+    },
+    {
+      "epoch": 1.5379648760330578,
+      "grad_norm": 3.640625,
+      "learning_rate": 0.00015714089400310912,
+      "loss": 4.3139,
+      "step": 178650
+    },
+    {
+      "epoch": 1.5383953168044076,
+      "grad_norm": 3.5625,
+      "learning_rate": 0.00015711865840663058,
+      "loss": 4.0926,
+      "step": 178700
+    },
+    {
+      "epoch": 1.5388257575757576,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.00015709641861776544,
+      "loss": 4.1119,
+      "step": 178750
+    },
+    {
+      "epoch": 1.5392561983471076,
+      "grad_norm": 3.125,
+      "learning_rate": 0.00015707417463814598,
+      "loss": 4.4247,
+      "step": 178800
+    },
+    {
+      "epoch": 1.5396866391184574,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.00015705192646940495,
+      "loss": 3.8769,
+      "step": 178850
+    },
+    {
+      "epoch": 1.5401170798898072,
+      "grad_norm": 2.9375,
+      "learning_rate": 0.0001570296741131752,
+      "loss": 4.9352,
+      "step": 178900
+    },
+    {
+      "epoch": 1.540547520661157,
+      "grad_norm": 3.078125,
+      "learning_rate": 0.00015700741757109011,
+      "loss": 4.5678,
+      "step": 178950
+    },
+    {
+      "epoch": 1.5409779614325068,
+      "grad_norm": 2.375,
+      "learning_rate": 0.00015698515684478323,
+      "loss": 4.179,
+      "step": 179000
+    },
+    {
+      "epoch": 1.5414084022038568,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.00015696289193588845,
+      "loss": 4.0854,
+      "step": 179050
+    },
+    {
+      "epoch": 1.5418388429752066,
+      "grad_norm": 1.5703125,
+      "learning_rate": 0.00015694062284603993,
+      "loss": 4.7211,
+      "step": 179100
+    },
+    {
+      "epoch": 1.5422692837465566,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00015691834957687223,
+      "loss": 4.9799,
+      "step": 179150
+    },
+    {
+      "epoch": 1.5426997245179064,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.00015689607213002015,
+      "loss": 4.777,
+      "step": 179200
+    },
+    {
+      "epoch": 1.5431301652892562,
+      "grad_norm": 3.921875,
+      "learning_rate": 0.00015687379050711878,
+      "loss": 4.6938,
+      "step": 179250
+    },
+    {
+      "epoch": 1.543560606060606,
+      "grad_norm": 0.859375,
+      "learning_rate": 0.00015685150470980355,
+      "loss": 4.4715,
+      "step": 179300
+    },
+    {
+      "epoch": 1.5439910468319558,
+      "grad_norm": 1.5390625,
+      "learning_rate": 0.00015682921473971018,
+      "loss": 4.6943,
+      "step": 179350
+    },
+    {
+      "epoch": 1.5444214876033058,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.0001568069205984747,
+      "loss": 4.0785,
+      "step": 179400
+    },
+    {
+      "epoch": 1.5448519283746558,
+      "grad_norm": 2.265625,
+      "learning_rate": 0.00015678462228773353,
+      "loss": 4.8702,
+      "step": 179450
+    },
+    {
+      "epoch": 1.5452823691460056,
+      "grad_norm": 3.09375,
+      "learning_rate": 0.0001567623198091232,
+      "loss": 4.5307,
+      "step": 179500
+    },
+    {
+      "epoch": 1.5457128099173554,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.00015674001316428075,
+      "loss": 4.3296,
+      "step": 179550
+    },
+    {
+      "epoch": 1.5461432506887052,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.0001567177023548434,
+      "loss": 4.3114,
+      "step": 179600
+    },
+    {
+      "epoch": 1.546573691460055,
+      "grad_norm": 3.21875,
+      "learning_rate": 0.00015669538738244875,
+      "loss": 4.4844,
+      "step": 179650
+    },
+    {
+      "epoch": 1.547004132231405,
+      "grad_norm": 3.53125,
+      "learning_rate": 0.00015667306824873465,
+      "loss": 4.9129,
+      "step": 179700
+    },
+    {
+      "epoch": 1.5474345730027548,
+      "grad_norm": 3.46875,
+      "learning_rate": 0.00015665074495533926,
+      "loss": 4.4515,
+      "step": 179750
+    },
+    {
+      "epoch": 1.5478650137741048,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.0001566284175039011,
+      "loss": 4.6278,
+      "step": 179800
+    },
+    {
+      "epoch": 1.5482954545454546,
+      "grad_norm": 1.3515625,
+      "learning_rate": 0.00015660608589605893,
+      "loss": 4.3313,
+      "step": 179850
+    },
+    {
+      "epoch": 1.5487258953168044,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.00015658375013345182,
+      "loss": 4.4498,
+      "step": 179900
+    },
+    {
+      "epoch": 1.5491563360881542,
+      "grad_norm": 3.140625,
+      "learning_rate": 0.0001565614102177192,
+      "loss": 4.3749,
+      "step": 179950
+    },
+    {
+      "epoch": 1.549586776859504,
+      "grad_norm": 3.453125,
+      "learning_rate": 0.00015653906615050077,
+      "loss": 4.1746,
+      "step": 180000
+    },
+    {
+      "epoch": 1.549586776859504,
+      "eval_loss": 5.1044020652771,
+      "eval_runtime": 21.9398,
+      "eval_samples_per_second": 29.171,
+      "eval_steps_per_second": 14.585,
+      "eval_tts_loss": 7.21328017280712,
+      "step": 180000
+    },
+    {
+      "epoch": 1.550017217630854,
+      "grad_norm": 2.296875,
+      "learning_rate": 0.00015651671793343654,
+      "loss": 4.3718,
+      "step": 180050
+    },
+    {
+      "epoch": 1.550447658402204,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.0001564943655681668,
+      "loss": 4.4271,
+      "step": 180100
+    },
+    {
+      "epoch": 1.5508780991735538,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.00015647200905633218,
+      "loss": 4.0395,
+      "step": 180150
+    },
+    {
+      "epoch": 1.5513085399449036,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.0001564496483995736,
+      "loss": 4.6101,
+      "step": 180200
+    },
+    {
+      "epoch": 1.5517389807162534,
+      "grad_norm": 2.234375,
+      "learning_rate": 0.00015642728359953226,
+      "loss": 4.5181,
+      "step": 180250
+    },
+    {
+      "epoch": 1.5521694214876032,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00015640491465784977,
+      "loss": 4.7049,
+      "step": 180300
+    },
+    {
+      "epoch": 1.5525998622589532,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.00015638254157616786,
+      "loss": 4.2113,
+      "step": 180350
+    },
+    {
+      "epoch": 1.553030303030303,
+      "grad_norm": 4.625,
+      "learning_rate": 0.00015636016435612872,
+      "loss": 4.4224,
+      "step": 180400
+    },
+    {
+      "epoch": 1.553460743801653,
+      "grad_norm": 1.921875,
+      "learning_rate": 0.00015633778299937475,
+      "loss": 4.4158,
+      "step": 180450
+    },
+    {
+      "epoch": 1.5538911845730028,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00015631539750754877,
+      "loss": 4.4066,
+      "step": 180500
+    },
+    {
+      "epoch": 1.5543216253443526,
+      "grad_norm": 1.34375,
+      "learning_rate": 0.00015629300788229378,
+      "loss": 4.6345,
+      "step": 180550
+    },
+    {
+      "epoch": 1.5547520661157024,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00015627061412525309,
+      "loss": 4.6628,
+      "step": 180600
+    },
+    {
+      "epoch": 1.5551825068870524,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.00015624821623807042,
+      "loss": 4.3681,
+      "step": 180650
+    },
+    {
+      "epoch": 1.5556129476584022,
+      "grad_norm": 1.3125,
+      "learning_rate": 0.00015622581422238967,
+      "loss": 4.7182,
+      "step": 180700
+    },
+    {
+      "epoch": 1.5560433884297522,
+      "grad_norm": 5.125,
+      "learning_rate": 0.0001562034080798551,
+      "loss": 4.6946,
+      "step": 180750
+    },
+    {
+      "epoch": 1.556473829201102,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00015618099781211138,
+      "loss": 4.4063,
+      "step": 180800
+    },
+    {
+      "epoch": 1.5569042699724518,
+      "grad_norm": 3.234375,
+      "learning_rate": 0.00015615858342080325,
+      "loss": 4.4838,
+      "step": 180850
+    },
+    {
+      "epoch": 1.5573347107438016,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.00015613616490757595,
+      "loss": 4.5482,
+      "step": 180900
+    },
+    {
+      "epoch": 1.5577651515151514,
+      "grad_norm": 1.625,
+      "learning_rate": 0.0001561137422740749,
+      "loss": 4.6375,
+      "step": 180950
+    },
+    {
+      "epoch": 1.5581955922865014,
+      "grad_norm": 3.734375,
+      "learning_rate": 0.00015609131552194594,
+      "loss": 4.6877,
+      "step": 181000
+    },
+    {
+      "epoch": 1.5586260330578512,
+      "grad_norm": 3.0,
+      "learning_rate": 0.00015606888465283507,
+      "loss": 4.3437,
+      "step": 181050
+    },
+    {
+      "epoch": 1.5590564738292012,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.0001560464496683887,
+      "loss": 4.664,
+      "step": 181100
+    },
+    {
+      "epoch": 1.559486914600551,
+      "grad_norm": 3.25,
+      "learning_rate": 0.00015602401057025352,
+      "loss": 4.3147,
+      "step": 181150
+    },
+    {
+      "epoch": 1.5599173553719008,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00015600156736007652,
+      "loss": 4.6034,
+      "step": 181200
+    },
+    {
+      "epoch": 1.5603477961432506,
+      "grad_norm": 1.2578125,
+      "learning_rate": 0.00015597912003950494,
+      "loss": 4.4191,
+      "step": 181250
+    },
+    {
+      "epoch": 1.5607782369146006,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.0001559566686101864,
+      "loss": 4.4776,
+      "step": 181300
+    },
+    {
+      "epoch": 1.5612086776859504,
+      "grad_norm": 1.96875,
+      "learning_rate": 0.0001559342130737688,
+      "loss": 4.4152,
+      "step": 181350
+    },
+    {
+      "epoch": 1.5616391184573004,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.00015591175343190026,
+      "loss": 4.5063,
+      "step": 181400
+    },
+    {
+      "epoch": 1.5620695592286502,
+      "grad_norm": 3.625,
+      "learning_rate": 0.00015588928968622936,
+      "loss": 4.3081,
+      "step": 181450
+    },
+    {
+      "epoch": 1.5625,
+      "grad_norm": 1.9453125,
+      "learning_rate": 0.0001558668218384048,
+      "loss": 4.6451,
+      "step": 181500
+    },
+    {
+      "epoch": 1.5629304407713498,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.00015584434989007577,
+      "loss": 4.6819,
+      "step": 181550
+    },
+    {
+      "epoch": 1.5633608815426996,
+      "grad_norm": 3.875,
+      "learning_rate": 0.00015582187384289157,
+      "loss": 4.5812,
+      "step": 181600
+    },
+    {
+      "epoch": 1.5637913223140496,
+      "grad_norm": 1.796875,
+      "learning_rate": 0.00015579939369850197,
+      "loss": 4.4379,
+      "step": 181650
+    },
+    {
+      "epoch": 1.5642217630853994,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.00015577690945855688,
+      "loss": 4.6937,
+      "step": 181700
+    },
+    {
+      "epoch": 1.5646522038567494,
+      "grad_norm": 3.1875,
+      "learning_rate": 0.00015575442112470667,
+      "loss": 4.32,
+      "step": 181750
+    },
+    {
+      "epoch": 1.5650826446280992,
+      "grad_norm": 1.7421875,
+      "learning_rate": 0.00015573192869860192,
+      "loss": 4.0504,
+      "step": 181800
+    },
+    {
+      "epoch": 1.565513085399449,
+      "grad_norm": 1.203125,
+      "learning_rate": 0.00015570943218189349,
+      "loss": 4.3327,
+      "step": 181850
+    },
+    {
+      "epoch": 1.5659435261707988,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.0001556869315762326,
+      "loss": 4.4529,
+      "step": 181900
+    },
+    {
+      "epoch": 1.5663739669421488,
+      "grad_norm": 3.21875,
+      "learning_rate": 0.00015566442688327078,
+      "loss": 4.5494,
+      "step": 181950
+    },
+    {
+      "epoch": 1.5668044077134986,
+      "grad_norm": 3.84375,
+      "learning_rate": 0.00015564191810465974,
+      "loss": 4.7399,
+      "step": 182000
+    },
+    {
+      "epoch": 1.5672348484848486,
+      "grad_norm": 3.15625,
+      "learning_rate": 0.00015561940524205166,
+      "loss": 4.5206,
+      "step": 182050
+    },
+    {
+      "epoch": 1.5676652892561984,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.00015559688829709894,
+      "loss": 4.1723,
+      "step": 182100
+    },
+    {
+      "epoch": 1.5680957300275482,
+      "grad_norm": 2.5,
+      "learning_rate": 0.00015557436727145415,
+      "loss": 4.7992,
+      "step": 182150
+    },
+    {
+      "epoch": 1.568526170798898,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.00015555184216677046,
+      "loss": 4.282,
+      "step": 182200
+    },
+    {
+      "epoch": 1.5689566115702478,
+      "grad_norm": 2.25,
+      "learning_rate": 0.00015552931298470102,
+      "loss": 4.5177,
+      "step": 182250
+    },
+    {
+      "epoch": 1.5693870523415978,
+      "grad_norm": 4.40625,
+      "learning_rate": 0.0001555067797268995,
+      "loss": 4.5946,
+      "step": 182300
+    },
+    {
+      "epoch": 1.5698174931129476,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.0001554842423950198,
+      "loss": 4.8026,
+      "step": 182350
+    },
+    {
+      "epoch": 1.5702479338842976,
+      "grad_norm": 1.3359375,
+      "learning_rate": 0.00015546170099071604,
+      "loss": 4.9309,
+      "step": 182400
+    },
+    {
+      "epoch": 1.5706783746556474,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.0001554391555156428,
+      "loss": 4.3014,
+      "step": 182450
+    },
+    {
+      "epoch": 1.5711088154269972,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.00015541660597145478,
+      "loss": 4.5463,
+      "step": 182500
+    },
+    {
+      "epoch": 1.571539256198347,
+      "grad_norm": 3.546875,
+      "learning_rate": 0.00015539405235980713,
+      "loss": 4.2778,
+      "step": 182550
+    },
+    {
+      "epoch": 1.571969696969697,
+      "grad_norm": 4.8125,
+      "learning_rate": 0.00015537149468235523,
+      "loss": 4.1458,
+      "step": 182600
+    },
+    {
+      "epoch": 1.5724001377410468,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.00015534893294075474,
+      "loss": 4.178,
+      "step": 182650
+    },
+    {
+      "epoch": 1.5728305785123968,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00015532636713666169,
+      "loss": 4.5607,
+      "step": 182700
+    },
+    {
+      "epoch": 1.5732610192837466,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.0001553037972717323,
+      "loss": 4.3971,
+      "step": 182750
+    },
+    {
+      "epoch": 1.5736914600550964,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.0001552812233476232,
+      "loss": 4.0441,
+      "step": 182800
+    },
+    {
+      "epoch": 1.5741219008264462,
+      "grad_norm": 6.8125,
+      "learning_rate": 0.00015525864536599122,
+      "loss": 4.2612,
+      "step": 182850
+    },
+    {
+      "epoch": 1.574552341597796,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00015523606332849357,
+      "loss": 4.0635,
+      "step": 182900
+    },
+    {
+      "epoch": 1.574982782369146,
+      "grad_norm": 0.87109375,
+      "learning_rate": 0.00015521347723678773,
+      "loss": 4.5155,
+      "step": 182950
+    },
+    {
+      "epoch": 1.575413223140496,
+      "grad_norm": 2.0,
+      "learning_rate": 0.00015519088709253147,
+      "loss": 4.4464,
+      "step": 183000
+    },
+    {
+      "epoch": 1.575413223140496,
+      "eval_loss": 5.101893424987793,
+      "eval_runtime": 21.716,
+      "eval_samples_per_second": 29.471,
+      "eval_steps_per_second": 14.736,
+      "eval_tts_loss": 7.200182005268934,
+      "step": 183000
+    },
+    {
+      "epoch": 1.5758436639118458,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.00015516829289738282,
+      "loss": 4.3117,
+      "step": 183050
+    },
+    {
+      "epoch": 1.5762741046831956,
+      "grad_norm": 1.34375,
+      "learning_rate": 0.00015514569465300017,
+      "loss": 4.495,
+      "step": 183100
+    },
+    {
+      "epoch": 1.5767045454545454,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.0001551230923610422,
+      "loss": 4.8702,
+      "step": 183150
+    },
+    {
+      "epoch": 1.5771349862258952,
+      "grad_norm": 1.9140625,
+      "learning_rate": 0.0001551004860231678,
+      "loss": 4.6704,
+      "step": 183200
+    },
+    {
+      "epoch": 1.5775654269972452,
+      "grad_norm": 1.6640625,
+      "learning_rate": 0.00015507787564103632,
+      "loss": 4.45,
+      "step": 183250
+    },
+    {
+      "epoch": 1.577995867768595,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.00015505526121630726,
+      "loss": 4.448,
+      "step": 183300
+    },
+    {
+      "epoch": 1.578426308539945,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00015503264275064052,
+      "loss": 4.4778,
+      "step": 183350
+    },
+    {
+      "epoch": 1.5788567493112948,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.00015501002024569615,
+      "loss": 4.317,
+      "step": 183400
+    },
+    {
+      "epoch": 1.5792871900826446,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00015498739370313467,
+      "loss": 4.7482,
+      "step": 183450
+    },
+    {
+      "epoch": 1.5797176308539944,
+      "grad_norm": 1.859375,
+      "learning_rate": 0.0001549647631246168,
+      "loss": 4.6417,
+      "step": 183500
+    },
+    {
+      "epoch": 1.5801480716253442,
+      "grad_norm": 4.46875,
+      "learning_rate": 0.00015494212851180358,
+      "loss": 4.3391,
+      "step": 183550
+    },
+    {
+      "epoch": 1.5805785123966942,
+      "grad_norm": 3.390625,
+      "learning_rate": 0.00015491948986635635,
+      "loss": 4.28,
+      "step": 183600
+    },
+    {
+      "epoch": 1.5810089531680442,
+      "grad_norm": 1.3984375,
+      "learning_rate": 0.0001548968471899367,
+      "loss": 4.6358,
+      "step": 183650
+    },
+    {
+      "epoch": 1.581439393939394,
+      "grad_norm": 4.0625,
+      "learning_rate": 0.00015487420048420657,
+      "loss": 4.6617,
+      "step": 183700
+    },
+    {
+      "epoch": 1.5818698347107438,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00015485154975082826,
+      "loss": 4.4271,
+      "step": 183750
+    },
+    {
+      "epoch": 1.5823002754820936,
+      "grad_norm": 3.03125,
+      "learning_rate": 0.00015482889499146416,
+      "loss": 4.4431,
+      "step": 183800
+    },
+    {
+      "epoch": 1.5827307162534434,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00015480623620777711,
+      "loss": 4.7784,
+      "step": 183850
+    },
+    {
+      "epoch": 1.5831611570247934,
+      "grad_norm": 3.875,
+      "learning_rate": 0.0001547835734014303,
+      "loss": 4.4549,
+      "step": 183900
+    },
+    {
+      "epoch": 1.5835915977961432,
+      "grad_norm": 0.98046875,
+      "learning_rate": 0.00015476090657408703,
+      "loss": 4.5275,
+      "step": 183950
+    },
+    {
+      "epoch": 1.5840220385674932,
+      "grad_norm": 1.46875,
+      "learning_rate": 0.00015473823572741108,
+      "loss": 4.622,
+      "step": 184000
+    },
+    {
+      "epoch": 1.584452479338843,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.00015471556086306637,
+      "loss": 4.5452,
+      "step": 184050
+    },
+    {
+      "epoch": 1.5848829201101928,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.0001546928819827172,
+      "loss": 4.5012,
+      "step": 184100
+    },
+    {
+      "epoch": 1.5853133608815426,
+      "grad_norm": 1.6171875,
+      "learning_rate": 0.0001546701990880282,
+      "loss": 4.6538,
+      "step": 184150
+    },
+    {
+      "epoch": 1.5857438016528924,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.0001546475121806642,
+      "loss": 4.3889,
+      "step": 184200
+    },
+    {
+      "epoch": 1.5861742424242424,
+      "grad_norm": 1.3828125,
+      "learning_rate": 0.00015462482126229042,
+      "loss": 4.3978,
+      "step": 184250
+    },
+    {
+      "epoch": 1.5866046831955924,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00015460212633457222,
+      "loss": 4.4821,
+      "step": 184300
+    },
+    {
+      "epoch": 1.5870351239669422,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.00015457942739917547,
+      "loss": 4.4233,
+      "step": 184350
+    },
+    {
+      "epoch": 1.587465564738292,
+      "grad_norm": 6.28125,
+      "learning_rate": 0.00015455672445776616,
+      "loss": 4.6526,
+      "step": 184400
+    },
+    {
+      "epoch": 1.5878960055096418,
+      "grad_norm": 4.84375,
+      "learning_rate": 0.00015453401751201068,
+      "loss": 4.3834,
+      "step": 184450
+    },
+    {
+      "epoch": 1.5883264462809916,
+      "grad_norm": 2.125,
+      "learning_rate": 0.00015451130656357563,
+      "loss": 4.5531,
+      "step": 184500
+    },
+    {
+      "epoch": 1.5887568870523416,
+      "grad_norm": 5.03125,
+      "learning_rate": 0.00015448859161412796,
+      "loss": 4.6788,
+      "step": 184550
+    },
+    {
+      "epoch": 1.5891873278236914,
+      "grad_norm": 1.96875,
+      "learning_rate": 0.0001544658726653349,
+      "loss": 4.2766,
+      "step": 184600
+    },
+    {
+      "epoch": 1.5896177685950414,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.00015444314971886397,
+      "loss": 4.7039,
+      "step": 184650
+    },
+    {
+      "epoch": 1.5900482093663912,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.000154420422776383,
+      "loss": 4.6241,
+      "step": 184700
+    },
+    {
+      "epoch": 1.590478650137741,
+      "grad_norm": 2.421875,
+      "learning_rate": 0.00015439769183956008,
+      "loss": 4.4409,
+      "step": 184750
+    },
+    {
+      "epoch": 1.5909090909090908,
+      "grad_norm": 3.328125,
+      "learning_rate": 0.00015437495691006362,
+      "loss": 4.2831,
+      "step": 184800
+    },
+    {
+      "epoch": 1.5913395316804406,
+      "grad_norm": 3.1875,
+      "learning_rate": 0.00015435221798956232,
+      "loss": 4.5333,
+      "step": 184850
+    },
+    {
+      "epoch": 1.5917699724517906,
+      "grad_norm": 1.796875,
+      "learning_rate": 0.00015432947507972516,
+      "loss": 4.4133,
+      "step": 184900
+    },
+    {
+      "epoch": 1.5922004132231407,
+      "grad_norm": 3.765625,
+      "learning_rate": 0.00015430672818222144,
+      "loss": 4.6597,
+      "step": 184950
+    },
+    {
+      "epoch": 1.5926308539944904,
+      "grad_norm": 2.984375,
+      "learning_rate": 0.00015428397729872067,
+      "loss": 4.7734,
+      "step": 185000
+    },
+    {
+      "epoch": 1.5930612947658402,
+      "grad_norm": 3.5625,
+      "learning_rate": 0.00015426122243089282,
+      "loss": 4.5071,
+      "step": 185050
+    },
+    {
+      "epoch": 1.59349173553719,
+      "grad_norm": 3.4375,
+      "learning_rate": 0.00015423846358040793,
+      "loss": 4.7066,
+      "step": 185100
+    },
+    {
+      "epoch": 1.5939221763085398,
+      "grad_norm": 3.546875,
+      "learning_rate": 0.00015421570074893655,
+      "loss": 4.4004,
+      "step": 185150
+    },
+    {
+      "epoch": 1.5943526170798898,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00015419293393814935,
+      "loss": 4.5699,
+      "step": 185200
+    },
+    {
+      "epoch": 1.5947830578512396,
+      "grad_norm": 1.59375,
+      "learning_rate": 0.00015417016314971742,
+      "loss": 4.4724,
+      "step": 185250
+    },
+    {
+      "epoch": 1.5952134986225897,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.00015414738838531207,
+      "loss": 4.2948,
+      "step": 185300
+    },
+    {
+      "epoch": 1.5956439393939394,
+      "grad_norm": 1.265625,
+      "learning_rate": 0.00015412460964660488,
+      "loss": 4.5494,
+      "step": 185350
+    },
+    {
+      "epoch": 1.5960743801652892,
+      "grad_norm": 0.86328125,
+      "learning_rate": 0.00015410182693526781,
+      "loss": 4.0839,
+      "step": 185400
+    },
+    {
+      "epoch": 1.596504820936639,
+      "grad_norm": 2.21875,
+      "learning_rate": 0.00015407904025297307,
+      "loss": 4.5799,
+      "step": 185450
+    },
+    {
+      "epoch": 1.5969352617079888,
+      "grad_norm": 4.84375,
+      "learning_rate": 0.00015405624960139312,
+      "loss": 4.5788,
+      "step": 185500
+    },
+    {
+      "epoch": 1.5973657024793388,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.00015403345498220077,
+      "loss": 4.4688,
+      "step": 185550
+    },
+    {
+      "epoch": 1.5977961432506889,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.00015401065639706905,
+      "loss": 4.4344,
+      "step": 185600
+    },
+    {
+      "epoch": 1.5982265840220387,
+      "grad_norm": 3.484375,
+      "learning_rate": 0.00015398785384767135,
+      "loss": 4.6916,
+      "step": 185650
+    },
+    {
+      "epoch": 1.5986570247933884,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.00015396504733568135,
+      "loss": 4.6307,
+      "step": 185700
+    },
+    {
+      "epoch": 1.5990874655647382,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00015394223686277298,
+      "loss": 4.7216,
+      "step": 185750
+    },
+    {
+      "epoch": 1.599517906336088,
+      "grad_norm": 3.9375,
+      "learning_rate": 0.00015391942243062051,
+      "loss": 4.4658,
+      "step": 185800
+    },
+    {
+      "epoch": 1.599948347107438,
+      "grad_norm": 3.203125,
+      "learning_rate": 0.0001538966040408984,
+      "loss": 4.5408,
+      "step": 185850
+    },
+    {
+      "epoch": 1.6003787878787878,
+      "grad_norm": 3.453125,
+      "learning_rate": 0.00015387378169528154,
+      "loss": 4.1321,
+      "step": 185900
+    },
+    {
+      "epoch": 1.6008092286501379,
+      "grad_norm": 3.921875,
+      "learning_rate": 0.00015385095539544502,
+      "loss": 4.4429,
+      "step": 185950
+    },
+    {
+      "epoch": 1.6012396694214877,
+      "grad_norm": 5.28125,
+      "learning_rate": 0.00015382812514306424,
+      "loss": 4.4706,
+      "step": 186000
+    },
+    {
+      "epoch": 1.6012396694214877,
+      "eval_loss": 5.098446846008301,
+      "eval_runtime": 22.2285,
+      "eval_samples_per_second": 28.792,
+      "eval_steps_per_second": 14.396,
+      "eval_tts_loss": 7.176906156351496,
+      "step": 186000
+    },
+    {
+      "epoch": 1.6016701101928374,
+      "grad_norm": 3.234375,
+      "learning_rate": 0.00015380529093981488,
+      "loss": 4.5768,
+      "step": 186050
+    },
+    {
+      "epoch": 1.6021005509641872,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00015378245278737293,
+      "loss": 4.2978,
+      "step": 186100
+    },
+    {
+      "epoch": 1.602530991735537,
+      "grad_norm": 2.421875,
+      "learning_rate": 0.00015375961068741468,
+      "loss": 4.6696,
+      "step": 186150
+    },
+    {
+      "epoch": 1.602961432506887,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.00015373676464161667,
+      "loss": 4.2091,
+      "step": 186200
+    },
+    {
+      "epoch": 1.603391873278237,
+      "grad_norm": 1.859375,
+      "learning_rate": 0.00015371391465165572,
+      "loss": 4.1663,
+      "step": 186250
+    },
+    {
+      "epoch": 1.6038223140495869,
+      "grad_norm": 2.125,
+      "learning_rate": 0.00015369106071920903,
+      "loss": 4.6291,
+      "step": 186300
+    },
+    {
+      "epoch": 1.6042527548209367,
+      "grad_norm": 1.8515625,
+      "learning_rate": 0.000153668202845954,
+      "loss": 3.8614,
+      "step": 186350
+    },
+    {
+      "epoch": 1.6046831955922864,
+      "grad_norm": 2.140625,
+      "learning_rate": 0.00015364534103356833,
+      "loss": 4.5498,
+      "step": 186400
+    },
+    {
+      "epoch": 1.6051136363636362,
+      "grad_norm": 4.09375,
+      "learning_rate": 0.00015362247528373008,
+      "loss": 4.4633,
+      "step": 186450
+    },
+    {
+      "epoch": 1.6055440771349863,
+      "grad_norm": 3.546875,
+      "learning_rate": 0.00015359960559811747,
+      "loss": 4.5751,
+      "step": 186500
+    },
+    {
+      "epoch": 1.605974517906336,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00015357673197840917,
+      "loss": 4.6192,
+      "step": 186550
+    },
+    {
+      "epoch": 1.606404958677686,
+      "grad_norm": 4.21875,
+      "learning_rate": 0.000153553854426284,
+      "loss": 4.9957,
+      "step": 186600
+    },
+    {
+      "epoch": 1.6068353994490359,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.00015353097294342117,
+      "loss": 4.5131,
+      "step": 186650
+    },
+    {
+      "epoch": 1.6072658402203857,
+      "grad_norm": 0.99609375,
+      "learning_rate": 0.00015350808753150007,
+      "loss": 4.3742,
+      "step": 186700
+    },
+    {
+      "epoch": 1.6076962809917354,
+      "grad_norm": 2.9375,
+      "learning_rate": 0.00015348519819220047,
+      "loss": 4.55,
+      "step": 186750
+    },
+    {
+      "epoch": 1.6081267217630852,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00015346230492720243,
+      "loss": 4.6985,
+      "step": 186800
+    },
+    {
+      "epoch": 1.6085571625344353,
+      "grad_norm": 4.25,
+      "learning_rate": 0.00015343940773818618,
+      "loss": 4.6359,
+      "step": 186850
+    },
+    {
+      "epoch": 1.6089876033057853,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00015341650662683241,
+      "loss": 4.5115,
+      "step": 186900
+    },
+    {
+      "epoch": 1.609418044077135,
+      "grad_norm": 0.9375,
+      "learning_rate": 0.000153393601594822,
+      "loss": 3.9881,
+      "step": 186950
+    },
+    {
+      "epoch": 1.6098484848484849,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.0001533706926438361,
+      "loss": 4.2635,
+      "step": 187000
+    },
+    {
+      "epoch": 1.6102789256198347,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.00015334777977555616,
+      "loss": 4.1501,
+      "step": 187050
+    },
+    {
+      "epoch": 1.6107093663911844,
+      "grad_norm": 1.609375,
+      "learning_rate": 0.000153324862991664,
+      "loss": 4.1152,
+      "step": 187100
+    },
+    {
+      "epoch": 1.6111398071625345,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.00015330194229384162,
+      "loss": 3.8501,
+      "step": 187150
+    },
+    {
+      "epoch": 1.6115702479338843,
+      "grad_norm": 3.546875,
+      "learning_rate": 0.00015327901768377137,
+      "loss": 4.6216,
+      "step": 187200
+    },
+    {
+      "epoch": 1.6120006887052343,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00015325608916313583,
+      "loss": 4.8409,
+      "step": 187250
+    },
+    {
+      "epoch": 1.612431129476584,
+      "grad_norm": 3.09375,
+      "learning_rate": 0.00015323315673361796,
+      "loss": 4.0541,
+      "step": 187300
+    },
+    {
+      "epoch": 1.6128615702479339,
+      "grad_norm": 3.421875,
+      "learning_rate": 0.0001532102203969009,
+      "loss": 4.5262,
+      "step": 187350
+    },
+    {
+      "epoch": 1.6132920110192837,
+      "grad_norm": 2.125,
+      "learning_rate": 0.00015318728015466815,
+      "loss": 4.6309,
+      "step": 187400
+    },
+    {
+      "epoch": 1.6137224517906334,
+      "grad_norm": 2.875,
+      "learning_rate": 0.00015316433600860347,
+      "loss": 4.4653,
+      "step": 187450
+    },
+    {
+      "epoch": 1.6141528925619835,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00015314138796039092,
+      "loss": 4.3707,
+      "step": 187500
+    },
+    {
+      "epoch": 1.6145833333333335,
+      "grad_norm": 3.546875,
+      "learning_rate": 0.00015311843601171482,
+      "loss": 4.5553,
+      "step": 187550
+    },
+    {
+      "epoch": 1.6150137741046833,
+      "grad_norm": 3.375,
+      "learning_rate": 0.00015309548016425982,
+      "loss": 4.2366,
+      "step": 187600
+    },
+    {
+      "epoch": 1.615444214876033,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.0001530725204197108,
+      "loss": 4.2739,
+      "step": 187650
+    },
+    {
+      "epoch": 1.6158746556473829,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.00015304955677975296,
+      "loss": 4.6743,
+      "step": 187700
+    },
+    {
+      "epoch": 1.6163050964187327,
+      "grad_norm": 3.1875,
+      "learning_rate": 0.0001530265892460718,
+      "loss": 4.9168,
+      "step": 187750
+    },
+    {
+      "epoch": 1.6167355371900827,
+      "grad_norm": 1.5234375,
+      "learning_rate": 0.00015300361782035312,
+      "loss": 4.8937,
+      "step": 187800
+    },
+    {
+      "epoch": 1.6171659779614325,
+      "grad_norm": 3.171875,
+      "learning_rate": 0.00015298064250428287,
+      "loss": 4.6395,
+      "step": 187850
+    },
+    {
+      "epoch": 1.6175964187327825,
+      "grad_norm": 5.375,
+      "learning_rate": 0.00015295766329954748,
+      "loss": 4.5475,
+      "step": 187900
+    },
+    {
+      "epoch": 1.6180268595041323,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.00015293468020783352,
+      "loss": 4.2386,
+      "step": 187950
+    },
+    {
+      "epoch": 1.618457300275482,
+      "grad_norm": 3.640625,
+      "learning_rate": 0.00015291169323082798,
+      "loss": 3.9456,
+      "step": 188000
+    },
+    {
+      "epoch": 1.6188877410468319,
+      "grad_norm": 1.3828125,
+      "learning_rate": 0.00015288870237021795,
+      "loss": 4.2349,
+      "step": 188050
+    },
+    {
+      "epoch": 1.6193181818181817,
+      "grad_norm": 1.890625,
+      "learning_rate": 0.00015286570762769098,
+      "loss": 4.4596,
+      "step": 188100
+    },
+    {
+      "epoch": 1.6197486225895317,
+      "grad_norm": 3.609375,
+      "learning_rate": 0.0001528427090049348,
+      "loss": 4.2438,
+      "step": 188150
+    },
+    {
+      "epoch": 1.6201790633608817,
+      "grad_norm": 1.28125,
+      "learning_rate": 0.00015281970650363747,
+      "loss": 4.5402,
+      "step": 188200
+    },
+    {
+      "epoch": 1.6206095041322315,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00015279670012548734,
+      "loss": 4.7755,
+      "step": 188250
+    },
+    {
+      "epoch": 1.6210399449035813,
+      "grad_norm": 3.953125,
+      "learning_rate": 0.00015277368987217301,
+      "loss": 4.5909,
+      "step": 188300
+    },
+    {
+      "epoch": 1.621470385674931,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.00015275067574538337,
+      "loss": 4.1169,
+      "step": 188350
+    },
+    {
+      "epoch": 1.6219008264462809,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.00015272765774680767,
+      "loss": 4.47,
+      "step": 188400
+    },
+    {
+      "epoch": 1.6223312672176309,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.0001527046358781353,
+      "loss": 4.5197,
+      "step": 188450
+    },
+    {
+      "epoch": 1.6227617079889807,
+      "grad_norm": 0.875,
+      "learning_rate": 0.0001526816101410561,
+      "loss": 4.3642,
+      "step": 188500
+    },
+    {
+      "epoch": 1.6231921487603307,
+      "grad_norm": 3.203125,
+      "learning_rate": 0.00015265858053726002,
+      "loss": 4.4812,
+      "step": 188550
+    },
+    {
+      "epoch": 1.6236225895316805,
+      "grad_norm": 2.625,
+      "learning_rate": 0.00015263554706843743,
+      "loss": 4.1564,
+      "step": 188600
+    },
+    {
+      "epoch": 1.6240530303030303,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.00015261250973627895,
+      "loss": 4.4735,
+      "step": 188650
+    },
+    {
+      "epoch": 1.62448347107438,
+      "grad_norm": 3.90625,
+      "learning_rate": 0.00015258946854247545,
+      "loss": 4.3295,
+      "step": 188700
+    },
+    {
+      "epoch": 1.6249139118457299,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.00015256642348871814,
+      "loss": 4.4475,
+      "step": 188750
+    },
+    {
+      "epoch": 1.6253443526170799,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00015254337457669837,
+      "loss": 4.5497,
+      "step": 188800
+    },
+    {
+      "epoch": 1.62577479338843,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.000152520321808108,
+      "loss": 4.4441,
+      "step": 188850
+    },
+    {
+      "epoch": 1.6262052341597797,
+      "grad_norm": 4.4375,
+      "learning_rate": 0.00015249726518463903,
+      "loss": 4.3371,
+      "step": 188900
+    },
+    {
+      "epoch": 1.6266356749311295,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00015247420470798368,
+      "loss": 4.3851,
+      "step": 188950
+    },
+    {
+      "epoch": 1.6270661157024793,
+      "grad_norm": 3.046875,
+      "learning_rate": 0.0001524511403798347,
+      "loss": 4.4025,
+      "step": 189000
+    },
+    {
+      "epoch": 1.6270661157024793,
+      "eval_loss": 5.09128475189209,
+      "eval_runtime": 21.7228,
+      "eval_samples_per_second": 29.462,
+      "eval_steps_per_second": 14.731,
+      "eval_tts_loss": 7.2094475086891014,
+      "step": 189000
+    },
+    {
+      "epoch": 1.627496556473829,
+      "grad_norm": 4.90625,
+      "learning_rate": 0.0001524280722018848,
+      "loss": 4.5485,
+      "step": 189050
+    },
+    {
+      "epoch": 1.627926997245179,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.0001524050001758272,
+      "loss": 4.8776,
+      "step": 189100
+    },
+    {
+      "epoch": 1.6283574380165289,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.00015238192430335536,
+      "loss": 4.4951,
+      "step": 189150
+    },
+    {
+      "epoch": 1.628787878787879,
+      "grad_norm": 4.625,
+      "learning_rate": 0.00015235884458616298,
+      "loss": 4.064,
+      "step": 189200
+    },
+    {
+      "epoch": 1.6292183195592287,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.00015233576102594406,
+      "loss": 4.4201,
+      "step": 189250
+    },
+    {
+      "epoch": 1.6296487603305785,
+      "grad_norm": 3.328125,
+      "learning_rate": 0.00015231267362439285,
+      "loss": 4.2791,
+      "step": 189300
+    },
+    {
+      "epoch": 1.6300792011019283,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.000152289582383204,
+      "loss": 4.7889,
+      "step": 189350
+    },
+    {
+      "epoch": 1.630509641873278,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.0001522664873040723,
+      "loss": 4.3743,
+      "step": 189400
+    },
+    {
+      "epoch": 1.630940082644628,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.00015224338838869286,
+      "loss": 4.4437,
+      "step": 189450
+    },
+    {
+      "epoch": 1.631370523415978,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.00015222028563876113,
+      "loss": 4.7092,
+      "step": 189500
+    },
+    {
+      "epoch": 1.631800964187328,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.00015219717905597281,
+      "loss": 4.5256,
+      "step": 189550
+    },
+    {
+      "epoch": 1.6322314049586777,
+      "grad_norm": 5.09375,
+      "learning_rate": 0.00015217406864202382,
+      "loss": 4.5503,
+      "step": 189600
+    },
+    {
+      "epoch": 1.6326618457300275,
+      "grad_norm": 3.21875,
+      "learning_rate": 0.00015215095439861047,
+      "loss": 4.4078,
+      "step": 189650
+    },
+    {
+      "epoch": 1.6330922865013773,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.00015212783632742927,
+      "loss": 4.8465,
+      "step": 189700
+    },
+    {
+      "epoch": 1.6335227272727273,
+      "grad_norm": 2.625,
+      "learning_rate": 0.000152104714430177,
+      "loss": 4.704,
+      "step": 189750
+    },
+    {
+      "epoch": 1.633953168044077,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.00015208158870855086,
+      "loss": 4.7097,
+      "step": 189800
+    },
+    {
+      "epoch": 1.634383608815427,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00015205845916424813,
+      "loss": 4.1001,
+      "step": 189850
+    },
+    {
+      "epoch": 1.634814049586777,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.0001520353257989665,
+      "loss": 4.47,
+      "step": 189900
+    },
+    {
+      "epoch": 1.6352444903581267,
+      "grad_norm": 1.6171875,
+      "learning_rate": 0.00015201218861440392,
+      "loss": 4.299,
+      "step": 189950
+    },
+    {
+      "epoch": 1.6356749311294765,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.00015198904761225856,
+      "loss": 4.4968,
+      "step": 190000
+    },
+    {
+      "epoch": 1.6361053719008265,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.00015196590279422905,
+      "loss": 4.8038,
+      "step": 190050
+    },
+    {
+      "epoch": 1.6365358126721763,
+      "grad_norm": 3.53125,
+      "learning_rate": 0.00015194275416201402,
+      "loss": 4.4972,
+      "step": 190100
+    },
+    {
+      "epoch": 1.6369662534435263,
+      "grad_norm": 3.671875,
+      "learning_rate": 0.00015191960171731262,
+      "loss": 4.456,
+      "step": 190150
+    },
+    {
+      "epoch": 1.637396694214876,
+      "grad_norm": 1.7890625,
+      "learning_rate": 0.00015189644546182416,
+      "loss": 4.8478,
+      "step": 190200
+    },
+    {
+      "epoch": 1.637827134986226,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.00015187328539724822,
+      "loss": 4.5881,
+      "step": 190250
+    },
+    {
+      "epoch": 1.6382575757575757,
+      "grad_norm": 2.625,
+      "learning_rate": 0.0001518501215252848,
+      "loss": 4.3995,
+      "step": 190300
+    },
+    {
+      "epoch": 1.6386880165289255,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.00015182695384763397,
+      "loss": 4.3681,
+      "step": 190350
+    },
+    {
+      "epoch": 1.6391184573002755,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.00015180378236599626,
+      "loss": 3.9152,
+      "step": 190400
+    },
+    {
+      "epoch": 1.6395488980716253,
+      "grad_norm": 1.9375,
+      "learning_rate": 0.00015178060708207237,
+      "loss": 4.5084,
+      "step": 190450
+    },
+    {
+      "epoch": 1.6399793388429753,
+      "grad_norm": 3.875,
+      "learning_rate": 0.00015175742799756337,
+      "loss": 4.7896,
+      "step": 190500
+    },
+    {
+      "epoch": 1.640409779614325,
+      "grad_norm": 4.09375,
+      "learning_rate": 0.0001517342451141705,
+      "loss": 4.5116,
+      "step": 190550
+    },
+    {
+      "epoch": 1.640840220385675,
+      "grad_norm": 1.765625,
+      "learning_rate": 0.00015171105843359532,
+      "loss": 4.305,
+      "step": 190600
+    },
+    {
+      "epoch": 1.6412706611570247,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.0001516878679575397,
+      "loss": 4.2678,
+      "step": 190650
+    },
+    {
+      "epoch": 1.6417011019283747,
+      "grad_norm": 3.140625,
+      "learning_rate": 0.0001516646736877058,
+      "loss": 4.5873,
+      "step": 190700
+    },
+    {
+      "epoch": 1.6421315426997245,
+      "grad_norm": 3.109375,
+      "learning_rate": 0.00015164147562579604,
+      "loss": 4.5,
+      "step": 190750
+    },
+    {
+      "epoch": 1.6425619834710745,
+      "grad_norm": 2.984375,
+      "learning_rate": 0.00015161827377351304,
+      "loss": 4.5827,
+      "step": 190800
+    },
+    {
+      "epoch": 1.6429924242424243,
+      "grad_norm": 2.234375,
+      "learning_rate": 0.00015159506813255982,
+      "loss": 4.2364,
+      "step": 190850
+    },
+    {
+      "epoch": 1.643422865013774,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.00015157185870463962,
+      "loss": 4.5109,
+      "step": 190900
+    },
+    {
+      "epoch": 1.643853305785124,
+      "grad_norm": 1.3125,
+      "learning_rate": 0.0001515486454914559,
+      "loss": 4.3653,
+      "step": 190950
+    },
+    {
+      "epoch": 1.6442837465564737,
+      "grad_norm": 1.3984375,
+      "learning_rate": 0.00015152542849471257,
+      "loss": 4.5856,
+      "step": 191000
+    },
+    {
+      "epoch": 1.6447141873278237,
+      "grad_norm": 1.9140625,
+      "learning_rate": 0.00015150220771611362,
+      "loss": 4.163,
+      "step": 191050
+    },
+    {
+      "epoch": 1.6451446280991735,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.00015147898315736344,
+      "loss": 4.4394,
+      "step": 191100
+    },
+    {
+      "epoch": 1.6455750688705235,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.00015145575482016665,
+      "loss": 4.8728,
+      "step": 191150
+    },
+    {
+      "epoch": 1.6460055096418733,
+      "grad_norm": 1.78125,
+      "learning_rate": 0.0001514325227062282,
+      "loss": 4.4098,
+      "step": 191200
+    },
+    {
+      "epoch": 1.646435950413223,
+      "grad_norm": 1.9140625,
+      "learning_rate": 0.0001514092868172532,
+      "loss": 4.7922,
+      "step": 191250
+    },
+    {
+      "epoch": 1.646866391184573,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.00015138604715494715,
+      "loss": 4.5203,
+      "step": 191300
+    },
+    {
+      "epoch": 1.647296831955923,
+      "grad_norm": 1.640625,
+      "learning_rate": 0.00015136280372101584,
+      "loss": 4.603,
+      "step": 191350
+    },
+    {
+      "epoch": 1.6477272727272727,
+      "grad_norm": 3.40625,
+      "learning_rate": 0.00015133955651716523,
+      "loss": 4.2557,
+      "step": 191400
+    },
+    {
+      "epoch": 1.6481577134986227,
+      "grad_norm": 2.140625,
+      "learning_rate": 0.00015131630554510164,
+      "loss": 4.4931,
+      "step": 191450
+    },
+    {
+      "epoch": 1.6485881542699725,
+      "grad_norm": 3.4375,
+      "learning_rate": 0.0001512930508065316,
+      "loss": 4.4463,
+      "step": 191500
+    },
+    {
+      "epoch": 1.6490185950413223,
+      "grad_norm": 1.9140625,
+      "learning_rate": 0.00015126979230316203,
+      "loss": 4.1648,
+      "step": 191550
+    },
+    {
+      "epoch": 1.649449035812672,
+      "grad_norm": 3.703125,
+      "learning_rate": 0.0001512465300367,
+      "loss": 4.6763,
+      "step": 191600
+    },
+    {
+      "epoch": 1.649879476584022,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.0001512232640088529,
+      "loss": 4.342,
+      "step": 191650
+    },
+    {
+      "epoch": 1.650309917355372,
+      "grad_norm": 3.75,
+      "learning_rate": 0.00015119999422132848,
+      "loss": 4.6833,
+      "step": 191700
+    },
+    {
+      "epoch": 1.650740358126722,
+      "grad_norm": 3.609375,
+      "learning_rate": 0.0001511767206758346,
+      "loss": 4.6345,
+      "step": 191750
+    },
+    {
+      "epoch": 1.6511707988980717,
+      "grad_norm": 3.0625,
+      "learning_rate": 0.00015115344337407954,
+      "loss": 4.7899,
+      "step": 191800
+    },
+    {
+      "epoch": 1.6516012396694215,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00015113016231777182,
+      "loss": 4.3758,
+      "step": 191850
+    },
+    {
+      "epoch": 1.6520316804407713,
+      "grad_norm": 1.671875,
+      "learning_rate": 0.00015110687750862018,
+      "loss": 4.597,
+      "step": 191900
+    },
+    {
+      "epoch": 1.652462121212121,
+      "grad_norm": 3.140625,
+      "learning_rate": 0.0001510835889483337,
+      "loss": 4.2432,
+      "step": 191950
+    },
+    {
+      "epoch": 1.6528925619834711,
+      "grad_norm": 4.46875,
+      "learning_rate": 0.0001510602966386217,
+      "loss": 4.4862,
+      "step": 192000
+    },
+    {
+      "epoch": 1.6528925619834711,
+      "eval_loss": 5.084532260894775,
+      "eval_runtime": 21.8396,
+      "eval_samples_per_second": 29.305,
+      "eval_steps_per_second": 14.652,
+      "eval_tts_loss": 7.2402833804910385,
+      "step": 192000
+    },
+    {
+      "epoch": 1.653323002754821,
+      "grad_norm": 3.40625,
+      "learning_rate": 0.00015103700058119375,
+      "loss": 4.6847,
+      "step": 192050
+    },
+    {
+      "epoch": 1.653753443526171,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.0001510137007777598,
+      "loss": 4.4774,
+      "step": 192100
+    },
+    {
+      "epoch": 1.6541838842975207,
+      "grad_norm": 3.671875,
+      "learning_rate": 0.00015099039723002997,
+      "loss": 4.2566,
+      "step": 192150
+    },
+    {
+      "epoch": 1.6546143250688705,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.00015096708993971467,
+      "loss": 4.6564,
+      "step": 192200
+    },
+    {
+      "epoch": 1.6550447658402203,
+      "grad_norm": 3.078125,
+      "learning_rate": 0.00015094377890852466,
+      "loss": 4.8912,
+      "step": 192250
+    },
+    {
+      "epoch": 1.65547520661157,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.00015092046413817087,
+      "loss": 4.1929,
+      "step": 192300
+    },
+    {
+      "epoch": 1.6559056473829201,
+      "grad_norm": 1.1328125,
+      "learning_rate": 0.00015089714563036458,
+      "loss": 4.3709,
+      "step": 192350
+    },
+    {
+      "epoch": 1.6563360881542701,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.00015087382338681727,
+      "loss": 4.2319,
+      "step": 192400
+    },
+    {
+      "epoch": 1.65676652892562,
+      "grad_norm": 3.28125,
+      "learning_rate": 0.00015085049740924084,
+      "loss": 4.5131,
+      "step": 192450
+    },
+    {
+      "epoch": 1.6571969696969697,
+      "grad_norm": 4.1875,
+      "learning_rate": 0.00015082716769934727,
+      "loss": 4.3805,
+      "step": 192500
+    },
+    {
+      "epoch": 1.6576274104683195,
+      "grad_norm": 3.40625,
+      "learning_rate": 0.00015080383425884896,
+      "loss": 4.6315,
+      "step": 192550
+    },
+    {
+      "epoch": 1.6580578512396693,
+      "grad_norm": 6.5,
+      "learning_rate": 0.0001507804970894585,
+      "loss": 4.4115,
+      "step": 192600
+    },
+    {
+      "epoch": 1.6584882920110193,
+      "grad_norm": 4.65625,
+      "learning_rate": 0.00015075715619288884,
+      "loss": 4.2834,
+      "step": 192650
+    },
+    {
+      "epoch": 1.6589187327823691,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.00015073381157085312,
+      "loss": 4.2,
+      "step": 192700
+    },
+    {
+      "epoch": 1.6593491735537191,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.00015071046322506478,
+      "loss": 4.5527,
+      "step": 192750
+    },
+    {
+      "epoch": 1.659779614325069,
+      "grad_norm": 3.234375,
+      "learning_rate": 0.00015068711115723754,
+      "loss": 4.6736,
+      "step": 192800
+    },
+    {
+      "epoch": 1.6602100550964187,
+      "grad_norm": 3.03125,
+      "learning_rate": 0.00015066375536908541,
+      "loss": 4.5932,
+      "step": 192850
+    },
+    {
+      "epoch": 1.6606404958677685,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00015064039586232264,
+      "loss": 4.5725,
+      "step": 192900
+    },
+    {
+      "epoch": 1.6610709366391183,
+      "grad_norm": 3.484375,
+      "learning_rate": 0.00015061703263866377,
+      "loss": 4.4514,
+      "step": 192950
+    },
+    {
+      "epoch": 1.6615013774104683,
+      "grad_norm": 3.15625,
+      "learning_rate": 0.0001505936656998236,
+      "loss": 4.2902,
+      "step": 193000
+    },
+    {
+      "epoch": 1.6619318181818183,
+      "grad_norm": 3.921875,
+      "learning_rate": 0.00015057029504751722,
+      "loss": 4.3745,
+      "step": 193050
+    },
+    {
+      "epoch": 1.6623622589531681,
+      "grad_norm": 3.03125,
+      "learning_rate": 0.00015054692068346,
+      "loss": 4.1348,
+      "step": 193100
+    },
+    {
+      "epoch": 1.662792699724518,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00015052354260936755,
+      "loss": 4.3627,
+      "step": 193150
+    },
+    {
+      "epoch": 1.6632231404958677,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00015050016082695578,
+      "loss": 4.3502,
+      "step": 193200
+    },
+    {
+      "epoch": 1.6636535812672175,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.00015047677533794085,
+      "loss": 4.0829,
+      "step": 193250
+    },
+    {
+      "epoch": 1.6640840220385675,
+      "grad_norm": 3.25,
+      "learning_rate": 0.00015045338614403923,
+      "loss": 4.6117,
+      "step": 193300
+    },
+    {
+      "epoch": 1.6645144628099173,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.00015042999324696758,
+      "loss": 4.1233,
+      "step": 193350
+    },
+    {
+      "epoch": 1.6649449035812673,
+      "grad_norm": 3.265625,
+      "learning_rate": 0.00015040659664844296,
+      "loss": 4.6023,
+      "step": 193400
+    },
+    {
+      "epoch": 1.6653753443526171,
+      "grad_norm": 3.328125,
+      "learning_rate": 0.00015038319635018264,
+      "loss": 4.8547,
+      "step": 193450
+    },
+    {
+      "epoch": 1.665805785123967,
+      "grad_norm": 4.46875,
+      "learning_rate": 0.00015035979235390406,
+      "loss": 4.1835,
+      "step": 193500
+    },
+    {
+      "epoch": 1.6662362258953167,
+      "grad_norm": 4.25,
+      "learning_rate": 0.00015033638466132508,
+      "loss": 4.5288,
+      "step": 193550
+    },
+    {
+      "epoch": 1.6666666666666665,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.00015031297327416376,
+      "loss": 4.5139,
+      "step": 193600
+    },
+    {
+      "epoch": 1.6670971074380165,
+      "grad_norm": 3.15625,
+      "learning_rate": 0.00015028955819413846,
+      "loss": 4.1563,
+      "step": 193650
+    },
+    {
+      "epoch": 1.6675275482093666,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.0001502661394229678,
+      "loss": 4.5309,
+      "step": 193700
+    },
+    {
+      "epoch": 1.6679579889807163,
+      "grad_norm": 3.453125,
+      "learning_rate": 0.00015024271696237063,
+      "loss": 4.2894,
+      "step": 193750
+    },
+    {
+      "epoch": 1.6683884297520661,
+      "grad_norm": 0.76171875,
+      "learning_rate": 0.00015021929081406618,
+      "loss": 4.2516,
+      "step": 193800
+    },
+    {
+      "epoch": 1.668818870523416,
+      "grad_norm": 3.859375,
+      "learning_rate": 0.00015019586097977377,
+      "loss": 4.7012,
+      "step": 193850
+    },
+    {
+      "epoch": 1.6692493112947657,
+      "grad_norm": 3.859375,
+      "learning_rate": 0.00015017242746121322,
+      "loss": 4.2681,
+      "step": 193900
+    },
+    {
+      "epoch": 1.6696797520661157,
+      "grad_norm": 2.296875,
+      "learning_rate": 0.0001501489902601044,
+      "loss": 4.5689,
+      "step": 193950
+    },
+    {
+      "epoch": 1.6701101928374655,
+      "grad_norm": 3.8125,
+      "learning_rate": 0.00015012554937816766,
+      "loss": 4.0734,
+      "step": 194000
+    },
+    {
+      "epoch": 1.6705406336088156,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.0001501021048171234,
+      "loss": 4.3694,
+      "step": 194050
+    },
+    {
+      "epoch": 1.6709710743801653,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.00015007865657869242,
+      "loss": 4.2628,
+      "step": 194100
+    },
+    {
+      "epoch": 1.6714015151515151,
+      "grad_norm": 4.34375,
+      "learning_rate": 0.00015005520466459583,
+      "loss": 4.424,
+      "step": 194150
+    },
+    {
+      "epoch": 1.671831955922865,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.0001500317490765549,
+      "loss": 4.4128,
+      "step": 194200
+    },
+    {
+      "epoch": 1.6722623966942147,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.0001500082898162912,
+      "loss": 4.1318,
+      "step": 194250
+    },
+    {
+      "epoch": 1.6726928374655647,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.0001499848268855267,
+      "loss": 4.6032,
+      "step": 194300
+    },
+    {
+      "epoch": 1.6731232782369148,
+      "grad_norm": 2.875,
+      "learning_rate": 0.0001499613602859834,
+      "loss": 4.4496,
+      "step": 194350
+    },
+    {
+      "epoch": 1.6735537190082646,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.00014993789001938377,
+      "loss": 4.098,
+      "step": 194400
+    },
+    {
+      "epoch": 1.6739841597796143,
+      "grad_norm": 3.34375,
+      "learning_rate": 0.00014991441608745043,
+      "loss": 4.5486,
+      "step": 194450
+    },
+    {
+      "epoch": 1.6744146005509641,
+      "grad_norm": 3.1875,
+      "learning_rate": 0.00014989093849190636,
+      "loss": 4.3258,
+      "step": 194500
+    },
+    {
+      "epoch": 1.674845041322314,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.00014986745723447476,
+      "loss": 4.0024,
+      "step": 194550
+    },
+    {
+      "epoch": 1.675275482093664,
+      "grad_norm": 3.90625,
+      "learning_rate": 0.00014984397231687906,
+      "loss": 4.1615,
+      "step": 194600
+    },
+    {
+      "epoch": 1.6757059228650137,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.00014982048374084306,
+      "loss": 4.5995,
+      "step": 194650
+    },
+    {
+      "epoch": 1.6761363636363638,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.00014979699150809075,
+      "loss": 4.1943,
+      "step": 194700
+    },
+    {
+      "epoch": 1.6765668044077136,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.0001497734956203464,
+      "loss": 4.5542,
+      "step": 194750
+    },
+    {
+      "epoch": 1.6769972451790633,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.00014974999607933454,
+      "loss": 4.0834,
+      "step": 194800
+    },
+    {
+      "epoch": 1.6774276859504131,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.00014972649288678,
+      "loss": 4.4329,
+      "step": 194850
+    },
+    {
+      "epoch": 1.677858126721763,
+      "grad_norm": 3.578125,
+      "learning_rate": 0.00014970298604440792,
+      "loss": 4.3069,
+      "step": 194900
+    },
+    {
+      "epoch": 1.678288567493113,
+      "grad_norm": 1.8984375,
+      "learning_rate": 0.00014967947555394358,
+      "loss": 4.1882,
+      "step": 194950
+    },
+    {
+      "epoch": 1.678719008264463,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00014965596141711262,
+      "loss": 4.1222,
+      "step": 195000
+    },
+    {
+      "epoch": 1.678719008264463,
+      "eval_loss": 5.083550453186035,
+      "eval_runtime": 21.8258,
+      "eval_samples_per_second": 29.323,
+      "eval_steps_per_second": 14.662,
+      "eval_tts_loss": 7.167528697308984,
+      "step": 195000
+    },
+    {
+      "epoch": 1.6791494490358128,
+      "grad_norm": 3.6875,
+      "learning_rate": 0.00014963244363564094,
+      "loss": 4.3102,
+      "step": 195050
+    },
+    {
+      "epoch": 1.6795798898071626,
+      "grad_norm": 4.03125,
+      "learning_rate": 0.00014960892221125468,
+      "loss": 4.3747,
+      "step": 195100
+    },
+    {
+      "epoch": 1.6800103305785123,
+      "grad_norm": 3.0,
+      "learning_rate": 0.00014958539714568026,
+      "loss": 4.2361,
+      "step": 195150
+    },
+    {
+      "epoch": 1.6804407713498621,
+      "grad_norm": 3.640625,
+      "learning_rate": 0.00014956186844064433,
+      "loss": 4.5603,
+      "step": 195200
+    },
+    {
+      "epoch": 1.6808712121212122,
+      "grad_norm": 2.296875,
+      "learning_rate": 0.00014953833609787396,
+      "loss": 4.4756,
+      "step": 195250
+    },
+    {
+      "epoch": 1.681301652892562,
+      "grad_norm": 3.65625,
+      "learning_rate": 0.00014951480011909628,
+      "loss": 4.6928,
+      "step": 195300
+    },
+    {
+      "epoch": 1.681732093663912,
+      "grad_norm": 1.5546875,
+      "learning_rate": 0.0001494912605060388,
+      "loss": 4.3995,
+      "step": 195350
+    },
+    {
+      "epoch": 1.6821625344352618,
+      "grad_norm": 1.8671875,
+      "learning_rate": 0.0001494677172604293,
+      "loss": 4.5686,
+      "step": 195400
+    },
+    {
+      "epoch": 1.6825929752066116,
+      "grad_norm": 3.40625,
+      "learning_rate": 0.00014944417038399577,
+      "loss": 4.3506,
+      "step": 195450
+    },
+    {
+      "epoch": 1.6830234159779613,
+      "grad_norm": 2.875,
+      "learning_rate": 0.0001494206198784665,
+      "loss": 4.6538,
+      "step": 195500
+    },
+    {
+      "epoch": 1.6834538567493111,
+      "grad_norm": 2.625,
+      "learning_rate": 0.00014939706574557007,
+      "loss": 4.3817,
+      "step": 195550
+    },
+    {
+      "epoch": 1.6838842975206612,
+      "grad_norm": 3.203125,
+      "learning_rate": 0.00014937350798703534,
+      "loss": 4.6453,
+      "step": 195600
+    },
+    {
+      "epoch": 1.6843147382920112,
+      "grad_norm": 1.8828125,
+      "learning_rate": 0.0001493499466045913,
+      "loss": 4.3793,
+      "step": 195650
+    },
+    {
+      "epoch": 1.684745179063361,
+      "grad_norm": 6.46875,
+      "learning_rate": 0.0001493263815999674,
+      "loss": 4.252,
+      "step": 195700
+    },
+    {
+      "epoch": 1.6851756198347108,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.0001493028129748932,
+      "loss": 4.5178,
+      "step": 195750
+    },
+    {
+      "epoch": 1.6856060606060606,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.00014927924073109857,
+      "loss": 4.6309,
+      "step": 195800
+    },
+    {
+      "epoch": 1.6860365013774103,
+      "grad_norm": 3.203125,
+      "learning_rate": 0.00014925566487031375,
+      "loss": 4.4545,
+      "step": 195850
+    },
+    {
+      "epoch": 1.6864669421487604,
+      "grad_norm": 0.87890625,
+      "learning_rate": 0.00014923208539426906,
+      "loss": 4.5176,
+      "step": 195900
+    },
+    {
+      "epoch": 1.6868973829201102,
+      "grad_norm": 4.71875,
+      "learning_rate": 0.00014920850230469527,
+      "loss": 4.4551,
+      "step": 195950
+    },
+    {
+      "epoch": 1.6873278236914602,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.00014918491560332326,
+      "loss": 4.4299,
+      "step": 196000
+    },
+    {
+      "epoch": 1.68775826446281,
+      "grad_norm": 6.03125,
+      "learning_rate": 0.00014916132529188428,
+      "loss": 4.6912,
+      "step": 196050
+    },
+    {
+      "epoch": 1.6881887052341598,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.0001491377313721098,
+      "loss": 4.459,
+      "step": 196100
+    },
+    {
+      "epoch": 1.6886191460055096,
+      "grad_norm": 1.8984375,
+      "learning_rate": 0.00014911413384573153,
+      "loss": 4.695,
+      "step": 196150
+    },
+    {
+      "epoch": 1.6890495867768593,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00014909053271448155,
+      "loss": 4.691,
+      "step": 196200
+    },
+    {
+      "epoch": 1.6894800275482094,
+      "grad_norm": 4.375,
+      "learning_rate": 0.00014906692798009207,
+      "loss": 4.4259,
+      "step": 196250
+    },
+    {
+      "epoch": 1.6899104683195594,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00014904331964429564,
+      "loss": 4.3836,
+      "step": 196300
+    },
+    {
+      "epoch": 1.6903409090909092,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.00014901970770882507,
+      "loss": 4.6083,
+      "step": 196350
+    },
+    {
+      "epoch": 1.690771349862259,
+      "grad_norm": 4.75,
+      "learning_rate": 0.00014899609217541344,
+      "loss": 4.3447,
+      "step": 196400
+    },
+    {
+      "epoch": 1.6912017906336088,
+      "grad_norm": 3.890625,
+      "learning_rate": 0.00014897247304579406,
+      "loss": 4.5977,
+      "step": 196450
+    },
+    {
+      "epoch": 1.6916322314049586,
+      "grad_norm": 3.4375,
+      "learning_rate": 0.00014894885032170051,
+      "loss": 4.5123,
+      "step": 196500
+    },
+    {
+      "epoch": 1.6920626721763086,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.00014892522400486669,
+      "loss": 4.6142,
+      "step": 196550
+    },
+    {
+      "epoch": 1.6924931129476584,
+      "grad_norm": 3.375,
+      "learning_rate": 0.00014890159409702665,
+      "loss": 4.3423,
+      "step": 196600
+    },
+    {
+      "epoch": 1.6929235537190084,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.00014887796059991483,
+      "loss": 4.5783,
+      "step": 196650
+    },
+    {
+      "epoch": 1.6933539944903582,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.00014885432351526587,
+      "loss": 4.1651,
+      "step": 196700
+    },
+    {
+      "epoch": 1.693784435261708,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.0001488306828448147,
+      "loss": 4.4109,
+      "step": 196750
+    },
+    {
+      "epoch": 1.6942148760330578,
+      "grad_norm": 3.03125,
+      "learning_rate": 0.00014880703859029645,
+      "loss": 4.3027,
+      "step": 196800
+    },
+    {
+      "epoch": 1.6946453168044076,
+      "grad_norm": 3.703125,
+      "learning_rate": 0.00014878339075344657,
+      "loss": 4.6697,
+      "step": 196850
+    },
+    {
+      "epoch": 1.6950757575757576,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00014875973933600078,
+      "loss": 4.5617,
+      "step": 196900
+    },
+    {
+      "epoch": 1.6955061983471076,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00014873608433969504,
+      "loss": 4.4808,
+      "step": 196950
+    },
+    {
+      "epoch": 1.6959366391184574,
+      "grad_norm": 1.8515625,
+      "learning_rate": 0.00014871242576626557,
+      "loss": 4.5233,
+      "step": 197000
+    },
+    {
+      "epoch": 1.6963670798898072,
+      "grad_norm": 5.03125,
+      "learning_rate": 0.00014868876361744882,
+      "loss": 4.802,
+      "step": 197050
+    },
+    {
+      "epoch": 1.696797520661157,
+      "grad_norm": 1.9375,
+      "learning_rate": 0.00014866509789498162,
+      "loss": 4.5043,
+      "step": 197100
+    },
+    {
+      "epoch": 1.6972279614325068,
+      "grad_norm": 2.25,
+      "learning_rate": 0.0001486414286006009,
+      "loss": 4.6736,
+      "step": 197150
+    },
+    {
+      "epoch": 1.6976584022038568,
+      "grad_norm": 1.921875,
+      "learning_rate": 0.000148617755736044,
+      "loss": 4.8234,
+      "step": 197200
+    },
+    {
+      "epoch": 1.6980888429752066,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.00014859407930304844,
+      "loss": 4.7812,
+      "step": 197250
+    },
+    {
+      "epoch": 1.6985192837465566,
+      "grad_norm": 1.1171875,
+      "learning_rate": 0.000148570399303352,
+      "loss": 4.5574,
+      "step": 197300
+    },
+    {
+      "epoch": 1.6989497245179064,
+      "grad_norm": 3.0,
+      "learning_rate": 0.00014854671573869276,
+      "loss": 4.3152,
+      "step": 197350
+    },
+    {
+      "epoch": 1.6993801652892562,
+      "grad_norm": 4.125,
+      "learning_rate": 0.00014852302861080905,
+      "loss": 4.6025,
+      "step": 197400
+    },
+    {
+      "epoch": 1.699810606060606,
+      "grad_norm": 1.96875,
+      "learning_rate": 0.0001484993379214394,
+      "loss": 4.721,
+      "step": 197450
+    },
+    {
+      "epoch": 1.7002410468319558,
+      "grad_norm": 3.265625,
+      "learning_rate": 0.00014847564367232274,
+      "loss": 4.4474,
+      "step": 197500
+    },
+    {
+      "epoch": 1.7006714876033058,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.0001484519458651981,
+      "loss": 4.3136,
+      "step": 197550
+    },
+    {
+      "epoch": 1.7011019283746558,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.00014842824450180488,
+      "loss": 4.4143,
+      "step": 197600
+    },
+    {
+      "epoch": 1.7015323691460056,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.0001484045395838827,
+      "loss": 4.3645,
+      "step": 197650
+    },
+    {
+      "epoch": 1.7019628099173554,
+      "grad_norm": 1.2421875,
+      "learning_rate": 0.0001483808311131715,
+      "loss": 4.5915,
+      "step": 197700
+    },
+    {
+      "epoch": 1.7023932506887052,
+      "grad_norm": 3.328125,
+      "learning_rate": 0.00014835711909141138,
+      "loss": 4.3323,
+      "step": 197750
+    },
+    {
+      "epoch": 1.702823691460055,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.00014833340352034274,
+      "loss": 4.3499,
+      "step": 197800
+    },
+    {
+      "epoch": 1.703254132231405,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.00014830968440170627,
+      "loss": 4.3072,
+      "step": 197850
+    },
+    {
+      "epoch": 1.7036845730027548,
+      "grad_norm": 3.859375,
+      "learning_rate": 0.00014828596173724294,
+      "loss": 4.3554,
+      "step": 197900
+    },
+    {
+      "epoch": 1.7041150137741048,
+      "grad_norm": 1.578125,
+      "learning_rate": 0.0001482622355286939,
+      "loss": 4.1934,
+      "step": 197950
+    },
+    {
+      "epoch": 1.7045454545454546,
+      "grad_norm": 2.265625,
+      "learning_rate": 0.0001482385057778006,
+      "loss": 4.1522,
+      "step": 198000
+    },
+    {
+      "epoch": 1.7045454545454546,
+      "eval_loss": 5.076416492462158,
+      "eval_runtime": 21.6723,
+      "eval_samples_per_second": 29.531,
+      "eval_steps_per_second": 14.765,
+      "eval_tts_loss": 7.241367056405855,
+      "step": 198000
+    },
+    {
+      "epoch": 1.7049758953168044,
+      "grad_norm": 3.0625,
+      "learning_rate": 0.00014821477248630474,
+      "loss": 4.1593,
+      "step": 198050
+    },
+    {
+      "epoch": 1.7054063360881542,
+      "grad_norm": 1.3359375,
+      "learning_rate": 0.00014819103565594837,
+      "loss": 4.7742,
+      "step": 198100
+    },
+    {
+      "epoch": 1.705836776859504,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.00014816729528847365,
+      "loss": 4.3037,
+      "step": 198150
+    },
+    {
+      "epoch": 1.706267217630854,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00014814355138562305,
+      "loss": 4.3153,
+      "step": 198200
+    },
+    {
+      "epoch": 1.706697658402204,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.0001481198039491394,
+      "loss": 4.4642,
+      "step": 198250
+    },
+    {
+      "epoch": 1.7071280991735538,
+      "grad_norm": 4.1875,
+      "learning_rate": 0.00014809605298076568,
+      "loss": 4.4042,
+      "step": 198300
+    },
+    {
+      "epoch": 1.7075585399449036,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00014807229848224512,
+      "loss": 4.6936,
+      "step": 198350
+    },
+    {
+      "epoch": 1.7079889807162534,
+      "grad_norm": 1.8046875,
+      "learning_rate": 0.0001480485404553213,
+      "loss": 4.2344,
+      "step": 198400
+    },
+    {
+      "epoch": 1.7084194214876032,
+      "grad_norm": 1.84375,
+      "learning_rate": 0.000148024778901738,
+      "loss": 4.0703,
+      "step": 198450
+    },
+    {
+      "epoch": 1.7088498622589532,
+      "grad_norm": 3.3125,
+      "learning_rate": 0.00014800101382323922,
+      "loss": 3.8407,
+      "step": 198500
+    },
+    {
+      "epoch": 1.709280303030303,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.00014797724522156935,
+      "loss": 4.6688,
+      "step": 198550
+    },
+    {
+      "epoch": 1.709710743801653,
+      "grad_norm": 3.578125,
+      "learning_rate": 0.00014795347309847287,
+      "loss": 4.6164,
+      "step": 198600
+    },
+    {
+      "epoch": 1.7101411845730028,
+      "grad_norm": 3.171875,
+      "learning_rate": 0.00014792969745569467,
+      "loss": 4.4741,
+      "step": 198650
+    },
+    {
+      "epoch": 1.7105716253443526,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00014790591829497976,
+      "loss": 4.261,
+      "step": 198700
+    },
+    {
+      "epoch": 1.7110020661157024,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.00014788213561807355,
+      "loss": 4.3749,
+      "step": 198750
+    },
+    {
+      "epoch": 1.7114325068870524,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.0001478583494267216,
+      "loss": 4.3727,
+      "step": 198800
+    },
+    {
+      "epoch": 1.7118629476584022,
+      "grad_norm": 1.46875,
+      "learning_rate": 0.00014783455972266975,
+      "loss": 4.1389,
+      "step": 198850
+    },
+    {
+      "epoch": 1.7122933884297522,
+      "grad_norm": 3.390625,
+      "learning_rate": 0.00014781076650766418,
+      "loss": 4.4756,
+      "step": 198900
+    },
+    {
+      "epoch": 1.712723829201102,
+      "grad_norm": 2.875,
+      "learning_rate": 0.0001477869697834512,
+      "loss": 4.4756,
+      "step": 198950
+    },
+    {
+      "epoch": 1.7131542699724518,
+      "grad_norm": 1.6875,
+      "learning_rate": 0.00014776316955177743,
+      "loss": 4.7666,
+      "step": 199000
+    },
+    {
+      "epoch": 1.7135847107438016,
+      "grad_norm": 3.4375,
+      "learning_rate": 0.0001477393658143898,
+      "loss": 4.5628,
+      "step": 199050
+    },
+    {
+      "epoch": 1.7140151515151514,
+      "grad_norm": 1.5859375,
+      "learning_rate": 0.0001477155585730354,
+      "loss": 4.4252,
+      "step": 199100
+    },
+    {
+      "epoch": 1.7144455922865014,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.00014769174782946175,
+      "loss": 4.5871,
+      "step": 199150
+    },
+    {
+      "epoch": 1.7148760330578512,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.00014766793358541634,
+      "loss": 4.3632,
+      "step": 199200
+    },
+    {
+      "epoch": 1.7153064738292012,
+      "grad_norm": 2.265625,
+      "learning_rate": 0.00014764411584264716,
+      "loss": 4.3026,
+      "step": 199250
+    },
+    {
+      "epoch": 1.715736914600551,
+      "grad_norm": 2.875,
+      "learning_rate": 0.00014762029460290243,
+      "loss": 4.6209,
+      "step": 199300
+    },
+    {
+      "epoch": 1.7161673553719008,
+      "grad_norm": 0.84765625,
+      "learning_rate": 0.0001475964698679305,
+      "loss": 3.9673,
+      "step": 199350
+    },
+    {
+      "epoch": 1.7165977961432506,
+      "grad_norm": 3.328125,
+      "learning_rate": 0.00014757264163948013,
+      "loss": 4.2442,
+      "step": 199400
+    },
+    {
+      "epoch": 1.7170282369146006,
+      "grad_norm": 3.609375,
+      "learning_rate": 0.00014754880991930017,
+      "loss": 4.4071,
+      "step": 199450
+    },
+    {
+      "epoch": 1.7174586776859504,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.0001475249747091399,
+      "loss": 4.3913,
+      "step": 199500
+    },
+    {
+      "epoch": 1.7178891184573004,
+      "grad_norm": 3.796875,
+      "learning_rate": 0.00014750113601074875,
+      "loss": 4.8984,
+      "step": 199550
+    },
+    {
+      "epoch": 1.7183195592286502,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.0001474772938258764,
+      "loss": 4.3203,
+      "step": 199600
+    },
+    {
+      "epoch": 1.71875,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.00014745344815627283,
+      "loss": 4.2185,
+      "step": 199650
+    },
+    {
+      "epoch": 1.7191804407713498,
+      "grad_norm": 3.0625,
+      "learning_rate": 0.00014742959900368826,
+      "loss": 4.3581,
+      "step": 199700
+    },
+    {
+      "epoch": 1.7196108815426996,
+      "grad_norm": 1.2890625,
+      "learning_rate": 0.00014740574636987316,
+      "loss": 4.3952,
+      "step": 199750
+    },
+    {
+      "epoch": 1.7200413223140496,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00014738189025657829,
+      "loss": 4.0192,
+      "step": 199800
+    },
+    {
+      "epoch": 1.7204717630853994,
+      "grad_norm": 2.625,
+      "learning_rate": 0.00014735803066555462,
+      "loss": 4.435,
+      "step": 199850
+    },
+    {
+      "epoch": 1.7209022038567494,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.00014733416759855339,
+      "loss": 4.6761,
+      "step": 199900
+    },
+    {
+      "epoch": 1.7213326446280992,
+      "grad_norm": 4.09375,
+      "learning_rate": 0.0001473103010573261,
+      "loss": 4.3119,
+      "step": 199950
+    },
+    {
+      "epoch": 1.721763085399449,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.0001472864310436245,
+      "loss": 4.5503,
+      "step": 200000
+    },
+    {
+      "epoch": 1.7221935261707988,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.00014726255755920063,
+      "loss": 4.4613,
+      "step": 200050
+    },
+    {
+      "epoch": 1.7226239669421488,
+      "grad_norm": 3.734375,
+      "learning_rate": 0.0001472386806058067,
+      "loss": 4.5029,
+      "step": 200100
+    },
+    {
+      "epoch": 1.7230544077134986,
+      "grad_norm": 3.609375,
+      "learning_rate": 0.00014721480018519523,
+      "loss": 4.5508,
+      "step": 200150
+    },
+    {
+      "epoch": 1.7234848484848486,
+      "grad_norm": 2.25,
+      "learning_rate": 0.00014719091629911905,
+      "loss": 4.2141,
+      "step": 200200
+    },
+    {
+      "epoch": 1.7239152892561984,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.00014716702894933113,
+      "loss": 4.1925,
+      "step": 200250
+    },
+    {
+      "epoch": 1.7243457300275482,
+      "grad_norm": 4.09375,
+      "learning_rate": 0.00014714313813758476,
+      "loss": 4.5075,
+      "step": 200300
+    },
+    {
+      "epoch": 1.724776170798898,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.00014711924386563347,
+      "loss": 4.4054,
+      "step": 200350
+    },
+    {
+      "epoch": 1.7252066115702478,
+      "grad_norm": 1.5703125,
+      "learning_rate": 0.00014709534613523107,
+      "loss": 3.9823,
+      "step": 200400
+    },
+    {
+      "epoch": 1.7256370523415978,
+      "grad_norm": 3.4375,
+      "learning_rate": 0.00014707144494813162,
+      "loss": 4.1744,
+      "step": 200450
+    },
+    {
+      "epoch": 1.7260674931129476,
+      "grad_norm": 3.703125,
+      "learning_rate": 0.00014704754030608937,
+      "loss": 4.2115,
+      "step": 200500
+    },
+    {
+      "epoch": 1.7264979338842976,
+      "grad_norm": 5.9375,
+      "learning_rate": 0.00014702363221085886,
+      "loss": 4.6492,
+      "step": 200550
+    },
+    {
+      "epoch": 1.7269283746556474,
+      "grad_norm": 3.328125,
+      "learning_rate": 0.00014699972066419497,
+      "loss": 4.589,
+      "step": 200600
+    },
+    {
+      "epoch": 1.7273588154269972,
+      "grad_norm": 4.65625,
+      "learning_rate": 0.00014697580566785267,
+      "loss": 4.4187,
+      "step": 200650
+    },
+    {
+      "epoch": 1.727789256198347,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.00014695188722358733,
+      "loss": 4.2957,
+      "step": 200700
+    },
+    {
+      "epoch": 1.728219696969697,
+      "grad_norm": 1.09375,
+      "learning_rate": 0.00014692796533315446,
+      "loss": 4.7696,
+      "step": 200750
+    },
+    {
+      "epoch": 1.7286501377410468,
+      "grad_norm": 3.71875,
+      "learning_rate": 0.0001469040399983099,
+      "loss": 4.7456,
+      "step": 200800
+    },
+    {
+      "epoch": 1.7290805785123968,
+      "grad_norm": 4.59375,
+      "learning_rate": 0.00014688011122080975,
+      "loss": 4.3568,
+      "step": 200850
+    },
+    {
+      "epoch": 1.7295110192837466,
+      "grad_norm": 1.265625,
+      "learning_rate": 0.00014685617900241026,
+      "loss": 4.2912,
+      "step": 200900
+    },
+    {
+      "epoch": 1.7299414600550964,
+      "grad_norm": 3.515625,
+      "learning_rate": 0.00014683224334486808,
+      "loss": 4.4503,
+      "step": 200950
+    },
+    {
+      "epoch": 1.7303719008264462,
+      "grad_norm": 3.46875,
+      "learning_rate": 0.00014680830424994,
+      "loss": 4.3639,
+      "step": 201000
+    },
+    {
+      "epoch": 1.7303719008264462,
+      "eval_loss": 5.069631576538086,
+      "eval_runtime": 21.9831,
+      "eval_samples_per_second": 29.113,
+      "eval_steps_per_second": 14.557,
+      "eval_tts_loss": 7.2692779077950265,
+      "step": 201000
+    },
+    {
+      "epoch": 1.730802341597796,
+      "grad_norm": 1.984375,
+      "learning_rate": 0.00014678436171938308,
+      "loss": 4.4839,
+      "step": 201050
+    },
+    {
+      "epoch": 1.731232782369146,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.0001467604157549547,
+      "loss": 4.4284,
+      "step": 201100
+    },
+    {
+      "epoch": 1.731663223140496,
+      "grad_norm": 1.1484375,
+      "learning_rate": 0.00014673646635841235,
+      "loss": 4.491,
+      "step": 201150
+    },
+    {
+      "epoch": 1.7320936639118458,
+      "grad_norm": 2.578125,
+      "learning_rate": 0.00014671251353151398,
+      "loss": 4.5108,
+      "step": 201200
+    },
+    {
+      "epoch": 1.7325241046831956,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.0001466885572760176,
+      "loss": 4.4842,
+      "step": 201250
+    },
+    {
+      "epoch": 1.7329545454545454,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.0001466645975936816,
+      "loss": 4.6091,
+      "step": 201300
+    },
+    {
+      "epoch": 1.7333849862258952,
+      "grad_norm": 2.921875,
+      "learning_rate": 0.00014664063448626452,
+      "loss": 4.4633,
+      "step": 201350
+    },
+    {
+      "epoch": 1.7338154269972452,
+      "grad_norm": 3.21875,
+      "learning_rate": 0.0001466166679555252,
+      "loss": 4.3776,
+      "step": 201400
+    },
+    {
+      "epoch": 1.734245867768595,
+      "grad_norm": 3.328125,
+      "learning_rate": 0.00014659269800322276,
+      "loss": 4.281,
+      "step": 201450
+    },
+    {
+      "epoch": 1.734676308539945,
+      "grad_norm": 4.21875,
+      "learning_rate": 0.00014656872463111653,
+      "loss": 4.2784,
+      "step": 201500
+    },
+    {
+      "epoch": 1.7351067493112948,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.00014654474784096613,
+      "loss": 4.501,
+      "step": 201550
+    },
+    {
+      "epoch": 1.7355371900826446,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.00014652076763453138,
+      "loss": 4.1394,
+      "step": 201600
+    },
+    {
+      "epoch": 1.7359676308539944,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00014649678401357235,
+      "loss": 4.3451,
+      "step": 201650
+    },
+    {
+      "epoch": 1.7363980716253442,
+      "grad_norm": 3.265625,
+      "learning_rate": 0.00014647279697984944,
+      "loss": 4.2643,
+      "step": 201700
+    },
+    {
+      "epoch": 1.7368285123966942,
+      "grad_norm": 2.375,
+      "learning_rate": 0.0001464488065351232,
+      "loss": 4.2888,
+      "step": 201750
+    },
+    {
+      "epoch": 1.7372589531680442,
+      "grad_norm": 5.4375,
+      "learning_rate": 0.0001464248126811545,
+      "loss": 4.6217,
+      "step": 201800
+    },
+    {
+      "epoch": 1.737689393939394,
+      "grad_norm": 3.203125,
+      "learning_rate": 0.00014640081541970442,
+      "loss": 4.5291,
+      "step": 201850
+    },
+    {
+      "epoch": 1.7381198347107438,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.00014637681475253438,
+      "loss": 4.3518,
+      "step": 201900
+    },
+    {
+      "epoch": 1.7385502754820936,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.00014635281068140586,
+      "loss": 4.1754,
+      "step": 201950
+    },
+    {
+      "epoch": 1.7389807162534434,
+      "grad_norm": 3.875,
+      "learning_rate": 0.0001463288032080808,
+      "loss": 4.7852,
+      "step": 202000
+    },
+    {
+      "epoch": 1.7394111570247934,
+      "grad_norm": 1.6953125,
+      "learning_rate": 0.00014630479233432127,
+      "loss": 4.4412,
+      "step": 202050
+    },
+    {
+      "epoch": 1.7398415977961432,
+      "grad_norm": 1.34375,
+      "learning_rate": 0.00014628077806188956,
+      "loss": 4.3081,
+      "step": 202100
+    },
+    {
+      "epoch": 1.7402720385674932,
+      "grad_norm": 3.359375,
+      "learning_rate": 0.00014625676039254834,
+      "loss": 4.1681,
+      "step": 202150
+    },
+    {
+      "epoch": 1.740702479338843,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.00014623273932806046,
+      "loss": 4.19,
+      "step": 202200
+    },
+    {
+      "epoch": 1.7411329201101928,
+      "grad_norm": 3.3125,
+      "learning_rate": 0.00014620871487018897,
+      "loss": 4.0179,
+      "step": 202250
+    },
+    {
+      "epoch": 1.7415633608815426,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00014618468702069722,
+      "loss": 4.6226,
+      "step": 202300
+    },
+    {
+      "epoch": 1.7419938016528924,
+      "grad_norm": 1.203125,
+      "learning_rate": 0.0001461606557813488,
+      "loss": 4.565,
+      "step": 202350
+    },
+    {
+      "epoch": 1.7424242424242424,
+      "grad_norm": 3.125,
+      "learning_rate": 0.0001461366211539076,
+      "loss": 4.8175,
+      "step": 202400
+    },
+    {
+      "epoch": 1.7428546831955924,
+      "grad_norm": 3.234375,
+      "learning_rate": 0.00014611258314013763,
+      "loss": 4.6324,
+      "step": 202450
+    },
+    {
+      "epoch": 1.7432851239669422,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.0001460885417418033,
+      "loss": 4.2822,
+      "step": 202500
+    },
+    {
+      "epoch": 1.743715564738292,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.00014606449696066918,
+      "loss": 4.1335,
+      "step": 202550
+    },
+    {
+      "epoch": 1.7441460055096418,
+      "grad_norm": 1.7890625,
+      "learning_rate": 0.00014604044879850005,
+      "loss": 4.3431,
+      "step": 202600
+    },
+    {
+      "epoch": 1.7445764462809916,
+      "grad_norm": 1.734375,
+      "learning_rate": 0.00014601639725706107,
+      "loss": 4.2001,
+      "step": 202650
+    },
+    {
+      "epoch": 1.7450068870523416,
+      "grad_norm": 1.984375,
+      "learning_rate": 0.00014599234233811752,
+      "loss": 4.0488,
+      "step": 202700
+    },
+    {
+      "epoch": 1.7454373278236914,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00014596828404343502,
+      "loss": 4.3412,
+      "step": 202750
+    },
+    {
+      "epoch": 1.7458677685950414,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.00014594422237477935,
+      "loss": 4.5877,
+      "step": 202800
+    },
+    {
+      "epoch": 1.7462982093663912,
+      "grad_norm": 1.171875,
+      "learning_rate": 0.0001459201573339166,
+      "loss": 4.2314,
+      "step": 202850
+    },
+    {
+      "epoch": 1.746728650137741,
+      "grad_norm": 1.6171875,
+      "learning_rate": 0.00014589608892261316,
+      "loss": 4.6578,
+      "step": 202900
+    },
+    {
+      "epoch": 1.7471590909090908,
+      "grad_norm": 3.171875,
+      "learning_rate": 0.0001458720171426355,
+      "loss": 4.5862,
+      "step": 202950
+    },
+    {
+      "epoch": 1.7475895316804406,
+      "grad_norm": 4.8125,
+      "learning_rate": 0.00014584794199575045,
+      "loss": 4.347,
+      "step": 203000
+    },
+    {
+      "epoch": 1.7480199724517906,
+      "grad_norm": 6.15625,
+      "learning_rate": 0.00014582386348372514,
+      "loss": 4.2003,
+      "step": 203050
+    },
+    {
+      "epoch": 1.7484504132231407,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00014579978160832684,
+      "loss": 4.5226,
+      "step": 203100
+    },
+    {
+      "epoch": 1.7488808539944904,
+      "grad_norm": 1.703125,
+      "learning_rate": 0.0001457756963713231,
+      "loss": 4.3868,
+      "step": 203150
+    },
+    {
+      "epoch": 1.7493112947658402,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00014575160777448174,
+      "loss": 4.0611,
+      "step": 203200
+    },
+    {
+      "epoch": 1.74974173553719,
+      "grad_norm": 2.9375,
+      "learning_rate": 0.00014572751581957082,
+      "loss": 4.3162,
+      "step": 203250
+    },
+    {
+      "epoch": 1.7501721763085398,
+      "grad_norm": 3.125,
+      "learning_rate": 0.00014570342050835861,
+      "loss": 4.0677,
+      "step": 203300
+    },
+    {
+      "epoch": 1.7506026170798898,
+      "grad_norm": 3.046875,
+      "learning_rate": 0.00014567932184261366,
+      "loss": 4.6464,
+      "step": 203350
+    },
+    {
+      "epoch": 1.7510330578512396,
+      "grad_norm": 3.609375,
+      "learning_rate": 0.0001456552198241048,
+      "loss": 4.2298,
+      "step": 203400
+    },
+    {
+      "epoch": 1.7514634986225897,
+      "grad_norm": 3.109375,
+      "learning_rate": 0.000145631114454601,
+      "loss": 4.4299,
+      "step": 203450
+    },
+    {
+      "epoch": 1.7518939393939394,
+      "grad_norm": 3.609375,
+      "learning_rate": 0.0001456070057358716,
+      "loss": 4.6576,
+      "step": 203500
+    },
+    {
+      "epoch": 1.7523243801652892,
+      "grad_norm": 1.984375,
+      "learning_rate": 0.00014558289366968608,
+      "loss": 4.7713,
+      "step": 203550
+    },
+    {
+      "epoch": 1.752754820936639,
+      "grad_norm": 3.515625,
+      "learning_rate": 0.00014555877825781426,
+      "loss": 4.8008,
+      "step": 203600
+    },
+    {
+      "epoch": 1.7531852617079888,
+      "grad_norm": 1.828125,
+      "learning_rate": 0.00014553465950202613,
+      "loss": 4.3375,
+      "step": 203650
+    },
+    {
+      "epoch": 1.7536157024793388,
+      "grad_norm": 4.5625,
+      "learning_rate": 0.00014551053740409197,
+      "loss": 4.4108,
+      "step": 203700
+    },
+    {
+      "epoch": 1.7540461432506889,
+      "grad_norm": 2.421875,
+      "learning_rate": 0.00014548641196578227,
+      "loss": 4.8124,
+      "step": 203750
+    },
+    {
+      "epoch": 1.7544765840220387,
+      "grad_norm": 3.234375,
+      "learning_rate": 0.00014546228318886782,
+      "loss": 4.6107,
+      "step": 203800
+    },
+    {
+      "epoch": 1.7549070247933884,
+      "grad_norm": 3.28125,
+      "learning_rate": 0.00014543815107511958,
+      "loss": 4.3705,
+      "step": 203850
+    },
+    {
+      "epoch": 1.7553374655647382,
+      "grad_norm": 4.34375,
+      "learning_rate": 0.0001454140156263088,
+      "loss": 4.3269,
+      "step": 203900
+    },
+    {
+      "epoch": 1.755767906336088,
+      "grad_norm": 1.9453125,
+      "learning_rate": 0.00014538987684420702,
+      "loss": 4.5044,
+      "step": 203950
+    },
+    {
+      "epoch": 1.756198347107438,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.00014536573473058592,
+      "loss": 4.1559,
+      "step": 204000
+    },
+    {
+      "epoch": 1.756198347107438,
+      "eval_loss": 5.066303253173828,
+      "eval_runtime": 21.8339,
+      "eval_samples_per_second": 29.312,
+      "eval_steps_per_second": 14.656,
+      "eval_tts_loss": 7.257139390627936,
+      "step": 204000
+    },
+    {
+      "epoch": 1.7566287878787878,
+      "grad_norm": 1.9453125,
+      "learning_rate": 0.0001453415892872175,
+      "loss": 3.9364,
+      "step": 204050
+    },
+    {
+      "epoch": 1.7570592286501379,
+      "grad_norm": 3.078125,
+      "learning_rate": 0.000145317440515874,
+      "loss": 4.7497,
+      "step": 204100
+    },
+    {
+      "epoch": 1.7574896694214877,
+      "grad_norm": 3.421875,
+      "learning_rate": 0.00014529328841832783,
+      "loss": 3.8812,
+      "step": 204150
+    },
+    {
+      "epoch": 1.7579201101928374,
+      "grad_norm": 3.4375,
+      "learning_rate": 0.00014526913299635176,
+      "loss": 4.3653,
+      "step": 204200
+    },
+    {
+      "epoch": 1.7583505509641872,
+      "grad_norm": 1.7421875,
+      "learning_rate": 0.00014524497425171874,
+      "loss": 4.3812,
+      "step": 204250
+    },
+    {
+      "epoch": 1.758780991735537,
+      "grad_norm": 5.8125,
+      "learning_rate": 0.00014522081218620194,
+      "loss": 4.8898,
+      "step": 204300
+    },
+    {
+      "epoch": 1.759211432506887,
+      "grad_norm": 2.421875,
+      "learning_rate": 0.0001451966468015748,
+      "loss": 4.2977,
+      "step": 204350
+    },
+    {
+      "epoch": 1.759641873278237,
+      "grad_norm": 3.546875,
+      "learning_rate": 0.00014517247809961106,
+      "loss": 4.728,
+      "step": 204400
+    },
+    {
+      "epoch": 1.7600723140495869,
+      "grad_norm": 3.109375,
+      "learning_rate": 0.00014514830608208462,
+      "loss": 4.5724,
+      "step": 204450
+    },
+    {
+      "epoch": 1.7605027548209367,
+      "grad_norm": 4.03125,
+      "learning_rate": 0.00014512413075076962,
+      "loss": 4.3678,
+      "step": 204500
+    },
+    {
+      "epoch": 1.7609331955922864,
+      "grad_norm": 3.3125,
+      "learning_rate": 0.00014509995210744056,
+      "loss": 4.2542,
+      "step": 204550
+    },
+    {
+      "epoch": 1.7613636363636362,
+      "grad_norm": 5.6875,
+      "learning_rate": 0.00014507577015387203,
+      "loss": 4.3262,
+      "step": 204600
+    },
+    {
+      "epoch": 1.7617940771349863,
+      "grad_norm": 3.125,
+      "learning_rate": 0.00014505158489183895,
+      "loss": 4.7296,
+      "step": 204650
+    },
+    {
+      "epoch": 1.762224517906336,
+      "grad_norm": 3.375,
+      "learning_rate": 0.00014502739632311646,
+      "loss": 4.7251,
+      "step": 204700
+    },
+    {
+      "epoch": 1.762654958677686,
+      "grad_norm": 4.8125,
+      "learning_rate": 0.00014500320444947994,
+      "loss": 4.4522,
+      "step": 204750
+    },
+    {
+      "epoch": 1.7630853994490359,
+      "grad_norm": 2.140625,
+      "learning_rate": 0.0001449790092727051,
+      "loss": 4.8276,
+      "step": 204800
+    },
+    {
+      "epoch": 1.7635158402203857,
+      "grad_norm": 1.4921875,
+      "learning_rate": 0.00014495481079456772,
+      "loss": 4.3861,
+      "step": 204850
+    },
+    {
+      "epoch": 1.7639462809917354,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00014493060901684394,
+      "loss": 4.711,
+      "step": 204900
+    },
+    {
+      "epoch": 1.7643767217630852,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.00014490640394131015,
+      "loss": 4.2437,
+      "step": 204950
+    },
+    {
+      "epoch": 1.7648071625344353,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00014488219556974292,
+      "loss": 4.2778,
+      "step": 205000
+    },
+    {
+      "epoch": 1.7652376033057853,
+      "grad_norm": 2.921875,
+      "learning_rate": 0.00014485798390391906,
+      "loss": 4.7075,
+      "step": 205050
+    },
+    {
+      "epoch": 1.765668044077135,
+      "grad_norm": 3.828125,
+      "learning_rate": 0.00014483376894561576,
+      "loss": 4.5107,
+      "step": 205100
+    },
+    {
+      "epoch": 1.7660984848484849,
+      "grad_norm": 5.1875,
+      "learning_rate": 0.00014480955069661028,
+      "loss": 4.893,
+      "step": 205150
+    },
+    {
+      "epoch": 1.7665289256198347,
+      "grad_norm": 3.1875,
+      "learning_rate": 0.00014478532915868016,
+      "loss": 4.4953,
+      "step": 205200
+    },
+    {
+      "epoch": 1.7669593663911844,
+      "grad_norm": 1.9296875,
+      "learning_rate": 0.00014476110433360323,
+      "loss": 4.2584,
+      "step": 205250
+    },
+    {
+      "epoch": 1.7673898071625345,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.00014473687622315758,
+      "loss": 4.6173,
+      "step": 205300
+    },
+    {
+      "epoch": 1.7678202479338843,
+      "grad_norm": 2.921875,
+      "learning_rate": 0.00014471264482912146,
+      "loss": 4.5259,
+      "step": 205350
+    },
+    {
+      "epoch": 1.7682506887052343,
+      "grad_norm": 4.1875,
+      "learning_rate": 0.0001446884101532734,
+      "loss": 4.5646,
+      "step": 205400
+    },
+    {
+      "epoch": 1.768681129476584,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.0001446641721973922,
+      "loss": 4.5003,
+      "step": 205450
+    },
+    {
+      "epoch": 1.7691115702479339,
+      "grad_norm": 1.3203125,
+      "learning_rate": 0.00014463993096325682,
+      "loss": 4.6141,
+      "step": 205500
+    },
+    {
+      "epoch": 1.7695420110192837,
+      "grad_norm": 2.9375,
+      "learning_rate": 0.00014461568645264658,
+      "loss": 4.5447,
+      "step": 205550
+    },
+    {
+      "epoch": 1.7699724517906334,
+      "grad_norm": 2.5,
+      "learning_rate": 0.00014459143866734096,
+      "loss": 4.262,
+      "step": 205600
+    },
+    {
+      "epoch": 1.7704028925619835,
+      "grad_norm": 3.28125,
+      "learning_rate": 0.00014456718760911966,
+      "loss": 4.6289,
+      "step": 205650
+    },
+    {
+      "epoch": 1.7708333333333335,
+      "grad_norm": 1.6328125,
+      "learning_rate": 0.0001445429332797627,
+      "loss": 4.2575,
+      "step": 205700
+    },
+    {
+      "epoch": 1.7712637741046833,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.00014451867568105026,
+      "loss": 4.4797,
+      "step": 205750
+    },
+    {
+      "epoch": 1.771694214876033,
+      "grad_norm": 3.734375,
+      "learning_rate": 0.0001444944148147628,
+      "loss": 4.5051,
+      "step": 205800
+    },
+    {
+      "epoch": 1.7721246556473829,
+      "grad_norm": 2.984375,
+      "learning_rate": 0.00014447015068268107,
+      "loss": 4.4409,
+      "step": 205850
+    },
+    {
+      "epoch": 1.7725550964187327,
+      "grad_norm": 3.9375,
+      "learning_rate": 0.0001444458832865859,
+      "loss": 4.6527,
+      "step": 205900
+    },
+    {
+      "epoch": 1.7729855371900827,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.00014442161262825853,
+      "loss": 4.8302,
+      "step": 205950
+    },
+    {
+      "epoch": 1.7734159779614325,
+      "grad_norm": 3.1875,
+      "learning_rate": 0.00014439733870948043,
+      "loss": 4.2343,
+      "step": 206000
+    },
+    {
+      "epoch": 1.7738464187327825,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.00014437306153203313,
+      "loss": 4.4547,
+      "step": 206050
+    },
+    {
+      "epoch": 1.7742768595041323,
+      "grad_norm": 2.75,
+      "learning_rate": 0.0001443487810976986,
+      "loss": 4.5386,
+      "step": 206100
+    },
+    {
+      "epoch": 1.774707300275482,
+      "grad_norm": 4.71875,
+      "learning_rate": 0.000144324497408259,
+      "loss": 4.71,
+      "step": 206150
+    },
+    {
+      "epoch": 1.7751377410468319,
+      "grad_norm": 4.34375,
+      "learning_rate": 0.00014430021046549658,
+      "loss": 4.3663,
+      "step": 206200
+    },
+    {
+      "epoch": 1.7755681818181817,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00014427592027119407,
+      "loss": 4.1685,
+      "step": 206250
+    },
+    {
+      "epoch": 1.7759986225895317,
+      "grad_norm": 3.203125,
+      "learning_rate": 0.00014425162682713428,
+      "loss": 4.1937,
+      "step": 206300
+    },
+    {
+      "epoch": 1.7764290633608817,
+      "grad_norm": 3.734375,
+      "learning_rate": 0.0001442273301351003,
+      "loss": 4.6254,
+      "step": 206350
+    },
+    {
+      "epoch": 1.7768595041322315,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.0001442030301968754,
+      "loss": 4.7547,
+      "step": 206400
+    },
+    {
+      "epoch": 1.7772899449035813,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.0001441787270142432,
+      "loss": 4.548,
+      "step": 206450
+    },
+    {
+      "epoch": 1.777720385674931,
+      "grad_norm": 5.03125,
+      "learning_rate": 0.00014415442058898754,
+      "loss": 4.6682,
+      "step": 206500
+    },
+    {
+      "epoch": 1.7781508264462809,
+      "grad_norm": 5.09375,
+      "learning_rate": 0.00014413011092289234,
+      "loss": 4.4968,
+      "step": 206550
+    },
+    {
+      "epoch": 1.7785812672176309,
+      "grad_norm": 1.9609375,
+      "learning_rate": 0.00014410579801774198,
+      "loss": 4.5569,
+      "step": 206600
+    },
+    {
+      "epoch": 1.7790117079889807,
+      "grad_norm": 3.109375,
+      "learning_rate": 0.00014408148187532096,
+      "loss": 4.5855,
+      "step": 206650
+    },
+    {
+      "epoch": 1.7794421487603307,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.000144057162497414,
+      "loss": 4.4967,
+      "step": 206700
+    },
+    {
+      "epoch": 1.7798725895316805,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.0001440328398858061,
+      "loss": 4.2051,
+      "step": 206750
+    },
+    {
+      "epoch": 1.7803030303030303,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.00014400851404228248,
+      "loss": 4.5489,
+      "step": 206800
+    },
+    {
+      "epoch": 1.78073347107438,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.00014398418496862863,
+      "loss": 4.4709,
+      "step": 206850
+    },
+    {
+      "epoch": 1.7811639118457299,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00014395985266663024,
+      "loss": 4.6817,
+      "step": 206900
+    },
+    {
+      "epoch": 1.7815943526170799,
+      "grad_norm": 1.5234375,
+      "learning_rate": 0.00014393551713807323,
+      "loss": 3.7956,
+      "step": 206950
+    },
+    {
+      "epoch": 1.78202479338843,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.0001439111783847438,
+      "loss": 4.3871,
+      "step": 207000
+    },
+    {
+      "epoch": 1.78202479338843,
+      "eval_loss": 5.065958499908447,
+      "eval_runtime": 21.7359,
+      "eval_samples_per_second": 29.444,
+      "eval_steps_per_second": 14.722,
+      "eval_tts_loss": 7.261567631698336,
+      "step": 207000
+    },
+    {
+      "epoch": 1.7824552341597797,
+      "grad_norm": 4.34375,
+      "learning_rate": 0.00014388683640842838,
+      "loss": 4.305,
+      "step": 207050
+    },
+    {
+      "epoch": 1.7828856749311295,
+      "grad_norm": 4.625,
+      "learning_rate": 0.00014386249121091355,
+      "loss": 4.181,
+      "step": 207100
+    },
+    {
+      "epoch": 1.7833161157024793,
+      "grad_norm": 3.078125,
+      "learning_rate": 0.00014383814279398627,
+      "loss": 4.5487,
+      "step": 207150
+    },
+    {
+      "epoch": 1.783746556473829,
+      "grad_norm": 3.3125,
+      "learning_rate": 0.00014381379115943363,
+      "loss": 4.2781,
+      "step": 207200
+    },
+    {
+      "epoch": 1.784176997245179,
+      "grad_norm": 4.78125,
+      "learning_rate": 0.00014378943630904297,
+      "loss": 4.693,
+      "step": 207250
+    },
+    {
+      "epoch": 1.7846074380165289,
+      "grad_norm": 2.296875,
+      "learning_rate": 0.0001437650782446019,
+      "loss": 4.6197,
+      "step": 207300
+    },
+    {
+      "epoch": 1.785037878787879,
+      "grad_norm": 3.203125,
+      "learning_rate": 0.00014374071696789825,
+      "loss": 4.8426,
+      "step": 207350
+    },
+    {
+      "epoch": 1.7854683195592287,
+      "grad_norm": 1.5546875,
+      "learning_rate": 0.0001437163524807201,
+      "loss": 4.4313,
+      "step": 207400
+    },
+    {
+      "epoch": 1.7858987603305785,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00014369198478485573,
+      "loss": 4.4223,
+      "step": 207450
+    },
+    {
+      "epoch": 1.7863292011019283,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.00014366761388209367,
+      "loss": 4.5474,
+      "step": 207500
+    },
+    {
+      "epoch": 1.786759641873278,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.00014364323977422271,
+      "loss": 4.8404,
+      "step": 207550
+    },
+    {
+      "epoch": 1.787190082644628,
+      "grad_norm": 3.90625,
+      "learning_rate": 0.00014361886246303185,
+      "loss": 4.3914,
+      "step": 207600
+    },
+    {
+      "epoch": 1.787620523415978,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.00014359448195031034,
+      "loss": 4.5706,
+      "step": 207650
+    },
+    {
+      "epoch": 1.788050964187328,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.00014357009823784766,
+      "loss": 4.4997,
+      "step": 207700
+    },
+    {
+      "epoch": 1.7884814049586777,
+      "grad_norm": 3.296875,
+      "learning_rate": 0.00014354571132743347,
+      "loss": 4.336,
+      "step": 207750
+    },
+    {
+      "epoch": 1.7889118457300275,
+      "grad_norm": 3.5625,
+      "learning_rate": 0.0001435213212208578,
+      "loss": 4.7328,
+      "step": 207800
+    },
+    {
+      "epoch": 1.7893422865013773,
+      "grad_norm": 1.5546875,
+      "learning_rate": 0.00014349692791991081,
+      "loss": 4.3584,
+      "step": 207850
+    },
+    {
+      "epoch": 1.7897727272727273,
+      "grad_norm": 3.703125,
+      "learning_rate": 0.00014347253142638286,
+      "loss": 4.5277,
+      "step": 207900
+    },
+    {
+      "epoch": 1.790203168044077,
+      "grad_norm": 3.125,
+      "learning_rate": 0.00014344813174206464,
+      "loss": 4.695,
+      "step": 207950
+    },
+    {
+      "epoch": 1.790633608815427,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.000143423728868747,
+      "loss": 4.5818,
+      "step": 208000
+    },
+    {
+      "epoch": 1.791064049586777,
+      "grad_norm": 1.9375,
+      "learning_rate": 0.00014339932280822118,
+      "loss": 4.3562,
+      "step": 208050
+    },
+    {
+      "epoch": 1.7914944903581267,
+      "grad_norm": 2.875,
+      "learning_rate": 0.00014337491356227836,
+      "loss": 4.7021,
+      "step": 208100
+    },
+    {
+      "epoch": 1.7919249311294765,
+      "grad_norm": 3.703125,
+      "learning_rate": 0.00014335050113271024,
+      "loss": 3.7902,
+      "step": 208150
+    },
+    {
+      "epoch": 1.7923553719008265,
+      "grad_norm": 3.140625,
+      "learning_rate": 0.00014332608552130858,
+      "loss": 4.5522,
+      "step": 208200
+    },
+    {
+      "epoch": 1.7927858126721763,
+      "grad_norm": 3.109375,
+      "learning_rate": 0.0001433016667298655,
+      "loss": 4.377,
+      "step": 208250
+    },
+    {
+      "epoch": 1.7932162534435263,
+      "grad_norm": 3.25,
+      "learning_rate": 0.00014327724476017326,
+      "loss": 4.3705,
+      "step": 208300
+    },
+    {
+      "epoch": 1.793646694214876,
+      "grad_norm": 1.828125,
+      "learning_rate": 0.0001432528196140243,
+      "loss": 4.4624,
+      "step": 208350
+    },
+    {
+      "epoch": 1.794077134986226,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.0001432283912932115,
+      "loss": 4.3584,
+      "step": 208400
+    },
+    {
+      "epoch": 1.7945075757575757,
+      "grad_norm": 1.203125,
+      "learning_rate": 0.00014320395979952778,
+      "loss": 4.2264,
+      "step": 208450
+    },
+    {
+      "epoch": 1.7949380165289255,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.00014317952513476633,
+      "loss": 4.3237,
+      "step": 208500
+    },
+    {
+      "epoch": 1.7953684573002755,
+      "grad_norm": 2.984375,
+      "learning_rate": 0.00014315508730072067,
+      "loss": 4.3732,
+      "step": 208550
+    },
+    {
+      "epoch": 1.7957988980716253,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.00014313064629918447,
+      "loss": 4.4379,
+      "step": 208600
+    },
+    {
+      "epoch": 1.7962293388429753,
+      "grad_norm": 4.65625,
+      "learning_rate": 0.00014310620213195158,
+      "loss": 4.611,
+      "step": 208650
+    },
+    {
+      "epoch": 1.796659779614325,
+      "grad_norm": 3.3125,
+      "learning_rate": 0.00014308175480081623,
+      "loss": 3.929,
+      "step": 208700
+    },
+    {
+      "epoch": 1.797090220385675,
+      "grad_norm": 3.625,
+      "learning_rate": 0.00014305730430757278,
+      "loss": 4.4699,
+      "step": 208750
+    },
+    {
+      "epoch": 1.7975206611570247,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.0001430328506540158,
+      "loss": 4.4323,
+      "step": 208800
+    },
+    {
+      "epoch": 1.7979511019283747,
+      "grad_norm": 3.421875,
+      "learning_rate": 0.0001430083938419402,
+      "loss": 4.6324,
+      "step": 208850
+    },
+    {
+      "epoch": 1.7983815426997245,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00014298393387314102,
+      "loss": 4.3147,
+      "step": 208900
+    },
+    {
+      "epoch": 1.7988119834710745,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00014295947074941358,
+      "loss": 3.8383,
+      "step": 208950
+    },
+    {
+      "epoch": 1.7992424242424243,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.0001429350044725534,
+      "loss": 4.5066,
+      "step": 209000
+    },
+    {
+      "epoch": 1.799672865013774,
+      "grad_norm": 4.59375,
+      "learning_rate": 0.00014291053504435629,
+      "loss": 4.4437,
+      "step": 209050
+    },
+    {
+      "epoch": 1.800103305785124,
+      "grad_norm": 3.171875,
+      "learning_rate": 0.00014288606246661822,
+      "loss": 4.6211,
+      "step": 209100
+    },
+    {
+      "epoch": 1.8005337465564737,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.0001428615867411354,
+      "loss": 4.4462,
+      "step": 209150
+    },
+    {
+      "epoch": 1.8009641873278237,
+      "grad_norm": 4.03125,
+      "learning_rate": 0.0001428371078697044,
+      "loss": 4.5513,
+      "step": 209200
+    },
+    {
+      "epoch": 1.8013946280991735,
+      "grad_norm": 3.234375,
+      "learning_rate": 0.0001428126258541218,
+      "loss": 4.5234,
+      "step": 209250
+    },
+    {
+      "epoch": 1.8018250688705235,
+      "grad_norm": 1.7578125,
+      "learning_rate": 0.00014278814069618463,
+      "loss": 4.546,
+      "step": 209300
+    },
+    {
+      "epoch": 1.8022555096418733,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00014276365239768995,
+      "loss": 4.3456,
+      "step": 209350
+    },
+    {
+      "epoch": 1.802685950413223,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00014273916096043518,
+      "loss": 4.4293,
+      "step": 209400
+    },
+    {
+      "epoch": 1.803116391184573,
+      "grad_norm": 3.53125,
+      "learning_rate": 0.000142714666386218,
+      "loss": 4.3945,
+      "step": 209450
+    },
+    {
+      "epoch": 1.803546831955923,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.00014269016867683615,
+      "loss": 4.497,
+      "step": 209500
+    },
+    {
+      "epoch": 1.8039772727272727,
+      "grad_norm": 2.578125,
+      "learning_rate": 0.0001426656678340878,
+      "loss": 4.576,
+      "step": 209550
+    },
+    {
+      "epoch": 1.8044077134986227,
+      "grad_norm": 3.390625,
+      "learning_rate": 0.00014264116385977123,
+      "loss": 4.0808,
+      "step": 209600
+    },
+    {
+      "epoch": 1.8048381542699725,
+      "grad_norm": 1.3671875,
+      "learning_rate": 0.000142616656755685,
+      "loss": 4.2861,
+      "step": 209650
+    },
+    {
+      "epoch": 1.8052685950413223,
+      "grad_norm": 3.3125,
+      "learning_rate": 0.00014259214652362782,
+      "loss": 4.4298,
+      "step": 209700
+    },
+    {
+      "epoch": 1.805699035812672,
+      "grad_norm": 3.9375,
+      "learning_rate": 0.00014256763316539874,
+      "loss": 4.3557,
+      "step": 209750
+    },
+    {
+      "epoch": 1.806129476584022,
+      "grad_norm": 3.515625,
+      "learning_rate": 0.00014254311668279696,
+      "loss": 4.4118,
+      "step": 209800
+    },
+    {
+      "epoch": 1.806559917355372,
+      "grad_norm": 3.0,
+      "learning_rate": 0.00014251859707762197,
+      "loss": 4.1832,
+      "step": 209850
+    },
+    {
+      "epoch": 1.806990358126722,
+      "grad_norm": 3.828125,
+      "learning_rate": 0.0001424940743516734,
+      "loss": 4.4702,
+      "step": 209900
+    },
+    {
+      "epoch": 1.8074207988980717,
+      "grad_norm": 3.203125,
+      "learning_rate": 0.00014246954850675127,
+      "loss": 4.1901,
+      "step": 209950
+    },
+    {
+      "epoch": 1.8078512396694215,
+      "grad_norm": 3.34375,
+      "learning_rate": 0.00014244501954465558,
+      "loss": 4.5261,
+      "step": 210000
+    },
+    {
+      "epoch": 1.8078512396694215,
+      "eval_loss": 5.058772087097168,
+      "eval_runtime": 21.8275,
+      "eval_samples_per_second": 29.321,
+      "eval_steps_per_second": 14.66,
+      "eval_tts_loss": 7.2322371725440675,
+      "step": 210000
+    },
+    {
+      "epoch": 1.8082816804407713,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.0001424204874671868,
+      "loss": 4.5437,
+      "step": 210050
+    },
+    {
+      "epoch": 1.808712121212121,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00014239595227614555,
+      "loss": 4.5104,
+      "step": 210100
+    },
+    {
+      "epoch": 1.8091425619834711,
+      "grad_norm": 4.375,
+      "learning_rate": 0.00014237141397333258,
+      "loss": 4.326,
+      "step": 210150
+    },
+    {
+      "epoch": 1.809573002754821,
+      "grad_norm": 4.84375,
+      "learning_rate": 0.000142346872560549,
+      "loss": 4.8106,
+      "step": 210200
+    },
+    {
+      "epoch": 1.810003443526171,
+      "grad_norm": 1.3984375,
+      "learning_rate": 0.00014232232803959607,
+      "loss": 3.9326,
+      "step": 210250
+    },
+    {
+      "epoch": 1.8104338842975207,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00014229778041227533,
+      "loss": 4.3245,
+      "step": 210300
+    },
+    {
+      "epoch": 1.8108643250688705,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00014227322968038853,
+      "loss": 4.7487,
+      "step": 210350
+    },
+    {
+      "epoch": 1.8112947658402203,
+      "grad_norm": 2.9375,
+      "learning_rate": 0.0001422486758457376,
+      "loss": 4.6411,
+      "step": 210400
+    },
+    {
+      "epoch": 1.81172520661157,
+      "grad_norm": 6.21875,
+      "learning_rate": 0.00014222411891012476,
+      "loss": 4.3358,
+      "step": 210450
+    },
+    {
+      "epoch": 1.8121556473829201,
+      "grad_norm": 2.578125,
+      "learning_rate": 0.0001421995588753525,
+      "loss": 4.1035,
+      "step": 210500
+    },
+    {
+      "epoch": 1.8125860881542701,
+      "grad_norm": 3.40625,
+      "learning_rate": 0.00014217499574322335,
+      "loss": 4.4316,
+      "step": 210550
+    },
+    {
+      "epoch": 1.81301652892562,
+      "grad_norm": 3.015625,
+      "learning_rate": 0.00014215042951554028,
+      "loss": 4.356,
+      "step": 210600
+    },
+    {
+      "epoch": 1.8134469696969697,
+      "grad_norm": 3.578125,
+      "learning_rate": 0.00014212586019410637,
+      "loss": 4.3285,
+      "step": 210650
+    },
+    {
+      "epoch": 1.8138774104683195,
+      "grad_norm": 4.28125,
+      "learning_rate": 0.00014210128778072492,
+      "loss": 4.4144,
+      "step": 210700
+    },
+    {
+      "epoch": 1.8143078512396693,
+      "grad_norm": 4.21875,
+      "learning_rate": 0.00014207671227719958,
+      "loss": 4.4266,
+      "step": 210750
+    },
+    {
+      "epoch": 1.8147382920110193,
+      "grad_norm": 3.46875,
+      "learning_rate": 0.00014205213368533408,
+      "loss": 4.7974,
+      "step": 210800
+    },
+    {
+      "epoch": 1.8151687327823691,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.00014202755200693246,
+      "loss": 4.4121,
+      "step": 210850
+    },
+    {
+      "epoch": 1.8155991735537191,
+      "grad_norm": 3.0625,
+      "learning_rate": 0.0001420029672437989,
+      "loss": 4.5135,
+      "step": 210900
+    },
+    {
+      "epoch": 1.816029614325069,
+      "grad_norm": 3.546875,
+      "learning_rate": 0.00014197837939773796,
+      "loss": 4.49,
+      "step": 210950
+    },
+    {
+      "epoch": 1.8164600550964187,
+      "grad_norm": 1.7890625,
+      "learning_rate": 0.00014195378847055428,
+      "loss": 4.3935,
+      "step": 211000
+    },
+    {
+      "epoch": 1.8168904958677685,
+      "grad_norm": 2.9375,
+      "learning_rate": 0.00014192919446405275,
+      "loss": 4.6175,
+      "step": 211050
+    },
+    {
+      "epoch": 1.8173209366391183,
+      "grad_norm": 3.203125,
+      "learning_rate": 0.0001419045973800386,
+      "loss": 4.268,
+      "step": 211100
+    },
+    {
+      "epoch": 1.8177513774104683,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00014187999722031717,
+      "loss": 4.5076,
+      "step": 211150
+    },
+    {
+      "epoch": 1.8181818181818183,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.00014185539398669402,
+      "loss": 4.4933,
+      "step": 211200
+    },
+    {
+      "epoch": 1.8186122589531681,
+      "grad_norm": 3.125,
+      "learning_rate": 0.00014183078768097504,
+      "loss": 4.7157,
+      "step": 211250
+    },
+    {
+      "epoch": 1.819042699724518,
+      "grad_norm": 3.296875,
+      "learning_rate": 0.00014180617830496624,
+      "loss": 4.4672,
+      "step": 211300
+    },
+    {
+      "epoch": 1.8194731404958677,
+      "grad_norm": 1.21875,
+      "learning_rate": 0.00014178156586047385,
+      "loss": 4.1881,
+      "step": 211350
+    },
+    {
+      "epoch": 1.8199035812672175,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.0001417569503493045,
+      "loss": 4.6438,
+      "step": 211400
+    },
+    {
+      "epoch": 1.8203340220385675,
+      "grad_norm": 2.578125,
+      "learning_rate": 0.0001417323317732648,
+      "loss": 4.2103,
+      "step": 211450
+    },
+    {
+      "epoch": 1.8207644628099173,
+      "grad_norm": 3.171875,
+      "learning_rate": 0.00014170771013416174,
+      "loss": 4.0585,
+      "step": 211500
+    },
+    {
+      "epoch": 1.8211949035812673,
+      "grad_norm": 3.25,
+      "learning_rate": 0.00014168308543380245,
+      "loss": 4.6537,
+      "step": 211550
+    },
+    {
+      "epoch": 1.8216253443526171,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.00014165845767399442,
+      "loss": 4.6123,
+      "step": 211600
+    },
+    {
+      "epoch": 1.822055785123967,
+      "grad_norm": 3.046875,
+      "learning_rate": 0.00014163382685654523,
+      "loss": 4.3003,
+      "step": 211650
+    },
+    {
+      "epoch": 1.8224862258953167,
+      "grad_norm": 3.65625,
+      "learning_rate": 0.0001416091929832627,
+      "loss": 4.2193,
+      "step": 211700
+    },
+    {
+      "epoch": 1.8229166666666665,
+      "grad_norm": 3.109375,
+      "learning_rate": 0.00014158455605595496,
+      "loss": 4.4595,
+      "step": 211750
+    },
+    {
+      "epoch": 1.8233471074380165,
+      "grad_norm": 1.5859375,
+      "learning_rate": 0.0001415599160764302,
+      "loss": 4.1979,
+      "step": 211800
+    },
+    {
+      "epoch": 1.8237775482093666,
+      "grad_norm": 3.59375,
+      "learning_rate": 0.00014153527304649712,
+      "loss": 4.7068,
+      "step": 211850
+    },
+    {
+      "epoch": 1.8242079889807163,
+      "grad_norm": 3.28125,
+      "learning_rate": 0.00014151062696796433,
+      "loss": 4.1244,
+      "step": 211900
+    },
+    {
+      "epoch": 1.8246384297520661,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00014148597784264082,
+      "loss": 4.2694,
+      "step": 211950
+    },
+    {
+      "epoch": 1.825068870523416,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.0001414613256723358,
+      "loss": 4.365,
+      "step": 212000
+    },
+    {
+      "epoch": 1.8254993112947657,
+      "grad_norm": 3.0625,
+      "learning_rate": 0.0001414366704588587,
+      "loss": 4.5256,
+      "step": 212050
+    },
+    {
+      "epoch": 1.8259297520661157,
+      "grad_norm": 3.953125,
+      "learning_rate": 0.0001414120122040191,
+      "loss": 4.5772,
+      "step": 212100
+    },
+    {
+      "epoch": 1.8263601928374655,
+      "grad_norm": 4.4375,
+      "learning_rate": 0.00014138735090962694,
+      "loss": 4.6309,
+      "step": 212150
+    },
+    {
+      "epoch": 1.8267906336088156,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.00014136268657749225,
+      "loss": 4.134,
+      "step": 212200
+    },
+    {
+      "epoch": 1.8272210743801653,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.00014133801920942536,
+      "loss": 4.503,
+      "step": 212250
+    },
+    {
+      "epoch": 1.8276515151515151,
+      "grad_norm": 3.375,
+      "learning_rate": 0.00014131334880723682,
+      "loss": 4.6735,
+      "step": 212300
+    },
+    {
+      "epoch": 1.828081955922865,
+      "grad_norm": 3.453125,
+      "learning_rate": 0.00014128867537273736,
+      "loss": 4.6235,
+      "step": 212350
+    },
+    {
+      "epoch": 1.8285123966942147,
+      "grad_norm": 3.484375,
+      "learning_rate": 0.00014126399890773797,
+      "loss": 4.3432,
+      "step": 212400
+    },
+    {
+      "epoch": 1.8289428374655647,
+      "grad_norm": 1.6953125,
+      "learning_rate": 0.00014123931941404982,
+      "loss": 4.8241,
+      "step": 212450
+    },
+    {
+      "epoch": 1.8293732782369148,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.00014121463689348434,
+      "loss": 4.5459,
+      "step": 212500
+    },
+    {
+      "epoch": 1.8298037190082646,
+      "grad_norm": 4.0625,
+      "learning_rate": 0.00014118995134785326,
+      "loss": 4.1721,
+      "step": 212550
+    },
+    {
+      "epoch": 1.8302341597796143,
+      "grad_norm": 4.5625,
+      "learning_rate": 0.0001411652627789683,
+      "loss": 4.4798,
+      "step": 212600
+    },
+    {
+      "epoch": 1.8306646005509641,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00014114057118864163,
+      "loss": 4.7275,
+      "step": 212650
+    },
+    {
+      "epoch": 1.831095041322314,
+      "grad_norm": 2.296875,
+      "learning_rate": 0.00014111587657868557,
+      "loss": 4.4862,
+      "step": 212700
+    },
+    {
+      "epoch": 1.831525482093664,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.0001410911789509126,
+      "loss": 4.8968,
+      "step": 212750
+    },
+    {
+      "epoch": 1.8319559228650137,
+      "grad_norm": 4.25,
+      "learning_rate": 0.00014106647830713556,
+      "loss": 4.4673,
+      "step": 212800
+    },
+    {
+      "epoch": 1.8323863636363638,
+      "grad_norm": 3.109375,
+      "learning_rate": 0.00014104177464916736,
+      "loss": 4.5319,
+      "step": 212850
+    },
+    {
+      "epoch": 1.8328168044077136,
+      "grad_norm": 1.5078125,
+      "learning_rate": 0.0001410170679788212,
+      "loss": 4.4805,
+      "step": 212900
+    },
+    {
+      "epoch": 1.8332472451790633,
+      "grad_norm": 4.25,
+      "learning_rate": 0.00014099235829791046,
+      "loss": 4.4092,
+      "step": 212950
+    },
+    {
+      "epoch": 1.8336776859504131,
+      "grad_norm": 3.265625,
+      "learning_rate": 0.00014096764560824886,
+      "loss": 4.5109,
+      "step": 213000
+    },
+    {
+      "epoch": 1.8336776859504131,
+      "eval_loss": 5.0560622215271,
+      "eval_runtime": 21.7274,
+      "eval_samples_per_second": 29.456,
+      "eval_steps_per_second": 14.728,
+      "eval_tts_loss": 7.280787373036879,
+      "step": 213000
+    },
+    {
+      "epoch": 1.834108126721763,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.0001409429299116502,
+      "loss": 4.5675,
+      "step": 213050
+    },
+    {
+      "epoch": 1.834538567493113,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.00014091821120992858,
+      "loss": 4.4511,
+      "step": 213100
+    },
+    {
+      "epoch": 1.834969008264463,
+      "grad_norm": 3.546875,
+      "learning_rate": 0.00014089348950489827,
+      "loss": 4.7129,
+      "step": 213150
+    },
+    {
+      "epoch": 1.8353994490358128,
+      "grad_norm": 1.7578125,
+      "learning_rate": 0.00014086876479837388,
+      "loss": 4.5207,
+      "step": 213200
+    },
+    {
+      "epoch": 1.8358298898071626,
+      "grad_norm": 1.9921875,
+      "learning_rate": 0.00014084403709217002,
+      "loss": 4.2102,
+      "step": 213250
+    },
+    {
+      "epoch": 1.8362603305785123,
+      "grad_norm": 4.375,
+      "learning_rate": 0.00014081930638810175,
+      "loss": 4.3774,
+      "step": 213300
+    },
+    {
+      "epoch": 1.8366907713498621,
+      "grad_norm": 1.4921875,
+      "learning_rate": 0.0001407945726879842,
+      "loss": 4.7277,
+      "step": 213350
+    },
+    {
+      "epoch": 1.8371212121212122,
+      "grad_norm": 0.93359375,
+      "learning_rate": 0.00014076983599363278,
+      "loss": 4.7125,
+      "step": 213400
+    },
+    {
+      "epoch": 1.837551652892562,
+      "grad_norm": 4.09375,
+      "learning_rate": 0.00014074509630686315,
+      "loss": 4.5953,
+      "step": 213450
+    },
+    {
+      "epoch": 1.837982093663912,
+      "grad_norm": 1.625,
+      "learning_rate": 0.00014072035362949108,
+      "loss": 4.4593,
+      "step": 213500
+    },
+    {
+      "epoch": 1.8384125344352618,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.0001406956079633327,
+      "loss": 4.235,
+      "step": 213550
+    },
+    {
+      "epoch": 1.8388429752066116,
+      "grad_norm": 3.34375,
+      "learning_rate": 0.00014067085931020423,
+      "loss": 4.6461,
+      "step": 213600
+    },
+    {
+      "epoch": 1.8392734159779613,
+      "grad_norm": 3.015625,
+      "learning_rate": 0.00014064610767192215,
+      "loss": 4.5389,
+      "step": 213650
+    },
+    {
+      "epoch": 1.8397038567493111,
+      "grad_norm": 4.09375,
+      "learning_rate": 0.00014062135305030333,
+      "loss": 4.4978,
+      "step": 213700
+    },
+    {
+      "epoch": 1.8401342975206612,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.0001405965954471645,
+      "loss": 4.0845,
+      "step": 213750
+    },
+    {
+      "epoch": 1.8405647382920112,
+      "grad_norm": 0.78125,
+      "learning_rate": 0.00014057183486432294,
+      "loss": 4.6071,
+      "step": 213800
+    },
+    {
+      "epoch": 1.840995179063361,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.00014054707130359598,
+      "loss": 4.5696,
+      "step": 213850
+    },
+    {
+      "epoch": 1.8414256198347108,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.00014052230476680125,
+      "loss": 4.5999,
+      "step": 213900
+    },
+    {
+      "epoch": 1.8418560606060606,
+      "grad_norm": 3.609375,
+      "learning_rate": 0.00014049753525575653,
+      "loss": 4.384,
+      "step": 213950
+    },
+    {
+      "epoch": 1.8422865013774103,
+      "grad_norm": 3.25,
+      "learning_rate": 0.00014047276277227985,
+      "loss": 4.5181,
+      "step": 214000
+    },
+    {
+      "epoch": 1.8427169421487604,
+      "grad_norm": 5.59375,
+      "learning_rate": 0.00014044798731818948,
+      "loss": 4.6399,
+      "step": 214050
+    },
+    {
+      "epoch": 1.8431473829201102,
+      "grad_norm": 4.03125,
+      "learning_rate": 0.00014042320889530384,
+      "loss": 4.6073,
+      "step": 214100
+    },
+    {
+      "epoch": 1.8435778236914602,
+      "grad_norm": 3.03125,
+      "learning_rate": 0.00014039842750544166,
+      "loss": 4.4025,
+      "step": 214150
+    },
+    {
+      "epoch": 1.84400826446281,
+      "grad_norm": 4.25,
+      "learning_rate": 0.00014037364315042181,
+      "loss": 4.5702,
+      "step": 214200
+    },
+    {
+      "epoch": 1.8444387052341598,
+      "grad_norm": 4.6875,
+      "learning_rate": 0.0001403488558320634,
+      "loss": 4.6863,
+      "step": 214250
+    },
+    {
+      "epoch": 1.8448691460055096,
+      "grad_norm": 3.453125,
+      "learning_rate": 0.0001403240655521858,
+      "loss": 4.5646,
+      "step": 214300
+    },
+    {
+      "epoch": 1.8452995867768593,
+      "grad_norm": 3.078125,
+      "learning_rate": 0.00014029927231260859,
+      "loss": 4.3433,
+      "step": 214350
+    },
+    {
+      "epoch": 1.8457300275482094,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00014027447611515145,
+      "loss": 4.3073,
+      "step": 214400
+    },
+    {
+      "epoch": 1.8461604683195594,
+      "grad_norm": 3.640625,
+      "learning_rate": 0.00014024967696163443,
+      "loss": 4.3947,
+      "step": 214450
+    },
+    {
+      "epoch": 1.8465909090909092,
+      "grad_norm": 3.59375,
+      "learning_rate": 0.0001402248748538777,
+      "loss": 4.3048,
+      "step": 214500
+    },
+    {
+      "epoch": 1.847021349862259,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.00014020006979370174,
+      "loss": 4.4919,
+      "step": 214550
+    },
+    {
+      "epoch": 1.8474517906336088,
+      "grad_norm": 3.40625,
+      "learning_rate": 0.00014017526178292713,
+      "loss": 4.3821,
+      "step": 214600
+    },
+    {
+      "epoch": 1.8478822314049586,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.00014015045082337472,
+      "loss": 4.3122,
+      "step": 214650
+    },
+    {
+      "epoch": 1.8483126721763086,
+      "grad_norm": 1.9140625,
+      "learning_rate": 0.00014012563691686563,
+      "loss": 4.1348,
+      "step": 214700
+    },
+    {
+      "epoch": 1.8487431129476584,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00014010082006522115,
+      "loss": 4.1318,
+      "step": 214750
+    },
+    {
+      "epoch": 1.8491735537190084,
+      "grad_norm": 2.984375,
+      "learning_rate": 0.0001400760002702627,
+      "loss": 4.6662,
+      "step": 214800
+    },
+    {
+      "epoch": 1.8496039944903582,
+      "grad_norm": 3.09375,
+      "learning_rate": 0.0001400511775338121,
+      "loss": 4.787,
+      "step": 214850
+    },
+    {
+      "epoch": 1.850034435261708,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00014002635185769122,
+      "loss": 4.3495,
+      "step": 214900
+    },
+    {
+      "epoch": 1.8504648760330578,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00014000152324372223,
+      "loss": 4.5445,
+      "step": 214950
+    },
+    {
+      "epoch": 1.8508953168044076,
+      "grad_norm": 3.15625,
+      "learning_rate": 0.0001399766916937275,
+      "loss": 4.73,
+      "step": 215000
+    },
+    {
+      "epoch": 1.8513257575757576,
+      "grad_norm": 3.75,
+      "learning_rate": 0.00013995185720952963,
+      "loss": 4.316,
+      "step": 215050
+    },
+    {
+      "epoch": 1.8517561983471076,
+      "grad_norm": 3.234375,
+      "learning_rate": 0.0001399270197929514,
+      "loss": 4.7898,
+      "step": 215100
+    },
+    {
+      "epoch": 1.8521866391184574,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.0001399021794458158,
+      "loss": 4.731,
+      "step": 215150
+    },
+    {
+      "epoch": 1.8526170798898072,
+      "grad_norm": 1.6640625,
+      "learning_rate": 0.00013987733616994608,
+      "loss": 4.2848,
+      "step": 215200
+    },
+    {
+      "epoch": 1.853047520661157,
+      "grad_norm": 1.9140625,
+      "learning_rate": 0.0001398524899671657,
+      "loss": 4.3253,
+      "step": 215250
+    },
+    {
+      "epoch": 1.8534779614325068,
+      "grad_norm": 3.109375,
+      "learning_rate": 0.00013982764083929832,
+      "loss": 4.6984,
+      "step": 215300
+    },
+    {
+      "epoch": 1.8539084022038568,
+      "grad_norm": 6.65625,
+      "learning_rate": 0.00013980278878816775,
+      "loss": 4.2792,
+      "step": 215350
+    },
+    {
+      "epoch": 1.8543388429752066,
+      "grad_norm": 2.984375,
+      "learning_rate": 0.00013977793381559814,
+      "loss": 4.4643,
+      "step": 215400
+    },
+    {
+      "epoch": 1.8547692837465566,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.00013975307592341377,
+      "loss": 4.3965,
+      "step": 215450
+    },
+    {
+      "epoch": 1.8551997245179064,
+      "grad_norm": 3.65625,
+      "learning_rate": 0.00013972821511343917,
+      "loss": 4.3163,
+      "step": 215500
+    },
+    {
+      "epoch": 1.8556301652892562,
+      "grad_norm": 4.75,
+      "learning_rate": 0.00013970335138749904,
+      "loss": 4.5453,
+      "step": 215550
+    },
+    {
+      "epoch": 1.856060606060606,
+      "grad_norm": 3.515625,
+      "learning_rate": 0.00013967848474741838,
+      "loss": 4.3151,
+      "step": 215600
+    },
+    {
+      "epoch": 1.8564910468319558,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.00013965361519502226,
+      "loss": 4.1963,
+      "step": 215650
+    },
+    {
+      "epoch": 1.8569214876033058,
+      "grad_norm": 3.203125,
+      "learning_rate": 0.00013962874273213611,
+      "loss": 4.1927,
+      "step": 215700
+    },
+    {
+      "epoch": 1.8573519283746558,
+      "grad_norm": 4.125,
+      "learning_rate": 0.00013960386736058553,
+      "loss": 4.1732,
+      "step": 215750
+    },
+    {
+      "epoch": 1.8577823691460056,
+      "grad_norm": 1.15625,
+      "learning_rate": 0.00013957898908219628,
+      "loss": 4.1553,
+      "step": 215800
+    },
+    {
+      "epoch": 1.8582128099173554,
+      "grad_norm": 1.8125,
+      "learning_rate": 0.0001395541078987944,
+      "loss": 4.4413,
+      "step": 215850
+    },
+    {
+      "epoch": 1.8586432506887052,
+      "grad_norm": 3.96875,
+      "learning_rate": 0.0001395292238122061,
+      "loss": 4.4824,
+      "step": 215900
+    },
+    {
+      "epoch": 1.859073691460055,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.00013950433682425781,
+      "loss": 4.5343,
+      "step": 215950
+    },
+    {
+      "epoch": 1.859504132231405,
+      "grad_norm": 1.640625,
+      "learning_rate": 0.00013947944693677621,
+      "loss": 4.6324,
+      "step": 216000
+    },
+    {
+      "epoch": 1.859504132231405,
+      "eval_loss": 5.0532026290893555,
+      "eval_runtime": 22.153,
+      "eval_samples_per_second": 28.89,
+      "eval_steps_per_second": 14.445,
+      "eval_tts_loss": 7.206952735664344,
+      "step": 216000
+    },
+    {
+      "epoch": 1.8599345730027548,
+      "grad_norm": 1.3671875,
+      "learning_rate": 0.00013945455415158813,
+      "loss": 4.1854,
+      "step": 216050
+    },
+    {
+      "epoch": 1.8603650137741048,
+      "grad_norm": 2.25,
+      "learning_rate": 0.00013942965847052067,
+      "loss": 4.4626,
+      "step": 216100
+    },
+    {
+      "epoch": 1.8607954545454546,
+      "grad_norm": 3.109375,
+      "learning_rate": 0.00013940475989540114,
+      "loss": 4.649,
+      "step": 216150
+    },
+    {
+      "epoch": 1.8612258953168044,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00013937985842805696,
+      "loss": 4.3841,
+      "step": 216200
+    },
+    {
+      "epoch": 1.8616563360881542,
+      "grad_norm": 3.03125,
+      "learning_rate": 0.00013935495407031592,
+      "loss": 4.8282,
+      "step": 216250
+    },
+    {
+      "epoch": 1.862086776859504,
+      "grad_norm": 3.390625,
+      "learning_rate": 0.00013933004682400597,
+      "loss": 4.7881,
+      "step": 216300
+    },
+    {
+      "epoch": 1.862517217630854,
+      "grad_norm": 0.7578125,
+      "learning_rate": 0.00013930513669095515,
+      "loss": 4.0291,
+      "step": 216350
+    },
+    {
+      "epoch": 1.862947658402204,
+      "grad_norm": 4.375,
+      "learning_rate": 0.0001392802236729919,
+      "loss": 4.6901,
+      "step": 216400
+    },
+    {
+      "epoch": 1.8633780991735538,
+      "grad_norm": 3.28125,
+      "learning_rate": 0.00013925530777194472,
+      "loss": 4.3857,
+      "step": 216450
+    },
+    {
+      "epoch": 1.8638085399449036,
+      "grad_norm": 3.625,
+      "learning_rate": 0.00013923038898964242,
+      "loss": 4.4497,
+      "step": 216500
+    },
+    {
+      "epoch": 1.8642389807162534,
+      "grad_norm": 3.296875,
+      "learning_rate": 0.00013920546732791396,
+      "loss": 4.2481,
+      "step": 216550
+    },
+    {
+      "epoch": 1.8646694214876032,
+      "grad_norm": 3.140625,
+      "learning_rate": 0.00013918054278858858,
+      "loss": 4.3764,
+      "step": 216600
+    },
+    {
+      "epoch": 1.8650998622589532,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.00013915561537349564,
+      "loss": 4.5323,
+      "step": 216650
+    },
+    {
+      "epoch": 1.865530303030303,
+      "grad_norm": 5.46875,
+      "learning_rate": 0.00013913068508446478,
+      "loss": 4.3292,
+      "step": 216700
+    },
+    {
+      "epoch": 1.865960743801653,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.0001391057519233258,
+      "loss": 4.1943,
+      "step": 216750
+    },
+    {
+      "epoch": 1.8663911845730028,
+      "grad_norm": 3.15625,
+      "learning_rate": 0.00013908081589190883,
+      "loss": 4.6298,
+      "step": 216800
+    },
+    {
+      "epoch": 1.8668216253443526,
+      "grad_norm": 3.546875,
+      "learning_rate": 0.000139055876992044,
+      "loss": 4.5038,
+      "step": 216850
+    },
+    {
+      "epoch": 1.8672520661157024,
+      "grad_norm": 2.21875,
+      "learning_rate": 0.00013903093522556187,
+      "loss": 4.3473,
+      "step": 216900
+    },
+    {
+      "epoch": 1.8676825068870524,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.00013900599059429305,
+      "loss": 4.7651,
+      "step": 216950
+    },
+    {
+      "epoch": 1.8681129476584022,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00013898104310006843,
+      "loss": 4.7159,
+      "step": 217000
+    },
+    {
+      "epoch": 1.8685433884297522,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.00013895609274471913,
+      "loss": 4.0896,
+      "step": 217050
+    },
+    {
+      "epoch": 1.868973829201102,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.00013893113953007645,
+      "loss": 4.5924,
+      "step": 217100
+    },
+    {
+      "epoch": 1.8694042699724518,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.0001389061834579719,
+      "loss": 4.5477,
+      "step": 217150
+    },
+    {
+      "epoch": 1.8698347107438016,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.00013888122453023716,
+      "loss": 4.7615,
+      "step": 217200
+    },
+    {
+      "epoch": 1.8702651515151514,
+      "grad_norm": 1.984375,
+      "learning_rate": 0.0001388562627487042,
+      "loss": 4.369,
+      "step": 217250
+    },
+    {
+      "epoch": 1.8706955922865014,
+      "grad_norm": 2.21875,
+      "learning_rate": 0.0001388312981152052,
+      "loss": 4.5305,
+      "step": 217300
+    },
+    {
+      "epoch": 1.8711260330578512,
+      "grad_norm": 2.0,
+      "learning_rate": 0.0001388063306315724,
+      "loss": 4.477,
+      "step": 217350
+    },
+    {
+      "epoch": 1.8715564738292012,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.0001387813602996385,
+      "loss": 4.3883,
+      "step": 217400
+    },
+    {
+      "epoch": 1.871986914600551,
+      "grad_norm": 2.921875,
+      "learning_rate": 0.0001387563871212361,
+      "loss": 4.6034,
+      "step": 217450
+    },
+    {
+      "epoch": 1.8724173553719008,
+      "grad_norm": 3.296875,
+      "learning_rate": 0.00013873141109819833,
+      "loss": 4.3873,
+      "step": 217500
+    },
+    {
+      "epoch": 1.8728477961432506,
+      "grad_norm": 2.25,
+      "learning_rate": 0.00013870643223235833,
+      "loss": 4.765,
+      "step": 217550
+    },
+    {
+      "epoch": 1.8732782369146006,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.00013868145052554947,
+      "loss": 4.6269,
+      "step": 217600
+    },
+    {
+      "epoch": 1.8737086776859504,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00013865646597960535,
+      "loss": 4.395,
+      "step": 217650
+    },
+    {
+      "epoch": 1.8741391184573004,
+      "grad_norm": 1.890625,
+      "learning_rate": 0.00013863147859635984,
+      "loss": 4.3026,
+      "step": 217700
+    },
+    {
+      "epoch": 1.8745695592286502,
+      "grad_norm": 3.90625,
+      "learning_rate": 0.00013860648837764686,
+      "loss": 4.5575,
+      "step": 217750
+    },
+    {
+      "epoch": 1.875,
+      "grad_norm": 1.7265625,
+      "learning_rate": 0.00013858149532530072,
+      "loss": 4.3687,
+      "step": 217800
+    },
+    {
+      "epoch": 1.8754304407713498,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.00013855649944115585,
+      "loss": 4.2108,
+      "step": 217850
+    },
+    {
+      "epoch": 1.8758608815426996,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.00013853150072704685,
+      "loss": 4.2543,
+      "step": 217900
+    },
+    {
+      "epoch": 1.8762913223140496,
+      "grad_norm": 2.265625,
+      "learning_rate": 0.00013850649918480863,
+      "loss": 4.3372,
+      "step": 217950
+    },
+    {
+      "epoch": 1.8767217630853994,
+      "grad_norm": 2.421875,
+      "learning_rate": 0.0001384814948162762,
+      "loss": 4.7017,
+      "step": 218000
+    },
+    {
+      "epoch": 1.8771522038567494,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.00013845648762328487,
+      "loss": 4.6242,
+      "step": 218050
+    },
+    {
+      "epoch": 1.8775826446280992,
+      "grad_norm": 3.359375,
+      "learning_rate": 0.00013843147760767005,
+      "loss": 4.213,
+      "step": 218100
+    },
+    {
+      "epoch": 1.878013085399449,
+      "grad_norm": 3.109375,
+      "learning_rate": 0.0001384064647712675,
+      "loss": 4.5054,
+      "step": 218150
+    },
+    {
+      "epoch": 1.8784435261707988,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.00013838144911591305,
+      "loss": 4.4854,
+      "step": 218200
+    },
+    {
+      "epoch": 1.8788739669421488,
+      "grad_norm": 3.96875,
+      "learning_rate": 0.0001383564306434428,
+      "loss": 4.0335,
+      "step": 218250
+    },
+    {
+      "epoch": 1.8793044077134986,
+      "grad_norm": 4.28125,
+      "learning_rate": 0.0001383314093556931,
+      "loss": 4.4872,
+      "step": 218300
+    },
+    {
+      "epoch": 1.8797348484848486,
+      "grad_norm": 2.0,
+      "learning_rate": 0.00013830638525450038,
+      "loss": 4.3535,
+      "step": 218350
+    },
+    {
+      "epoch": 1.8801652892561984,
+      "grad_norm": 1.578125,
+      "learning_rate": 0.00013828135834170144,
+      "loss": 4.3971,
+      "step": 218400
+    },
+    {
+      "epoch": 1.8805957300275482,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.00013825632861913317,
+      "loss": 4.0632,
+      "step": 218450
+    },
+    {
+      "epoch": 1.881026170798898,
+      "grad_norm": 3.46875,
+      "learning_rate": 0.00013823129608863267,
+      "loss": 4.2979,
+      "step": 218500
+    },
+    {
+      "epoch": 1.8814566115702478,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.00013820626075203727,
+      "loss": 4.8421,
+      "step": 218550
+    },
+    {
+      "epoch": 1.8818870523415978,
+      "grad_norm": 2.9375,
+      "learning_rate": 0.00013818122261118458,
+      "loss": 4.5021,
+      "step": 218600
+    },
+    {
+      "epoch": 1.8823174931129476,
+      "grad_norm": 3.140625,
+      "learning_rate": 0.00013815618166791228,
+      "loss": 4.3414,
+      "step": 218650
+    },
+    {
+      "epoch": 1.8827479338842976,
+      "grad_norm": 2.140625,
+      "learning_rate": 0.00013813113792405835,
+      "loss": 4.5276,
+      "step": 218700
+    },
+    {
+      "epoch": 1.8831783746556474,
+      "grad_norm": 3.765625,
+      "learning_rate": 0.00013810609138146092,
+      "loss": 4.3439,
+      "step": 218750
+    },
+    {
+      "epoch": 1.8836088154269972,
+      "grad_norm": 3.78125,
+      "learning_rate": 0.00013808104204195834,
+      "loss": 4.6214,
+      "step": 218800
+    },
+    {
+      "epoch": 1.884039256198347,
+      "grad_norm": 3.109375,
+      "learning_rate": 0.00013805598990738926,
+      "loss": 4.6648,
+      "step": 218850
+    },
+    {
+      "epoch": 1.884469696969697,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.0001380309349795924,
+      "loss": 4.4938,
+      "step": 218900
+    },
+    {
+      "epoch": 1.8849001377410468,
+      "grad_norm": 3.765625,
+      "learning_rate": 0.0001380058772604067,
+      "loss": 4.5276,
+      "step": 218950
+    },
+    {
+      "epoch": 1.8853305785123968,
+      "grad_norm": 2.921875,
+      "learning_rate": 0.0001379808167516714,
+      "loss": 4.408,
+      "step": 219000
+    },
+    {
+      "epoch": 1.8853305785123968,
+      "eval_loss": 5.048763275146484,
+      "eval_runtime": 21.6468,
+      "eval_samples_per_second": 29.566,
+      "eval_steps_per_second": 14.783,
+      "eval_tts_loss": 7.257337959361985,
+      "step": 219000
+    },
+    {
+      "epoch": 1.8857610192837466,
+      "grad_norm": 1.3984375,
+      "learning_rate": 0.00013795575345522588,
+      "loss": 4.0965,
+      "step": 219050
+    },
+    {
+      "epoch": 1.8861914600550964,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00013793068737290973,
+      "loss": 4.3708,
+      "step": 219100
+    },
+    {
+      "epoch": 1.8866219008264462,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00013790561850656269,
+      "loss": 4.6446,
+      "step": 219150
+    },
+    {
+      "epoch": 1.887052341597796,
+      "grad_norm": 1.8359375,
+      "learning_rate": 0.00013788054685802487,
+      "loss": 4.6007,
+      "step": 219200
+    },
+    {
+      "epoch": 1.887482782369146,
+      "grad_norm": 3.890625,
+      "learning_rate": 0.00013785547242913638,
+      "loss": 4.075,
+      "step": 219250
+    },
+    {
+      "epoch": 1.887913223140496,
+      "grad_norm": 1.015625,
+      "learning_rate": 0.0001378303952217376,
+      "loss": 4.1568,
+      "step": 219300
+    },
+    {
+      "epoch": 1.8883436639118458,
+      "grad_norm": 1.7109375,
+      "learning_rate": 0.00013780531523766931,
+      "loss": 4.7238,
+      "step": 219350
+    },
+    {
+      "epoch": 1.8887741046831956,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.00013778023247877215,
+      "loss": 4.2892,
+      "step": 219400
+    },
+    {
+      "epoch": 1.8892045454545454,
+      "grad_norm": 1.625,
+      "learning_rate": 0.00013775514694688726,
+      "loss": 4.0651,
+      "step": 219450
+    },
+    {
+      "epoch": 1.8896349862258952,
+      "grad_norm": 4.0,
+      "learning_rate": 0.00013773005864385577,
+      "loss": 4.6462,
+      "step": 219500
+    },
+    {
+      "epoch": 1.8900654269972452,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.0001377049675715192,
+      "loss": 3.9879,
+      "step": 219550
+    },
+    {
+      "epoch": 1.890495867768595,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.00013767987373171914,
+      "loss": 4.7724,
+      "step": 219600
+    },
+    {
+      "epoch": 1.890926308539945,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.00013765477712629736,
+      "loss": 4.7653,
+      "step": 219650
+    },
+    {
+      "epoch": 1.8913567493112948,
+      "grad_norm": 4.0625,
+      "learning_rate": 0.000137629677757096,
+      "loss": 4.3554,
+      "step": 219700
+    },
+    {
+      "epoch": 1.8917871900826446,
+      "grad_norm": 4.0625,
+      "learning_rate": 0.00013760457562595724,
+      "loss": 4.0532,
+      "step": 219750
+    },
+    {
+      "epoch": 1.8922176308539944,
+      "grad_norm": 3.546875,
+      "learning_rate": 0.00013757947073472353,
+      "loss": 4.4743,
+      "step": 219800
+    },
+    {
+      "epoch": 1.8926480716253442,
+      "grad_norm": 2.296875,
+      "learning_rate": 0.00013755436308523754,
+      "loss": 4.5887,
+      "step": 219850
+    },
+    {
+      "epoch": 1.8930785123966942,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.00013752925267934208,
+      "loss": 4.2288,
+      "step": 219900
+    },
+    {
+      "epoch": 1.8935089531680442,
+      "grad_norm": 1.3671875,
+      "learning_rate": 0.00013750413951888024,
+      "loss": 4.4342,
+      "step": 219950
+    },
+    {
+      "epoch": 1.893939393939394,
+      "grad_norm": 3.96875,
+      "learning_rate": 0.00013747902360569526,
+      "loss": 4.5701,
+      "step": 220000
+    },
+    {
+      "epoch": 1.8943698347107438,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00013745390494163058,
+      "loss": 4.4464,
+      "step": 220050
+    },
+    {
+      "epoch": 1.8948002754820936,
+      "grad_norm": 2.578125,
+      "learning_rate": 0.00013742878352852983,
+      "loss": 4.3805,
+      "step": 220100
+    },
+    {
+      "epoch": 1.8952307162534434,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.0001374036593682369,
+      "loss": 4.1921,
+      "step": 220150
+    },
+    {
+      "epoch": 1.8956611570247934,
+      "grad_norm": 3.984375,
+      "learning_rate": 0.0001373785324625959,
+      "loss": 4.7629,
+      "step": 220200
+    },
+    {
+      "epoch": 1.8960915977961432,
+      "grad_norm": 1.234375,
+      "learning_rate": 0.00013735340281345102,
+      "loss": 4.3835,
+      "step": 220250
+    },
+    {
+      "epoch": 1.8965220385674932,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00013732827042264672,
+      "loss": 4.2564,
+      "step": 220300
+    },
+    {
+      "epoch": 1.896952479338843,
+      "grad_norm": 3.46875,
+      "learning_rate": 0.00013730313529202769,
+      "loss": 4.6334,
+      "step": 220350
+    },
+    {
+      "epoch": 1.8973829201101928,
+      "grad_norm": 3.796875,
+      "learning_rate": 0.00013727799742343878,
+      "loss": 4.1464,
+      "step": 220400
+    },
+    {
+      "epoch": 1.8978133608815426,
+      "grad_norm": 3.015625,
+      "learning_rate": 0.00013725285681872508,
+      "loss": 4.4526,
+      "step": 220450
+    },
+    {
+      "epoch": 1.8982438016528924,
+      "grad_norm": 3.140625,
+      "learning_rate": 0.00013722771347973186,
+      "loss": 4.5367,
+      "step": 220500
+    },
+    {
+      "epoch": 1.8986742424242424,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00013720256740830454,
+      "loss": 4.0994,
+      "step": 220550
+    },
+    {
+      "epoch": 1.8991046831955924,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.0001371774186062888,
+      "loss": 4.7075,
+      "step": 220600
+    },
+    {
+      "epoch": 1.8995351239669422,
+      "grad_norm": 2.9375,
+      "learning_rate": 0.00013715226707553056,
+      "loss": 4.3286,
+      "step": 220650
+    },
+    {
+      "epoch": 1.899965564738292,
+      "grad_norm": 3.6875,
+      "learning_rate": 0.00013712711281787583,
+      "loss": 4.2165,
+      "step": 220700
+    },
+    {
+      "epoch": 1.9003960055096418,
+      "grad_norm": 3.0,
+      "learning_rate": 0.00013710195583517093,
+      "loss": 3.693,
+      "step": 220750
+    },
+    {
+      "epoch": 1.9008264462809916,
+      "grad_norm": 2.921875,
+      "learning_rate": 0.00013707679612926226,
+      "loss": 4.5443,
+      "step": 220800
+    },
+    {
+      "epoch": 1.9012568870523416,
+      "grad_norm": 3.578125,
+      "learning_rate": 0.00013705163370199652,
+      "loss": 4.3232,
+      "step": 220850
+    },
+    {
+      "epoch": 1.9016873278236914,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.00013702646855522064,
+      "loss": 4.0892,
+      "step": 220900
+    },
+    {
+      "epoch": 1.9021177685950414,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.00013700130069078158,
+      "loss": 4.2569,
+      "step": 220950
+    },
+    {
+      "epoch": 1.9025482093663912,
+      "grad_norm": 3.0,
+      "learning_rate": 0.0001369761301105267,
+      "loss": 4.4526,
+      "step": 221000
+    },
+    {
+      "epoch": 1.902978650137741,
+      "grad_norm": 3.828125,
+      "learning_rate": 0.0001369509568163034,
+      "loss": 4.0168,
+      "step": 221050
+    },
+    {
+      "epoch": 1.9034090909090908,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.00013692578080995937,
+      "loss": 4.5407,
+      "step": 221100
+    },
+    {
+      "epoch": 1.9038395316804406,
+      "grad_norm": 1.9140625,
+      "learning_rate": 0.0001369006020933425,
+      "loss": 4.3903,
+      "step": 221150
+    },
+    {
+      "epoch": 1.9042699724517906,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.00013687542066830082,
+      "loss": 4.6166,
+      "step": 221200
+    },
+    {
+      "epoch": 1.9047004132231407,
+      "grad_norm": 3.046875,
+      "learning_rate": 0.0001368502365366826,
+      "loss": 4.5691,
+      "step": 221250
+    },
+    {
+      "epoch": 1.9051308539944904,
+      "grad_norm": 3.453125,
+      "learning_rate": 0.0001368250497003363,
+      "loss": 4.5418,
+      "step": 221300
+    },
+    {
+      "epoch": 1.9055612947658402,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00013679986016111055,
+      "loss": 3.9199,
+      "step": 221350
+    },
+    {
+      "epoch": 1.90599173553719,
+      "grad_norm": 3.40625,
+      "learning_rate": 0.0001367746679208543,
+      "loss": 4.613,
+      "step": 221400
+    },
+    {
+      "epoch": 1.9064221763085398,
+      "grad_norm": 3.0,
+      "learning_rate": 0.0001367494729814165,
+      "loss": 4.3766,
+      "step": 221450
+    },
+    {
+      "epoch": 1.9068526170798898,
+      "grad_norm": 3.0625,
+      "learning_rate": 0.00013672427534464648,
+      "loss": 4.7049,
+      "step": 221500
+    },
+    {
+      "epoch": 1.9072830578512396,
+      "grad_norm": 2.375,
+      "learning_rate": 0.00013669907501239368,
+      "loss": 4.6986,
+      "step": 221550
+    },
+    {
+      "epoch": 1.9077134986225897,
+      "grad_norm": 2.921875,
+      "learning_rate": 0.0001366738719865077,
+      "loss": 4.4958,
+      "step": 221600
+    },
+    {
+      "epoch": 1.9081439393939394,
+      "grad_norm": 2.0,
+      "learning_rate": 0.00013664866626883845,
+      "loss": 4.5369,
+      "step": 221650
+    },
+    {
+      "epoch": 1.9085743801652892,
+      "grad_norm": 1.6171875,
+      "learning_rate": 0.00013662345786123593,
+      "loss": 4.4422,
+      "step": 221700
+    },
+    {
+      "epoch": 1.909004820936639,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00013659824676555039,
+      "loss": 4.3376,
+      "step": 221750
+    },
+    {
+      "epoch": 1.9094352617079888,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.0001365730329836323,
+      "loss": 4.6718,
+      "step": 221800
+    },
+    {
+      "epoch": 1.9098657024793388,
+      "grad_norm": 2.875,
+      "learning_rate": 0.00013654781651733225,
+      "loss": 4.5326,
+      "step": 221850
+    },
+    {
+      "epoch": 1.9102961432506889,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.00013652259736850116,
+      "loss": 4.5264,
+      "step": 221900
+    },
+    {
+      "epoch": 1.9107265840220387,
+      "grad_norm": 3.078125,
+      "learning_rate": 0.00013649737553898995,
+      "loss": 4.2224,
+      "step": 221950
+    },
+    {
+      "epoch": 1.9111570247933884,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.00013647215103064993,
+      "loss": 4.3864,
+      "step": 222000
+    },
+    {
+      "epoch": 1.9111570247933884,
+      "eval_loss": 5.045414924621582,
+      "eval_runtime": 22.036,
+      "eval_samples_per_second": 29.043,
+      "eval_steps_per_second": 14.522,
+      "eval_tts_loss": 7.304645935329636,
+      "step": 222000
+    },
+    {
+      "epoch": 1.9115874655647382,
+      "grad_norm": 3.28125,
+      "learning_rate": 0.00013644692384533246,
+      "loss": 4.4359,
+      "step": 222050
+    },
+    {
+      "epoch": 1.912017906336088,
+      "grad_norm": 4.53125,
+      "learning_rate": 0.00013642169398488924,
+      "loss": 4.819,
+      "step": 222100
+    },
+    {
+      "epoch": 1.912448347107438,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.00013639646145117203,
+      "loss": 4.5313,
+      "step": 222150
+    },
+    {
+      "epoch": 1.9128787878787878,
+      "grad_norm": 2.125,
+      "learning_rate": 0.00013637122624603285,
+      "loss": 4.2293,
+      "step": 222200
+    },
+    {
+      "epoch": 1.9133092286501379,
+      "grad_norm": 3.65625,
+      "learning_rate": 0.0001363459883713239,
+      "loss": 4.1539,
+      "step": 222250
+    },
+    {
+      "epoch": 1.9137396694214877,
+      "grad_norm": 3.15625,
+      "learning_rate": 0.00013632074782889765,
+      "loss": 4.4515,
+      "step": 222300
+    },
+    {
+      "epoch": 1.9141701101928374,
+      "grad_norm": 3.390625,
+      "learning_rate": 0.00013629550462060662,
+      "loss": 4.6567,
+      "step": 222350
+    },
+    {
+      "epoch": 1.9146005509641872,
+      "grad_norm": 3.484375,
+      "learning_rate": 0.00013627025874830365,
+      "loss": 4.243,
+      "step": 222400
+    },
+    {
+      "epoch": 1.915030991735537,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.00013624501021384174,
+      "loss": 4.2362,
+      "step": 222450
+    },
+    {
+      "epoch": 1.915461432506887,
+      "grad_norm": 1.59375,
+      "learning_rate": 0.00013621975901907405,
+      "loss": 4.3312,
+      "step": 222500
+    },
+    {
+      "epoch": 1.915891873278237,
+      "grad_norm": 1.6875,
+      "learning_rate": 0.000136194505165854,
+      "loss": 4.2704,
+      "step": 222550
+    },
+    {
+      "epoch": 1.9163223140495869,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.0001361692486560351,
+      "loss": 4.2205,
+      "step": 222600
+    },
+    {
+      "epoch": 1.9167527548209367,
+      "grad_norm": 4.3125,
+      "learning_rate": 0.0001361439894914712,
+      "loss": 4.1049,
+      "step": 222650
+    },
+    {
+      "epoch": 1.9171831955922864,
+      "grad_norm": 3.171875,
+      "learning_rate": 0.00013611872767401624,
+      "loss": 4.3834,
+      "step": 222700
+    },
+    {
+      "epoch": 1.9176136363636362,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00013609346320552437,
+      "loss": 4.3754,
+      "step": 222750
+    },
+    {
+      "epoch": 1.9180440771349863,
+      "grad_norm": 3.546875,
+      "learning_rate": 0.00013606819608785,
+      "loss": 4.8799,
+      "step": 222800
+    },
+    {
+      "epoch": 1.918474517906336,
+      "grad_norm": 4.15625,
+      "learning_rate": 0.00013604292632284757,
+      "loss": 4.5783,
+      "step": 222850
+    },
+    {
+      "epoch": 1.918904958677686,
+      "grad_norm": 3.625,
+      "learning_rate": 0.00013601765391237192,
+      "loss": 4.1957,
+      "step": 222900
+    },
+    {
+      "epoch": 1.9193353994490359,
+      "grad_norm": 4.5625,
+      "learning_rate": 0.00013599237885827798,
+      "loss": 4.4157,
+      "step": 222950
+    },
+    {
+      "epoch": 1.9197658402203857,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.00013596710116242086,
+      "loss": 4.7131,
+      "step": 223000
+    },
+    {
+      "epoch": 1.9201962809917354,
+      "grad_norm": 2.421875,
+      "learning_rate": 0.00013594182082665593,
+      "loss": 4.839,
+      "step": 223050
+    },
+    {
+      "epoch": 1.9206267217630852,
+      "grad_norm": 1.7265625,
+      "learning_rate": 0.00013591653785283863,
+      "loss": 4.655,
+      "step": 223100
+    },
+    {
+      "epoch": 1.9210571625344353,
+      "grad_norm": 2.234375,
+      "learning_rate": 0.00013589125224282476,
+      "loss": 4.4033,
+      "step": 223150
+    },
+    {
+      "epoch": 1.9214876033057853,
+      "grad_norm": 4.03125,
+      "learning_rate": 0.0001358659639984702,
+      "loss": 4.9092,
+      "step": 223200
+    },
+    {
+      "epoch": 1.921918044077135,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00013584067312163103,
+      "loss": 4.5733,
+      "step": 223250
+    },
+    {
+      "epoch": 1.9223484848484849,
+      "grad_norm": 3.015625,
+      "learning_rate": 0.0001358153796141636,
+      "loss": 4.4031,
+      "step": 223300
+    },
+    {
+      "epoch": 1.9227789256198347,
+      "grad_norm": 4.15625,
+      "learning_rate": 0.00013579008347792433,
+      "loss": 4.2904,
+      "step": 223350
+    },
+    {
+      "epoch": 1.9232093663911844,
+      "grad_norm": 3.9375,
+      "learning_rate": 0.00013576478471476994,
+      "loss": 4.2628,
+      "step": 223400
+    },
+    {
+      "epoch": 1.9236398071625345,
+      "grad_norm": 2.578125,
+      "learning_rate": 0.00013573948332655733,
+      "loss": 4.1737,
+      "step": 223450
+    },
+    {
+      "epoch": 1.9240702479338843,
+      "grad_norm": 1.8515625,
+      "learning_rate": 0.0001357141793151435,
+      "loss": 4.3229,
+      "step": 223500
+    },
+    {
+      "epoch": 1.9245006887052343,
+      "grad_norm": 3.8125,
+      "learning_rate": 0.00013568887268238577,
+      "loss": 4.4558,
+      "step": 223550
+    },
+    {
+      "epoch": 1.924931129476584,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.00013566356343014158,
+      "loss": 4.7687,
+      "step": 223600
+    },
+    {
+      "epoch": 1.9253615702479339,
+      "grad_norm": 3.296875,
+      "learning_rate": 0.00013563825156026857,
+      "loss": 4.3129,
+      "step": 223650
+    },
+    {
+      "epoch": 1.9257920110192837,
+      "grad_norm": 1.4921875,
+      "learning_rate": 0.00013561293707462456,
+      "loss": 4.2717,
+      "step": 223700
+    },
+    {
+      "epoch": 1.9262224517906334,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.0001355876199750676,
+      "loss": 4.4938,
+      "step": 223750
+    },
+    {
+      "epoch": 1.9266528925619835,
+      "grad_norm": 1.578125,
+      "learning_rate": 0.00013556230026345588,
+      "loss": 4.3552,
+      "step": 223800
+    },
+    {
+      "epoch": 1.9270833333333335,
+      "grad_norm": 3.03125,
+      "learning_rate": 0.00013553697794164787,
+      "loss": 4.5425,
+      "step": 223850
+    },
+    {
+      "epoch": 1.9275137741046833,
+      "grad_norm": 1.4765625,
+      "learning_rate": 0.0001355116530115021,
+      "loss": 4.5694,
+      "step": 223900
+    },
+    {
+      "epoch": 1.927944214876033,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00013548632547487745,
+      "loss": 4.297,
+      "step": 223950
+    },
+    {
+      "epoch": 1.9283746556473829,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.00013546099533363282,
+      "loss": 4.1554,
+      "step": 224000
+    },
+    {
+      "epoch": 1.9288050964187327,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.00013543566258962745,
+      "loss": 4.0928,
+      "step": 224050
+    },
+    {
+      "epoch": 1.9292355371900827,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.0001354103272447207,
+      "loss": 4.5334,
+      "step": 224100
+    },
+    {
+      "epoch": 1.9296659779614325,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.00013538498930077208,
+      "loss": 4.562,
+      "step": 224150
+    },
+    {
+      "epoch": 1.9300964187327825,
+      "grad_norm": 1.46875,
+      "learning_rate": 0.0001353596487596414,
+      "loss": 4.3581,
+      "step": 224200
+    },
+    {
+      "epoch": 1.9305268595041323,
+      "grad_norm": 4.09375,
+      "learning_rate": 0.00013533430562318857,
+      "loss": 4.2937,
+      "step": 224250
+    },
+    {
+      "epoch": 1.930957300275482,
+      "grad_norm": 1.765625,
+      "learning_rate": 0.00013530895989327373,
+      "loss": 4.7866,
+      "step": 224300
+    },
+    {
+      "epoch": 1.9313877410468319,
+      "grad_norm": 2.25,
+      "learning_rate": 0.00013528361157175725,
+      "loss": 4.1617,
+      "step": 224350
+    },
+    {
+      "epoch": 1.9318181818181817,
+      "grad_norm": 3.765625,
+      "learning_rate": 0.00013525826066049953,
+      "loss": 4.3879,
+      "step": 224400
+    },
+    {
+      "epoch": 1.9322486225895317,
+      "grad_norm": 3.1875,
+      "learning_rate": 0.00013523290716136135,
+      "loss": 4.3734,
+      "step": 224450
+    },
+    {
+      "epoch": 1.9326790633608817,
+      "grad_norm": 3.65625,
+      "learning_rate": 0.00013520755107620359,
+      "loss": 4.5149,
+      "step": 224500
+    },
+    {
+      "epoch": 1.9331095041322315,
+      "grad_norm": 1.25,
+      "learning_rate": 0.00013518219240688737,
+      "loss": 4.6907,
+      "step": 224550
+    },
+    {
+      "epoch": 1.9335399449035813,
+      "grad_norm": 2.21875,
+      "learning_rate": 0.0001351568311552739,
+      "loss": 4.4333,
+      "step": 224600
+    },
+    {
+      "epoch": 1.933970385674931,
+      "grad_norm": 3.140625,
+      "learning_rate": 0.00013513146732322466,
+      "loss": 4.5083,
+      "step": 224650
+    },
+    {
+      "epoch": 1.9344008264462809,
+      "grad_norm": 2.9375,
+      "learning_rate": 0.00013510610091260132,
+      "loss": 4.499,
+      "step": 224700
+    },
+    {
+      "epoch": 1.9348312672176309,
+      "grad_norm": 2.875,
+      "learning_rate": 0.0001350807319252657,
+      "loss": 4.4249,
+      "step": 224750
+    },
+    {
+      "epoch": 1.9352617079889807,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00013505536036307982,
+      "loss": 4.2964,
+      "step": 224800
+    },
+    {
+      "epoch": 1.9356921487603307,
+      "grad_norm": 1.7265625,
+      "learning_rate": 0.00013502998622790593,
+      "loss": 4.1366,
+      "step": 224850
+    },
+    {
+      "epoch": 1.9361225895316805,
+      "grad_norm": 3.34375,
+      "learning_rate": 0.00013500460952160643,
+      "loss": 4.278,
+      "step": 224900
+    },
+    {
+      "epoch": 1.9365530303030303,
+      "grad_norm": 3.40625,
+      "learning_rate": 0.00013497923024604385,
+      "loss": 4.2314,
+      "step": 224950
+    },
+    {
+      "epoch": 1.93698347107438,
+      "grad_norm": 4.21875,
+      "learning_rate": 0.0001349538484030811,
+      "loss": 4.2852,
+      "step": 225000
+    },
+    {
+      "epoch": 1.93698347107438,
+      "eval_loss": 5.044693470001221,
+      "eval_runtime": 21.8144,
+      "eval_samples_per_second": 29.338,
+      "eval_steps_per_second": 14.669,
+      "eval_tts_loss": 7.318111731969122,
+      "step": 225000
+    },
+    {
+      "epoch": 1.9374139118457299,
+      "grad_norm": 3.78125,
+      "learning_rate": 0.00013492846399458104,
+      "loss": 4.4199,
+      "step": 225050
+    },
+    {
+      "epoch": 1.9378443526170799,
+      "grad_norm": 1.8671875,
+      "learning_rate": 0.00013490307702240686,
+      "loss": 4.5072,
+      "step": 225100
+    },
+    {
+      "epoch": 1.93827479338843,
+      "grad_norm": 1.375,
+      "learning_rate": 0.00013487768748842197,
+      "loss": 4.442,
+      "step": 225150
+    },
+    {
+      "epoch": 1.9387052341597797,
+      "grad_norm": 3.40625,
+      "learning_rate": 0.00013485229539448985,
+      "loss": 4.8625,
+      "step": 225200
+    },
+    {
+      "epoch": 1.9391356749311295,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.0001348269007424742,
+      "loss": 4.3945,
+      "step": 225250
+    },
+    {
+      "epoch": 1.9395661157024793,
+      "grad_norm": 3.09375,
+      "learning_rate": 0.000134801503534239,
+      "loss": 3.9488,
+      "step": 225300
+    },
+    {
+      "epoch": 1.939996556473829,
+      "grad_norm": 4.1875,
+      "learning_rate": 0.0001347761037716483,
+      "loss": 4.631,
+      "step": 225350
+    },
+    {
+      "epoch": 1.940426997245179,
+      "grad_norm": 3.25,
+      "learning_rate": 0.00013475070145656644,
+      "loss": 4.5675,
+      "step": 225400
+    },
+    {
+      "epoch": 1.9408574380165289,
+      "grad_norm": 1.0078125,
+      "learning_rate": 0.0001347252965908578,
+      "loss": 3.746,
+      "step": 225450
+    },
+    {
+      "epoch": 1.941287878787879,
+      "grad_norm": 2.25,
+      "learning_rate": 0.00013469988917638718,
+      "loss": 4.2301,
+      "step": 225500
+    },
+    {
+      "epoch": 1.9417183195592287,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.0001346744792150193,
+      "loss": 4.2385,
+      "step": 225550
+    },
+    {
+      "epoch": 1.9421487603305785,
+      "grad_norm": 7.15625,
+      "learning_rate": 0.00013464906670861924,
+      "loss": 4.552,
+      "step": 225600
+    },
+    {
+      "epoch": 1.9425792011019283,
+      "grad_norm": 3.75,
+      "learning_rate": 0.00013462365165905227,
+      "loss": 4.8005,
+      "step": 225650
+    },
+    {
+      "epoch": 1.943009641873278,
+      "grad_norm": 1.7734375,
+      "learning_rate": 0.00013459823406818374,
+      "loss": 4.6385,
+      "step": 225700
+    },
+    {
+      "epoch": 1.943440082644628,
+      "grad_norm": 3.484375,
+      "learning_rate": 0.0001345728139378793,
+      "loss": 4.3224,
+      "step": 225750
+    },
+    {
+      "epoch": 1.943870523415978,
+      "grad_norm": 3.109375,
+      "learning_rate": 0.00013454739127000465,
+      "loss": 4.242,
+      "step": 225800
+    },
+    {
+      "epoch": 1.944300964187328,
+      "grad_norm": 1.453125,
+      "learning_rate": 0.00013452196606642583,
+      "loss": 4.5361,
+      "step": 225850
+    },
+    {
+      "epoch": 1.9447314049586777,
+      "grad_norm": 3.953125,
+      "learning_rate": 0.000134496538329009,
+      "loss": 4.65,
+      "step": 225900
+    },
+    {
+      "epoch": 1.9451618457300275,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.00013447110805962046,
+      "loss": 4.4073,
+      "step": 225950
+    },
+    {
+      "epoch": 1.9455922865013773,
+      "grad_norm": 2.578125,
+      "learning_rate": 0.00013444567526012675,
+      "loss": 4.4841,
+      "step": 226000
+    },
+    {
+      "epoch": 1.9460227272727273,
+      "grad_norm": 3.125,
+      "learning_rate": 0.00013442023993239462,
+      "loss": 4.1725,
+      "step": 226050
+    },
+    {
+      "epoch": 1.946453168044077,
+      "grad_norm": 3.109375,
+      "learning_rate": 0.00013439480207829088,
+      "loss": 4.1709,
+      "step": 226100
+    },
+    {
+      "epoch": 1.946883608815427,
+      "grad_norm": 3.125,
+      "learning_rate": 0.00013436936169968272,
+      "loss": 4.2776,
+      "step": 226150
+    },
+    {
+      "epoch": 1.947314049586777,
+      "grad_norm": 4.6875,
+      "learning_rate": 0.00013434391879843733,
+      "loss": 4.5293,
+      "step": 226200
+    },
+    {
+      "epoch": 1.9477444903581267,
+      "grad_norm": 3.6875,
+      "learning_rate": 0.00013431847337642223,
+      "loss": 4.7208,
+      "step": 226250
+    },
+    {
+      "epoch": 1.9481749311294765,
+      "grad_norm": 4.625,
+      "learning_rate": 0.000134293025435505,
+      "loss": 4.4366,
+      "step": 226300
+    },
+    {
+      "epoch": 1.9486053719008265,
+      "grad_norm": 3.046875,
+      "learning_rate": 0.00013426757497755345,
+      "loss": 4.2549,
+      "step": 226350
+    },
+    {
+      "epoch": 1.9490358126721763,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00013424212200443568,
+      "loss": 4.5855,
+      "step": 226400
+    },
+    {
+      "epoch": 1.9494662534435263,
+      "grad_norm": 3.546875,
+      "learning_rate": 0.0001342166665180198,
+      "loss": 4.449,
+      "step": 226450
+    },
+    {
+      "epoch": 1.949896694214876,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00013419120852017422,
+      "loss": 4.3388,
+      "step": 226500
+    },
+    {
+      "epoch": 1.950327134986226,
+      "grad_norm": 4.1875,
+      "learning_rate": 0.0001341657480127675,
+      "loss": 4.229,
+      "step": 226550
+    },
+    {
+      "epoch": 1.9507575757575757,
+      "grad_norm": 1.7890625,
+      "learning_rate": 0.0001341402849976684,
+      "loss": 4.3056,
+      "step": 226600
+    },
+    {
+      "epoch": 1.9511880165289255,
+      "grad_norm": 0.99609375,
+      "learning_rate": 0.00013411481947674583,
+      "loss": 4.2969,
+      "step": 226650
+    },
+    {
+      "epoch": 1.9516184573002755,
+      "grad_norm": 4.09375,
+      "learning_rate": 0.0001340893514518689,
+      "loss": 4.2659,
+      "step": 226700
+    },
+    {
+      "epoch": 1.9520488980716253,
+      "grad_norm": 3.25,
+      "learning_rate": 0.0001340638809249069,
+      "loss": 4.3612,
+      "step": 226750
+    },
+    {
+      "epoch": 1.9524793388429753,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00013403840789772938,
+      "loss": 4.516,
+      "step": 226800
+    },
+    {
+      "epoch": 1.952909779614325,
+      "grad_norm": 6.53125,
+      "learning_rate": 0.00013401293237220592,
+      "loss": 4.211,
+      "step": 226850
+    },
+    {
+      "epoch": 1.953340220385675,
+      "grad_norm": 3.921875,
+      "learning_rate": 0.0001339874543502064,
+      "loss": 4.338,
+      "step": 226900
+    },
+    {
+      "epoch": 1.9537706611570247,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.0001339619738336009,
+      "loss": 4.6376,
+      "step": 226950
+    },
+    {
+      "epoch": 1.9542011019283747,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.00013393649082425955,
+      "loss": 4.4115,
+      "step": 227000
+    },
+    {
+      "epoch": 1.9546315426997245,
+      "grad_norm": 3.046875,
+      "learning_rate": 0.0001339110053240528,
+      "loss": 4.6546,
+      "step": 227050
+    },
+    {
+      "epoch": 1.9550619834710745,
+      "grad_norm": 1.2578125,
+      "learning_rate": 0.0001338855173348512,
+      "loss": 4.6077,
+      "step": 227100
+    },
+    {
+      "epoch": 1.9554924242424243,
+      "grad_norm": 4.09375,
+      "learning_rate": 0.00013386002685852557,
+      "loss": 4.2572,
+      "step": 227150
+    },
+    {
+      "epoch": 1.955922865013774,
+      "grad_norm": 2.375,
+      "learning_rate": 0.0001338345338969468,
+      "loss": 4.1396,
+      "step": 227200
+    },
+    {
+      "epoch": 1.956353305785124,
+      "grad_norm": 1.734375,
+      "learning_rate": 0.00013380903845198607,
+      "loss": 4.3685,
+      "step": 227250
+    },
+    {
+      "epoch": 1.9567837465564737,
+      "grad_norm": 1.8125,
+      "learning_rate": 0.00013378354052551462,
+      "loss": 4.4616,
+      "step": 227300
+    },
+    {
+      "epoch": 1.9572141873278237,
+      "grad_norm": 4.28125,
+      "learning_rate": 0.00013375804011940402,
+      "loss": 4.7125,
+      "step": 227350
+    },
+    {
+      "epoch": 1.9576446280991735,
+      "grad_norm": 3.09375,
+      "learning_rate": 0.00013373253723552588,
+      "loss": 4.434,
+      "step": 227400
+    },
+    {
+      "epoch": 1.9580750688705235,
+      "grad_norm": 5.46875,
+      "learning_rate": 0.0001337070318757521,
+      "loss": 4.1449,
+      "step": 227450
+    },
+    {
+      "epoch": 1.9585055096418733,
+      "grad_norm": 3.515625,
+      "learning_rate": 0.00013368152404195466,
+      "loss": 4.6336,
+      "step": 227500
+    },
+    {
+      "epoch": 1.958935950413223,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.00013365601373600586,
+      "loss": 4.3233,
+      "step": 227550
+    },
+    {
+      "epoch": 1.959366391184573,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.00013363050095977806,
+      "loss": 4.2766,
+      "step": 227600
+    },
+    {
+      "epoch": 1.959796831955923,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00013360498571514384,
+      "loss": 4.3526,
+      "step": 227650
+    },
+    {
+      "epoch": 1.9602272727272727,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.000133579468003976,
+      "loss": 4.6202,
+      "step": 227700
+    },
+    {
+      "epoch": 1.9606577134986227,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.00013355394782814742,
+      "loss": 4.5862,
+      "step": 227750
+    },
+    {
+      "epoch": 1.9610881542699725,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.0001335284251895313,
+      "loss": 4.2518,
+      "step": 227800
+    },
+    {
+      "epoch": 1.9615185950413223,
+      "grad_norm": 2.375,
+      "learning_rate": 0.00013350290009000087,
+      "loss": 4.0589,
+      "step": 227850
+    },
+    {
+      "epoch": 1.961949035812672,
+      "grad_norm": 1.4453125,
+      "learning_rate": 0.00013347737253142968,
+      "loss": 4.4233,
+      "step": 227900
+    },
+    {
+      "epoch": 1.962379476584022,
+      "grad_norm": 3.109375,
+      "learning_rate": 0.0001334518425156914,
+      "loss": 4.3156,
+      "step": 227950
+    },
+    {
+      "epoch": 1.962809917355372,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.00013342631004465978,
+      "loss": 4.3603,
+      "step": 228000
+    },
+    {
+      "epoch": 1.962809917355372,
+      "eval_loss": 5.03814697265625,
+      "eval_runtime": 21.7978,
+      "eval_samples_per_second": 29.361,
+      "eval_steps_per_second": 14.68,
+      "eval_tts_loss": 7.288137049212746,
+      "step": 228000
+    },
+    {
+      "epoch": 1.963240358126722,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.00013340077512020897,
+      "loss": 4.3022,
+      "step": 228050
+    },
+    {
+      "epoch": 1.9636707988980717,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.00013337523774421313,
+      "loss": 4.5097,
+      "step": 228100
+    },
+    {
+      "epoch": 1.9641012396694215,
+      "grad_norm": 2.265625,
+      "learning_rate": 0.00013334969791854663,
+      "loss": 4.3277,
+      "step": 228150
+    },
+    {
+      "epoch": 1.9645316804407713,
+      "grad_norm": 3.328125,
+      "learning_rate": 0.00013332415564508407,
+      "loss": 4.3447,
+      "step": 228200
+    },
+    {
+      "epoch": 1.964962121212121,
+      "grad_norm": 1.90625,
+      "learning_rate": 0.00013329861092570018,
+      "loss": 4.6091,
+      "step": 228250
+    },
+    {
+      "epoch": 1.9653925619834711,
+      "grad_norm": 2.421875,
+      "learning_rate": 0.0001332730637622699,
+      "loss": 4.4401,
+      "step": 228300
+    },
+    {
+      "epoch": 1.965823002754821,
+      "grad_norm": 3.78125,
+      "learning_rate": 0.00013324751415666832,
+      "loss": 4.6507,
+      "step": 228350
+    },
+    {
+      "epoch": 1.966253443526171,
+      "grad_norm": 4.21875,
+      "learning_rate": 0.00013322196211077074,
+      "loss": 4.6663,
+      "step": 228400
+    },
+    {
+      "epoch": 1.9666838842975207,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.00013319640762645261,
+      "loss": 4.3732,
+      "step": 228450
+    },
+    {
+      "epoch": 1.9671143250688705,
+      "grad_norm": 3.59375,
+      "learning_rate": 0.00013317085070558958,
+      "loss": 4.6632,
+      "step": 228500
+    },
+    {
+      "epoch": 1.9675447658402203,
+      "grad_norm": 4.21875,
+      "learning_rate": 0.00013314529135005748,
+      "loss": 4.3959,
+      "step": 228550
+    },
+    {
+      "epoch": 1.96797520661157,
+      "grad_norm": 2.234375,
+      "learning_rate": 0.00013311972956173235,
+      "loss": 4.5491,
+      "step": 228600
+    },
+    {
+      "epoch": 1.9684056473829201,
+      "grad_norm": 4.59375,
+      "learning_rate": 0.0001330941653424903,
+      "loss": 4.4633,
+      "step": 228650
+    },
+    {
+      "epoch": 1.9688360881542701,
+      "grad_norm": 1.609375,
+      "learning_rate": 0.0001330685986942077,
+      "loss": 4.3238,
+      "step": 228700
+    },
+    {
+      "epoch": 1.96926652892562,
+      "grad_norm": 3.046875,
+      "learning_rate": 0.00013304302961876112,
+      "loss": 4.624,
+      "step": 228750
+    },
+    {
+      "epoch": 1.9696969696969697,
+      "grad_norm": 3.578125,
+      "learning_rate": 0.00013301745811802727,
+      "loss": 4.3781,
+      "step": 228800
+    },
+    {
+      "epoch": 1.9701274104683195,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.00013299188419388302,
+      "loss": 4.391,
+      "step": 228850
+    },
+    {
+      "epoch": 1.9705578512396693,
+      "grad_norm": 3.78125,
+      "learning_rate": 0.00013296630784820546,
+      "loss": 4.3632,
+      "step": 228900
+    },
+    {
+      "epoch": 1.9709882920110193,
+      "grad_norm": 1.75,
+      "learning_rate": 0.00013294072908287183,
+      "loss": 4.452,
+      "step": 228950
+    },
+    {
+      "epoch": 1.9714187327823691,
+      "grad_norm": 1.1953125,
+      "learning_rate": 0.00013291514789975958,
+      "loss": 4.5408,
+      "step": 229000
+    },
+    {
+      "epoch": 1.9718491735537191,
+      "grad_norm": 3.109375,
+      "learning_rate": 0.00013288956430074623,
+      "loss": 4.357,
+      "step": 229050
+    },
+    {
+      "epoch": 1.972279614325069,
+      "grad_norm": 4.6875,
+      "learning_rate": 0.00013286397828770968,
+      "loss": 4.6441,
+      "step": 229100
+    },
+    {
+      "epoch": 1.9727100550964187,
+      "grad_norm": 1.6328125,
+      "learning_rate": 0.00013283838986252782,
+      "loss": 3.9945,
+      "step": 229150
+    },
+    {
+      "epoch": 1.9731404958677685,
+      "grad_norm": 4.1875,
+      "learning_rate": 0.00013281279902707877,
+      "loss": 4.6092,
+      "step": 229200
+    },
+    {
+      "epoch": 1.9735709366391183,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.0001327872057832409,
+      "loss": 4.447,
+      "step": 229250
+    },
+    {
+      "epoch": 1.9740013774104683,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.00013276161013289268,
+      "loss": 4.6184,
+      "step": 229300
+    },
+    {
+      "epoch": 1.9744318181818183,
+      "grad_norm": 1.8359375,
+      "learning_rate": 0.00013273601207791275,
+      "loss": 4.5302,
+      "step": 229350
+    },
+    {
+      "epoch": 1.9748622589531681,
+      "grad_norm": 3.625,
+      "learning_rate": 0.00013271041162017997,
+      "loss": 4.0027,
+      "step": 229400
+    },
+    {
+      "epoch": 1.975292699724518,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.00013268480876157333,
+      "loss": 4.07,
+      "step": 229450
+    },
+    {
+      "epoch": 1.9757231404958677,
+      "grad_norm": 2.5,
+      "learning_rate": 0.0001326592035039721,
+      "loss": 4.5722,
+      "step": 229500
+    },
+    {
+      "epoch": 1.9761535812672175,
+      "grad_norm": 3.125,
+      "learning_rate": 0.00013263359584925554,
+      "loss": 4.256,
+      "step": 229550
+    },
+    {
+      "epoch": 1.9765840220385675,
+      "grad_norm": 2.875,
+      "learning_rate": 0.0001326079857993033,
+      "loss": 4.0357,
+      "step": 229600
+    },
+    {
+      "epoch": 1.9770144628099173,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.00013258237335599506,
+      "loss": 4.5307,
+      "step": 229650
+    },
+    {
+      "epoch": 1.9774449035812673,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.0001325567585212107,
+      "loss": 4.2672,
+      "step": 229700
+    },
+    {
+      "epoch": 1.9778753443526171,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.00013253114129683032,
+      "loss": 4.1747,
+      "step": 229750
+    },
+    {
+      "epoch": 1.978305785123967,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00013250552168473417,
+      "loss": 4.5218,
+      "step": 229800
+    },
+    {
+      "epoch": 1.9787362258953167,
+      "grad_norm": 4.53125,
+      "learning_rate": 0.00013247989968680264,
+      "loss": 4.4838,
+      "step": 229850
+    },
+    {
+      "epoch": 1.9791666666666665,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.0001324542753049164,
+      "loss": 4.1306,
+      "step": 229900
+    },
+    {
+      "epoch": 1.9795971074380165,
+      "grad_norm": 3.53125,
+      "learning_rate": 0.00013242864854095613,
+      "loss": 4.5694,
+      "step": 229950
+    },
+    {
+      "epoch": 1.9800275482093666,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.0001324030193968029,
+      "loss": 4.3008,
+      "step": 230000
+    },
+    {
+      "epoch": 1.9804579889807163,
+      "grad_norm": 1.1484375,
+      "learning_rate": 0.00013237738787433767,
+      "loss": 4.2187,
+      "step": 230050
+    },
+    {
+      "epoch": 1.9808884297520661,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.00013235175397544185,
+      "loss": 4.4121,
+      "step": 230100
+    },
+    {
+      "epoch": 1.981318870523416,
+      "grad_norm": 4.03125,
+      "learning_rate": 0.00013232611770199693,
+      "loss": 4.1831,
+      "step": 230150
+    },
+    {
+      "epoch": 1.9817493112947657,
+      "grad_norm": 4.46875,
+      "learning_rate": 0.00013230047905588455,
+      "loss": 4.4687,
+      "step": 230200
+    },
+    {
+      "epoch": 1.9821797520661157,
+      "grad_norm": 1.3984375,
+      "learning_rate": 0.00013227483803898645,
+      "loss": 4.5879,
+      "step": 230250
+    },
+    {
+      "epoch": 1.9826101928374655,
+      "grad_norm": 3.328125,
+      "learning_rate": 0.00013224919465318468,
+      "loss": 4.6668,
+      "step": 230300
+    },
+    {
+      "epoch": 1.9830406336088156,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.00013222354890036144,
+      "loss": 4.1026,
+      "step": 230350
+    },
+    {
+      "epoch": 1.9834710743801653,
+      "grad_norm": 1.890625,
+      "learning_rate": 0.00013219790078239902,
+      "loss": 4.3291,
+      "step": 230400
+    },
+    {
+      "epoch": 1.9839015151515151,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00013217225030117996,
+      "loss": 4.6322,
+      "step": 230450
+    },
+    {
+      "epoch": 1.984331955922865,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.00013214659745858698,
+      "loss": 4.8035,
+      "step": 230500
+    },
+    {
+      "epoch": 1.9847623966942147,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.00013212094225650286,
+      "loss": 4.4787,
+      "step": 230550
+    },
+    {
+      "epoch": 1.9851928374655647,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.0001320952846968107,
+      "loss": 4.4221,
+      "step": 230600
+    },
+    {
+      "epoch": 1.9856232782369148,
+      "grad_norm": 1.796875,
+      "learning_rate": 0.00013206962478139373,
+      "loss": 4.1563,
+      "step": 230650
+    },
+    {
+      "epoch": 1.9860537190082646,
+      "grad_norm": 1.90625,
+      "learning_rate": 0.0001320439625121353,
+      "loss": 4.6338,
+      "step": 230700
+    },
+    {
+      "epoch": 1.9864841597796143,
+      "grad_norm": 1.5078125,
+      "learning_rate": 0.00013201829789091893,
+      "loss": 4.6254,
+      "step": 230750
+    },
+    {
+      "epoch": 1.9869146005509641,
+      "grad_norm": 4.65625,
+      "learning_rate": 0.00013199263091962838,
+      "loss": 4.6582,
+      "step": 230800
+    },
+    {
+      "epoch": 1.987345041322314,
+      "grad_norm": 1.0234375,
+      "learning_rate": 0.0001319669616001476,
+      "loss": 4.4411,
+      "step": 230850
+    },
+    {
+      "epoch": 1.987775482093664,
+      "grad_norm": 3.65625,
+      "learning_rate": 0.00013194128993436062,
+      "loss": 4.4935,
+      "step": 230900
+    },
+    {
+      "epoch": 1.9882059228650137,
+      "grad_norm": 3.21875,
+      "learning_rate": 0.00013191561592415163,
+      "loss": 4.2984,
+      "step": 230950
+    },
+    {
+      "epoch": 1.9886363636363638,
+      "grad_norm": 3.296875,
+      "learning_rate": 0.00013188993957140513,
+      "loss": 4.3819,
+      "step": 231000
+    },
+    {
+      "epoch": 1.9886363636363638,
+      "eval_loss": 5.036816120147705,
+      "eval_runtime": 21.8132,
+      "eval_samples_per_second": 29.34,
+      "eval_steps_per_second": 14.67,
+      "eval_tts_loss": 7.275183283766534,
+      "step": 231000
+    },
+    {
+      "epoch": 1.9890668044077136,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.00013186426087800568,
+      "loss": 4.5505,
+      "step": 231050
+    },
+    {
+      "epoch": 1.9894972451790633,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00013183857984583802,
+      "loss": 4.0794,
+      "step": 231100
+    },
+    {
+      "epoch": 1.9899276859504131,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.00013181289647678712,
+      "loss": 4.3185,
+      "step": 231150
+    },
+    {
+      "epoch": 1.990358126721763,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.00013178721077273807,
+      "loss": 4.5574,
+      "step": 231200
+    },
+    {
+      "epoch": 1.990788567493113,
+      "grad_norm": 3.78125,
+      "learning_rate": 0.00013176152273557609,
+      "loss": 4.4835,
+      "step": 231250
+    },
+    {
+      "epoch": 1.991219008264463,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00013173583236718674,
+      "loss": 4.383,
+      "step": 231300
+    },
+    {
+      "epoch": 1.9916494490358128,
+      "grad_norm": 2.984375,
+      "learning_rate": 0.00013171013966945552,
+      "loss": 4.5291,
+      "step": 231350
+    },
+    {
+      "epoch": 1.9920798898071626,
+      "grad_norm": 1.3984375,
+      "learning_rate": 0.00013168444464426834,
+      "loss": 4.2791,
+      "step": 231400
+    },
+    {
+      "epoch": 1.9925103305785123,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00013165874729351105,
+      "loss": 4.3751,
+      "step": 231450
+    },
+    {
+      "epoch": 1.9929407713498621,
+      "grad_norm": 2.234375,
+      "learning_rate": 0.00013163304761906983,
+      "loss": 4.714,
+      "step": 231500
+    },
+    {
+      "epoch": 1.9933712121212122,
+      "grad_norm": 4.09375,
+      "learning_rate": 0.000131607345622831,
+      "loss": 4.8913,
+      "step": 231550
+    },
+    {
+      "epoch": 1.993801652892562,
+      "grad_norm": 3.453125,
+      "learning_rate": 0.000131581641306681,
+      "loss": 4.1482,
+      "step": 231600
+    },
+    {
+      "epoch": 1.994232093663912,
+      "grad_norm": 4.46875,
+      "learning_rate": 0.00013155593467250643,
+      "loss": 4.7114,
+      "step": 231650
+    },
+    {
+      "epoch": 1.9946625344352618,
+      "grad_norm": 4.0625,
+      "learning_rate": 0.0001315302257221942,
+      "loss": 4.5604,
+      "step": 231700
+    },
+    {
+      "epoch": 1.9950929752066116,
+      "grad_norm": 3.5,
+      "learning_rate": 0.00013150451445763123,
+      "loss": 4.2361,
+      "step": 231750
+    },
+    {
+      "epoch": 1.9955234159779613,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.0001314788008807047,
+      "loss": 4.4734,
+      "step": 231800
+    },
+    {
+      "epoch": 1.9959538567493111,
+      "grad_norm": 2.875,
+      "learning_rate": 0.00013145308499330186,
+      "loss": 4.2232,
+      "step": 231850
+    },
+    {
+      "epoch": 1.9963842975206612,
+      "grad_norm": 1.4296875,
+      "learning_rate": 0.0001314273667973103,
+      "loss": 4.4652,
+      "step": 231900
+    },
+    {
+      "epoch": 1.9968147382920112,
+      "grad_norm": 3.046875,
+      "learning_rate": 0.00013140164629461763,
+      "loss": 4.4417,
+      "step": 231950
+    },
+    {
+      "epoch": 1.997245179063361,
+      "grad_norm": 3.171875,
+      "learning_rate": 0.0001313759234871117,
+      "loss": 4.5901,
+      "step": 232000
+    },
+    {
+      "epoch": 1.9976756198347108,
+      "grad_norm": 4.09375,
+      "learning_rate": 0.00013135019837668045,
+      "loss": 4.3243,
+      "step": 232050
+    },
+    {
+      "epoch": 1.9981060606060606,
+      "grad_norm": 3.875,
+      "learning_rate": 0.00013132447096521213,
+      "loss": 4.2767,
+      "step": 232100
+    },
+    {
+      "epoch": 1.9985365013774103,
+      "grad_norm": 3.234375,
+      "learning_rate": 0.000131298741254595,
+      "loss": 4.7821,
+      "step": 232150
+    },
+    {
+      "epoch": 1.9989669421487604,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00013127300924671762,
+      "loss": 4.6669,
+      "step": 232200
+    },
+    {
+      "epoch": 1.9993973829201102,
+      "grad_norm": 4.03125,
+      "learning_rate": 0.00013124727494346865,
+      "loss": 4.0649,
+      "step": 232250
+    },
+    {
+      "epoch": 1.9998278236914602,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.00013122153834673692,
+      "loss": 4.6496,
+      "step": 232300
+    },
+    {
+      "epoch": 2.00025826446281,
+      "grad_norm": 1.9609375,
+      "learning_rate": 0.00013119579945841144,
+      "loss": 4.3524,
+      "step": 232350
+    },
+    {
+      "epoch": 2.0006887052341598,
+      "grad_norm": 3.296875,
+      "learning_rate": 0.00013117005828038142,
+      "loss": 4.4407,
+      "step": 232400
+    },
+    {
+      "epoch": 2.0011191460055096,
+      "grad_norm": 3.921875,
+      "learning_rate": 0.0001311443148145362,
+      "loss": 4.6714,
+      "step": 232450
+    },
+    {
+      "epoch": 2.0015495867768593,
+      "grad_norm": 3.46875,
+      "learning_rate": 0.00013111856906276525,
+      "loss": 4.2322,
+      "step": 232500
+    },
+    {
+      "epoch": 2.0019800275482096,
+      "grad_norm": 0.703125,
+      "learning_rate": 0.00013109282102695833,
+      "loss": 4.5528,
+      "step": 232550
+    },
+    {
+      "epoch": 2.0024104683195594,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.0001310670707090052,
+      "loss": 4.3319,
+      "step": 232600
+    },
+    {
+      "epoch": 2.002840909090909,
+      "grad_norm": 1.734375,
+      "learning_rate": 0.0001310413181107959,
+      "loss": 4.4109,
+      "step": 232650
+    },
+    {
+      "epoch": 2.003271349862259,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00013101556323422072,
+      "loss": 4.5476,
+      "step": 232700
+    },
+    {
+      "epoch": 2.0037017906336088,
+      "grad_norm": 3.984375,
+      "learning_rate": 0.00013098980608116986,
+      "loss": 4.3996,
+      "step": 232750
+    },
+    {
+      "epoch": 2.0041322314049586,
+      "grad_norm": 2.296875,
+      "learning_rate": 0.00013096404665353392,
+      "loss": 4.4785,
+      "step": 232800
+    },
+    {
+      "epoch": 2.0045626721763083,
+      "grad_norm": 3.84375,
+      "learning_rate": 0.00013093828495320362,
+      "loss": 4.2747,
+      "step": 232850
+    },
+    {
+      "epoch": 2.0049931129476586,
+      "grad_norm": 3.015625,
+      "learning_rate": 0.00013091252098206975,
+      "loss": 4.3942,
+      "step": 232900
+    },
+    {
+      "epoch": 2.0054235537190084,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.00013088675474202333,
+      "loss": 3.7228,
+      "step": 232950
+    },
+    {
+      "epoch": 2.005853994490358,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.00013086098623495555,
+      "loss": 4.5769,
+      "step": 233000
+    },
+    {
+      "epoch": 2.006284435261708,
+      "grad_norm": 1.359375,
+      "learning_rate": 0.00013083521546275781,
+      "loss": 4.2394,
+      "step": 233050
+    },
+    {
+      "epoch": 2.0067148760330578,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.00013080944242732163,
+      "loss": 4.2471,
+      "step": 233100
+    },
+    {
+      "epoch": 2.0071453168044076,
+      "grad_norm": 2.0,
+      "learning_rate": 0.0001307836671305386,
+      "loss": 4.3415,
+      "step": 233150
+    },
+    {
+      "epoch": 2.007575757575758,
+      "grad_norm": 3.140625,
+      "learning_rate": 0.00013075788957430065,
+      "loss": 4.6546,
+      "step": 233200
+    },
+    {
+      "epoch": 2.0080061983471076,
+      "grad_norm": 3.4375,
+      "learning_rate": 0.00013073210976049984,
+      "loss": 4.2723,
+      "step": 233250
+    },
+    {
+      "epoch": 2.0084366391184574,
+      "grad_norm": 3.1875,
+      "learning_rate": 0.00013070632769102822,
+      "loss": 4.6354,
+      "step": 233300
+    },
+    {
+      "epoch": 2.008867079889807,
+      "grad_norm": 3.96875,
+      "learning_rate": 0.00013068054336777825,
+      "loss": 4.5433,
+      "step": 233350
+    },
+    {
+      "epoch": 2.009297520661157,
+      "grad_norm": 3.15625,
+      "learning_rate": 0.00013065475679264237,
+      "loss": 4.3098,
+      "step": 233400
+    },
+    {
+      "epoch": 2.0097279614325068,
+      "grad_norm": 4.40625,
+      "learning_rate": 0.00013062896796751336,
+      "loss": 4.2104,
+      "step": 233450
+    },
+    {
+      "epoch": 2.0101584022038566,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.000130603176894284,
+      "loss": 4.7517,
+      "step": 233500
+    },
+    {
+      "epoch": 2.010588842975207,
+      "grad_norm": 5.0625,
+      "learning_rate": 0.00013057738357484723,
+      "loss": 4.4149,
+      "step": 233550
+    },
+    {
+      "epoch": 2.0110192837465566,
+      "grad_norm": 5.75,
+      "learning_rate": 0.00013055158801109635,
+      "loss": 4.2466,
+      "step": 233600
+    },
+    {
+      "epoch": 2.0114497245179064,
+      "grad_norm": 2.421875,
+      "learning_rate": 0.00013052579020492465,
+      "loss": 4.4027,
+      "step": 233650
+    },
+    {
+      "epoch": 2.011880165289256,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.00013049999015822558,
+      "loss": 4.196,
+      "step": 233700
+    },
+    {
+      "epoch": 2.012310606060606,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00013047418787289286,
+      "loss": 4.2452,
+      "step": 233750
+    },
+    {
+      "epoch": 2.0127410468319558,
+      "grad_norm": 4.53125,
+      "learning_rate": 0.00013044838335082038,
+      "loss": 4.4962,
+      "step": 233800
+    },
+    {
+      "epoch": 2.013171487603306,
+      "grad_norm": 1.7890625,
+      "learning_rate": 0.000130422576593902,
+      "loss": 4.1795,
+      "step": 233850
+    },
+    {
+      "epoch": 2.013601928374656,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.00013039676760403198,
+      "loss": 4.1406,
+      "step": 233900
+    },
+    {
+      "epoch": 2.0140323691460056,
+      "grad_norm": 2.265625,
+      "learning_rate": 0.00013037095638310464,
+      "loss": 4.5681,
+      "step": 233950
+    },
+    {
+      "epoch": 2.0144628099173554,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.00013034514293301446,
+      "loss": 4.1719,
+      "step": 234000
+    },
+    {
+      "epoch": 2.0144628099173554,
+      "eval_loss": 5.037001609802246,
+      "eval_runtime": 22.0979,
+      "eval_samples_per_second": 28.962,
+      "eval_steps_per_second": 14.481,
+      "eval_tts_loss": 7.315695545191628,
+      "step": 234000
+    },
+    {
+      "epoch": 2.014893250688705,
+      "grad_norm": 3.15625,
+      "learning_rate": 0.00013031932725565603,
+      "loss": 4.1027,
+      "step": 234050
+    },
+    {
+      "epoch": 2.015323691460055,
+      "grad_norm": 3.578125,
+      "learning_rate": 0.00013029350935292423,
+      "loss": 4.6561,
+      "step": 234100
+    },
+    {
+      "epoch": 2.0157541322314048,
+      "grad_norm": 1.9609375,
+      "learning_rate": 0.00013026768922671402,
+      "loss": 4.5103,
+      "step": 234150
+    },
+    {
+      "epoch": 2.016184573002755,
+      "grad_norm": 5.96875,
+      "learning_rate": 0.00013024186687892052,
+      "loss": 4.2305,
+      "step": 234200
+    },
+    {
+      "epoch": 2.016615013774105,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.0001302160423114391,
+      "loss": 3.971,
+      "step": 234250
+    },
+    {
+      "epoch": 2.0170454545454546,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.0001301902155261652,
+      "loss": 4.8936,
+      "step": 234300
+    },
+    {
+      "epoch": 2.0174758953168044,
+      "grad_norm": 4.625,
+      "learning_rate": 0.00013016438652499438,
+      "loss": 4.6421,
+      "step": 234350
+    },
+    {
+      "epoch": 2.017906336088154,
+      "grad_norm": 4.71875,
+      "learning_rate": 0.00013013855530982253,
+      "loss": 4.1621,
+      "step": 234400
+    },
+    {
+      "epoch": 2.018336776859504,
+      "grad_norm": 2.125,
+      "learning_rate": 0.00013011272188254557,
+      "loss": 4.1475,
+      "step": 234450
+    },
+    {
+      "epoch": 2.018767217630854,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.00013008688624505959,
+      "loss": 4.0935,
+      "step": 234500
+    },
+    {
+      "epoch": 2.019197658402204,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.00013006104839926086,
+      "loss": 4.6147,
+      "step": 234550
+    },
+    {
+      "epoch": 2.019628099173554,
+      "grad_norm": 4.15625,
+      "learning_rate": 0.0001300352083470459,
+      "loss": 4.7151,
+      "step": 234600
+    },
+    {
+      "epoch": 2.0200585399449036,
+      "grad_norm": 1.75,
+      "learning_rate": 0.0001300093660903113,
+      "loss": 4.2808,
+      "step": 234650
+    },
+    {
+      "epoch": 2.0204889807162534,
+      "grad_norm": 3.78125,
+      "learning_rate": 0.00012998352163095373,
+      "loss": 4.1704,
+      "step": 234700
+    },
+    {
+      "epoch": 2.020919421487603,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.0001299576749708702,
+      "loss": 4.2644,
+      "step": 234750
+    },
+    {
+      "epoch": 2.021349862258953,
+      "grad_norm": 1.5234375,
+      "learning_rate": 0.00012993182611195783,
+      "loss": 4.1252,
+      "step": 234800
+    },
+    {
+      "epoch": 2.021780303030303,
+      "grad_norm": 3.46875,
+      "learning_rate": 0.00012990597505611374,
+      "loss": 4.6791,
+      "step": 234850
+    },
+    {
+      "epoch": 2.022210743801653,
+      "grad_norm": 5.03125,
+      "learning_rate": 0.0001298801218052355,
+      "loss": 4.3523,
+      "step": 234900
+    },
+    {
+      "epoch": 2.022641184573003,
+      "grad_norm": 3.921875,
+      "learning_rate": 0.00012985426636122056,
+      "loss": 4.272,
+      "step": 234950
+    },
+    {
+      "epoch": 2.0230716253443526,
+      "grad_norm": 1.5,
+      "learning_rate": 0.00012982840872596675,
+      "loss": 4.784,
+      "step": 235000
+    },
+    {
+      "epoch": 2.0235020661157024,
+      "grad_norm": 3.75,
+      "learning_rate": 0.00012980254890137188,
+      "loss": 4.5586,
+      "step": 235050
+    },
+    {
+      "epoch": 2.023932506887052,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00012977668688933405,
+      "loss": 4.5532,
+      "step": 235100
+    },
+    {
+      "epoch": 2.0243629476584024,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.00012975082269175148,
+      "loss": 4.5497,
+      "step": 235150
+    },
+    {
+      "epoch": 2.024793388429752,
+      "grad_norm": 3.21875,
+      "learning_rate": 0.00012972495631052253,
+      "loss": 4.0533,
+      "step": 235200
+    },
+    {
+      "epoch": 2.025223829201102,
+      "grad_norm": 1.734375,
+      "learning_rate": 0.00012969908774754574,
+      "loss": 4.3129,
+      "step": 235250
+    },
+    {
+      "epoch": 2.025654269972452,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.0001296732170047198,
+      "loss": 4.37,
+      "step": 235300
+    },
+    {
+      "epoch": 2.0260847107438016,
+      "grad_norm": 3.390625,
+      "learning_rate": 0.0001296473440839436,
+      "loss": 4.4195,
+      "step": 235350
+    },
+    {
+      "epoch": 2.0265151515151514,
+      "grad_norm": 3.921875,
+      "learning_rate": 0.0001296214689871161,
+      "loss": 4.2989,
+      "step": 235400
+    },
+    {
+      "epoch": 2.026945592286501,
+      "grad_norm": 4.34375,
+      "learning_rate": 0.00012959559171613652,
+      "loss": 4.688,
+      "step": 235450
+    },
+    {
+      "epoch": 2.0273760330578514,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00012956971227290418,
+      "loss": 4.5615,
+      "step": 235500
+    },
+    {
+      "epoch": 2.027806473829201,
+      "grad_norm": 3.53125,
+      "learning_rate": 0.0001295438306593186,
+      "loss": 4.2132,
+      "step": 235550
+    },
+    {
+      "epoch": 2.028236914600551,
+      "grad_norm": 2.21875,
+      "learning_rate": 0.00012951794687727938,
+      "loss": 4.336,
+      "step": 235600
+    },
+    {
+      "epoch": 2.028667355371901,
+      "grad_norm": 1.953125,
+      "learning_rate": 0.0001294920609286864,
+      "loss": 4.3663,
+      "step": 235650
+    },
+    {
+      "epoch": 2.0290977961432506,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.0001294661728154396,
+      "loss": 4.6438,
+      "step": 235700
+    },
+    {
+      "epoch": 2.0295282369146004,
+      "grad_norm": 2.984375,
+      "learning_rate": 0.0001294402825394391,
+      "loss": 4.5965,
+      "step": 235750
+    },
+    {
+      "epoch": 2.0299586776859506,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.00012941439010258522,
+      "loss": 4.0466,
+      "step": 235800
+    },
+    {
+      "epoch": 2.0303891184573004,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.00012938849550677838,
+      "loss": 4.269,
+      "step": 235850
+    },
+    {
+      "epoch": 2.03081955922865,
+      "grad_norm": 3.046875,
+      "learning_rate": 0.00012936259875391922,
+      "loss": 4.2691,
+      "step": 235900
+    },
+    {
+      "epoch": 2.03125,
+      "grad_norm": 3.421875,
+      "learning_rate": 0.00012933669984590846,
+      "loss": 4.3433,
+      "step": 235950
+    },
+    {
+      "epoch": 2.03168044077135,
+      "grad_norm": 3.171875,
+      "learning_rate": 0.00012931079878464706,
+      "loss": 4.4711,
+      "step": 236000
+    },
+    {
+      "epoch": 2.0321108815426996,
+      "grad_norm": 4.21875,
+      "learning_rate": 0.0001292848955720361,
+      "loss": 4.4537,
+      "step": 236050
+    },
+    {
+      "epoch": 2.0325413223140494,
+      "grad_norm": 1.796875,
+      "learning_rate": 0.0001292589902099768,
+      "loss": 4.1053,
+      "step": 236100
+    },
+    {
+      "epoch": 2.0329717630853996,
+      "grad_norm": 3.734375,
+      "learning_rate": 0.0001292330827003706,
+      "loss": 4.6452,
+      "step": 236150
+    },
+    {
+      "epoch": 2.0334022038567494,
+      "grad_norm": 4.15625,
+      "learning_rate": 0.000129207173045119,
+      "loss": 4.4227,
+      "step": 236200
+    },
+    {
+      "epoch": 2.033832644628099,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.00012918126124612374,
+      "loss": 4.2074,
+      "step": 236250
+    },
+    {
+      "epoch": 2.034263085399449,
+      "grad_norm": 3.203125,
+      "learning_rate": 0.00012915534730528668,
+      "loss": 4.4036,
+      "step": 236300
+    },
+    {
+      "epoch": 2.034693526170799,
+      "grad_norm": 1.6484375,
+      "learning_rate": 0.0001291294312245099,
+      "loss": 4.1443,
+      "step": 236350
+    },
+    {
+      "epoch": 2.0351239669421486,
+      "grad_norm": 3.4375,
+      "learning_rate": 0.00012910351300569547,
+      "loss": 4.5357,
+      "step": 236400
+    },
+    {
+      "epoch": 2.035554407713499,
+      "grad_norm": 3.8125,
+      "learning_rate": 0.00012907759265074587,
+      "loss": 4.3159,
+      "step": 236450
+    },
+    {
+      "epoch": 2.0359848484848486,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.0001290516701615635,
+      "loss": 4.1391,
+      "step": 236500
+    },
+    {
+      "epoch": 2.0364152892561984,
+      "grad_norm": 4.40625,
+      "learning_rate": 0.00012902574554005106,
+      "loss": 4.4062,
+      "step": 236550
+    },
+    {
+      "epoch": 2.036845730027548,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.00012899981878811134,
+      "loss": 4.3044,
+      "step": 236600
+    },
+    {
+      "epoch": 2.037276170798898,
+      "grad_norm": 2.625,
+      "learning_rate": 0.0001289738899076473,
+      "loss": 4.2948,
+      "step": 236650
+    },
+    {
+      "epoch": 2.037706611570248,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.00012894795890056211,
+      "loss": 4.7039,
+      "step": 236700
+    },
+    {
+      "epoch": 2.0381370523415976,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.000128922025768759,
+      "loss": 4.5328,
+      "step": 236750
+    },
+    {
+      "epoch": 2.038567493112948,
+      "grad_norm": 3.453125,
+      "learning_rate": 0.00012889609051414144,
+      "loss": 4.2461,
+      "step": 236800
+    },
+    {
+      "epoch": 2.0389979338842976,
+      "grad_norm": 4.75,
+      "learning_rate": 0.00012887015313861301,
+      "loss": 4.005,
+      "step": 236850
+    },
+    {
+      "epoch": 2.0394283746556474,
+      "grad_norm": 4.5625,
+      "learning_rate": 0.00012884421364407745,
+      "loss": 4.2708,
+      "step": 236900
+    },
+    {
+      "epoch": 2.039858815426997,
+      "grad_norm": 3.078125,
+      "learning_rate": 0.00012881827203243865,
+      "loss": 4.4036,
+      "step": 236950
+    },
+    {
+      "epoch": 2.040289256198347,
+      "grad_norm": 2.296875,
+      "learning_rate": 0.0001287923283056007,
+      "loss": 4.6053,
+      "step": 237000
+    },
+    {
+      "epoch": 2.040289256198347,
+      "eval_loss": 5.032306671142578,
+      "eval_runtime": 21.7709,
+      "eval_samples_per_second": 29.397,
+      "eval_steps_per_second": 14.699,
+      "eval_tts_loss": 7.360729168440964,
+      "step": 237000
+    },
+    {
+      "epoch": 2.040719696969697,
+      "grad_norm": 2.9375,
+      "learning_rate": 0.00012876638246546782,
+      "loss": 4.4661,
+      "step": 237050
+    },
+    {
+      "epoch": 2.041150137741047,
+      "grad_norm": 3.09375,
+      "learning_rate": 0.00012874043451394432,
+      "loss": 4.5194,
+      "step": 237100
+    },
+    {
+      "epoch": 2.041580578512397,
+      "grad_norm": 3.296875,
+      "learning_rate": 0.0001287144844529348,
+      "loss": 4.674,
+      "step": 237150
+    },
+    {
+      "epoch": 2.0420110192837466,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00012868853228434387,
+      "loss": 4.6751,
+      "step": 237200
+    },
+    {
+      "epoch": 2.0424414600550964,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.00012866257801007644,
+      "loss": 4.5797,
+      "step": 237250
+    },
+    {
+      "epoch": 2.042871900826446,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.0001286366216320374,
+      "loss": 4.3509,
+      "step": 237300
+    },
+    {
+      "epoch": 2.043302341597796,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.00012861066315213198,
+      "loss": 4.2642,
+      "step": 237350
+    },
+    {
+      "epoch": 2.043732782369146,
+      "grad_norm": 3.34375,
+      "learning_rate": 0.00012858470257226542,
+      "loss": 4.7571,
+      "step": 237400
+    },
+    {
+      "epoch": 2.044163223140496,
+      "grad_norm": 6.125,
+      "learning_rate": 0.00012855873989434325,
+      "loss": 3.9321,
+      "step": 237450
+    },
+    {
+      "epoch": 2.044593663911846,
+      "grad_norm": 1.328125,
+      "learning_rate": 0.00012853277512027094,
+      "loss": 4.1185,
+      "step": 237500
+    },
+    {
+      "epoch": 2.0450241046831956,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00012850680825195437,
+      "loss": 4.7012,
+      "step": 237550
+    },
+    {
+      "epoch": 2.0454545454545454,
+      "grad_norm": 1.5390625,
+      "learning_rate": 0.0001284808392912994,
+      "loss": 4.3283,
+      "step": 237600
+    },
+    {
+      "epoch": 2.045884986225895,
+      "grad_norm": 1.7265625,
+      "learning_rate": 0.00012845486824021208,
+      "loss": 4.4183,
+      "step": 237650
+    },
+    {
+      "epoch": 2.046315426997245,
+      "grad_norm": 3.28125,
+      "learning_rate": 0.00012842889510059871,
+      "loss": 4.6658,
+      "step": 237700
+    },
+    {
+      "epoch": 2.0467458677685952,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.00012840291987436557,
+      "loss": 4.4295,
+      "step": 237750
+    },
+    {
+      "epoch": 2.047176308539945,
+      "grad_norm": 1.84375,
+      "learning_rate": 0.0001283769425634192,
+      "loss": 4.0736,
+      "step": 237800
+    },
+    {
+      "epoch": 2.047606749311295,
+      "grad_norm": 3.46875,
+      "learning_rate": 0.00012835096316966631,
+      "loss": 4.2706,
+      "step": 237850
+    },
+    {
+      "epoch": 2.0480371900826446,
+      "grad_norm": 3.484375,
+      "learning_rate": 0.00012832498169501375,
+      "loss": 4.3896,
+      "step": 237900
+    },
+    {
+      "epoch": 2.0484676308539944,
+      "grad_norm": 3.3125,
+      "learning_rate": 0.00012829899814136846,
+      "loss": 4.7148,
+      "step": 237950
+    },
+    {
+      "epoch": 2.048898071625344,
+      "grad_norm": 1.1953125,
+      "learning_rate": 0.00012827301251063758,
+      "loss": 4.3816,
+      "step": 238000
+    },
+    {
+      "epoch": 2.049328512396694,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.00012824702480472846,
+      "loss": 4.5911,
+      "step": 238050
+    },
+    {
+      "epoch": 2.0497589531680442,
+      "grad_norm": 4.25,
+      "learning_rate": 0.00012822103502554847,
+      "loss": 4.0037,
+      "step": 238100
+    },
+    {
+      "epoch": 2.050189393939394,
+      "grad_norm": 3.421875,
+      "learning_rate": 0.00012819504317500519,
+      "loss": 4.3853,
+      "step": 238150
+    },
+    {
+      "epoch": 2.050619834710744,
+      "grad_norm": 1.359375,
+      "learning_rate": 0.0001281690492550064,
+      "loss": 4.4044,
+      "step": 238200
+    },
+    {
+      "epoch": 2.0510502754820936,
+      "grad_norm": 6.15625,
+      "learning_rate": 0.00012814305326746007,
+      "loss": 4.6058,
+      "step": 238250
+    },
+    {
+      "epoch": 2.0514807162534434,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.00012811705521427415,
+      "loss": 4.4907,
+      "step": 238300
+    },
+    {
+      "epoch": 2.051911157024793,
+      "grad_norm": 3.671875,
+      "learning_rate": 0.00012809105509735687,
+      "loss": 4.401,
+      "step": 238350
+    },
+    {
+      "epoch": 2.0523415977961434,
+      "grad_norm": 3.921875,
+      "learning_rate": 0.0001280650529186166,
+      "loss": 4.2815,
+      "step": 238400
+    },
+    {
+      "epoch": 2.0527720385674932,
+      "grad_norm": 3.40625,
+      "learning_rate": 0.00012803904867996182,
+      "loss": 4.486,
+      "step": 238450
+    },
+    {
+      "epoch": 2.053202479338843,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.0001280130423833012,
+      "loss": 4.2161,
+      "step": 238500
+    },
+    {
+      "epoch": 2.053632920110193,
+      "grad_norm": 1.7109375,
+      "learning_rate": 0.00012798703403054347,
+      "loss": 4.0947,
+      "step": 238550
+    },
+    {
+      "epoch": 2.0540633608815426,
+      "grad_norm": 5.21875,
+      "learning_rate": 0.00012796102362359776,
+      "loss": 4.047,
+      "step": 238600
+    },
+    {
+      "epoch": 2.0544938016528924,
+      "grad_norm": 3.75,
+      "learning_rate": 0.00012793501116437303,
+      "loss": 4.5175,
+      "step": 238650
+    },
+    {
+      "epoch": 2.054924242424242,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.00012790899665477857,
+      "loss": 4.5597,
+      "step": 238700
+    },
+    {
+      "epoch": 2.0553546831955924,
+      "grad_norm": 1.625,
+      "learning_rate": 0.00012788298009672387,
+      "loss": 4.3696,
+      "step": 238750
+    },
+    {
+      "epoch": 2.0557851239669422,
+      "grad_norm": 3.09375,
+      "learning_rate": 0.00012785696149211835,
+      "loss": 4.3517,
+      "step": 238800
+    },
+    {
+      "epoch": 2.056215564738292,
+      "grad_norm": 1.7265625,
+      "learning_rate": 0.0001278309408428718,
+      "loss": 3.9974,
+      "step": 238850
+    },
+    {
+      "epoch": 2.056646005509642,
+      "grad_norm": 5.03125,
+      "learning_rate": 0.0001278049181508941,
+      "loss": 4.7621,
+      "step": 238900
+    },
+    {
+      "epoch": 2.0570764462809916,
+      "grad_norm": 1.8984375,
+      "learning_rate": 0.00012777889341809522,
+      "loss": 4.3698,
+      "step": 238950
+    },
+    {
+      "epoch": 2.0575068870523414,
+      "grad_norm": 4.21875,
+      "learning_rate": 0.00012775286664638532,
+      "loss": 4.456,
+      "step": 239000
+    },
+    {
+      "epoch": 2.0579373278236917,
+      "grad_norm": 1.015625,
+      "learning_rate": 0.00012772683783767474,
+      "loss": 4.093,
+      "step": 239050
+    },
+    {
+      "epoch": 2.0583677685950414,
+      "grad_norm": 3.734375,
+      "learning_rate": 0.0001277008069938739,
+      "loss": 4.5206,
+      "step": 239100
+    },
+    {
+      "epoch": 2.0587982093663912,
+      "grad_norm": 3.140625,
+      "learning_rate": 0.00012767477411689346,
+      "loss": 3.8517,
+      "step": 239150
+    },
+    {
+      "epoch": 2.059228650137741,
+      "grad_norm": 3.21875,
+      "learning_rate": 0.00012764873920864412,
+      "loss": 4.3505,
+      "step": 239200
+    },
+    {
+      "epoch": 2.059659090909091,
+      "grad_norm": 3.046875,
+      "learning_rate": 0.0001276227022710368,
+      "loss": 4.3288,
+      "step": 239250
+    },
+    {
+      "epoch": 2.0600895316804406,
+      "grad_norm": 3.359375,
+      "learning_rate": 0.00012759666330598257,
+      "loss": 4.4192,
+      "step": 239300
+    },
+    {
+      "epoch": 2.0605199724517904,
+      "grad_norm": 4.3125,
+      "learning_rate": 0.00012757062231539263,
+      "loss": 4.2234,
+      "step": 239350
+    },
+    {
+      "epoch": 2.0609504132231407,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00012754457930117834,
+      "loss": 4.259,
+      "step": 239400
+    },
+    {
+      "epoch": 2.0613808539944904,
+      "grad_norm": 2.140625,
+      "learning_rate": 0.00012751853426525117,
+      "loss": 4.2679,
+      "step": 239450
+    },
+    {
+      "epoch": 2.0618112947658402,
+      "grad_norm": 4.25,
+      "learning_rate": 0.00012749248720952285,
+      "loss": 4.2597,
+      "step": 239500
+    },
+    {
+      "epoch": 2.06224173553719,
+      "grad_norm": 3.765625,
+      "learning_rate": 0.00012746643813590507,
+      "loss": 4.5196,
+      "step": 239550
+    },
+    {
+      "epoch": 2.06267217630854,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00012744038704630984,
+      "loss": 4.4117,
+      "step": 239600
+    },
+    {
+      "epoch": 2.0631026170798896,
+      "grad_norm": 3.78125,
+      "learning_rate": 0.00012741433394264923,
+      "loss": 4.2253,
+      "step": 239650
+    },
+    {
+      "epoch": 2.06353305785124,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.0001273882788268355,
+      "loss": 4.6198,
+      "step": 239700
+    },
+    {
+      "epoch": 2.0639634986225897,
+      "grad_norm": 3.359375,
+      "learning_rate": 0.00012736222170078103,
+      "loss": 4.7058,
+      "step": 239750
+    },
+    {
+      "epoch": 2.0643939393939394,
+      "grad_norm": 4.28125,
+      "learning_rate": 0.0001273361625663984,
+      "loss": 4.3315,
+      "step": 239800
+    },
+    {
+      "epoch": 2.0648243801652892,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.00012731010142560015,
+      "loss": 3.9417,
+      "step": 239850
+    },
+    {
+      "epoch": 2.065254820936639,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.00012728403828029928,
+      "loss": 4.8033,
+      "step": 239900
+    },
+    {
+      "epoch": 2.065685261707989,
+      "grad_norm": 1.6796875,
+      "learning_rate": 0.00012725797313240873,
+      "loss": 4.5976,
+      "step": 239950
+    },
+    {
+      "epoch": 2.0661157024793386,
+      "grad_norm": 4.5625,
+      "learning_rate": 0.00012723190598384155,
+      "loss": 4.4649,
+      "step": 240000
+    },
+    {
+      "epoch": 2.0661157024793386,
+      "eval_loss": 5.033450126647949,
+      "eval_runtime": 21.8456,
+      "eval_samples_per_second": 29.296,
+      "eval_steps_per_second": 14.648,
+      "eval_tts_loss": 7.34856754614598,
+      "step": 240000
+    },
+    {
+      "epoch": 2.066546143250689,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.00012720583683651108,
+      "loss": 4.1617,
+      "step": 240050
+    },
+    {
+      "epoch": 2.0669765840220387,
+      "grad_norm": 1.203125,
+      "learning_rate": 0.00012717976569233074,
+      "loss": 4.3832,
+      "step": 240100
+    },
+    {
+      "epoch": 2.0674070247933884,
+      "grad_norm": 3.078125,
+      "learning_rate": 0.00012715369255321404,
+      "loss": 4.5626,
+      "step": 240150
+    },
+    {
+      "epoch": 2.0678374655647382,
+      "grad_norm": 2.9375,
+      "learning_rate": 0.00012712761742107476,
+      "loss": 4.1653,
+      "step": 240200
+    },
+    {
+      "epoch": 2.068267906336088,
+      "grad_norm": 2.234375,
+      "learning_rate": 0.0001271015402978267,
+      "loss": 4.2125,
+      "step": 240250
+    },
+    {
+      "epoch": 2.068698347107438,
+      "grad_norm": 3.453125,
+      "learning_rate": 0.00012707546118538395,
+      "loss": 4.2864,
+      "step": 240300
+    },
+    {
+      "epoch": 2.069128787878788,
+      "grad_norm": 1.8359375,
+      "learning_rate": 0.00012704938008566055,
+      "loss": 4.1236,
+      "step": 240350
+    },
+    {
+      "epoch": 2.069559228650138,
+      "grad_norm": 3.203125,
+      "learning_rate": 0.00012702329700057085,
+      "loss": 4.3714,
+      "step": 240400
+    },
+    {
+      "epoch": 2.0699896694214877,
+      "grad_norm": 1.6640625,
+      "learning_rate": 0.0001269972119320293,
+      "loss": 4.2462,
+      "step": 240450
+    },
+    {
+      "epoch": 2.0704201101928374,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.00012697112488195048,
+      "loss": 4.0073,
+      "step": 240500
+    },
+    {
+      "epoch": 2.0708505509641872,
+      "grad_norm": 1.9296875,
+      "learning_rate": 0.00012694503585224915,
+      "loss": 4.2194,
+      "step": 240550
+    },
+    {
+      "epoch": 2.071280991735537,
+      "grad_norm": 2.625,
+      "learning_rate": 0.00012691894484484012,
+      "loss": 4.3994,
+      "step": 240600
+    },
+    {
+      "epoch": 2.071711432506887,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.0001268928518616385,
+      "loss": 4.6783,
+      "step": 240650
+    },
+    {
+      "epoch": 2.072141873278237,
+      "grad_norm": 0.88671875,
+      "learning_rate": 0.00012686675690455938,
+      "loss": 4.1149,
+      "step": 240700
+    },
+    {
+      "epoch": 2.072572314049587,
+      "grad_norm": 4.59375,
+      "learning_rate": 0.0001268406599755181,
+      "loss": 4.3801,
+      "step": 240750
+    },
+    {
+      "epoch": 2.0730027548209367,
+      "grad_norm": 3.75,
+      "learning_rate": 0.00012681456107643016,
+      "loss": 4.1964,
+      "step": 240800
+    },
+    {
+      "epoch": 2.0734331955922864,
+      "grad_norm": 1.5625,
+      "learning_rate": 0.00012678846020921115,
+      "loss": 4.2355,
+      "step": 240850
+    },
+    {
+      "epoch": 2.0738636363636362,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00012676235737577672,
+      "loss": 4.3201,
+      "step": 240900
+    },
+    {
+      "epoch": 2.074294077134986,
+      "grad_norm": 4.59375,
+      "learning_rate": 0.0001267362525780429,
+      "loss": 4.825,
+      "step": 240950
+    },
+    {
+      "epoch": 2.0747245179063363,
+      "grad_norm": 2.671875,
+      "learning_rate": 0.00012671014581792567,
+      "loss": 4.1755,
+      "step": 241000
+    },
+    {
+      "epoch": 2.075154958677686,
+      "grad_norm": 4.5625,
+      "learning_rate": 0.00012668403709734119,
+      "loss": 4.5968,
+      "step": 241050
+    },
+    {
+      "epoch": 2.075585399449036,
+      "grad_norm": 3.640625,
+      "learning_rate": 0.00012665792641820578,
+      "loss": 4.3305,
+      "step": 241100
+    },
+    {
+      "epoch": 2.0760158402203857,
+      "grad_norm": 1.53125,
+      "learning_rate": 0.00012663181378243593,
+      "loss": 4.5825,
+      "step": 241150
+    },
+    {
+      "epoch": 2.0764462809917354,
+      "grad_norm": 4.3125,
+      "learning_rate": 0.00012660569919194827,
+      "loss": 4.4997,
+      "step": 241200
+    },
+    {
+      "epoch": 2.0768767217630852,
+      "grad_norm": 3.234375,
+      "learning_rate": 0.0001265795826486595,
+      "loss": 4.6448,
+      "step": 241250
+    },
+    {
+      "epoch": 2.077307162534435,
+      "grad_norm": 1.25,
+      "learning_rate": 0.00012655346415448658,
+      "loss": 4.1226,
+      "step": 241300
+    },
+    {
+      "epoch": 2.0777376033057853,
+      "grad_norm": 3.5625,
+      "learning_rate": 0.00012652734371134652,
+      "loss": 4.3799,
+      "step": 241350
+    },
+    {
+      "epoch": 2.078168044077135,
+      "grad_norm": 1.96875,
+      "learning_rate": 0.0001265012213211565,
+      "loss": 4.4141,
+      "step": 241400
+    },
+    {
+      "epoch": 2.078598484848485,
+      "grad_norm": 3.25,
+      "learning_rate": 0.00012647509698583386,
+      "loss": 4.4535,
+      "step": 241450
+    },
+    {
+      "epoch": 2.0790289256198347,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.00012644897070729607,
+      "loss": 4.254,
+      "step": 241500
+    },
+    {
+      "epoch": 2.0794593663911844,
+      "grad_norm": 1.7734375,
+      "learning_rate": 0.0001264228424874607,
+      "loss": 4.7315,
+      "step": 241550
+    },
+    {
+      "epoch": 2.0798898071625342,
+      "grad_norm": 1.2734375,
+      "learning_rate": 0.0001263967123282456,
+      "loss": 4.4795,
+      "step": 241600
+    },
+    {
+      "epoch": 2.0803202479338845,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.00012637058023156855,
+      "loss": 4.3565,
+      "step": 241650
+    },
+    {
+      "epoch": 2.0807506887052343,
+      "grad_norm": 1.625,
+      "learning_rate": 0.0001263444461993477,
+      "loss": 4.2859,
+      "step": 241700
+    },
+    {
+      "epoch": 2.081181129476584,
+      "grad_norm": 3.65625,
+      "learning_rate": 0.00012631831023350113,
+      "loss": 4.4222,
+      "step": 241750
+    },
+    {
+      "epoch": 2.081611570247934,
+      "grad_norm": 8.75,
+      "learning_rate": 0.00012629217233594723,
+      "loss": 4.4645,
+      "step": 241800
+    },
+    {
+      "epoch": 2.0820420110192837,
+      "grad_norm": 2.921875,
+      "learning_rate": 0.0001262660325086045,
+      "loss": 4.6954,
+      "step": 241850
+    },
+    {
+      "epoch": 2.0824724517906334,
+      "grad_norm": 3.8125,
+      "learning_rate": 0.00012623989075339145,
+      "loss": 4.1415,
+      "step": 241900
+    },
+    {
+      "epoch": 2.0829028925619832,
+      "grad_norm": 6.125,
+      "learning_rate": 0.00012621374707222688,
+      "loss": 3.9138,
+      "step": 241950
+    },
+    {
+      "epoch": 2.0833333333333335,
+      "grad_norm": 3.265625,
+      "learning_rate": 0.00012618760146702968,
+      "loss": 4.4802,
+      "step": 242000
+    },
+    {
+      "epoch": 2.0837637741046833,
+      "grad_norm": 1.8828125,
+      "learning_rate": 0.00012616145393971886,
+      "loss": 4.4809,
+      "step": 242050
+    },
+    {
+      "epoch": 2.084194214876033,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.00012613530449221363,
+      "loss": 4.0511,
+      "step": 242100
+    },
+    {
+      "epoch": 2.084624655647383,
+      "grad_norm": 1.6796875,
+      "learning_rate": 0.00012610915312643325,
+      "loss": 4.2237,
+      "step": 242150
+    },
+    {
+      "epoch": 2.0850550964187327,
+      "grad_norm": 3.09375,
+      "learning_rate": 0.00012608299984429724,
+      "loss": 4.3035,
+      "step": 242200
+    },
+    {
+      "epoch": 2.0854855371900825,
+      "grad_norm": 3.53125,
+      "learning_rate": 0.00012605684464772515,
+      "loss": 4.531,
+      "step": 242250
+    },
+    {
+      "epoch": 2.0859159779614327,
+      "grad_norm": 3.625,
+      "learning_rate": 0.0001260306875386367,
+      "loss": 4.5293,
+      "step": 242300
+    },
+    {
+      "epoch": 2.0863464187327825,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.00012600452851895186,
+      "loss": 4.4599,
+      "step": 242350
+    },
+    {
+      "epoch": 2.0867768595041323,
+      "grad_norm": 4.25,
+      "learning_rate": 0.0001259783675905905,
+      "loss": 4.1468,
+      "step": 242400
+    },
+    {
+      "epoch": 2.087207300275482,
+      "grad_norm": 0.921875,
+      "learning_rate": 0.00012595220475547292,
+      "loss": 4.2179,
+      "step": 242450
+    },
+    {
+      "epoch": 2.087637741046832,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.0001259260400155193,
+      "loss": 3.8571,
+      "step": 242500
+    },
+    {
+      "epoch": 2.0880681818181817,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.00012589987337265014,
+      "loss": 4.7711,
+      "step": 242550
+    },
+    {
+      "epoch": 2.088498622589532,
+      "grad_norm": 3.546875,
+      "learning_rate": 0.00012587370482878603,
+      "loss": 4.2143,
+      "step": 242600
+    },
+    {
+      "epoch": 2.0889290633608817,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.0001258475343858476,
+      "loss": 4.3296,
+      "step": 242650
+    },
+    {
+      "epoch": 2.0893595041322315,
+      "grad_norm": 3.34375,
+      "learning_rate": 0.0001258213620457558,
+      "loss": 4.4828,
+      "step": 242700
+    },
+    {
+      "epoch": 2.0897899449035813,
+      "grad_norm": 4.59375,
+      "learning_rate": 0.00012579518781043157,
+      "loss": 4.7032,
+      "step": 242750
+    },
+    {
+      "epoch": 2.090220385674931,
+      "grad_norm": 1.953125,
+      "learning_rate": 0.00012576901168179601,
+      "loss": 4.3263,
+      "step": 242800
+    },
+    {
+      "epoch": 2.090650826446281,
+      "grad_norm": 3.421875,
+      "learning_rate": 0.00012574283366177048,
+      "loss": 4.3343,
+      "step": 242850
+    },
+    {
+      "epoch": 2.0910812672176307,
+      "grad_norm": 3.46875,
+      "learning_rate": 0.00012571665375227635,
+      "loss": 4.6081,
+      "step": 242900
+    },
+    {
+      "epoch": 2.091511707988981,
+      "grad_norm": 3.28125,
+      "learning_rate": 0.00012569047195523514,
+      "loss": 4.062,
+      "step": 242950
+    },
+    {
+      "epoch": 2.0919421487603307,
+      "grad_norm": 4.78125,
+      "learning_rate": 0.0001256642882725686,
+      "loss": 4.4874,
+      "step": 243000
+    },
+    {
+      "epoch": 2.0919421487603307,
+      "eval_loss": 5.032301425933838,
+      "eval_runtime": 21.7671,
+      "eval_samples_per_second": 29.402,
+      "eval_steps_per_second": 14.701,
+      "eval_tts_loss": 7.369165301432359,
+      "step": 243000
+    },
+    {
+      "epoch": 2.0923725895316805,
+      "grad_norm": 3.140625,
+      "learning_rate": 0.0001256381027061985,
+      "loss": 4.187,
+      "step": 243050
+    },
+    {
+      "epoch": 2.0928030303030303,
+      "grad_norm": 3.859375,
+      "learning_rate": 0.00012561191525804684,
+      "loss": 4.2545,
+      "step": 243100
+    },
+    {
+      "epoch": 2.09323347107438,
+      "grad_norm": 3.515625,
+      "learning_rate": 0.00012558572593003564,
+      "loss": 4.5492,
+      "step": 243150
+    },
+    {
+      "epoch": 2.09366391184573,
+      "grad_norm": 3.90625,
+      "learning_rate": 0.00012555953472408727,
+      "loss": 4.3254,
+      "step": 243200
+    },
+    {
+      "epoch": 2.09409435261708,
+      "grad_norm": 4.25,
+      "learning_rate": 0.00012553334164212403,
+      "loss": 4.7374,
+      "step": 243250
+    },
+    {
+      "epoch": 2.09452479338843,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.00012550714668606845,
+      "loss": 4.067,
+      "step": 243300
+    },
+    {
+      "epoch": 2.0949552341597797,
+      "grad_norm": 2.265625,
+      "learning_rate": 0.00012548094985784318,
+      "loss": 4.5255,
+      "step": 243350
+    },
+    {
+      "epoch": 2.0953856749311295,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.00012545475115937103,
+      "loss": 4.436,
+      "step": 243400
+    },
+    {
+      "epoch": 2.0958161157024793,
+      "grad_norm": 3.28125,
+      "learning_rate": 0.0001254285505925749,
+      "loss": 4.7913,
+      "step": 243450
+    },
+    {
+      "epoch": 2.096246556473829,
+      "grad_norm": 4.15625,
+      "learning_rate": 0.00012540234815937784,
+      "loss": 4.4039,
+      "step": 243500
+    },
+    {
+      "epoch": 2.096676997245179,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.0001253761438617031,
+      "loss": 4.3459,
+      "step": 243550
+    },
+    {
+      "epoch": 2.097107438016529,
+      "grad_norm": 4.3125,
+      "learning_rate": 0.000125349937701474,
+      "loss": 4.4695,
+      "step": 243600
+    },
+    {
+      "epoch": 2.097537878787879,
+      "grad_norm": 3.0625,
+      "learning_rate": 0.00012532372968061402,
+      "loss": 4.1749,
+      "step": 243650
+    },
+    {
+      "epoch": 2.0979683195592287,
+      "grad_norm": 1.6171875,
+      "learning_rate": 0.00012529751980104679,
+      "loss": 4.5597,
+      "step": 243700
+    },
+    {
+      "epoch": 2.0983987603305785,
+      "grad_norm": 1.65625,
+      "learning_rate": 0.00012527130806469597,
+      "loss": 4.0737,
+      "step": 243750
+    },
+    {
+      "epoch": 2.0988292011019283,
+      "grad_norm": 3.15625,
+      "learning_rate": 0.00012524509447348555,
+      "loss": 4.762,
+      "step": 243800
+    },
+    {
+      "epoch": 2.099259641873278,
+      "grad_norm": 0.44140625,
+      "learning_rate": 0.00012521887902933952,
+      "loss": 4.3837,
+      "step": 243850
+    },
+    {
+      "epoch": 2.0996900826446283,
+      "grad_norm": 2.984375,
+      "learning_rate": 0.00012519266173418203,
+      "loss": 4.8533,
+      "step": 243900
+    },
+    {
+      "epoch": 2.100120523415978,
+      "grad_norm": 3.296875,
+      "learning_rate": 0.00012516644258993734,
+      "loss": 4.2632,
+      "step": 243950
+    },
+    {
+      "epoch": 2.100550964187328,
+      "grad_norm": 3.796875,
+      "learning_rate": 0.00012514022159852996,
+      "loss": 4.3544,
+      "step": 244000
+    },
+    {
+      "epoch": 2.1009814049586777,
+      "grad_norm": 2.578125,
+      "learning_rate": 0.00012511399876188433,
+      "loss": 4.5486,
+      "step": 244050
+    },
+    {
+      "epoch": 2.1014118457300275,
+      "grad_norm": 1.046875,
+      "learning_rate": 0.0001250877740819253,
+      "loss": 4.4981,
+      "step": 244100
+    },
+    {
+      "epoch": 2.1018422865013773,
+      "grad_norm": 3.046875,
+      "learning_rate": 0.00012506154756057762,
+      "loss": 4.4835,
+      "step": 244150
+    },
+    {
+      "epoch": 2.102272727272727,
+      "grad_norm": 3.328125,
+      "learning_rate": 0.00012503531919976624,
+      "loss": 4.0523,
+      "step": 244200
+    },
+    {
+      "epoch": 2.1027031680440773,
+      "grad_norm": 2.5,
+      "learning_rate": 0.0001250090890014163,
+      "loss": 4.4435,
+      "step": 244250
+    },
+    {
+      "epoch": 2.103133608815427,
+      "grad_norm": 3.0625,
+      "learning_rate": 0.00012498285696745307,
+      "loss": 4.3182,
+      "step": 244300
+    },
+    {
+      "epoch": 2.103564049586777,
+      "grad_norm": 3.5,
+      "learning_rate": 0.00012495662309980186,
+      "loss": 4.3892,
+      "step": 244350
+    },
+    {
+      "epoch": 2.1039944903581267,
+      "grad_norm": 5.15625,
+      "learning_rate": 0.0001249303874003882,
+      "loss": 4.4317,
+      "step": 244400
+    },
+    {
+      "epoch": 2.1044249311294765,
+      "grad_norm": 1.4140625,
+      "learning_rate": 0.00012490414987113775,
+      "loss": 4.2303,
+      "step": 244450
+    },
+    {
+      "epoch": 2.1048553719008263,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.0001248779105139763,
+      "loss": 4.5734,
+      "step": 244500
+    },
+    {
+      "epoch": 2.1052858126721765,
+      "grad_norm": 4.34375,
+      "learning_rate": 0.00012485166933082974,
+      "loss": 4.6328,
+      "step": 244550
+    },
+    {
+      "epoch": 2.1057162534435263,
+      "grad_norm": 3.84375,
+      "learning_rate": 0.00012482542632362412,
+      "loss": 4.0772,
+      "step": 244600
+    },
+    {
+      "epoch": 2.106146694214876,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00012479918149428565,
+      "loss": 4.3002,
+      "step": 244650
+    },
+    {
+      "epoch": 2.106577134986226,
+      "grad_norm": 1.4375,
+      "learning_rate": 0.00012477293484474056,
+      "loss": 4.1697,
+      "step": 244700
+    },
+    {
+      "epoch": 2.1070075757575757,
+      "grad_norm": 3.765625,
+      "learning_rate": 0.0001247466863769154,
+      "loss": 4.3076,
+      "step": 244750
+    },
+    {
+      "epoch": 2.1074380165289255,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00012472043609273667,
+      "loss": 4.2345,
+      "step": 244800
+    },
+    {
+      "epoch": 2.1078684573002753,
+      "grad_norm": 4.375,
+      "learning_rate": 0.0001246941839941311,
+      "loss": 4.3255,
+      "step": 244850
+    },
+    {
+      "epoch": 2.1082988980716255,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.0001246679300830256,
+      "loss": 4.5231,
+      "step": 244900
+    },
+    {
+      "epoch": 2.1087293388429753,
+      "grad_norm": 5.5625,
+      "learning_rate": 0.0001246416743613471,
+      "loss": 4.045,
+      "step": 244950
+    },
+    {
+      "epoch": 2.109159779614325,
+      "grad_norm": 3.734375,
+      "learning_rate": 0.0001246154168310227,
+      "loss": 4.467,
+      "step": 245000
+    },
+    {
+      "epoch": 2.109590220385675,
+      "grad_norm": 4.125,
+      "learning_rate": 0.00012458915749397967,
+      "loss": 4.4776,
+      "step": 245050
+    },
+    {
+      "epoch": 2.1100206611570247,
+      "grad_norm": 4.375,
+      "learning_rate": 0.00012456289635214538,
+      "loss": 3.9669,
+      "step": 245100
+    },
+    {
+      "epoch": 2.1104511019283745,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.00012453663340744736,
+      "loss": 4.4018,
+      "step": 245150
+    },
+    {
+      "epoch": 2.1108815426997247,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.00012451036866181325,
+      "loss": 4.3882,
+      "step": 245200
+    },
+    {
+      "epoch": 2.1113119834710745,
+      "grad_norm": 3.671875,
+      "learning_rate": 0.00012448410211717082,
+      "loss": 4.0448,
+      "step": 245250
+    },
+    {
+      "epoch": 2.1117424242424243,
+      "grad_norm": 3.140625,
+      "learning_rate": 0.00012445783377544795,
+      "loss": 4.309,
+      "step": 245300
+    },
+    {
+      "epoch": 2.112172865013774,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.0001244315636385727,
+      "loss": 3.9951,
+      "step": 245350
+    },
+    {
+      "epoch": 2.112603305785124,
+      "grad_norm": 3.453125,
+      "learning_rate": 0.00012440529170847323,
+      "loss": 4.1266,
+      "step": 245400
+    },
+    {
+      "epoch": 2.1130337465564737,
+      "grad_norm": 1.6015625,
+      "learning_rate": 0.00012437901798707793,
+      "loss": 4.203,
+      "step": 245450
+    },
+    {
+      "epoch": 2.1134641873278235,
+      "grad_norm": 3.484375,
+      "learning_rate": 0.0001243527424763151,
+      "loss": 4.4071,
+      "step": 245500
+    },
+    {
+      "epoch": 2.1138946280991737,
+      "grad_norm": 3.90625,
+      "learning_rate": 0.00012432646517811338,
+      "loss": 4.8537,
+      "step": 245550
+    },
+    {
+      "epoch": 2.1143250688705235,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.00012430018609440146,
+      "loss": 4.7115,
+      "step": 245600
+    },
+    {
+      "epoch": 2.1147555096418733,
+      "grad_norm": 2.375,
+      "learning_rate": 0.00012427390522710816,
+      "loss": 4.1112,
+      "step": 245650
+    },
+    {
+      "epoch": 2.115185950413223,
+      "grad_norm": 4.90625,
+      "learning_rate": 0.00012424762257816244,
+      "loss": 4.5935,
+      "step": 245700
+    },
+    {
+      "epoch": 2.115616391184573,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.0001242213381494934,
+      "loss": 4.5632,
+      "step": 245750
+    },
+    {
+      "epoch": 2.1160468319559227,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00012419505194303027,
+      "loss": 4.3616,
+      "step": 245800
+    },
+    {
+      "epoch": 2.116477272727273,
+      "grad_norm": 1.9609375,
+      "learning_rate": 0.00012416876396070236,
+      "loss": 4.2153,
+      "step": 245850
+    },
+    {
+      "epoch": 2.1169077134986227,
+      "grad_norm": 2.421875,
+      "learning_rate": 0.00012414247420443916,
+      "loss": 4.4025,
+      "step": 245900
+    },
+    {
+      "epoch": 2.1173381542699725,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00012411618267617035,
+      "loss": 4.4464,
+      "step": 245950
+    },
+    {
+      "epoch": 2.1177685950413223,
+      "grad_norm": 4.59375,
+      "learning_rate": 0.0001240898893778256,
+      "loss": 4.3501,
+      "step": 246000
+    },
+    {
+      "epoch": 2.1177685950413223,
+      "eval_loss": 5.029580116271973,
+      "eval_runtime": 22.1853,
+      "eval_samples_per_second": 28.848,
+      "eval_steps_per_second": 14.424,
+      "eval_tts_loss": 7.369533711766074,
+      "step": 246000
+    },
+    {
+      "epoch": 2.118199035812672,
+      "grad_norm": 3.125,
+      "learning_rate": 0.00012406359431133473,
+      "loss": 4.3839,
+      "step": 246050
+    },
+    {
+      "epoch": 2.118629476584022,
+      "grad_norm": 5.0,
+      "learning_rate": 0.00012403729747862788,
+      "loss": 4.2713,
+      "step": 246100
+    },
+    {
+      "epoch": 2.1190599173553717,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.00012401099888163512,
+      "loss": 4.363,
+      "step": 246150
+    },
+    {
+      "epoch": 2.119490358126722,
+      "grad_norm": 3.015625,
+      "learning_rate": 0.0001239846985222867,
+      "loss": 4.3895,
+      "step": 246200
+    },
+    {
+      "epoch": 2.1199207988980717,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.000123958396402513,
+      "loss": 4.2328,
+      "step": 246250
+    },
+    {
+      "epoch": 2.1203512396694215,
+      "grad_norm": 3.375,
+      "learning_rate": 0.00012393209252424452,
+      "loss": 4.5248,
+      "step": 246300
+    },
+    {
+      "epoch": 2.1207816804407713,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.000123905786889412,
+      "loss": 4.1609,
+      "step": 246350
+    },
+    {
+      "epoch": 2.121212121212121,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00012387947949994612,
+      "loss": 4.3887,
+      "step": 246400
+    },
+    {
+      "epoch": 2.121642561983471,
+      "grad_norm": 4.28125,
+      "learning_rate": 0.00012385317035777785,
+      "loss": 4.3359,
+      "step": 246450
+    },
+    {
+      "epoch": 2.122073002754821,
+      "grad_norm": 1.390625,
+      "learning_rate": 0.00012382685946483817,
+      "loss": 4.5074,
+      "step": 246500
+    },
+    {
+      "epoch": 2.122503443526171,
+      "grad_norm": 2.578125,
+      "learning_rate": 0.00012380054682305827,
+      "loss": 4.2195,
+      "step": 246550
+    },
+    {
+      "epoch": 2.1229338842975207,
+      "grad_norm": 3.96875,
+      "learning_rate": 0.00012377423243436947,
+      "loss": 3.7267,
+      "step": 246600
+    },
+    {
+      "epoch": 2.1233643250688705,
+      "grad_norm": 2.125,
+      "learning_rate": 0.00012374791630070313,
+      "loss": 4.2385,
+      "step": 246650
+    },
+    {
+      "epoch": 2.1237947658402203,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00012372159842399084,
+      "loss": 4.2527,
+      "step": 246700
+    },
+    {
+      "epoch": 2.12422520661157,
+      "grad_norm": 3.28125,
+      "learning_rate": 0.00012369527880616425,
+      "loss": 4.003,
+      "step": 246750
+    },
+    {
+      "epoch": 2.12465564738292,
+      "grad_norm": 3.015625,
+      "learning_rate": 0.00012366895744915518,
+      "loss": 4.3187,
+      "step": 246800
+    },
+    {
+      "epoch": 2.12508608815427,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00012364263435489558,
+      "loss": 4.2765,
+      "step": 246850
+    },
+    {
+      "epoch": 2.12551652892562,
+      "grad_norm": 1.859375,
+      "learning_rate": 0.00012361630952531746,
+      "loss": 4.5432,
+      "step": 246900
+    },
+    {
+      "epoch": 2.1259469696969697,
+      "grad_norm": 2.421875,
+      "learning_rate": 0.00012358998296235307,
+      "loss": 4.543,
+      "step": 246950
+    },
+    {
+      "epoch": 2.1263774104683195,
+      "grad_norm": 4.53125,
+      "learning_rate": 0.00012356365466793467,
+      "loss": 4.3545,
+      "step": 247000
+    },
+    {
+      "epoch": 2.1268078512396693,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.0001235373246439947,
+      "loss": 4.393,
+      "step": 247050
+    },
+    {
+      "epoch": 2.127238292011019,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.00012351099289246574,
+      "loss": 4.5236,
+      "step": 247100
+    },
+    {
+      "epoch": 2.1276687327823693,
+      "grad_norm": 3.921875,
+      "learning_rate": 0.0001234846594152805,
+      "loss": 4.2457,
+      "step": 247150
+    },
+    {
+      "epoch": 2.128099173553719,
+      "grad_norm": 3.0625,
+      "learning_rate": 0.00012345832421437178,
+      "loss": 4.2482,
+      "step": 247200
+    },
+    {
+      "epoch": 2.128529614325069,
+      "grad_norm": 1.7734375,
+      "learning_rate": 0.00012343198729167256,
+      "loss": 4.3731,
+      "step": 247250
+    },
+    {
+      "epoch": 2.1289600550964187,
+      "grad_norm": 1.75,
+      "learning_rate": 0.00012340564864911585,
+      "loss": 4.2384,
+      "step": 247300
+    },
+    {
+      "epoch": 2.1293904958677685,
+      "grad_norm": 4.6875,
+      "learning_rate": 0.00012337930828863492,
+      "loss": 4.0476,
+      "step": 247350
+    },
+    {
+      "epoch": 2.1298209366391183,
+      "grad_norm": 1.7734375,
+      "learning_rate": 0.00012335296621216306,
+      "loss": 4.4302,
+      "step": 247400
+    },
+    {
+      "epoch": 2.1302513774104685,
+      "grad_norm": 3.0625,
+      "learning_rate": 0.00012332662242163367,
+      "loss": 4.2237,
+      "step": 247450
+    },
+    {
+      "epoch": 2.1306818181818183,
+      "grad_norm": 3.921875,
+      "learning_rate": 0.00012330027691898046,
+      "loss": 4.4013,
+      "step": 247500
+    },
+    {
+      "epoch": 2.131112258953168,
+      "grad_norm": 3.140625,
+      "learning_rate": 0.00012327392970613702,
+      "loss": 4.2359,
+      "step": 247550
+    },
+    {
+      "epoch": 2.131542699724518,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00012324758078503724,
+      "loss": 4.0828,
+      "step": 247600
+    },
+    {
+      "epoch": 2.1319731404958677,
+      "grad_norm": 3.140625,
+      "learning_rate": 0.000123221230157615,
+      "loss": 4.4656,
+      "step": 247650
+    },
+    {
+      "epoch": 2.1324035812672175,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00012319487782580448,
+      "loss": 4.7965,
+      "step": 247700
+    },
+    {
+      "epoch": 2.1328340220385673,
+      "grad_norm": 3.1875,
+      "learning_rate": 0.00012316852379153982,
+      "loss": 4.5846,
+      "step": 247750
+    },
+    {
+      "epoch": 2.1332644628099175,
+      "grad_norm": 1.703125,
+      "learning_rate": 0.0001231421680567553,
+      "loss": 4.3269,
+      "step": 247800
+    },
+    {
+      "epoch": 2.1336949035812673,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.0001231158106233855,
+      "loss": 4.6014,
+      "step": 247850
+    },
+    {
+      "epoch": 2.134125344352617,
+      "grad_norm": 4.78125,
+      "learning_rate": 0.00012308945149336494,
+      "loss": 4.5338,
+      "step": 247900
+    },
+    {
+      "epoch": 2.134555785123967,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.00012306309066862828,
+      "loss": 4.6047,
+      "step": 247950
+    },
+    {
+      "epoch": 2.1349862258953167,
+      "grad_norm": 3.296875,
+      "learning_rate": 0.00012303672815111043,
+      "loss": 4.171,
+      "step": 248000
+    },
+    {
+      "epoch": 2.1354166666666665,
+      "grad_norm": 3.8125,
+      "learning_rate": 0.00012301036394274628,
+      "loss": 4.1846,
+      "step": 248050
+    },
+    {
+      "epoch": 2.1358471074380168,
+      "grad_norm": 4.0625,
+      "learning_rate": 0.0001229839980454709,
+      "loss": 4.416,
+      "step": 248100
+    },
+    {
+      "epoch": 2.1362775482093666,
+      "grad_norm": 3.015625,
+      "learning_rate": 0.00012295763046121956,
+      "loss": 4.4332,
+      "step": 248150
+    },
+    {
+      "epoch": 2.1367079889807163,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.0001229312611919275,
+      "loss": 4.3031,
+      "step": 248200
+    },
+    {
+      "epoch": 2.137138429752066,
+      "grad_norm": 0.88671875,
+      "learning_rate": 0.00012290489023953025,
+      "loss": 4.3623,
+      "step": 248250
+    },
+    {
+      "epoch": 2.137568870523416,
+      "grad_norm": 3.203125,
+      "learning_rate": 0.00012287851760596328,
+      "loss": 4.352,
+      "step": 248300
+    },
+    {
+      "epoch": 2.1379993112947657,
+      "grad_norm": 4.71875,
+      "learning_rate": 0.00012285214329316237,
+      "loss": 4.5255,
+      "step": 248350
+    },
+    {
+      "epoch": 2.1384297520661155,
+      "grad_norm": 3.703125,
+      "learning_rate": 0.0001228257673030633,
+      "loss": 4.334,
+      "step": 248400
+    },
+    {
+      "epoch": 2.1388601928374658,
+      "grad_norm": 5.03125,
+      "learning_rate": 0.00012279938963760204,
+      "loss": 4.6355,
+      "step": 248450
+    },
+    {
+      "epoch": 2.1392906336088156,
+      "grad_norm": 3.0625,
+      "learning_rate": 0.00012277301029871464,
+      "loss": 4.587,
+      "step": 248500
+    },
+    {
+      "epoch": 2.1397210743801653,
+      "grad_norm": 3.15625,
+      "learning_rate": 0.00012274662928833728,
+      "loss": 4.3958,
+      "step": 248550
+    },
+    {
+      "epoch": 2.140151515151515,
+      "grad_norm": 3.203125,
+      "learning_rate": 0.00012272024660840624,
+      "loss": 4.2842,
+      "step": 248600
+    },
+    {
+      "epoch": 2.140581955922865,
+      "grad_norm": 1.4609375,
+      "learning_rate": 0.00012269386226085798,
+      "loss": 3.9295,
+      "step": 248650
+    },
+    {
+      "epoch": 2.1410123966942147,
+      "grad_norm": 3.40625,
+      "learning_rate": 0.0001226674762476291,
+      "loss": 4.4017,
+      "step": 248700
+    },
+    {
+      "epoch": 2.141442837465565,
+      "grad_norm": 3.34375,
+      "learning_rate": 0.00012264108857065623,
+      "loss": 4.5864,
+      "step": 248750
+    },
+    {
+      "epoch": 2.1418732782369148,
+      "grad_norm": 4.25,
+      "learning_rate": 0.00012261469923187615,
+      "loss": 4.2633,
+      "step": 248800
+    },
+    {
+      "epoch": 2.1423037190082646,
+      "grad_norm": 1.765625,
+      "learning_rate": 0.0001225883082332258,
+      "loss": 4.6043,
+      "step": 248850
+    },
+    {
+      "epoch": 2.1427341597796143,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.00012256191557664226,
+      "loss": 4.4089,
+      "step": 248900
+    },
+    {
+      "epoch": 2.143164600550964,
+      "grad_norm": 3.03125,
+      "learning_rate": 0.0001225355212640626,
+      "loss": 4.4099,
+      "step": 248950
+    },
+    {
+      "epoch": 2.143595041322314,
+      "grad_norm": 4.5625,
+      "learning_rate": 0.00012250912529742419,
+      "loss": 4.3886,
+      "step": 249000
+    },
+    {
+      "epoch": 2.143595041322314,
+      "eval_loss": 5.030261039733887,
+      "eval_runtime": 22.19,
+      "eval_samples_per_second": 28.842,
+      "eval_steps_per_second": 14.421,
+      "eval_tts_loss": 7.399030686667878,
+      "step": 249000
+    },
+    {
+      "epoch": 2.1440254820936637,
+      "grad_norm": 3.4375,
+      "learning_rate": 0.00012248272767866443,
+      "loss": 4.7167,
+      "step": 249050
+    },
+    {
+      "epoch": 2.144455922865014,
+      "grad_norm": 3.34375,
+      "learning_rate": 0.0001224563284097208,
+      "loss": 4.3494,
+      "step": 249100
+    },
+    {
+      "epoch": 2.1448863636363638,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00012242992749253104,
+      "loss": 3.9664,
+      "step": 249150
+    },
+    {
+      "epoch": 2.1453168044077136,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.00012240352492903282,
+      "loss": 4.3974,
+      "step": 249200
+    },
+    {
+      "epoch": 2.1457472451790633,
+      "grad_norm": 3.78125,
+      "learning_rate": 0.00012237712072116406,
+      "loss": 4.3286,
+      "step": 249250
+    },
+    {
+      "epoch": 2.146177685950413,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.00012235071487086283,
+      "loss": 4.3842,
+      "step": 249300
+    },
+    {
+      "epoch": 2.146608126721763,
+      "grad_norm": 3.765625,
+      "learning_rate": 0.00012232430738006716,
+      "loss": 4.6674,
+      "step": 249350
+    },
+    {
+      "epoch": 2.147038567493113,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.00012229789825071542,
+      "loss": 4.3949,
+      "step": 249400
+    },
+    {
+      "epoch": 2.147469008264463,
+      "grad_norm": 5.3125,
+      "learning_rate": 0.00012227148748474587,
+      "loss": 3.5729,
+      "step": 249450
+    },
+    {
+      "epoch": 2.1478994490358128,
+      "grad_norm": 4.25,
+      "learning_rate": 0.00012224507508409705,
+      "loss": 4.5456,
+      "step": 249500
+    },
+    {
+      "epoch": 2.1483298898071626,
+      "grad_norm": 3.890625,
+      "learning_rate": 0.00012221866105070766,
+      "loss": 4.4121,
+      "step": 249550
+    },
+    {
+      "epoch": 2.1487603305785123,
+      "grad_norm": 3.3125,
+      "learning_rate": 0.0001221922453865163,
+      "loss": 4.3994,
+      "step": 249600
+    },
+    {
+      "epoch": 2.149190771349862,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.0001221658280934619,
+      "loss": 4.1728,
+      "step": 249650
+    },
+    {
+      "epoch": 2.149621212121212,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.00012213940917348335,
+      "loss": 3.9191,
+      "step": 249700
+    },
+    {
+      "epoch": 2.150051652892562,
+      "grad_norm": 1.9453125,
+      "learning_rate": 0.00012211298862851985,
+      "loss": 4.5975,
+      "step": 249750
+    },
+    {
+      "epoch": 2.150482093663912,
+      "grad_norm": 3.734375,
+      "learning_rate": 0.00012208656646051056,
+      "loss": 4.6434,
+      "step": 249800
+    },
+    {
+      "epoch": 2.1509125344352618,
+      "grad_norm": 2.21875,
+      "learning_rate": 0.00012206014267139478,
+      "loss": 4.3724,
+      "step": 249850
+    },
+    {
+      "epoch": 2.1513429752066116,
+      "grad_norm": 3.8125,
+      "learning_rate": 0.00012203371726311204,
+      "loss": 4.6818,
+      "step": 249900
+    },
+    {
+      "epoch": 2.1517734159779613,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00012200729023760184,
+      "loss": 4.5516,
+      "step": 249950
+    },
+    {
+      "epoch": 2.152203856749311,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.00012198086159680387,
+      "loss": 4.4988,
+      "step": 250000
+    },
+    {
+      "epoch": 2.1526342975206614,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.00012195443134265795,
+      "loss": 4.3184,
+      "step": 250050
+    },
+    {
+      "epoch": 2.153064738292011,
+      "grad_norm": 3.390625,
+      "learning_rate": 0.00012192799947710405,
+      "loss": 4.4347,
+      "step": 250100
+    },
+    {
+      "epoch": 2.153495179063361,
+      "grad_norm": 2.875,
+      "learning_rate": 0.00012190156600208211,
+      "loss": 4.5737,
+      "step": 250150
+    },
+    {
+      "epoch": 2.1539256198347108,
+      "grad_norm": 1.9609375,
+      "learning_rate": 0.00012187513091953237,
+      "loss": 4.5308,
+      "step": 250200
+    },
+    {
+      "epoch": 2.1543560606060606,
+      "grad_norm": 1.828125,
+      "learning_rate": 0.00012184869423139511,
+      "loss": 3.9899,
+      "step": 250250
+    },
+    {
+      "epoch": 2.1547865013774103,
+      "grad_norm": 4.0,
+      "learning_rate": 0.00012182225593961069,
+      "loss": 4.4631,
+      "step": 250300
+    },
+    {
+      "epoch": 2.15521694214876,
+      "grad_norm": 5.40625,
+      "learning_rate": 0.00012179581604611963,
+      "loss": 4.515,
+      "step": 250350
+    },
+    {
+      "epoch": 2.1556473829201104,
+      "grad_norm": 3.078125,
+      "learning_rate": 0.00012176937455286257,
+      "loss": 4.2414,
+      "step": 250400
+    },
+    {
+      "epoch": 2.15607782369146,
+      "grad_norm": 4.25,
+      "learning_rate": 0.00012174293146178027,
+      "loss": 4.5981,
+      "step": 250450
+    },
+    {
+      "epoch": 2.15650826446281,
+      "grad_norm": 3.25,
+      "learning_rate": 0.00012171648677481353,
+      "loss": 4.5456,
+      "step": 250500
+    },
+    {
+      "epoch": 2.1569387052341598,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00012169004049390342,
+      "loss": 4.3493,
+      "step": 250550
+    },
+    {
+      "epoch": 2.1573691460055096,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00012166359262099103,
+      "loss": 4.3432,
+      "step": 250600
+    },
+    {
+      "epoch": 2.1577995867768593,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.0001216371431580175,
+      "loss": 4.2171,
+      "step": 250650
+    },
+    {
+      "epoch": 2.1582300275482096,
+      "grad_norm": 1.9921875,
+      "learning_rate": 0.00012161069210692426,
+      "loss": 4.3573,
+      "step": 250700
+    },
+    {
+      "epoch": 2.1586604683195594,
+      "grad_norm": 4.53125,
+      "learning_rate": 0.00012158423946965269,
+      "loss": 4.4684,
+      "step": 250750
+    },
+    {
+      "epoch": 2.159090909090909,
+      "grad_norm": 3.96875,
+      "learning_rate": 0.00012155778524814438,
+      "loss": 4.4885,
+      "step": 250800
+    },
+    {
+      "epoch": 2.159521349862259,
+      "grad_norm": 3.0625,
+      "learning_rate": 0.00012153132944434103,
+      "loss": 4.1797,
+      "step": 250850
+    },
+    {
+      "epoch": 2.1599517906336088,
+      "grad_norm": 5.1875,
+      "learning_rate": 0.00012150487206018442,
+      "loss": 4.2264,
+      "step": 250900
+    },
+    {
+      "epoch": 2.1603822314049586,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00012147841309761649,
+      "loss": 4.1449,
+      "step": 250950
+    },
+    {
+      "epoch": 2.1608126721763083,
+      "grad_norm": 2.875,
+      "learning_rate": 0.00012145195255857922,
+      "loss": 4.4546,
+      "step": 251000
+    },
+    {
+      "epoch": 2.1612431129476586,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00012142549044501477,
+      "loss": 4.461,
+      "step": 251050
+    },
+    {
+      "epoch": 2.1616735537190084,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.00012139902675886548,
+      "loss": 4.5465,
+      "step": 251100
+    },
+    {
+      "epoch": 2.162103994490358,
+      "grad_norm": 3.15625,
+      "learning_rate": 0.00012137256150207364,
+      "loss": 4.4751,
+      "step": 251150
+    },
+    {
+      "epoch": 2.162534435261708,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.00012134609467658176,
+      "loss": 4.5424,
+      "step": 251200
+    },
+    {
+      "epoch": 2.1629648760330578,
+      "grad_norm": 3.09375,
+      "learning_rate": 0.00012131962628433244,
+      "loss": 4.7024,
+      "step": 251250
+    },
+    {
+      "epoch": 2.1633953168044076,
+      "grad_norm": 2.421875,
+      "learning_rate": 0.00012129315632726845,
+      "loss": 4.3189,
+      "step": 251300
+    },
+    {
+      "epoch": 2.163825757575758,
+      "grad_norm": 1.8125,
+      "learning_rate": 0.00012126668480733261,
+      "loss": 4.0801,
+      "step": 251350
+    },
+    {
+      "epoch": 2.1642561983471076,
+      "grad_norm": 3.4375,
+      "learning_rate": 0.00012124021172646782,
+      "loss": 4.7692,
+      "step": 251400
+    },
+    {
+      "epoch": 2.1646866391184574,
+      "grad_norm": 3.1875,
+      "learning_rate": 0.00012121373708661725,
+      "loss": 4.2609,
+      "step": 251450
+    },
+    {
+      "epoch": 2.165117079889807,
+      "grad_norm": 3.65625,
+      "learning_rate": 0.00012118726088972396,
+      "loss": 4.1146,
+      "step": 251500
+    },
+    {
+      "epoch": 2.165547520661157,
+      "grad_norm": 2.21875,
+      "learning_rate": 0.00012116078313773131,
+      "loss": 4.3169,
+      "step": 251550
+    },
+    {
+      "epoch": 2.1659779614325068,
+      "grad_norm": 4.84375,
+      "learning_rate": 0.00012113430383258275,
+      "loss": 4.5276,
+      "step": 251600
+    },
+    {
+      "epoch": 2.1664084022038566,
+      "grad_norm": 2.921875,
+      "learning_rate": 0.00012110782297622173,
+      "loss": 4.1889,
+      "step": 251650
+    },
+    {
+      "epoch": 2.166838842975207,
+      "grad_norm": 1.7890625,
+      "learning_rate": 0.00012108134057059193,
+      "loss": 4.3648,
+      "step": 251700
+    },
+    {
+      "epoch": 2.1672692837465566,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.00012105485661763709,
+      "loss": 4.3965,
+      "step": 251750
+    },
+    {
+      "epoch": 2.1676997245179064,
+      "grad_norm": 0.7109375,
+      "learning_rate": 0.00012102837111930108,
+      "loss": 3.9214,
+      "step": 251800
+    },
+    {
+      "epoch": 2.168130165289256,
+      "grad_norm": 3.65625,
+      "learning_rate": 0.00012100188407752786,
+      "loss": 4.3281,
+      "step": 251850
+    },
+    {
+      "epoch": 2.168560606060606,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.00012097539549426151,
+      "loss": 4.2369,
+      "step": 251900
+    },
+    {
+      "epoch": 2.1689910468319558,
+      "grad_norm": 3.5,
+      "learning_rate": 0.00012094890537144632,
+      "loss": 4.2376,
+      "step": 251950
+    },
+    {
+      "epoch": 2.169421487603306,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.0001209224137110265,
+      "loss": 4.2486,
+      "step": 252000
+    },
+    {
+      "epoch": 2.169421487603306,
+      "eval_loss": 5.0270843505859375,
+      "eval_runtime": 21.9413,
+      "eval_samples_per_second": 29.169,
+      "eval_steps_per_second": 14.584,
+      "eval_tts_loss": 7.36063719755612,
+      "step": 252000
+    },
+    {
+      "epoch": 2.169851928374656,
+      "grad_norm": 1.4609375,
+      "learning_rate": 0.00012089592051494654,
+      "loss": 4.2264,
+      "step": 252050
+    },
+    {
+      "epoch": 2.1702823691460056,
+      "grad_norm": 1.796875,
+      "learning_rate": 0.00012086942578515099,
+      "loss": 4.2205,
+      "step": 252100
+    },
+    {
+      "epoch": 2.1707128099173554,
+      "grad_norm": 3.265625,
+      "learning_rate": 0.00012084292952358449,
+      "loss": 4.4578,
+      "step": 252150
+    },
+    {
+      "epoch": 2.171143250688705,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.00012081643173219178,
+      "loss": 4.4302,
+      "step": 252200
+    },
+    {
+      "epoch": 2.171573691460055,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.00012078993241291779,
+      "loss": 4.379,
+      "step": 252250
+    },
+    {
+      "epoch": 2.1720041322314048,
+      "grad_norm": 3.21875,
+      "learning_rate": 0.00012076343156770748,
+      "loss": 4.3006,
+      "step": 252300
+    },
+    {
+      "epoch": 2.172434573002755,
+      "grad_norm": 3.40625,
+      "learning_rate": 0.00012073692919850598,
+      "loss": 4.5328,
+      "step": 252350
+    },
+    {
+      "epoch": 2.172865013774105,
+      "grad_norm": 5.8125,
+      "learning_rate": 0.00012071042530725849,
+      "loss": 4.3117,
+      "step": 252400
+    },
+    {
+      "epoch": 2.1732954545454546,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.00012068391989591035,
+      "loss": 4.3748,
+      "step": 252450
+    },
+    {
+      "epoch": 2.1737258953168044,
+      "grad_norm": 3.9375,
+      "learning_rate": 0.00012065741296640698,
+      "loss": 4.3187,
+      "step": 252500
+    },
+    {
+      "epoch": 2.174156336088154,
+      "grad_norm": 2.984375,
+      "learning_rate": 0.00012063090452069392,
+      "loss": 4.6749,
+      "step": 252550
+    },
+    {
+      "epoch": 2.174586776859504,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.00012060439456071689,
+      "loss": 4.2573,
+      "step": 252600
+    },
+    {
+      "epoch": 2.175017217630854,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00012057788308842164,
+      "loss": 4.1282,
+      "step": 252650
+    },
+    {
+      "epoch": 2.175447658402204,
+      "grad_norm": 1.671875,
+      "learning_rate": 0.00012055137010575402,
+      "loss": 4.4505,
+      "step": 252700
+    },
+    {
+      "epoch": 2.175878099173554,
+      "grad_norm": 3.390625,
+      "learning_rate": 0.00012052485561466004,
+      "loss": 4.2804,
+      "step": 252750
+    },
+    {
+      "epoch": 2.1763085399449036,
+      "grad_norm": 2.9375,
+      "learning_rate": 0.00012049833961708588,
+      "loss": 4.221,
+      "step": 252800
+    },
+    {
+      "epoch": 2.1767389807162534,
+      "grad_norm": 4.03125,
+      "learning_rate": 0.00012047182211497762,
+      "loss": 4.4549,
+      "step": 252850
+    },
+    {
+      "epoch": 2.177169421487603,
+      "grad_norm": 2.875,
+      "learning_rate": 0.00012044530311028171,
+      "loss": 4.488,
+      "step": 252900
+    },
+    {
+      "epoch": 2.177599862258953,
+      "grad_norm": 3.375,
+      "learning_rate": 0.0001204187826049445,
+      "loss": 4.2177,
+      "step": 252950
+    },
+    {
+      "epoch": 2.178030303030303,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00012039226060091264,
+      "loss": 4.2626,
+      "step": 253000
+    },
+    {
+      "epoch": 2.178460743801653,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00012036573710013268,
+      "loss": 4.4399,
+      "step": 253050
+    },
+    {
+      "epoch": 2.178891184573003,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00012033921210455144,
+      "loss": 4.427,
+      "step": 253100
+    },
+    {
+      "epoch": 2.1793216253443526,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.0001203126856161158,
+      "loss": 4.6042,
+      "step": 253150
+    },
+    {
+      "epoch": 2.1797520661157024,
+      "grad_norm": 3.703125,
+      "learning_rate": 0.00012028615763677276,
+      "loss": 4.1871,
+      "step": 253200
+    },
+    {
+      "epoch": 2.180182506887052,
+      "grad_norm": 3.234375,
+      "learning_rate": 0.00012025962816846936,
+      "loss": 4.4172,
+      "step": 253250
+    },
+    {
+      "epoch": 2.1806129476584024,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.00012023309721315285,
+      "loss": 4.6173,
+      "step": 253300
+    },
+    {
+      "epoch": 2.181043388429752,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.00012020656477277056,
+      "loss": 4.1272,
+      "step": 253350
+    },
+    {
+      "epoch": 2.181473829201102,
+      "grad_norm": 0.70703125,
+      "learning_rate": 0.00012018003084926988,
+      "loss": 3.895,
+      "step": 253400
+    },
+    {
+      "epoch": 2.181904269972452,
+      "grad_norm": 1.78125,
+      "learning_rate": 0.00012015349544459832,
+      "loss": 4.3051,
+      "step": 253450
+    },
+    {
+      "epoch": 2.1823347107438016,
+      "grad_norm": 3.578125,
+      "learning_rate": 0.00012012695856070361,
+      "loss": 4.7618,
+      "step": 253500
+    },
+    {
+      "epoch": 2.1827651515151514,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00012010042019953343,
+      "loss": 3.9043,
+      "step": 253550
+    },
+    {
+      "epoch": 2.183195592286501,
+      "grad_norm": 1.3203125,
+      "learning_rate": 0.00012007388036303565,
+      "loss": 4.1297,
+      "step": 253600
+    },
+    {
+      "epoch": 2.1836260330578514,
+      "grad_norm": 3.21875,
+      "learning_rate": 0.00012004733905315824,
+      "loss": 4.2037,
+      "step": 253650
+    },
+    {
+      "epoch": 2.184056473829201,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00012002079627184927,
+      "loss": 4.3872,
+      "step": 253700
+    },
+    {
+      "epoch": 2.184486914600551,
+      "grad_norm": 1.75,
+      "learning_rate": 0.00011999425202105691,
+      "loss": 4.1089,
+      "step": 253750
+    },
+    {
+      "epoch": 2.184917355371901,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.0001199677063027295,
+      "loss": 4.3687,
+      "step": 253800
+    },
+    {
+      "epoch": 2.1853477961432506,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00011994115911881542,
+      "loss": 4.0994,
+      "step": 253850
+    },
+    {
+      "epoch": 2.1857782369146004,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.00011991461047126316,
+      "loss": 4.5209,
+      "step": 253900
+    },
+    {
+      "epoch": 2.1862086776859506,
+      "grad_norm": 5.8125,
+      "learning_rate": 0.00011988806036202132,
+      "loss": 4.366,
+      "step": 253950
+    },
+    {
+      "epoch": 2.1866391184573004,
+      "grad_norm": 1.7109375,
+      "learning_rate": 0.00011986150879303866,
+      "loss": 4.3154,
+      "step": 254000
+    },
+    {
+      "epoch": 2.18706955922865,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00011983495576626398,
+      "loss": 4.2244,
+      "step": 254050
+    },
+    {
+      "epoch": 2.1875,
+      "grad_norm": 4.4375,
+      "learning_rate": 0.00011980840128364619,
+      "loss": 4.1018,
+      "step": 254100
+    },
+    {
+      "epoch": 2.18793044077135,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.00011978184534713441,
+      "loss": 4.4684,
+      "step": 254150
+    },
+    {
+      "epoch": 2.1883608815426996,
+      "grad_norm": 3.609375,
+      "learning_rate": 0.00011975528795867774,
+      "loss": 4.0454,
+      "step": 254200
+    },
+    {
+      "epoch": 2.1887913223140494,
+      "grad_norm": 3.171875,
+      "learning_rate": 0.00011972872912022541,
+      "loss": 4.4463,
+      "step": 254250
+    },
+    {
+      "epoch": 2.1892217630853996,
+      "grad_norm": 3.046875,
+      "learning_rate": 0.00011970216883372683,
+      "loss": 4.3833,
+      "step": 254300
+    },
+    {
+      "epoch": 2.1896522038567494,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.00011967560710113143,
+      "loss": 4.3826,
+      "step": 254350
+    },
+    {
+      "epoch": 2.190082644628099,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00011964904392438883,
+      "loss": 4.7634,
+      "step": 254400
+    },
+    {
+      "epoch": 2.190513085399449,
+      "grad_norm": 3.328125,
+      "learning_rate": 0.00011962247930544866,
+      "loss": 4.87,
+      "step": 254450
+    },
+    {
+      "epoch": 2.190943526170799,
+      "grad_norm": 4.65625,
+      "learning_rate": 0.00011959591324626071,
+      "loss": 4.3389,
+      "step": 254500
+    },
+    {
+      "epoch": 2.1913739669421486,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.00011956934574877494,
+      "loss": 4.2565,
+      "step": 254550
+    },
+    {
+      "epoch": 2.191804407713499,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00011954277681494126,
+      "loss": 4.7334,
+      "step": 254600
+    },
+    {
+      "epoch": 2.1922348484848486,
+      "grad_norm": 2.296875,
+      "learning_rate": 0.00011951620644670982,
+      "loss": 4.2053,
+      "step": 254650
+    },
+    {
+      "epoch": 2.1926652892561984,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.00011948963464603084,
+      "loss": 4.2323,
+      "step": 254700
+    },
+    {
+      "epoch": 2.193095730027548,
+      "grad_norm": 3.40625,
+      "learning_rate": 0.00011946306141485457,
+      "loss": 4.4563,
+      "step": 254750
+    },
+    {
+      "epoch": 2.193526170798898,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00011943648675513148,
+      "loss": 4.5507,
+      "step": 254800
+    },
+    {
+      "epoch": 2.193956611570248,
+      "grad_norm": 3.859375,
+      "learning_rate": 0.00011940991066881208,
+      "loss": 4.3949,
+      "step": 254850
+    },
+    {
+      "epoch": 2.1943870523415976,
+      "grad_norm": 1.2578125,
+      "learning_rate": 0.00011938333315784702,
+      "loss": 4.3002,
+      "step": 254900
+    },
+    {
+      "epoch": 2.194817493112948,
+      "grad_norm": 4.0625,
+      "learning_rate": 0.000119356754224187,
+      "loss": 4.2587,
+      "step": 254950
+    },
+    {
+      "epoch": 2.1952479338842976,
+      "grad_norm": 7.03125,
+      "learning_rate": 0.00011933017386978284,
+      "loss": 4.3259,
+      "step": 255000
+    },
+    {
+      "epoch": 2.1952479338842976,
+      "eval_loss": 5.025729656219482,
+      "eval_runtime": 22.1733,
+      "eval_samples_per_second": 28.864,
+      "eval_steps_per_second": 14.432,
+      "eval_tts_loss": 7.315380834432901,
+      "step": 255000
+    },
+    {
+      "epoch": 2.1956783746556474,
+      "grad_norm": 3.65625,
+      "learning_rate": 0.00011930359209658555,
+      "loss": 4.4452,
+      "step": 255050
+    },
+    {
+      "epoch": 2.196108815426997,
+      "grad_norm": 1.78125,
+      "learning_rate": 0.00011927700890654609,
+      "loss": 4.5726,
+      "step": 255100
+    },
+    {
+      "epoch": 2.196539256198347,
+      "grad_norm": 3.453125,
+      "learning_rate": 0.00011925042430161568,
+      "loss": 4.2654,
+      "step": 255150
+    },
+    {
+      "epoch": 2.196969696969697,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.00011922383828374555,
+      "loss": 4.309,
+      "step": 255200
+    },
+    {
+      "epoch": 2.197400137741047,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00011919725085488704,
+      "loss": 4.2618,
+      "step": 255250
+    },
+    {
+      "epoch": 2.197830578512397,
+      "grad_norm": 1.3515625,
+      "learning_rate": 0.00011917066201699161,
+      "loss": 3.8487,
+      "step": 255300
+    },
+    {
+      "epoch": 2.1982610192837466,
+      "grad_norm": 3.15625,
+      "learning_rate": 0.00011914407177201083,
+      "loss": 4.6028,
+      "step": 255350
+    },
+    {
+      "epoch": 2.1986914600550964,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.0001191174801218964,
+      "loss": 3.9963,
+      "step": 255400
+    },
+    {
+      "epoch": 2.199121900826446,
+      "grad_norm": 2.0,
+      "learning_rate": 0.00011909088706860005,
+      "loss": 4.3256,
+      "step": 255450
+    },
+    {
+      "epoch": 2.199552341597796,
+      "grad_norm": 1.2578125,
+      "learning_rate": 0.00011906429261407366,
+      "loss": 4.2647,
+      "step": 255500
+    },
+    {
+      "epoch": 2.199982782369146,
+      "grad_norm": 3.265625,
+      "learning_rate": 0.00011903769676026921,
+      "loss": 4.1903,
+      "step": 255550
+    },
+    {
+      "epoch": 2.200413223140496,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.0001190110995091388,
+      "loss": 4.6628,
+      "step": 255600
+    },
+    {
+      "epoch": 2.200843663911846,
+      "grad_norm": 4.21875,
+      "learning_rate": 0.00011898450086263455,
+      "loss": 4.3471,
+      "step": 255650
+    },
+    {
+      "epoch": 2.2012741046831956,
+      "grad_norm": 2.625,
+      "learning_rate": 0.0001189579008227088,
+      "loss": 4.4032,
+      "step": 255700
+    },
+    {
+      "epoch": 2.2017045454545454,
+      "grad_norm": 4.875,
+      "learning_rate": 0.00011893129939131395,
+      "loss": 4.3063,
+      "step": 255750
+    },
+    {
+      "epoch": 2.202134986225895,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.00011890469657040242,
+      "loss": 4.4818,
+      "step": 255800
+    },
+    {
+      "epoch": 2.202565426997245,
+      "grad_norm": 4.5,
+      "learning_rate": 0.00011887809236192685,
+      "loss": 4.196,
+      "step": 255850
+    },
+    {
+      "epoch": 2.2029958677685952,
+      "grad_norm": 3.421875,
+      "learning_rate": 0.00011885148676783991,
+      "loss": 4.292,
+      "step": 255900
+    },
+    {
+      "epoch": 2.203426308539945,
+      "grad_norm": 3.640625,
+      "learning_rate": 0.00011882487979009443,
+      "loss": 4.293,
+      "step": 255950
+    },
+    {
+      "epoch": 2.203856749311295,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.00011879827143064322,
+      "loss": 4.5005,
+      "step": 256000
+    },
+    {
+      "epoch": 2.2042871900826446,
+      "grad_norm": 4.78125,
+      "learning_rate": 0.00011877166169143936,
+      "loss": 4.572,
+      "step": 256050
+    },
+    {
+      "epoch": 2.2047176308539944,
+      "grad_norm": 4.03125,
+      "learning_rate": 0.00011874505057443594,
+      "loss": 4.3167,
+      "step": 256100
+    },
+    {
+      "epoch": 2.205148071625344,
+      "grad_norm": 4.46875,
+      "learning_rate": 0.00011871843808158609,
+      "loss": 4.5634,
+      "step": 256150
+    },
+    {
+      "epoch": 2.205578512396694,
+      "grad_norm": 4.5,
+      "learning_rate": 0.00011869182421484319,
+      "loss": 4.4854,
+      "step": 256200
+    },
+    {
+      "epoch": 2.2060089531680442,
+      "grad_norm": 5.59375,
+      "learning_rate": 0.00011866520897616063,
+      "loss": 4.331,
+      "step": 256250
+    },
+    {
+      "epoch": 2.206439393939394,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.00011863859236749186,
+      "loss": 4.5912,
+      "step": 256300
+    },
+    {
+      "epoch": 2.206869834710744,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.00011861197439079053,
+      "loss": 4.2557,
+      "step": 256350
+    },
+    {
+      "epoch": 2.2073002754820936,
+      "grad_norm": 2.25,
+      "learning_rate": 0.00011858535504801031,
+      "loss": 4.0516,
+      "step": 256400
+    },
+    {
+      "epoch": 2.2077307162534434,
+      "grad_norm": 3.078125,
+      "learning_rate": 0.00011855873434110503,
+      "loss": 4.2145,
+      "step": 256450
+    },
+    {
+      "epoch": 2.208161157024793,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.00011853211227202856,
+      "loss": 4.7137,
+      "step": 256500
+    },
+    {
+      "epoch": 2.2085915977961434,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.00011850548884273493,
+      "loss": 4.6057,
+      "step": 256550
+    },
+    {
+      "epoch": 2.2090220385674932,
+      "grad_norm": 3.078125,
+      "learning_rate": 0.00011847886405517828,
+      "loss": 4.1449,
+      "step": 256600
+    },
+    {
+      "epoch": 2.209452479338843,
+      "grad_norm": 3.046875,
+      "learning_rate": 0.0001184522379113127,
+      "loss": 4.3742,
+      "step": 256650
+    },
+    {
+      "epoch": 2.209882920110193,
+      "grad_norm": 4.59375,
+      "learning_rate": 0.00011842561041309257,
+      "loss": 4.622,
+      "step": 256700
+    },
+    {
+      "epoch": 2.2103133608815426,
+      "grad_norm": 1.890625,
+      "learning_rate": 0.00011839898156247231,
+      "loss": 4.4993,
+      "step": 256750
+    },
+    {
+      "epoch": 2.2107438016528924,
+      "grad_norm": 5.375,
+      "learning_rate": 0.00011837235136140636,
+      "loss": 4.4557,
+      "step": 256800
+    },
+    {
+      "epoch": 2.211174242424242,
+      "grad_norm": 3.390625,
+      "learning_rate": 0.00011834571981184939,
+      "loss": 3.9743,
+      "step": 256850
+    },
+    {
+      "epoch": 2.2116046831955924,
+      "grad_norm": 3.65625,
+      "learning_rate": 0.00011831908691575602,
+      "loss": 4.6455,
+      "step": 256900
+    },
+    {
+      "epoch": 2.2120351239669422,
+      "grad_norm": 4.71875,
+      "learning_rate": 0.00011829245267508109,
+      "loss": 4.3311,
+      "step": 256950
+    },
+    {
+      "epoch": 2.212465564738292,
+      "grad_norm": 0.87109375,
+      "learning_rate": 0.00011826581709177951,
+      "loss": 4.3819,
+      "step": 257000
+    },
+    {
+      "epoch": 2.212896005509642,
+      "grad_norm": 3.140625,
+      "learning_rate": 0.00011823918016780622,
+      "loss": 4.3464,
+      "step": 257050
+    },
+    {
+      "epoch": 2.2133264462809916,
+      "grad_norm": 3.09375,
+      "learning_rate": 0.00011821254190511637,
+      "loss": 4.4691,
+      "step": 257100
+    },
+    {
+      "epoch": 2.2137568870523414,
+      "grad_norm": 1.7578125,
+      "learning_rate": 0.00011818590230566516,
+      "loss": 4.3192,
+      "step": 257150
+    },
+    {
+      "epoch": 2.2141873278236917,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.00011815926137140778,
+      "loss": 4.4428,
+      "step": 257200
+    },
+    {
+      "epoch": 2.2146177685950414,
+      "grad_norm": 2.875,
+      "learning_rate": 0.00011813261910429974,
+      "loss": 4.4558,
+      "step": 257250
+    },
+    {
+      "epoch": 2.2150482093663912,
+      "grad_norm": 3.453125,
+      "learning_rate": 0.00011810597550629645,
+      "loss": 4.2951,
+      "step": 257300
+    },
+    {
+      "epoch": 2.215478650137741,
+      "grad_norm": 4.125,
+      "learning_rate": 0.00011807933057935352,
+      "loss": 4.1378,
+      "step": 257350
+    },
+    {
+      "epoch": 2.215909090909091,
+      "grad_norm": 4.6875,
+      "learning_rate": 0.00011805268432542662,
+      "loss": 4.3002,
+      "step": 257400
+    },
+    {
+      "epoch": 2.2163395316804406,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.00011802603674647155,
+      "loss": 4.3459,
+      "step": 257450
+    },
+    {
+      "epoch": 2.2167699724517904,
+      "grad_norm": 3.375,
+      "learning_rate": 0.00011799938784444415,
+      "loss": 4.1584,
+      "step": 257500
+    },
+    {
+      "epoch": 2.2172004132231407,
+      "grad_norm": 2.578125,
+      "learning_rate": 0.00011797273762130041,
+      "loss": 4.0454,
+      "step": 257550
+    },
+    {
+      "epoch": 2.2176308539944904,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.0001179460860789964,
+      "loss": 4.5808,
+      "step": 257600
+    },
+    {
+      "epoch": 2.2180612947658402,
+      "grad_norm": 3.90625,
+      "learning_rate": 0.00011791943321948828,
+      "loss": 4.4137,
+      "step": 257650
+    },
+    {
+      "epoch": 2.21849173553719,
+      "grad_norm": 3.125,
+      "learning_rate": 0.0001178927790447323,
+      "loss": 4.6564,
+      "step": 257700
+    },
+    {
+      "epoch": 2.21892217630854,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00011786612355668488,
+      "loss": 4.3404,
+      "step": 257750
+    },
+    {
+      "epoch": 2.2193526170798896,
+      "grad_norm": 4.96875,
+      "learning_rate": 0.0001178394667573024,
+      "loss": 4.0181,
+      "step": 257800
+    },
+    {
+      "epoch": 2.21978305785124,
+      "grad_norm": 2.375,
+      "learning_rate": 0.00011781280864854143,
+      "loss": 4.3928,
+      "step": 257850
+    },
+    {
+      "epoch": 2.2202134986225897,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00011778614923235866,
+      "loss": 4.2034,
+      "step": 257900
+    },
+    {
+      "epoch": 2.2206439393939394,
+      "grad_norm": 1.8359375,
+      "learning_rate": 0.00011775948851071079,
+      "loss": 4.3024,
+      "step": 257950
+    },
+    {
+      "epoch": 2.2210743801652892,
+      "grad_norm": 3.265625,
+      "learning_rate": 0.00011773282648555465,
+      "loss": 4.1851,
+      "step": 258000
+    },
+    {
+      "epoch": 2.2210743801652892,
+      "eval_loss": 5.025103569030762,
+      "eval_runtime": 24.1477,
+      "eval_samples_per_second": 26.504,
+      "eval_steps_per_second": 13.252,
+      "eval_tts_loss": 7.391507342525869,
+      "step": 258000
+    },
+    {
+      "epoch": 2.221504820936639,
+      "grad_norm": 3.234375,
+      "learning_rate": 0.00011770616315884726,
+      "loss": 4.6283,
+      "step": 258050
+    },
+    {
+      "epoch": 2.221935261707989,
+      "grad_norm": 3.25,
+      "learning_rate": 0.00011767949853254554,
+      "loss": 4.5709,
+      "step": 258100
+    },
+    {
+      "epoch": 2.2223657024793386,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.00011765283260860671,
+      "loss": 4.2582,
+      "step": 258150
+    },
+    {
+      "epoch": 2.222796143250689,
+      "grad_norm": 3.359375,
+      "learning_rate": 0.0001176261653889879,
+      "loss": 4.1063,
+      "step": 258200
+    },
+    {
+      "epoch": 2.2232265840220387,
+      "grad_norm": 0.9453125,
+      "learning_rate": 0.00011759949687564646,
+      "loss": 4.2151,
+      "step": 258250
+    },
+    {
+      "epoch": 2.2236570247933884,
+      "grad_norm": 5.3125,
+      "learning_rate": 0.0001175728270705399,
+      "loss": 4.4871,
+      "step": 258300
+    },
+    {
+      "epoch": 2.2240874655647382,
+      "grad_norm": 4.25,
+      "learning_rate": 0.00011754615597562557,
+      "loss": 4.1392,
+      "step": 258350
+    },
+    {
+      "epoch": 2.224517906336088,
+      "grad_norm": 4.59375,
+      "learning_rate": 0.00011751948359286117,
+      "loss": 4.1695,
+      "step": 258400
+    },
+    {
+      "epoch": 2.224948347107438,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.00011749280992420436,
+      "loss": 4.7462,
+      "step": 258450
+    },
+    {
+      "epoch": 2.225378787878788,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00011746613497161294,
+      "loss": 4.5731,
+      "step": 258500
+    },
+    {
+      "epoch": 2.225809228650138,
+      "grad_norm": 3.359375,
+      "learning_rate": 0.00011743945873704478,
+      "loss": 4.0189,
+      "step": 258550
+    },
+    {
+      "epoch": 2.2262396694214877,
+      "grad_norm": 4.15625,
+      "learning_rate": 0.00011741278122245789,
+      "loss": 4.1192,
+      "step": 258600
+    },
+    {
+      "epoch": 2.2266701101928374,
+      "grad_norm": 1.4453125,
+      "learning_rate": 0.00011738610242981034,
+      "loss": 4.2686,
+      "step": 258650
+    },
+    {
+      "epoch": 2.2271005509641872,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.00011735942236106023,
+      "loss": 4.2397,
+      "step": 258700
+    },
+    {
+      "epoch": 2.227530991735537,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00011733274101816589,
+      "loss": 4.4109,
+      "step": 258750
+    },
+    {
+      "epoch": 2.227961432506887,
+      "grad_norm": 3.578125,
+      "learning_rate": 0.00011730605840308566,
+      "loss": 4.3267,
+      "step": 258800
+    },
+    {
+      "epoch": 2.228391873278237,
+      "grad_norm": 0.7265625,
+      "learning_rate": 0.00011727937451777797,
+      "loss": 4.4647,
+      "step": 258850
+    },
+    {
+      "epoch": 2.228822314049587,
+      "grad_norm": 4.84375,
+      "learning_rate": 0.00011725268936420138,
+      "loss": 4.3085,
+      "step": 258900
+    },
+    {
+      "epoch": 2.2292527548209367,
+      "grad_norm": 4.71875,
+      "learning_rate": 0.00011722600294431449,
+      "loss": 4.588,
+      "step": 258950
+    },
+    {
+      "epoch": 2.2296831955922864,
+      "grad_norm": 3.90625,
+      "learning_rate": 0.00011719931526007606,
+      "loss": 4.2792,
+      "step": 259000
+    },
+    {
+      "epoch": 2.2301136363636362,
+      "grad_norm": 2.4375,
+      "learning_rate": 0.00011717262631344492,
+      "loss": 4.5527,
+      "step": 259050
+    },
+    {
+      "epoch": 2.230544077134986,
+      "grad_norm": 1.6796875,
+      "learning_rate": 0.00011714593610637992,
+      "loss": 4.4178,
+      "step": 259100
+    },
+    {
+      "epoch": 2.2309745179063363,
+      "grad_norm": 1.7109375,
+      "learning_rate": 0.0001171192446408401,
+      "loss": 4.8266,
+      "step": 259150
+    },
+    {
+      "epoch": 2.231404958677686,
+      "grad_norm": 4.90625,
+      "learning_rate": 0.00011709255191878458,
+      "loss": 3.7358,
+      "step": 259200
+    },
+    {
+      "epoch": 2.231835399449036,
+      "grad_norm": 3.15625,
+      "learning_rate": 0.00011706585794217248,
+      "loss": 4.3325,
+      "step": 259250
+    },
+    {
+      "epoch": 2.2322658402203857,
+      "grad_norm": 5.53125,
+      "learning_rate": 0.00011703916271296318,
+      "loss": 4.5312,
+      "step": 259300
+    },
+    {
+      "epoch": 2.2326962809917354,
+      "grad_norm": 1.7890625,
+      "learning_rate": 0.00011701246623311597,
+      "loss": 4.2548,
+      "step": 259350
+    },
+    {
+      "epoch": 2.2331267217630852,
+      "grad_norm": 4.15625,
+      "learning_rate": 0.00011698576850459036,
+      "loss": 4.7601,
+      "step": 259400
+    },
+    {
+      "epoch": 2.233557162534435,
+      "grad_norm": 2.5,
+      "learning_rate": 0.00011695906952934585,
+      "loss": 4.361,
+      "step": 259450
+    },
+    {
+      "epoch": 2.2339876033057853,
+      "grad_norm": 3.21875,
+      "learning_rate": 0.00011693236930934217,
+      "loss": 4.277,
+      "step": 259500
+    },
+    {
+      "epoch": 2.234418044077135,
+      "grad_norm": 1.78125,
+      "learning_rate": 0.000116905667846539,
+      "loss": 4.5201,
+      "step": 259550
+    },
+    {
+      "epoch": 2.234848484848485,
+      "grad_norm": 4.6875,
+      "learning_rate": 0.00011687896514289616,
+      "loss": 4.5688,
+      "step": 259600
+    },
+    {
+      "epoch": 2.2352789256198347,
+      "grad_norm": 3.4375,
+      "learning_rate": 0.00011685226120037363,
+      "loss": 4.4471,
+      "step": 259650
+    },
+    {
+      "epoch": 2.2357093663911844,
+      "grad_norm": 4.46875,
+      "learning_rate": 0.00011682555602093134,
+      "loss": 4.4415,
+      "step": 259700
+    },
+    {
+      "epoch": 2.2361398071625342,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00011679884960652947,
+      "loss": 4.5308,
+      "step": 259750
+    },
+    {
+      "epoch": 2.2365702479338845,
+      "grad_norm": 3.03125,
+      "learning_rate": 0.00011677214195912815,
+      "loss": 4.6011,
+      "step": 259800
+    },
+    {
+      "epoch": 2.2370006887052343,
+      "grad_norm": 3.34375,
+      "learning_rate": 0.00011674543308068777,
+      "loss": 4.5214,
+      "step": 259850
+    },
+    {
+      "epoch": 2.237431129476584,
+      "grad_norm": 6.3125,
+      "learning_rate": 0.00011671872297316857,
+      "loss": 4.1105,
+      "step": 259900
+    },
+    {
+      "epoch": 2.237861570247934,
+      "grad_norm": 3.671875,
+      "learning_rate": 0.0001166920116385311,
+      "loss": 4.0888,
+      "step": 259950
+    },
+    {
+      "epoch": 2.2382920110192837,
+      "grad_norm": 3.0,
+      "learning_rate": 0.00011666529907873587,
+      "loss": 4.1547,
+      "step": 260000
+    },
+    {
+      "epoch": 2.2387224517906334,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.00011663858529574354,
+      "loss": 4.5505,
+      "step": 260050
+    },
+    {
+      "epoch": 2.2391528925619832,
+      "grad_norm": 3.71875,
+      "learning_rate": 0.00011661187029151488,
+      "loss": 4.7665,
+      "step": 260100
+    },
+    {
+      "epoch": 2.2395833333333335,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00011658515406801065,
+      "loss": 4.4342,
+      "step": 260150
+    },
+    {
+      "epoch": 2.2400137741046833,
+      "grad_norm": 0.99609375,
+      "learning_rate": 0.00011655843662719183,
+      "loss": 4.1525,
+      "step": 260200
+    },
+    {
+      "epoch": 2.240444214876033,
+      "grad_norm": 4.6875,
+      "learning_rate": 0.00011653171797101936,
+      "loss": 4.8385,
+      "step": 260250
+    },
+    {
+      "epoch": 2.240874655647383,
+      "grad_norm": 3.8125,
+      "learning_rate": 0.00011650499810145437,
+      "loss": 4.3841,
+      "step": 260300
+    },
+    {
+      "epoch": 2.2413050964187327,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00011647827702045801,
+      "loss": 4.4282,
+      "step": 260350
+    },
+    {
+      "epoch": 2.2417355371900825,
+      "grad_norm": 3.5,
+      "learning_rate": 0.00011645155472999161,
+      "loss": 4.2831,
+      "step": 260400
+    },
+    {
+      "epoch": 2.2421659779614327,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.00011642483123201647,
+      "loss": 4.017,
+      "step": 260450
+    },
+    {
+      "epoch": 2.2425964187327825,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00011639810652849403,
+      "loss": 3.8839,
+      "step": 260500
+    },
+    {
+      "epoch": 2.2430268595041323,
+      "grad_norm": 1.890625,
+      "learning_rate": 0.00011637138062138588,
+      "loss": 3.9701,
+      "step": 260550
+    },
+    {
+      "epoch": 2.243457300275482,
+      "grad_norm": 3.5625,
+      "learning_rate": 0.00011634465351265364,
+      "loss": 4.311,
+      "step": 260600
+    },
+    {
+      "epoch": 2.243887741046832,
+      "grad_norm": 4.375,
+      "learning_rate": 0.00011631792520425895,
+      "loss": 4.4238,
+      "step": 260650
+    },
+    {
+      "epoch": 2.2443181818181817,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00011629119569816369,
+      "loss": 4.2005,
+      "step": 260700
+    },
+    {
+      "epoch": 2.2447486225895315,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.0001162644649963297,
+      "loss": 4.2634,
+      "step": 260750
+    },
+    {
+      "epoch": 2.2451790633608817,
+      "grad_norm": 3.0625,
+      "learning_rate": 0.00011623773310071897,
+      "loss": 4.2966,
+      "step": 260800
+    },
+    {
+      "epoch": 2.2456095041322315,
+      "grad_norm": 3.1875,
+      "learning_rate": 0.0001162110000132936,
+      "loss": 4.3746,
+      "step": 260850
+    },
+    {
+      "epoch": 2.2460399449035813,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.00011618426573601568,
+      "loss": 4.6821,
+      "step": 260900
+    },
+    {
+      "epoch": 2.246470385674931,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.00011615753027084747,
+      "loss": 4.4889,
+      "step": 260950
+    },
+    {
+      "epoch": 2.246900826446281,
+      "grad_norm": 3.40625,
+      "learning_rate": 0.00011613079361975132,
+      "loss": 4.1965,
+      "step": 261000
+    },
+    {
+      "epoch": 2.246900826446281,
+      "eval_loss": 5.025378227233887,
+      "eval_runtime": 25.0094,
+      "eval_samples_per_second": 25.59,
+      "eval_steps_per_second": 12.795,
+      "eval_tts_loss": 7.371634570443995,
+      "step": 261000
+    },
+    {
+      "epoch": 2.2473312672176307,
+      "grad_norm": 1.5703125,
+      "learning_rate": 0.00011610405578468964,
+      "loss": 4.2573,
+      "step": 261050
+    },
+    {
+      "epoch": 2.247761707988981,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.00011607731676762487,
+      "loss": 4.4789,
+      "step": 261100
+    },
+    {
+      "epoch": 2.2481921487603307,
+      "grad_norm": 2.625,
+      "learning_rate": 0.00011605057657051967,
+      "loss": 4.5357,
+      "step": 261150
+    },
+    {
+      "epoch": 2.2486225895316805,
+      "grad_norm": 1.0546875,
+      "learning_rate": 0.00011602383519533669,
+      "loss": 4.294,
+      "step": 261200
+    },
+    {
+      "epoch": 2.2490530303030303,
+      "grad_norm": 0.6953125,
+      "learning_rate": 0.00011599709264403866,
+      "loss": 4.2997,
+      "step": 261250
+    },
+    {
+      "epoch": 2.24948347107438,
+      "grad_norm": 2.5,
+      "learning_rate": 0.00011597034891858844,
+      "loss": 4.3767,
+      "step": 261300
+    },
+    {
+      "epoch": 2.24991391184573,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00011594360402094899,
+      "loss": 4.4321,
+      "step": 261350
+    },
+    {
+      "epoch": 2.2503443526170797,
+      "grad_norm": 3.234375,
+      "learning_rate": 0.00011591685795308334,
+      "loss": 4.4887,
+      "step": 261400
+    },
+    {
+      "epoch": 2.25077479338843,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.0001158901107169545,
+      "loss": 4.2412,
+      "step": 261450
+    },
+    {
+      "epoch": 2.2512052341597797,
+      "grad_norm": 3.125,
+      "learning_rate": 0.00011586336231452576,
+      "loss": 4.3327,
+      "step": 261500
+    },
+    {
+      "epoch": 2.2516356749311295,
+      "grad_norm": 4.8125,
+      "learning_rate": 0.00011583661274776036,
+      "loss": 4.2752,
+      "step": 261550
+    },
+    {
+      "epoch": 2.2520661157024793,
+      "grad_norm": 6.875,
+      "learning_rate": 0.00011580986201862163,
+      "loss": 4.8654,
+      "step": 261600
+    },
+    {
+      "epoch": 2.252496556473829,
+      "grad_norm": 3.65625,
+      "learning_rate": 0.00011578311012907306,
+      "loss": 4.4581,
+      "step": 261650
+    },
+    {
+      "epoch": 2.2529269972451793,
+      "grad_norm": 1.6796875,
+      "learning_rate": 0.00011575635708107815,
+      "loss": 3.9214,
+      "step": 261700
+    },
+    {
+      "epoch": 2.253357438016529,
+      "grad_norm": 3.125,
+      "learning_rate": 0.00011572960287660057,
+      "loss": 4.491,
+      "step": 261750
+    },
+    {
+      "epoch": 2.253787878787879,
+      "grad_norm": 3.234375,
+      "learning_rate": 0.00011570284751760395,
+      "loss": 4.59,
+      "step": 261800
+    },
+    {
+      "epoch": 2.2542183195592287,
+      "grad_norm": 3.546875,
+      "learning_rate": 0.0001156760910060521,
+      "loss": 4.3888,
+      "step": 261850
+    },
+    {
+      "epoch": 2.2546487603305785,
+      "grad_norm": 3.1875,
+      "learning_rate": 0.00011564933334390893,
+      "loss": 4.4528,
+      "step": 261900
+    },
+    {
+      "epoch": 2.2550792011019283,
+      "grad_norm": 4.40625,
+      "learning_rate": 0.00011562257453313833,
+      "loss": 4.2057,
+      "step": 261950
+    },
+    {
+      "epoch": 2.255509641873278,
+      "grad_norm": 4.5625,
+      "learning_rate": 0.00011559581457570441,
+      "loss": 4.2181,
+      "step": 262000
+    },
+    {
+      "epoch": 2.255940082644628,
+      "grad_norm": 3.921875,
+      "learning_rate": 0.00011556905347357121,
+      "loss": 4.5553,
+      "step": 262050
+    },
+    {
+      "epoch": 2.256370523415978,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.00011554229122870302,
+      "loss": 4.434,
+      "step": 262100
+    },
+    {
+      "epoch": 2.256800964187328,
+      "grad_norm": 3.21875,
+      "learning_rate": 0.00011551552784306408,
+      "loss": 4.2748,
+      "step": 262150
+    },
+    {
+      "epoch": 2.2572314049586777,
+      "grad_norm": 4.03125,
+      "learning_rate": 0.00011548876331861875,
+      "loss": 4.2321,
+      "step": 262200
+    },
+    {
+      "epoch": 2.2576618457300275,
+      "grad_norm": 2.296875,
+      "learning_rate": 0.00011546199765733158,
+      "loss": 4.338,
+      "step": 262250
+    },
+    {
+      "epoch": 2.2580922865013773,
+      "grad_norm": 3.140625,
+      "learning_rate": 0.00011543523086116701,
+      "loss": 4.031,
+      "step": 262300
+    },
+    {
+      "epoch": 2.2585227272727275,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.00011540846293208969,
+      "loss": 4.0706,
+      "step": 262350
+    },
+    {
+      "epoch": 2.2589531680440773,
+      "grad_norm": 3.40625,
+      "learning_rate": 0.00011538169387206436,
+      "loss": 4.5887,
+      "step": 262400
+    },
+    {
+      "epoch": 2.259383608815427,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00011535492368305583,
+      "loss": 4.5115,
+      "step": 262450
+    },
+    {
+      "epoch": 2.259814049586777,
+      "grad_norm": 2.75,
+      "learning_rate": 0.0001153281523670289,
+      "loss": 4.2577,
+      "step": 262500
+    },
+    {
+      "epoch": 2.2602444903581267,
+      "grad_norm": 3.75,
+      "learning_rate": 0.00011530137992594856,
+      "loss": 4.6653,
+      "step": 262550
+    },
+    {
+      "epoch": 2.2606749311294765,
+      "grad_norm": 4.6875,
+      "learning_rate": 0.00011527460636177987,
+      "loss": 4.4077,
+      "step": 262600
+    },
+    {
+      "epoch": 2.2611053719008263,
+      "grad_norm": 2.390625,
+      "learning_rate": 0.00011524783167648793,
+      "loss": 4.6283,
+      "step": 262650
+    },
+    {
+      "epoch": 2.261535812672176,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00011522105587203795,
+      "loss": 4.2903,
+      "step": 262700
+    },
+    {
+      "epoch": 2.2619662534435263,
+      "grad_norm": 4.34375,
+      "learning_rate": 0.00011519427895039524,
+      "loss": 4.0598,
+      "step": 262750
+    },
+    {
+      "epoch": 2.262396694214876,
+      "grad_norm": 2.875,
+      "learning_rate": 0.00011516750091352513,
+      "loss": 4.5757,
+      "step": 262800
+    },
+    {
+      "epoch": 2.262827134986226,
+      "grad_norm": 3.609375,
+      "learning_rate": 0.00011514072176339308,
+      "loss": 4.6233,
+      "step": 262850
+    },
+    {
+      "epoch": 2.2632575757575757,
+      "grad_norm": 3.375,
+      "learning_rate": 0.00011511394150196464,
+      "loss": 4.5223,
+      "step": 262900
+    },
+    {
+      "epoch": 2.2636880165289255,
+      "grad_norm": 3.1875,
+      "learning_rate": 0.00011508716013120539,
+      "loss": 4.5717,
+      "step": 262950
+    },
+    {
+      "epoch": 2.2641184573002757,
+      "grad_norm": 1.875,
+      "learning_rate": 0.00011506037765308106,
+      "loss": 4.1961,
+      "step": 263000
+    },
+    {
+      "epoch": 2.2645488980716255,
+      "grad_norm": 2.375,
+      "learning_rate": 0.00011503359406955743,
+      "loss": 4.1396,
+      "step": 263050
+    },
+    {
+      "epoch": 2.2649793388429753,
+      "grad_norm": 3.953125,
+      "learning_rate": 0.00011500680938260033,
+      "loss": 4.0465,
+      "step": 263100
+    },
+    {
+      "epoch": 2.265409779614325,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00011498002359417573,
+      "loss": 4.1431,
+      "step": 263150
+    },
+    {
+      "epoch": 2.265840220385675,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.00011495323670624958,
+      "loss": 4.1064,
+      "step": 263200
+    },
+    {
+      "epoch": 2.2662706611570247,
+      "grad_norm": 4.90625,
+      "learning_rate": 0.00011492644872078804,
+      "loss": 4.2736,
+      "step": 263250
+    },
+    {
+      "epoch": 2.2667011019283745,
+      "grad_norm": 4.1875,
+      "learning_rate": 0.00011489965963975733,
+      "loss": 4.3057,
+      "step": 263300
+    },
+    {
+      "epoch": 2.2671315426997243,
+      "grad_norm": 3.671875,
+      "learning_rate": 0.00011487286946512362,
+      "loss": 4.1887,
+      "step": 263350
+    },
+    {
+      "epoch": 2.2675619834710745,
+      "grad_norm": 3.6875,
+      "learning_rate": 0.0001148460781988533,
+      "loss": 4.6514,
+      "step": 263400
+    },
+    {
+      "epoch": 2.2679924242424243,
+      "grad_norm": 3.46875,
+      "learning_rate": 0.00011481928584291277,
+      "loss": 4.422,
+      "step": 263450
+    },
+    {
+      "epoch": 2.268422865013774,
+      "grad_norm": 3.046875,
+      "learning_rate": 0.00011479249239926854,
+      "loss": 4.3908,
+      "step": 263500
+    },
+    {
+      "epoch": 2.268853305785124,
+      "grad_norm": 3.359375,
+      "learning_rate": 0.00011476569786988722,
+      "loss": 4.4813,
+      "step": 263550
+    },
+    {
+      "epoch": 2.2692837465564737,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.0001147389022567354,
+      "loss": 4.2909,
+      "step": 263600
+    },
+    {
+      "epoch": 2.269714187327824,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.00011471210556177992,
+      "loss": 4.3768,
+      "step": 263650
+    },
+    {
+      "epoch": 2.2701446280991737,
+      "grad_norm": 1.5078125,
+      "learning_rate": 0.00011468530778698751,
+      "loss": 4.2772,
+      "step": 263700
+    },
+    {
+      "epoch": 2.2705750688705235,
+      "grad_norm": 3.703125,
+      "learning_rate": 0.0001146585089343251,
+      "loss": 4.3767,
+      "step": 263750
+    },
+    {
+      "epoch": 2.2710055096418733,
+      "grad_norm": 4.59375,
+      "learning_rate": 0.00011463170900575972,
+      "loss": 4.4768,
+      "step": 263800
+    },
+    {
+      "epoch": 2.271435950413223,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.00011460490800325836,
+      "loss": 4.2622,
+      "step": 263850
+    },
+    {
+      "epoch": 2.271866391184573,
+      "grad_norm": 2.625,
+      "learning_rate": 0.00011457810592878815,
+      "loss": 4.4093,
+      "step": 263900
+    },
+    {
+      "epoch": 2.2722968319559227,
+      "grad_norm": 3.765625,
+      "learning_rate": 0.00011455130278431635,
+      "loss": 4.5669,
+      "step": 263950
+    },
+    {
+      "epoch": 2.2727272727272725,
+      "grad_norm": 4.59375,
+      "learning_rate": 0.00011452449857181022,
+      "loss": 4.2765,
+      "step": 264000
+    },
+    {
+      "epoch": 2.2727272727272725,
+      "eval_loss": 5.020726203918457,
+      "eval_runtime": 24.0931,
+      "eval_samples_per_second": 26.564,
+      "eval_steps_per_second": 13.282,
+      "eval_tts_loss": 7.375918081878187,
+      "step": 264000
+    },
+    {
+      "epoch": 2.2731577134986227,
+      "grad_norm": 2.3125,
+      "learning_rate": 0.00011449769329323719,
+      "loss": 4.144,
+      "step": 264050
+    },
+    {
+      "epoch": 2.2735881542699725,
+      "grad_norm": 2.5,
+      "learning_rate": 0.00011447088695056461,
+      "loss": 4.5033,
+      "step": 264100
+    },
+    {
+      "epoch": 2.2740185950413223,
+      "grad_norm": 0.78515625,
+      "learning_rate": 0.00011444407954576008,
+      "loss": 4.4551,
+      "step": 264150
+    },
+    {
+      "epoch": 2.274449035812672,
+      "grad_norm": 2.0,
+      "learning_rate": 0.00011441727108079118,
+      "loss": 4.5068,
+      "step": 264200
+    },
+    {
+      "epoch": 2.274879476584022,
+      "grad_norm": 2.421875,
+      "learning_rate": 0.00011439046155762562,
+      "loss": 4.0674,
+      "step": 264250
+    },
+    {
+      "epoch": 2.275309917355372,
+      "grad_norm": 3.515625,
+      "learning_rate": 0.00011436365097823113,
+      "loss": 4.3386,
+      "step": 264300
+    },
+    {
+      "epoch": 2.275740358126722,
+      "grad_norm": 3.34375,
+      "learning_rate": 0.0001143368393445756,
+      "loss": 4.5215,
+      "step": 264350
+    },
+    {
+      "epoch": 2.2761707988980717,
+      "grad_norm": 2.296875,
+      "learning_rate": 0.00011431002665862684,
+      "loss": 4.0957,
+      "step": 264400
+    },
+    {
+      "epoch": 2.2766012396694215,
+      "grad_norm": 2.578125,
+      "learning_rate": 0.00011428321292235298,
+      "loss": 4.5167,
+      "step": 264450
+    },
+    {
+      "epoch": 2.2770316804407713,
+      "grad_norm": 3.34375,
+      "learning_rate": 0.000114256398137722,
+      "loss": 4.1995,
+      "step": 264500
+    },
+    {
+      "epoch": 2.277462121212121,
+      "grad_norm": 3.125,
+      "learning_rate": 0.00011422958230670204,
+      "loss": 4.5269,
+      "step": 264550
+    },
+    {
+      "epoch": 2.277892561983471,
+      "grad_norm": 3.078125,
+      "learning_rate": 0.00011420276543126139,
+      "loss": 4.659,
+      "step": 264600
+    },
+    {
+      "epoch": 2.2783230027548207,
+      "grad_norm": 3.234375,
+      "learning_rate": 0.00011417594751336831,
+      "loss": 4.4911,
+      "step": 264650
+    },
+    {
+      "epoch": 2.278753443526171,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00011414912855499118,
+      "loss": 4.1821,
+      "step": 264700
+    },
+    {
+      "epoch": 2.2791838842975207,
+      "grad_norm": 2.140625,
+      "learning_rate": 0.00011412230855809843,
+      "loss": 4.5744,
+      "step": 264750
+    },
+    {
+      "epoch": 2.2796143250688705,
+      "grad_norm": 2.0,
+      "learning_rate": 0.00011409548752465866,
+      "loss": 4.463,
+      "step": 264800
+    },
+    {
+      "epoch": 2.2800447658402203,
+      "grad_norm": 3.046875,
+      "learning_rate": 0.00011406866545664041,
+      "loss": 4.1471,
+      "step": 264850
+    },
+    {
+      "epoch": 2.28047520661157,
+      "grad_norm": 3.078125,
+      "learning_rate": 0.00011404184235601236,
+      "loss": 4.2949,
+      "step": 264900
+    },
+    {
+      "epoch": 2.2809056473829203,
+      "grad_norm": 2.125,
+      "learning_rate": 0.00011401501822474334,
+      "loss": 4.3699,
+      "step": 264950
+    },
+    {
+      "epoch": 2.28133608815427,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00011398819306480214,
+      "loss": 4.3122,
+      "step": 265000
+    },
+    {
+      "epoch": 2.28176652892562,
+      "grad_norm": 3.671875,
+      "learning_rate": 0.00011396136687815762,
+      "loss": 4.395,
+      "step": 265050
+    },
+    {
+      "epoch": 2.2821969696969697,
+      "grad_norm": 4.09375,
+      "learning_rate": 0.00011393453966677883,
+      "loss": 4.5916,
+      "step": 265100
+    },
+    {
+      "epoch": 2.2826274104683195,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.00011390771143263481,
+      "loss": 4.291,
+      "step": 265150
+    },
+    {
+      "epoch": 2.2830578512396693,
+      "grad_norm": 3.25,
+      "learning_rate": 0.0001138808821776947,
+      "loss": 4.1604,
+      "step": 265200
+    },
+    {
+      "epoch": 2.283488292011019,
+      "grad_norm": 3.296875,
+      "learning_rate": 0.00011385405190392769,
+      "loss": 4.2222,
+      "step": 265250
+    },
+    {
+      "epoch": 2.283918732782369,
+      "grad_norm": 4.40625,
+      "learning_rate": 0.00011382722061330306,
+      "loss": 4.3219,
+      "step": 265300
+    },
+    {
+      "epoch": 2.284349173553719,
+      "grad_norm": 2.625,
+      "learning_rate": 0.00011380038830779025,
+      "loss": 4.4364,
+      "step": 265350
+    },
+    {
+      "epoch": 2.284779614325069,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.00011377355498935857,
+      "loss": 3.9465,
+      "step": 265400
+    },
+    {
+      "epoch": 2.2852100550964187,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00011374672065997761,
+      "loss": 4.1925,
+      "step": 265450
+    },
+    {
+      "epoch": 2.2856404958677685,
+      "grad_norm": 3.796875,
+      "learning_rate": 0.00011371988532161697,
+      "loss": 3.9821,
+      "step": 265500
+    },
+    {
+      "epoch": 2.2860709366391183,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.00011369304897624622,
+      "loss": 4.3449,
+      "step": 265550
+    },
+    {
+      "epoch": 2.2865013774104685,
+      "grad_norm": 5.625,
+      "learning_rate": 0.00011366621162583515,
+      "loss": 4.4995,
+      "step": 265600
+    },
+    {
+      "epoch": 2.2869318181818183,
+      "grad_norm": 4.875,
+      "learning_rate": 0.00011363937327235352,
+      "loss": 4.4606,
+      "step": 265650
+    },
+    {
+      "epoch": 2.287362258953168,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.00011361253391777126,
+      "loss": 4.1662,
+      "step": 265700
+    },
+    {
+      "epoch": 2.287792699724518,
+      "grad_norm": 4.59375,
+      "learning_rate": 0.00011358569356405832,
+      "loss": 4.0176,
+      "step": 265750
+    },
+    {
+      "epoch": 2.2882231404958677,
+      "grad_norm": 2.921875,
+      "learning_rate": 0.00011355885221318468,
+      "loss": 4.3935,
+      "step": 265800
+    },
+    {
+      "epoch": 2.2886535812672175,
+      "grad_norm": 1.859375,
+      "learning_rate": 0.00011353200986712046,
+      "loss": 4.544,
+      "step": 265850
+    },
+    {
+      "epoch": 2.2890840220385673,
+      "grad_norm": 6.25,
+      "learning_rate": 0.00011350516652783585,
+      "loss": 4.2888,
+      "step": 265900
+    },
+    {
+      "epoch": 2.289514462809917,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.00011347832219730104,
+      "loss": 4.6411,
+      "step": 265950
+    },
+    {
+      "epoch": 2.2899449035812673,
+      "grad_norm": 3.921875,
+      "learning_rate": 0.0001134514768774864,
+      "loss": 3.9569,
+      "step": 266000
+    },
+    {
+      "epoch": 2.290375344352617,
+      "grad_norm": 3.453125,
+      "learning_rate": 0.00011342463057036233,
+      "loss": 4.2417,
+      "step": 266050
+    },
+    {
+      "epoch": 2.290805785123967,
+      "grad_norm": 3.53125,
+      "learning_rate": 0.00011339778327789923,
+      "loss": 4.1836,
+      "step": 266100
+    },
+    {
+      "epoch": 2.2912362258953167,
+      "grad_norm": 1.6875,
+      "learning_rate": 0.00011337093500206764,
+      "loss": 4.4778,
+      "step": 266150
+    },
+    {
+      "epoch": 2.2916666666666665,
+      "grad_norm": 2.296875,
+      "learning_rate": 0.00011334408574483821,
+      "loss": 4.4123,
+      "step": 266200
+    },
+    {
+      "epoch": 2.2920971074380168,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.00011331723550818161,
+      "loss": 4.4007,
+      "step": 266250
+    },
+    {
+      "epoch": 2.2925275482093666,
+      "grad_norm": 3.71875,
+      "learning_rate": 0.00011329038429406855,
+      "loss": 4.6705,
+      "step": 266300
+    },
+    {
+      "epoch": 2.2929579889807163,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00011326353210446992,
+      "loss": 4.2245,
+      "step": 266350
+    },
+    {
+      "epoch": 2.293388429752066,
+      "grad_norm": 3.484375,
+      "learning_rate": 0.00011323667894135654,
+      "loss": 4.6366,
+      "step": 266400
+    },
+    {
+      "epoch": 2.293818870523416,
+      "grad_norm": 3.34375,
+      "learning_rate": 0.0001132098248066994,
+      "loss": 4.3999,
+      "step": 266450
+    },
+    {
+      "epoch": 2.2942493112947657,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00011318296970246956,
+      "loss": 4.2901,
+      "step": 266500
+    },
+    {
+      "epoch": 2.2946797520661155,
+      "grad_norm": 1.484375,
+      "learning_rate": 0.0001131561136306381,
+      "loss": 4.2358,
+      "step": 266550
+    },
+    {
+      "epoch": 2.2951101928374658,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.00011312925659317617,
+      "loss": 4.4223,
+      "step": 266600
+    },
+    {
+      "epoch": 2.2955406336088156,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.00011310239859205509,
+      "loss": 4.399,
+      "step": 266650
+    },
+    {
+      "epoch": 2.2959710743801653,
+      "grad_norm": 2.25,
+      "learning_rate": 0.00011307553962924618,
+      "loss": 4.3867,
+      "step": 266700
+    },
+    {
+      "epoch": 2.296401515151515,
+      "grad_norm": 1.3046875,
+      "learning_rate": 0.00011304867970672075,
+      "loss": 4.3279,
+      "step": 266750
+    },
+    {
+      "epoch": 2.296831955922865,
+      "grad_norm": 3.78125,
+      "learning_rate": 0.00011302181882645029,
+      "loss": 4.3662,
+      "step": 266800
+    },
+    {
+      "epoch": 2.2972623966942147,
+      "grad_norm": 4.40625,
+      "learning_rate": 0.00011299495699040634,
+      "loss": 3.9623,
+      "step": 266850
+    },
+    {
+      "epoch": 2.297692837465565,
+      "grad_norm": 1.4296875,
+      "learning_rate": 0.00011296809420056055,
+      "loss": 4.2957,
+      "step": 266900
+    },
+    {
+      "epoch": 2.2981232782369148,
+      "grad_norm": 3.390625,
+      "learning_rate": 0.00011294123045888454,
+      "loss": 4.1539,
+      "step": 266950
+    },
+    {
+      "epoch": 2.2985537190082646,
+      "grad_norm": 4.0625,
+      "learning_rate": 0.00011291436576735004,
+      "loss": 4.5285,
+      "step": 267000
+    },
+    {
+      "epoch": 2.2985537190082646,
+      "eval_loss": 5.019803524017334,
+      "eval_runtime": 24.235,
+      "eval_samples_per_second": 26.408,
+      "eval_steps_per_second": 13.204,
+      "eval_tts_loss": 7.400831753020406,
+      "step": 267000
+    },
+    {
+      "epoch": 2.2989841597796143,
+      "grad_norm": 6.71875,
+      "learning_rate": 0.0001128875001279289,
+      "loss": 4.3166,
+      "step": 267050
+    },
+    {
+      "epoch": 2.299414600550964,
+      "grad_norm": 3.390625,
+      "learning_rate": 0.00011286063354259295,
+      "loss": 4.0372,
+      "step": 267100
+    },
+    {
+      "epoch": 2.299845041322314,
+      "grad_norm": 1.6484375,
+      "learning_rate": 0.00011283376601331419,
+      "loss": 4.228,
+      "step": 267150
+    },
+    {
+      "epoch": 2.3002754820936637,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00011280689754206459,
+      "loss": 4.1248,
+      "step": 267200
+    },
+    {
+      "epoch": 2.300705922865014,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.0001127800281308163,
+      "loss": 4.3304,
+      "step": 267250
+    },
+    {
+      "epoch": 2.3011363636363638,
+      "grad_norm": 2.875,
+      "learning_rate": 0.00011275315778154144,
+      "loss": 4.3871,
+      "step": 267300
+    },
+    {
+      "epoch": 2.3015668044077136,
+      "grad_norm": 1.7890625,
+      "learning_rate": 0.0001127262864962122,
+      "loss": 4.7217,
+      "step": 267350
+    },
+    {
+      "epoch": 2.3019972451790633,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00011269941427680097,
+      "loss": 4.4358,
+      "step": 267400
+    },
+    {
+      "epoch": 2.302427685950413,
+      "grad_norm": 3.75,
+      "learning_rate": 0.00011267254112528001,
+      "loss": 4.6437,
+      "step": 267450
+    },
+    {
+      "epoch": 2.302858126721763,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.00011264566704362178,
+      "loss": 4.4117,
+      "step": 267500
+    },
+    {
+      "epoch": 2.303288567493113,
+      "grad_norm": 3.109375,
+      "learning_rate": 0.00011261879203379884,
+      "loss": 4.1981,
+      "step": 267550
+    },
+    {
+      "epoch": 2.303719008264463,
+      "grad_norm": 4.15625,
+      "learning_rate": 0.0001125919160977837,
+      "loss": 4.7515,
+      "step": 267600
+    },
+    {
+      "epoch": 2.3041494490358128,
+      "grad_norm": 3.875,
+      "learning_rate": 0.00011256503923754902,
+      "loss": 4.0131,
+      "step": 267650
+    },
+    {
+      "epoch": 2.3045798898071626,
+      "grad_norm": 3.453125,
+      "learning_rate": 0.00011253816145506748,
+      "loss": 4.292,
+      "step": 267700
+    },
+    {
+      "epoch": 2.3050103305785123,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00011251128275231189,
+      "loss": 4.2676,
+      "step": 267750
+    },
+    {
+      "epoch": 2.305440771349862,
+      "grad_norm": 3.265625,
+      "learning_rate": 0.00011248440313125504,
+      "loss": 4.3254,
+      "step": 267800
+    },
+    {
+      "epoch": 2.305871212121212,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.00011245752259386985,
+      "loss": 4.4629,
+      "step": 267850
+    },
+    {
+      "epoch": 2.306301652892562,
+      "grad_norm": 3.390625,
+      "learning_rate": 0.00011243064114212933,
+      "loss": 4.1554,
+      "step": 267900
+    },
+    {
+      "epoch": 2.306732093663912,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.0001124037587780065,
+      "loss": 4.3493,
+      "step": 267950
+    },
+    {
+      "epoch": 2.3071625344352618,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00011237687550347444,
+      "loss": 4.0415,
+      "step": 268000
+    },
+    {
+      "epoch": 2.3075929752066116,
+      "grad_norm": 3.21875,
+      "learning_rate": 0.00011234999132050636,
+      "loss": 4.3151,
+      "step": 268050
+    },
+    {
+      "epoch": 2.3080234159779613,
+      "grad_norm": 3.6875,
+      "learning_rate": 0.00011232310623107549,
+      "loss": 4.3389,
+      "step": 268100
+    },
+    {
+      "epoch": 2.308453856749311,
+      "grad_norm": 3.15625,
+      "learning_rate": 0.00011229622023715512,
+      "loss": 4.6592,
+      "step": 268150
+    },
+    {
+      "epoch": 2.3088842975206614,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00011226933334071868,
+      "loss": 4.2798,
+      "step": 268200
+    },
+    {
+      "epoch": 2.309314738292011,
+      "grad_norm": 1.9609375,
+      "learning_rate": 0.00011224244554373953,
+      "loss": 4.4876,
+      "step": 268250
+    },
+    {
+      "epoch": 2.309745179063361,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00011221555684819125,
+      "loss": 4.323,
+      "step": 268300
+    },
+    {
+      "epoch": 2.3101756198347108,
+      "grad_norm": 0.9921875,
+      "learning_rate": 0.00011218866725604735,
+      "loss": 4.522,
+      "step": 268350
+    },
+    {
+      "epoch": 2.3106060606060606,
+      "grad_norm": 3.078125,
+      "learning_rate": 0.00011216177676928152,
+      "loss": 4.3923,
+      "step": 268400
+    },
+    {
+      "epoch": 2.3110365013774103,
+      "grad_norm": 1.6484375,
+      "learning_rate": 0.00011213488538986743,
+      "loss": 4.6585,
+      "step": 268450
+    },
+    {
+      "epoch": 2.31146694214876,
+      "grad_norm": 2.421875,
+      "learning_rate": 0.00011210799311977884,
+      "loss": 4.6634,
+      "step": 268500
+    },
+    {
+      "epoch": 2.3118973829201104,
+      "grad_norm": 3.53125,
+      "learning_rate": 0.00011208109996098965,
+      "loss": 4.2848,
+      "step": 268550
+    },
+    {
+      "epoch": 2.31232782369146,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.00011205420591547371,
+      "loss": 4.0101,
+      "step": 268600
+    },
+    {
+      "epoch": 2.31275826446281,
+      "grad_norm": 3.046875,
+      "learning_rate": 0.00011202731098520496,
+      "loss": 4.5398,
+      "step": 268650
+    },
+    {
+      "epoch": 2.3131887052341598,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.00011200041517215747,
+      "loss": 4.1761,
+      "step": 268700
+    },
+    {
+      "epoch": 2.3136191460055096,
+      "grad_norm": 2.265625,
+      "learning_rate": 0.00011197351847830535,
+      "loss": 4.2692,
+      "step": 268750
+    },
+    {
+      "epoch": 2.3140495867768593,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.00011194662090562274,
+      "loss": 4.3893,
+      "step": 268800
+    },
+    {
+      "epoch": 2.3144800275482096,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.00011191972245608382,
+      "loss": 4.1133,
+      "step": 268850
+    },
+    {
+      "epoch": 2.3149104683195594,
+      "grad_norm": 2.984375,
+      "learning_rate": 0.00011189282313166292,
+      "loss": 4.4693,
+      "step": 268900
+    },
+    {
+      "epoch": 2.315340909090909,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00011186592293433443,
+      "loss": 4.1404,
+      "step": 268950
+    },
+    {
+      "epoch": 2.315771349862259,
+      "grad_norm": 4.90625,
+      "learning_rate": 0.00011183902186607268,
+      "loss": 4.6434,
+      "step": 269000
+    },
+    {
+      "epoch": 2.3162017906336088,
+      "grad_norm": 4.40625,
+      "learning_rate": 0.00011181211992885222,
+      "loss": 4.5971,
+      "step": 269050
+    },
+    {
+      "epoch": 2.3166322314049586,
+      "grad_norm": 3.1875,
+      "learning_rate": 0.00011178521712464758,
+      "loss": 4.4282,
+      "step": 269100
+    },
+    {
+      "epoch": 2.3170626721763083,
+      "grad_norm": 5.28125,
+      "learning_rate": 0.00011175831345543335,
+      "loss": 4.1139,
+      "step": 269150
+    },
+    {
+      "epoch": 2.3174931129476586,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.0001117314089231842,
+      "loss": 4.3137,
+      "step": 269200
+    },
+    {
+      "epoch": 2.3179235537190084,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.0001117045035298749,
+      "loss": 4.0266,
+      "step": 269250
+    },
+    {
+      "epoch": 2.318353994490358,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00011167759727748022,
+      "loss": 4.4134,
+      "step": 269300
+    },
+    {
+      "epoch": 2.318784435261708,
+      "grad_norm": 3.109375,
+      "learning_rate": 0.00011165069016797502,
+      "loss": 4.4916,
+      "step": 269350
+    },
+    {
+      "epoch": 2.3192148760330578,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.00011162378220333421,
+      "loss": 4.1318,
+      "step": 269400
+    },
+    {
+      "epoch": 2.3196453168044076,
+      "grad_norm": 3.96875,
+      "learning_rate": 0.00011159687338553282,
+      "loss": 4.3287,
+      "step": 269450
+    },
+    {
+      "epoch": 2.320075757575758,
+      "grad_norm": 3.203125,
+      "learning_rate": 0.00011156996371654587,
+      "loss": 4.0735,
+      "step": 269500
+    },
+    {
+      "epoch": 2.3205061983471076,
+      "grad_norm": 1.9140625,
+      "learning_rate": 0.00011154305319834846,
+      "loss": 4.0443,
+      "step": 269550
+    },
+    {
+      "epoch": 2.3209366391184574,
+      "grad_norm": 3.234375,
+      "learning_rate": 0.0001115161418329158,
+      "loss": 5.0241,
+      "step": 269600
+    },
+    {
+      "epoch": 2.321367079889807,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.00011148922962222311,
+      "loss": 4.2335,
+      "step": 269650
+    },
+    {
+      "epoch": 2.321797520661157,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00011146231656824568,
+      "loss": 4.7398,
+      "step": 269700
+    },
+    {
+      "epoch": 2.3222279614325068,
+      "grad_norm": 4.21875,
+      "learning_rate": 0.00011143540267295887,
+      "loss": 4.1892,
+      "step": 269750
+    },
+    {
+      "epoch": 2.3226584022038566,
+      "grad_norm": 4.03125,
+      "learning_rate": 0.00011140848793833812,
+      "loss": 4.2207,
+      "step": 269800
+    },
+    {
+      "epoch": 2.323088842975207,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.0001113815723663589,
+      "loss": 4.3673,
+      "step": 269850
+    },
+    {
+      "epoch": 2.3235192837465566,
+      "grad_norm": 2.984375,
+      "learning_rate": 0.00011135465595899672,
+      "loss": 4.5524,
+      "step": 269900
+    },
+    {
+      "epoch": 2.3239497245179064,
+      "grad_norm": 1.9375,
+      "learning_rate": 0.00011132773871822725,
+      "loss": 4.3183,
+      "step": 269950
+    },
+    {
+      "epoch": 2.324380165289256,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.00011130082064602613,
+      "loss": 3.9652,
+      "step": 270000
+    },
+    {
+      "epoch": 2.324380165289256,
+      "eval_loss": 5.020001411437988,
+      "eval_runtime": 24.2049,
+      "eval_samples_per_second": 26.441,
+      "eval_steps_per_second": 13.22,
+      "eval_tts_loss": 7.4423839900824404,
+      "step": 270000
+    },
+    {
+      "epoch": 2.324810606060606,
+      "grad_norm": 1.96875,
+      "learning_rate": 0.00011127390174436909,
+      "loss": 4.2847,
+      "step": 270050
+    },
+    {
+      "epoch": 2.3252410468319558,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.00011124698201523192,
+      "loss": 4.2435,
+      "step": 270100
+    },
+    {
+      "epoch": 2.325671487603306,
+      "grad_norm": 4.25,
+      "learning_rate": 0.00011122006146059048,
+      "loss": 4.2181,
+      "step": 270150
+    },
+    {
+      "epoch": 2.326101928374656,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.00011119314008242063,
+      "loss": 4.3434,
+      "step": 270200
+    },
+    {
+      "epoch": 2.3265323691460056,
+      "grad_norm": 1.5546875,
+      "learning_rate": 0.00011116621788269841,
+      "loss": 4.2894,
+      "step": 270250
+    },
+    {
+      "epoch": 2.3269628099173554,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00011113929486339982,
+      "loss": 4.6308,
+      "step": 270300
+    },
+    {
+      "epoch": 2.327393250688705,
+      "grad_norm": 4.21875,
+      "learning_rate": 0.00011111237102650096,
+      "loss": 4.503,
+      "step": 270350
+    },
+    {
+      "epoch": 2.327823691460055,
+      "grad_norm": 3.140625,
+      "learning_rate": 0.00011108544637397798,
+      "loss": 4.5456,
+      "step": 270400
+    },
+    {
+      "epoch": 2.3282541322314048,
+      "grad_norm": 1.3359375,
+      "learning_rate": 0.00011105852090780703,
+      "loss": 4.0705,
+      "step": 270450
+    },
+    {
+      "epoch": 2.328684573002755,
+      "grad_norm": 1.3984375,
+      "learning_rate": 0.00011103159462996453,
+      "loss": 4.5411,
+      "step": 270500
+    },
+    {
+      "epoch": 2.329115013774105,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.00011100466754242665,
+      "loss": 4.6688,
+      "step": 270550
+    },
+    {
+      "epoch": 2.3295454545454546,
+      "grad_norm": 1.7578125,
+      "learning_rate": 0.00011097773964716989,
+      "loss": 4.7192,
+      "step": 270600
+    },
+    {
+      "epoch": 2.3299758953168044,
+      "grad_norm": 2.21875,
+      "learning_rate": 0.00011095081094617063,
+      "loss": 4.3651,
+      "step": 270650
+    },
+    {
+      "epoch": 2.330406336088154,
+      "grad_norm": 3.796875,
+      "learning_rate": 0.00011092388144140541,
+      "loss": 4.3971,
+      "step": 270700
+    },
+    {
+      "epoch": 2.330836776859504,
+      "grad_norm": 4.375,
+      "learning_rate": 0.00011089695113485081,
+      "loss": 4.0952,
+      "step": 270750
+    },
+    {
+      "epoch": 2.331267217630854,
+      "grad_norm": 5.25,
+      "learning_rate": 0.00011087002002848343,
+      "loss": 4.1595,
+      "step": 270800
+    },
+    {
+      "epoch": 2.331697658402204,
+      "grad_norm": 1.9921875,
+      "learning_rate": 0.00011084308812427999,
+      "loss": 4.3974,
+      "step": 270850
+    },
+    {
+      "epoch": 2.332128099173554,
+      "grad_norm": 3.203125,
+      "learning_rate": 0.00011081615542421719,
+      "loss": 4.5905,
+      "step": 270900
+    },
+    {
+      "epoch": 2.3325585399449036,
+      "grad_norm": 5.0,
+      "learning_rate": 0.00011078922193027186,
+      "loss": 4.2248,
+      "step": 270950
+    },
+    {
+      "epoch": 2.3329889807162534,
+      "grad_norm": 3.703125,
+      "learning_rate": 0.00011076228764442088,
+      "loss": 4.3198,
+      "step": 271000
+    },
+    {
+      "epoch": 2.333419421487603,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.00011073535256864109,
+      "loss": 4.2229,
+      "step": 271050
+    },
+    {
+      "epoch": 2.333849862258953,
+      "grad_norm": 5.25,
+      "learning_rate": 0.00011070841670490952,
+      "loss": 4.4069,
+      "step": 271100
+    },
+    {
+      "epoch": 2.334280303030303,
+      "grad_norm": 4.78125,
+      "learning_rate": 0.00011068148005520325,
+      "loss": 4.3388,
+      "step": 271150
+    },
+    {
+      "epoch": 2.334710743801653,
+      "grad_norm": 3.25,
+      "learning_rate": 0.0001106545426214993,
+      "loss": 3.9887,
+      "step": 271200
+    },
+    {
+      "epoch": 2.335141184573003,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00011062760440577487,
+      "loss": 4.433,
+      "step": 271250
+    },
+    {
+      "epoch": 2.3355716253443526,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.00011060066541000712,
+      "loss": 4.2246,
+      "step": 271300
+    },
+    {
+      "epoch": 2.3360020661157024,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.00011057372563617331,
+      "loss": 4.4175,
+      "step": 271350
+    },
+    {
+      "epoch": 2.336432506887052,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.00011054678508625085,
+      "loss": 4.5034,
+      "step": 271400
+    },
+    {
+      "epoch": 2.3368629476584024,
+      "grad_norm": 3.578125,
+      "learning_rate": 0.000110519843762217,
+      "loss": 4.4328,
+      "step": 271450
+    },
+    {
+      "epoch": 2.337293388429752,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.00011049290166604929,
+      "loss": 4.7694,
+      "step": 271500
+    },
+    {
+      "epoch": 2.337723829201102,
+      "grad_norm": 1.84375,
+      "learning_rate": 0.00011046595879972517,
+      "loss": 4.4014,
+      "step": 271550
+    },
+    {
+      "epoch": 2.338154269972452,
+      "grad_norm": 1.3515625,
+      "learning_rate": 0.00011043901516522215,
+      "loss": 4.517,
+      "step": 271600
+    },
+    {
+      "epoch": 2.3385847107438016,
+      "grad_norm": 3.71875,
+      "learning_rate": 0.00011041207076451792,
+      "loss": 4.0097,
+      "step": 271650
+    },
+    {
+      "epoch": 2.3390151515151514,
+      "grad_norm": 3.109375,
+      "learning_rate": 0.00011038512559959009,
+      "loss": 4.3613,
+      "step": 271700
+    },
+    {
+      "epoch": 2.339445592286501,
+      "grad_norm": 4.625,
+      "learning_rate": 0.00011035817967241638,
+      "loss": 3.7529,
+      "step": 271750
+    },
+    {
+      "epoch": 2.3398760330578514,
+      "grad_norm": 3.828125,
+      "learning_rate": 0.00011033123298497456,
+      "loss": 4.2193,
+      "step": 271800
+    },
+    {
+      "epoch": 2.340306473829201,
+      "grad_norm": 4.6875,
+      "learning_rate": 0.00011030428553924248,
+      "loss": 4.3163,
+      "step": 271850
+    },
+    {
+      "epoch": 2.340736914600551,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.00011027733733719803,
+      "loss": 4.4093,
+      "step": 271900
+    },
+    {
+      "epoch": 2.341167355371901,
+      "grad_norm": 1.8671875,
+      "learning_rate": 0.00011025038838081909,
+      "loss": 4.5671,
+      "step": 271950
+    },
+    {
+      "epoch": 2.3415977961432506,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.00011022343867208373,
+      "loss": 4.2148,
+      "step": 272000
+    },
+    {
+      "epoch": 2.3420282369146004,
+      "grad_norm": 3.4375,
+      "learning_rate": 0.00011019648821296996,
+      "loss": 4.5127,
+      "step": 272050
+    },
+    {
+      "epoch": 2.3424586776859506,
+      "grad_norm": 3.0625,
+      "learning_rate": 0.00011016953700545587,
+      "loss": 4.1278,
+      "step": 272100
+    },
+    {
+      "epoch": 2.3428891184573004,
+      "grad_norm": 1.5,
+      "learning_rate": 0.00011014258505151966,
+      "loss": 4.1266,
+      "step": 272150
+    },
+    {
+      "epoch": 2.34331955922865,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00011011563235313955,
+      "loss": 4.3159,
+      "step": 272200
+    },
+    {
+      "epoch": 2.34375,
+      "grad_norm": 1.28125,
+      "learning_rate": 0.00011008867891229375,
+      "loss": 4.4189,
+      "step": 272250
+    },
+    {
+      "epoch": 2.34418044077135,
+      "grad_norm": 3.515625,
+      "learning_rate": 0.00011006172473096066,
+      "loss": 4.4422,
+      "step": 272300
+    },
+    {
+      "epoch": 2.3446108815426996,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00011003476981111862,
+      "loss": 4.1728,
+      "step": 272350
+    },
+    {
+      "epoch": 2.3450413223140494,
+      "grad_norm": 3.65625,
+      "learning_rate": 0.00011000781415474605,
+      "loss": 4.4759,
+      "step": 272400
+    },
+    {
+      "epoch": 2.3454717630853996,
+      "grad_norm": 3.8125,
+      "learning_rate": 0.00010998085776382145,
+      "loss": 4.4576,
+      "step": 272450
+    },
+    {
+      "epoch": 2.3459022038567494,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00010995390064032336,
+      "loss": 4.6311,
+      "step": 272500
+    },
+    {
+      "epoch": 2.346332644628099,
+      "grad_norm": 3.28125,
+      "learning_rate": 0.00010992694278623041,
+      "loss": 4.5312,
+      "step": 272550
+    },
+    {
+      "epoch": 2.346763085399449,
+      "grad_norm": 1.4921875,
+      "learning_rate": 0.00010989998420352118,
+      "loss": 4.2317,
+      "step": 272600
+    },
+    {
+      "epoch": 2.347193526170799,
+      "grad_norm": 4.0,
+      "learning_rate": 0.00010987302489417443,
+      "loss": 4.4326,
+      "step": 272650
+    },
+    {
+      "epoch": 2.3476239669421486,
+      "grad_norm": 4.5,
+      "learning_rate": 0.0001098460648601689,
+      "loss": 4.163,
+      "step": 272700
+    },
+    {
+      "epoch": 2.348054407713499,
+      "grad_norm": 3.96875,
+      "learning_rate": 0.00010981910410348339,
+      "loss": 4.1036,
+      "step": 272750
+    },
+    {
+      "epoch": 2.3484848484848486,
+      "grad_norm": 3.609375,
+      "learning_rate": 0.00010979214262609679,
+      "loss": 4.1812,
+      "step": 272800
+    },
+    {
+      "epoch": 2.3489152892561984,
+      "grad_norm": 2.875,
+      "learning_rate": 0.00010976518042998793,
+      "loss": 4.122,
+      "step": 272850
+    },
+    {
+      "epoch": 2.349345730027548,
+      "grad_norm": 1.8515625,
+      "learning_rate": 0.0001097382175171359,
+      "loss": 4.3523,
+      "step": 272900
+    },
+    {
+      "epoch": 2.349776170798898,
+      "grad_norm": 3.671875,
+      "learning_rate": 0.00010971125388951962,
+      "loss": 4.5591,
+      "step": 272950
+    },
+    {
+      "epoch": 2.350206611570248,
+      "grad_norm": 6.34375,
+      "learning_rate": 0.0001096842895491182,
+      "loss": 4.3939,
+      "step": 273000
+    },
+    {
+      "epoch": 2.350206611570248,
+      "eval_loss": 5.015591621398926,
+      "eval_runtime": 23.9688,
+      "eval_samples_per_second": 26.701,
+      "eval_steps_per_second": 13.351,
+      "eval_tts_loss": 7.413518178344671,
+      "step": 273000
+    },
+    {
+      "epoch": 2.3506370523415976,
+      "grad_norm": 2.578125,
+      "learning_rate": 0.00010965732449791077,
+      "loss": 4.45,
+      "step": 273050
+    },
+    {
+      "epoch": 2.351067493112948,
+      "grad_norm": 4.25,
+      "learning_rate": 0.0001096303587378765,
+      "loss": 4.5807,
+      "step": 273100
+    },
+    {
+      "epoch": 2.3514979338842976,
+      "grad_norm": 5.09375,
+      "learning_rate": 0.0001096033922709946,
+      "loss": 4.5846,
+      "step": 273150
+    },
+    {
+      "epoch": 2.3519283746556474,
+      "grad_norm": 3.234375,
+      "learning_rate": 0.00010957642509924442,
+      "loss": 4.8317,
+      "step": 273200
+    },
+    {
+      "epoch": 2.352358815426997,
+      "grad_norm": 1.078125,
+      "learning_rate": 0.00010954945722460521,
+      "loss": 4.3433,
+      "step": 273250
+    },
+    {
+      "epoch": 2.352789256198347,
+      "grad_norm": 3.28125,
+      "learning_rate": 0.00010952248864905638,
+      "loss": 4.3074,
+      "step": 273300
+    },
+    {
+      "epoch": 2.353219696969697,
+      "grad_norm": 3.203125,
+      "learning_rate": 0.00010949551937457738,
+      "loss": 4.3437,
+      "step": 273350
+    },
+    {
+      "epoch": 2.353650137741047,
+      "grad_norm": 4.46875,
+      "learning_rate": 0.00010946854940314768,
+      "loss": 4.5188,
+      "step": 273400
+    },
+    {
+      "epoch": 2.354080578512397,
+      "grad_norm": 3.65625,
+      "learning_rate": 0.00010944157873674682,
+      "loss": 4.1051,
+      "step": 273450
+    },
+    {
+      "epoch": 2.3545110192837466,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.0001094146073773544,
+      "loss": 4.2475,
+      "step": 273500
+    },
+    {
+      "epoch": 2.3549414600550964,
+      "grad_norm": 1.8984375,
+      "learning_rate": 0.0001093876353269501,
+      "loss": 4.3901,
+      "step": 273550
+    },
+    {
+      "epoch": 2.355371900826446,
+      "grad_norm": 1.6484375,
+      "learning_rate": 0.00010936066258751351,
+      "loss": 4.4052,
+      "step": 273600
+    },
+    {
+      "epoch": 2.355802341597796,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00010933368916102442,
+      "loss": 4.4126,
+      "step": 273650
+    },
+    {
+      "epoch": 2.356232782369146,
+      "grad_norm": 3.1875,
+      "learning_rate": 0.00010930671504946266,
+      "loss": 3.9544,
+      "step": 273700
+    },
+    {
+      "epoch": 2.356663223140496,
+      "grad_norm": 3.78125,
+      "learning_rate": 0.00010927974025480805,
+      "loss": 4.4966,
+      "step": 273750
+    },
+    {
+      "epoch": 2.357093663911846,
+      "grad_norm": 1.828125,
+      "learning_rate": 0.00010925276477904043,
+      "loss": 4.0124,
+      "step": 273800
+    },
+    {
+      "epoch": 2.3575241046831956,
+      "grad_norm": 0.90234375,
+      "learning_rate": 0.00010922578862413981,
+      "loss": 4.6764,
+      "step": 273850
+    },
+    {
+      "epoch": 2.3579545454545454,
+      "grad_norm": 1.3359375,
+      "learning_rate": 0.00010919881179208616,
+      "loss": 3.9896,
+      "step": 273900
+    },
+    {
+      "epoch": 2.358384986225895,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.00010917183428485951,
+      "loss": 4.0156,
+      "step": 273950
+    },
+    {
+      "epoch": 2.358815426997245,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00010914485610443991,
+      "loss": 4.4215,
+      "step": 274000
+    },
+    {
+      "epoch": 2.3592458677685952,
+      "grad_norm": 2.875,
+      "learning_rate": 0.00010911787725280759,
+      "loss": 3.8769,
+      "step": 274050
+    },
+    {
+      "epoch": 2.359676308539945,
+      "grad_norm": 3.0625,
+      "learning_rate": 0.00010909089773194273,
+      "loss": 4.6535,
+      "step": 274100
+    },
+    {
+      "epoch": 2.360106749311295,
+      "grad_norm": 2.65625,
+      "learning_rate": 0.00010906391754382547,
+      "loss": 4.4374,
+      "step": 274150
+    },
+    {
+      "epoch": 2.3605371900826446,
+      "grad_norm": 1.421875,
+      "learning_rate": 0.00010903693669043622,
+      "loss": 4.5197,
+      "step": 274200
+    },
+    {
+      "epoch": 2.3609676308539944,
+      "grad_norm": 4.40625,
+      "learning_rate": 0.00010900995517375523,
+      "loss": 4.3713,
+      "step": 274250
+    },
+    {
+      "epoch": 2.361398071625344,
+      "grad_norm": 3.359375,
+      "learning_rate": 0.00010898297299576292,
+      "loss": 4.3623,
+      "step": 274300
+    },
+    {
+      "epoch": 2.361828512396694,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.00010895599015843971,
+      "loss": 4.3241,
+      "step": 274350
+    },
+    {
+      "epoch": 2.3622589531680442,
+      "grad_norm": 2.765625,
+      "learning_rate": 0.00010892900666376611,
+      "loss": 4.4181,
+      "step": 274400
+    },
+    {
+      "epoch": 2.362689393939394,
+      "grad_norm": 5.21875,
+      "learning_rate": 0.00010890202251372261,
+      "loss": 4.2732,
+      "step": 274450
+    },
+    {
+      "epoch": 2.363119834710744,
+      "grad_norm": 0.98828125,
+      "learning_rate": 0.00010887503771028984,
+      "loss": 4.3896,
+      "step": 274500
+    },
+    {
+      "epoch": 2.3635502754820936,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.00010884805225544837,
+      "loss": 4.2706,
+      "step": 274550
+    },
+    {
+      "epoch": 2.3639807162534434,
+      "grad_norm": 5.53125,
+      "learning_rate": 0.00010882106615117896,
+      "loss": 4.5979,
+      "step": 274600
+    },
+    {
+      "epoch": 2.364411157024793,
+      "grad_norm": 4.28125,
+      "learning_rate": 0.0001087940793994622,
+      "loss": 4.3754,
+      "step": 274650
+    },
+    {
+      "epoch": 2.3648415977961434,
+      "grad_norm": 2.9375,
+      "learning_rate": 0.00010876709200227895,
+      "loss": 4.6806,
+      "step": 274700
+    },
+    {
+      "epoch": 2.3652720385674932,
+      "grad_norm": 3.703125,
+      "learning_rate": 0.00010874010396161002,
+      "loss": 4.1623,
+      "step": 274750
+    },
+    {
+      "epoch": 2.365702479338843,
+      "grad_norm": 5.625,
+      "learning_rate": 0.00010871311527943628,
+      "loss": 4.3409,
+      "step": 274800
+    },
+    {
+      "epoch": 2.366132920110193,
+      "grad_norm": 5.375,
+      "learning_rate": 0.00010868612595773862,
+      "loss": 4.3937,
+      "step": 274850
+    },
+    {
+      "epoch": 2.3665633608815426,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.00010865913599849798,
+      "loss": 4.5467,
+      "step": 274900
+    },
+    {
+      "epoch": 2.3669938016528924,
+      "grad_norm": 1.59375,
+      "learning_rate": 0.00010863214540369539,
+      "loss": 4.1924,
+      "step": 274950
+    },
+    {
+      "epoch": 2.367424242424242,
+      "grad_norm": 2.578125,
+      "learning_rate": 0.00010860515417531192,
+      "loss": 4.3766,
+      "step": 275000
+    },
+    {
+      "epoch": 2.3678546831955924,
+      "grad_norm": 5.0,
+      "learning_rate": 0.00010857816231532862,
+      "loss": 4.6195,
+      "step": 275050
+    },
+    {
+      "epoch": 2.3682851239669422,
+      "grad_norm": 4.90625,
+      "learning_rate": 0.00010855116982572671,
+      "loss": 4.5974,
+      "step": 275100
+    },
+    {
+      "epoch": 2.368715564738292,
+      "grad_norm": 4.09375,
+      "learning_rate": 0.00010852417670848731,
+      "loss": 4.1674,
+      "step": 275150
+    },
+    {
+      "epoch": 2.369146005509642,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.00010849718296559165,
+      "loss": 4.3197,
+      "step": 275200
+    },
+    {
+      "epoch": 2.3695764462809916,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.00010847018859902107,
+      "loss": 4.2036,
+      "step": 275250
+    },
+    {
+      "epoch": 2.3700068870523414,
+      "grad_norm": 2.140625,
+      "learning_rate": 0.00010844319361075688,
+      "loss": 4.4151,
+      "step": 275300
+    },
+    {
+      "epoch": 2.3704373278236917,
+      "grad_norm": 1.3671875,
+      "learning_rate": 0.00010841619800278045,
+      "loss": 4.282,
+      "step": 275350
+    },
+    {
+      "epoch": 2.3708677685950414,
+      "grad_norm": 3.453125,
+      "learning_rate": 0.00010838920177707317,
+      "loss": 4.363,
+      "step": 275400
+    },
+    {
+      "epoch": 2.3712982093663912,
+      "grad_norm": 3.1875,
+      "learning_rate": 0.00010836220493561654,
+      "loss": 4.0632,
+      "step": 275450
+    },
+    {
+      "epoch": 2.371728650137741,
+      "grad_norm": 2.296875,
+      "learning_rate": 0.00010833520748039205,
+      "loss": 4.602,
+      "step": 275500
+    },
+    {
+      "epoch": 2.372159090909091,
+      "grad_norm": 4.96875,
+      "learning_rate": 0.00010830820941338127,
+      "loss": 4.0662,
+      "step": 275550
+    },
+    {
+      "epoch": 2.3725895316804406,
+      "grad_norm": 5.21875,
+      "learning_rate": 0.00010828121073656579,
+      "loss": 4.3643,
+      "step": 275600
+    },
+    {
+      "epoch": 2.3730199724517904,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.00010825421145192727,
+      "loss": 4.6803,
+      "step": 275650
+    },
+    {
+      "epoch": 2.3734504132231407,
+      "grad_norm": 3.828125,
+      "learning_rate": 0.00010822721156144738,
+      "loss": 4.3229,
+      "step": 275700
+    },
+    {
+      "epoch": 2.3738808539944904,
+      "grad_norm": 3.640625,
+      "learning_rate": 0.00010820021106710787,
+      "loss": 4.3691,
+      "step": 275750
+    },
+    {
+      "epoch": 2.3743112947658402,
+      "grad_norm": 3.140625,
+      "learning_rate": 0.00010817320997089051,
+      "loss": 4.4262,
+      "step": 275800
+    },
+    {
+      "epoch": 2.37474173553719,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.00010814620827477711,
+      "loss": 4.344,
+      "step": 275850
+    },
+    {
+      "epoch": 2.37517217630854,
+      "grad_norm": 2.328125,
+      "learning_rate": 0.00010811920598074956,
+      "loss": 4.1819,
+      "step": 275900
+    },
+    {
+      "epoch": 2.3756026170798896,
+      "grad_norm": 4.3125,
+      "learning_rate": 0.00010809220309078976,
+      "loss": 4.3527,
+      "step": 275950
+    },
+    {
+      "epoch": 2.37603305785124,
+      "grad_norm": 2.96875,
+      "learning_rate": 0.00010806519960687968,
+      "loss": 4.1432,
+      "step": 276000
+    },
+    {
+      "epoch": 2.37603305785124,
+      "eval_loss": 5.014104843139648,
+      "eval_runtime": 24.4776,
+      "eval_samples_per_second": 26.146,
+      "eval_steps_per_second": 13.073,
+      "eval_tts_loss": 7.396875379435267,
+      "step": 276000
+    },
+    {
+      "epoch": 2.3764634986225897,
+      "grad_norm": 3.296875,
+      "learning_rate": 0.00010803819553100127,
+      "loss": 4.2043,
+      "step": 276050
+    },
+    {
+      "epoch": 2.3768939393939394,
+      "grad_norm": 3.453125,
+      "learning_rate": 0.00010801119086513662,
+      "loss": 4.3224,
+      "step": 276100
+    },
+    {
+      "epoch": 2.3773243801652892,
+      "grad_norm": 1.65625,
+      "learning_rate": 0.00010798418561126781,
+      "loss": 4.1211,
+      "step": 276150
+    },
+    {
+      "epoch": 2.377754820936639,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00010795717977137694,
+      "loss": 4.2971,
+      "step": 276200
+    },
+    {
+      "epoch": 2.378185261707989,
+      "grad_norm": 3.546875,
+      "learning_rate": 0.0001079301733474462,
+      "loss": 4.2575,
+      "step": 276250
+    },
+    {
+      "epoch": 2.3786157024793386,
+      "grad_norm": 3.59375,
+      "learning_rate": 0.00010790316634145778,
+      "loss": 4.6188,
+      "step": 276300
+    },
+    {
+      "epoch": 2.379046143250689,
+      "grad_norm": 3.8125,
+      "learning_rate": 0.00010787615875539396,
+      "loss": 4.2032,
+      "step": 276350
+    },
+    {
+      "epoch": 2.3794765840220387,
+      "grad_norm": 1.8828125,
+      "learning_rate": 0.00010784915059123706,
+      "loss": 4.2535,
+      "step": 276400
+    },
+    {
+      "epoch": 2.3799070247933884,
+      "grad_norm": 1.8828125,
+      "learning_rate": 0.00010782214185096938,
+      "loss": 4.4132,
+      "step": 276450
+    },
+    {
+      "epoch": 2.3803374655647382,
+      "grad_norm": 5.8125,
+      "learning_rate": 0.00010779513253657329,
+      "loss": 4.7309,
+      "step": 276500
+    },
+    {
+      "epoch": 2.380767906336088,
+      "grad_norm": 3.75,
+      "learning_rate": 0.00010776812265003129,
+      "loss": 4.5798,
+      "step": 276550
+    },
+    {
+      "epoch": 2.381198347107438,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00010774111219332577,
+      "loss": 4.4564,
+      "step": 276600
+    },
+    {
+      "epoch": 2.381628787878788,
+      "grad_norm": 3.25,
+      "learning_rate": 0.00010771410116843927,
+      "loss": 4.3156,
+      "step": 276650
+    },
+    {
+      "epoch": 2.382059228650138,
+      "grad_norm": 5.4375,
+      "learning_rate": 0.00010768708957735436,
+      "loss": 4.523,
+      "step": 276700
+    },
+    {
+      "epoch": 2.3824896694214877,
+      "grad_norm": 3.625,
+      "learning_rate": 0.0001076600774220536,
+      "loss": 4.3499,
+      "step": 276750
+    },
+    {
+      "epoch": 2.3829201101928374,
+      "grad_norm": 5.46875,
+      "learning_rate": 0.00010763306470451966,
+      "loss": 4.1044,
+      "step": 276800
+    },
+    {
+      "epoch": 2.3833505509641872,
+      "grad_norm": 4.3125,
+      "learning_rate": 0.00010760605142673517,
+      "loss": 4.4604,
+      "step": 276850
+    },
+    {
+      "epoch": 2.383780991735537,
+      "grad_norm": 3.53125,
+      "learning_rate": 0.00010757903759068288,
+      "loss": 4.3727,
+      "step": 276900
+    },
+    {
+      "epoch": 2.384211432506887,
+      "grad_norm": 3.546875,
+      "learning_rate": 0.00010755202319834552,
+      "loss": 3.8865,
+      "step": 276950
+    },
+    {
+      "epoch": 2.384641873278237,
+      "grad_norm": 4.34375,
+      "learning_rate": 0.00010752500825170593,
+      "loss": 4.707,
+      "step": 277000
+    },
+    {
+      "epoch": 2.385072314049587,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.00010749799275274693,
+      "loss": 4.1706,
+      "step": 277050
+    },
+    {
+      "epoch": 2.3855027548209367,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00010747097670345136,
+      "loss": 4.4873,
+      "step": 277100
+    },
+    {
+      "epoch": 2.3859331955922864,
+      "grad_norm": 4.28125,
+      "learning_rate": 0.0001074439601058022,
+      "loss": 3.9552,
+      "step": 277150
+    },
+    {
+      "epoch": 2.3863636363636362,
+      "grad_norm": 1.53125,
+      "learning_rate": 0.00010741694296178239,
+      "loss": 4.073,
+      "step": 277200
+    },
+    {
+      "epoch": 2.386794077134986,
+      "grad_norm": 3.03125,
+      "learning_rate": 0.0001073899252733749,
+      "loss": 4.7507,
+      "step": 277250
+    },
+    {
+      "epoch": 2.3872245179063363,
+      "grad_norm": 3.234375,
+      "learning_rate": 0.0001073629070425628,
+      "loss": 4.4815,
+      "step": 277300
+    },
+    {
+      "epoch": 2.387654958677686,
+      "grad_norm": 3.90625,
+      "learning_rate": 0.00010733588827132919,
+      "loss": 4.5385,
+      "step": 277350
+    },
+    {
+      "epoch": 2.388085399449036,
+      "grad_norm": 4.40625,
+      "learning_rate": 0.00010730886896165713,
+      "loss": 4.4679,
+      "step": 277400
+    },
+    {
+      "epoch": 2.3885158402203857,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.00010728184911552983,
+      "loss": 4.3386,
+      "step": 277450
+    },
+    {
+      "epoch": 2.3889462809917354,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.00010725482873493049,
+      "loss": 4.0653,
+      "step": 277500
+    },
+    {
+      "epoch": 2.3893767217630852,
+      "grad_norm": 3.6875,
+      "learning_rate": 0.00010722780782184232,
+      "loss": 4.215,
+      "step": 277550
+    },
+    {
+      "epoch": 2.389807162534435,
+      "grad_norm": 2.6875,
+      "learning_rate": 0.00010720078637824857,
+      "loss": 4.1995,
+      "step": 277600
+    },
+    {
+      "epoch": 2.3902376033057853,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.00010717376440613263,
+      "loss": 4.4691,
+      "step": 277650
+    },
+    {
+      "epoch": 2.390668044077135,
+      "grad_norm": 4.71875,
+      "learning_rate": 0.00010714674190747782,
+      "loss": 4.4872,
+      "step": 277700
+    },
+    {
+      "epoch": 2.391098484848485,
+      "grad_norm": 3.9375,
+      "learning_rate": 0.0001071197188842675,
+      "loss": 4.4085,
+      "step": 277750
+    },
+    {
+      "epoch": 2.3915289256198347,
+      "grad_norm": 1.25,
+      "learning_rate": 0.00010709269533848518,
+      "loss": 4.1456,
+      "step": 277800
+    },
+    {
+      "epoch": 2.3919593663911844,
+      "grad_norm": 4.28125,
+      "learning_rate": 0.00010706567127211426,
+      "loss": 4.1615,
+      "step": 277850
+    },
+    {
+      "epoch": 2.3923898071625342,
+      "grad_norm": 2.9375,
+      "learning_rate": 0.00010703864668713826,
+      "loss": 4.3844,
+      "step": 277900
+    },
+    {
+      "epoch": 2.3928202479338845,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00010701162158554074,
+      "loss": 4.4029,
+      "step": 277950
+    },
+    {
+      "epoch": 2.3932506887052343,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00010698459596930526,
+      "loss": 4.4122,
+      "step": 278000
+    },
+    {
+      "epoch": 2.393681129476584,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00010695756984041552,
+      "loss": 4.4777,
+      "step": 278050
+    },
+    {
+      "epoch": 2.394111570247934,
+      "grad_norm": 3.609375,
+      "learning_rate": 0.00010693054320085507,
+      "loss": 4.6485,
+      "step": 278100
+    },
+    {
+      "epoch": 2.3945420110192837,
+      "grad_norm": 3.734375,
+      "learning_rate": 0.00010690351605260769,
+      "loss": 4.5897,
+      "step": 278150
+    },
+    {
+      "epoch": 2.3949724517906334,
+      "grad_norm": 4.0,
+      "learning_rate": 0.0001068764883976571,
+      "loss": 4.115,
+      "step": 278200
+    },
+    {
+      "epoch": 2.3954028925619832,
+      "grad_norm": 3.5625,
+      "learning_rate": 0.00010684946023798701,
+      "loss": 4.7298,
+      "step": 278250
+    },
+    {
+      "epoch": 2.3958333333333335,
+      "grad_norm": 3.0625,
+      "learning_rate": 0.00010682243157558129,
+      "loss": 4.6303,
+      "step": 278300
+    },
+    {
+      "epoch": 2.3962637741046833,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00010679540241242379,
+      "loss": 4.5638,
+      "step": 278350
+    },
+    {
+      "epoch": 2.396694214876033,
+      "grad_norm": 3.453125,
+      "learning_rate": 0.00010676837275049836,
+      "loss": 3.9242,
+      "step": 278400
+    },
+    {
+      "epoch": 2.397124655647383,
+      "grad_norm": 2.375,
+      "learning_rate": 0.00010674134259178896,
+      "loss": 4.4327,
+      "step": 278450
+    },
+    {
+      "epoch": 2.3975550964187327,
+      "grad_norm": 3.1875,
+      "learning_rate": 0.00010671431193827948,
+      "loss": 4.3538,
+      "step": 278500
+    },
+    {
+      "epoch": 2.397985537190083,
+      "grad_norm": 2.03125,
+      "learning_rate": 0.00010668728079195396,
+      "loss": 4.1395,
+      "step": 278550
+    },
+    {
+      "epoch": 2.3984159779614327,
+      "grad_norm": 2.875,
+      "learning_rate": 0.00010666024915479642,
+      "loss": 4.4853,
+      "step": 278600
+    },
+    {
+      "epoch": 2.3988464187327825,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.00010663321702879091,
+      "loss": 4.739,
+      "step": 278650
+    },
+    {
+      "epoch": 2.3992768595041323,
+      "grad_norm": 3.078125,
+      "learning_rate": 0.00010660618441592158,
+      "loss": 4.3426,
+      "step": 278700
+    },
+    {
+      "epoch": 2.399707300275482,
+      "grad_norm": 3.4375,
+      "learning_rate": 0.00010657915131817251,
+      "loss": 4.4866,
+      "step": 278750
+    },
+    {
+      "epoch": 2.400137741046832,
+      "grad_norm": 1.7734375,
+      "learning_rate": 0.00010655211773752786,
+      "loss": 4.0232,
+      "step": 278800
+    },
+    {
+      "epoch": 2.4005681818181817,
+      "grad_norm": 4.0625,
+      "learning_rate": 0.00010652508367597189,
+      "loss": 4.3151,
+      "step": 278850
+    },
+    {
+      "epoch": 2.4009986225895315,
+      "grad_norm": 2.125,
+      "learning_rate": 0.00010649804913548885,
+      "loss": 4.55,
+      "step": 278900
+    },
+    {
+      "epoch": 2.4014290633608817,
+      "grad_norm": 1.8671875,
+      "learning_rate": 0.00010647101411806292,
+      "loss": 4.3104,
+      "step": 278950
+    },
+    {
+      "epoch": 2.4018595041322315,
+      "grad_norm": 3.125,
+      "learning_rate": 0.00010644397862567852,
+      "loss": 4.1131,
+      "step": 279000
+    },
+    {
+      "epoch": 2.4018595041322315,
+      "eval_loss": 5.014746189117432,
+      "eval_runtime": 24.8643,
+      "eval_samples_per_second": 25.74,
+      "eval_steps_per_second": 12.87,
+      "eval_tts_loss": 7.389301508190825,
+      "step": 279000
+    },
+    {
+      "epoch": 2.4022899449035813,
+      "grad_norm": 1.8046875,
+      "learning_rate": 0.00010641694266031994,
+      "loss": 4.4027,
+      "step": 279050
+    },
+    {
+      "epoch": 2.402720385674931,
+      "grad_norm": 1.0625,
+      "learning_rate": 0.00010638990622397158,
+      "loss": 3.9444,
+      "step": 279100
+    },
+    {
+      "epoch": 2.403150826446281,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.00010636286931861783,
+      "loss": 4.1661,
+      "step": 279150
+    },
+    {
+      "epoch": 2.403581267217631,
+      "grad_norm": 1.953125,
+      "learning_rate": 0.0001063358319462432,
+      "loss": 4.4234,
+      "step": 279200
+    },
+    {
+      "epoch": 2.404011707988981,
+      "grad_norm": 3.140625,
+      "learning_rate": 0.0001063087941088321,
+      "loss": 4.3915,
+      "step": 279250
+    },
+    {
+      "epoch": 2.4044421487603307,
+      "grad_norm": 6.34375,
+      "learning_rate": 0.0001062817558083691,
+      "loss": 4.5487,
+      "step": 279300
+    },
+    {
+      "epoch": 2.4048725895316805,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00010625471704683874,
+      "loss": 3.7869,
+      "step": 279350
+    },
+    {
+      "epoch": 2.4053030303030303,
+      "grad_norm": 1.8515625,
+      "learning_rate": 0.00010622767782622562,
+      "loss": 4.2739,
+      "step": 279400
+    },
+    {
+      "epoch": 2.40573347107438,
+      "grad_norm": 4.3125,
+      "learning_rate": 0.0001062006381485143,
+      "loss": 4.0549,
+      "step": 279450
+    },
+    {
+      "epoch": 2.40616391184573,
+      "grad_norm": 2.5,
+      "learning_rate": 0.00010617359801568954,
+      "loss": 4.56,
+      "step": 279500
+    },
+    {
+      "epoch": 2.4065943526170797,
+      "grad_norm": 2.578125,
+      "learning_rate": 0.00010614655742973591,
+      "loss": 4.1112,
+      "step": 279550
+    },
+    {
+      "epoch": 2.40702479338843,
+      "grad_norm": 2.5625,
+      "learning_rate": 0.00010611951639263826,
+      "loss": 4.2552,
+      "step": 279600
+    },
+    {
+      "epoch": 2.4074552341597797,
+      "grad_norm": 5.5625,
+      "learning_rate": 0.00010609247490638123,
+      "loss": 4.2132,
+      "step": 279650
+    },
+    {
+      "epoch": 2.4078856749311295,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.00010606543297294964,
+      "loss": 4.5925,
+      "step": 279700
+    },
+    {
+      "epoch": 2.4083161157024793,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.00010603839059432837,
+      "loss": 4.1446,
+      "step": 279750
+    },
+    {
+      "epoch": 2.408746556473829,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.00010601134777250217,
+      "loss": 4.5189,
+      "step": 279800
+    },
+    {
+      "epoch": 2.4091769972451793,
+      "grad_norm": 3.375,
+      "learning_rate": 0.000105984304509456,
+      "loss": 4.4627,
+      "step": 279850
+    },
+    {
+      "epoch": 2.409607438016529,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.00010595726080717474,
+      "loss": 4.1316,
+      "step": 279900
+    },
+    {
+      "epoch": 2.410037878787879,
+      "grad_norm": 3.78125,
+      "learning_rate": 0.00010593021666764335,
+      "loss": 4.3968,
+      "step": 279950
+    },
+    {
+      "epoch": 2.4104683195592287,
+      "grad_norm": 3.453125,
+      "learning_rate": 0.00010590317209284686,
+      "loss": 4.5615,
+      "step": 280000
+    },
+    {
+      "epoch": 2.4108987603305785,
+      "grad_norm": 1.78125,
+      "learning_rate": 0.00010587612708477017,
+      "loss": 4.2917,
+      "step": 280050
+    },
+    {
+      "epoch": 2.4113292011019283,
+      "grad_norm": 3.40625,
+      "learning_rate": 0.00010584908164539846,
+      "loss": 4.3689,
+      "step": 280100
+    },
+    {
+      "epoch": 2.411759641873278,
+      "grad_norm": 3.21875,
+      "learning_rate": 0.00010582203577671672,
+      "loss": 4.3412,
+      "step": 280150
+    },
+    {
+      "epoch": 2.412190082644628,
+      "grad_norm": 2.140625,
+      "learning_rate": 0.00010579498948071005,
+      "loss": 4.328,
+      "step": 280200
+    },
+    {
+      "epoch": 2.412620523415978,
+      "grad_norm": 6.25,
+      "learning_rate": 0.00010576794275936366,
+      "loss": 4.4093,
+      "step": 280250
+    },
+    {
+      "epoch": 2.413050964187328,
+      "grad_norm": 3.078125,
+      "learning_rate": 0.00010574089561466267,
+      "loss": 4.3293,
+      "step": 280300
+    },
+    {
+      "epoch": 2.4134814049586777,
+      "grad_norm": 5.0625,
+      "learning_rate": 0.00010571384804859228,
+      "loss": 4.4423,
+      "step": 280350
+    },
+    {
+      "epoch": 2.4139118457300275,
+      "grad_norm": 5.15625,
+      "learning_rate": 0.00010568680006313776,
+      "loss": 4.4732,
+      "step": 280400
+    },
+    {
+      "epoch": 2.4143422865013773,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00010565975166028435,
+      "loss": 4.7185,
+      "step": 280450
+    },
+    {
+      "epoch": 2.4147727272727275,
+      "grad_norm": 2.9375,
+      "learning_rate": 0.00010563270284201734,
+      "loss": 4.1013,
+      "step": 280500
+    },
+    {
+      "epoch": 2.4152031680440773,
+      "grad_norm": 1.828125,
+      "learning_rate": 0.00010560565361032204,
+      "loss": 4.4158,
+      "step": 280550
+    },
+    {
+      "epoch": 2.415633608815427,
+      "grad_norm": 3.8125,
+      "learning_rate": 0.00010557860396718384,
+      "loss": 4.1939,
+      "step": 280600
+    },
+    {
+      "epoch": 2.416064049586777,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.00010555155391458813,
+      "loss": 3.9688,
+      "step": 280650
+    },
+    {
+      "epoch": 2.4164944903581267,
+      "grad_norm": 4.15625,
+      "learning_rate": 0.00010552450345452028,
+      "loss": 4.1904,
+      "step": 280700
+    },
+    {
+      "epoch": 2.4169249311294765,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.00010549745258896577,
+      "loss": 4.2963,
+      "step": 280750
+    },
+    {
+      "epoch": 2.4173553719008263,
+      "grad_norm": 3.859375,
+      "learning_rate": 0.00010547040131991009,
+      "loss": 4.2181,
+      "step": 280800
+    },
+    {
+      "epoch": 2.417785812672176,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.00010544334964933866,
+      "loss": 4.1302,
+      "step": 280850
+    },
+    {
+      "epoch": 2.4182162534435263,
+      "grad_norm": 2.921875,
+      "learning_rate": 0.00010541629757923714,
+      "loss": 4.4083,
+      "step": 280900
+    },
+    {
+      "epoch": 2.418646694214876,
+      "grad_norm": 3.515625,
+      "learning_rate": 0.00010538924511159103,
+      "loss": 4.5556,
+      "step": 280950
+    },
+    {
+      "epoch": 2.419077134986226,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.00010536219224838588,
+      "loss": 4.2889,
+      "step": 281000
+    },
+    {
+      "epoch": 2.4195075757575757,
+      "grad_norm": 4.5,
+      "learning_rate": 0.00010533513899160739,
+      "loss": 4.4081,
+      "step": 281050
+    },
+    {
+      "epoch": 2.4199380165289255,
+      "grad_norm": 3.875,
+      "learning_rate": 0.00010530808534324117,
+      "loss": 4.0653,
+      "step": 281100
+    },
+    {
+      "epoch": 2.4203684573002757,
+      "grad_norm": 3.234375,
+      "learning_rate": 0.00010528103130527292,
+      "loss": 4.5311,
+      "step": 281150
+    },
+    {
+      "epoch": 2.4207988980716255,
+      "grad_norm": 3.5625,
+      "learning_rate": 0.0001052539768796883,
+      "loss": 4.4613,
+      "step": 281200
+    },
+    {
+      "epoch": 2.4212293388429753,
+      "grad_norm": 4.0625,
+      "learning_rate": 0.00010522692206847309,
+      "loss": 4.507,
+      "step": 281250
+    },
+    {
+      "epoch": 2.421659779614325,
+      "grad_norm": 5.84375,
+      "learning_rate": 0.0001051998668736131,
+      "loss": 4.244,
+      "step": 281300
+    },
+    {
+      "epoch": 2.422090220385675,
+      "grad_norm": 4.1875,
+      "learning_rate": 0.000105172811297094,
+      "loss": 4.127,
+      "step": 281350
+    },
+    {
+      "epoch": 2.4225206611570247,
+      "grad_norm": 3.0,
+      "learning_rate": 0.00010514575534090172,
+      "loss": 4.4536,
+      "step": 281400
+    },
+    {
+      "epoch": 2.4229511019283745,
+      "grad_norm": 3.0625,
+      "learning_rate": 0.00010511869900702204,
+      "loss": 4.4533,
+      "step": 281450
+    },
+    {
+      "epoch": 2.4233815426997243,
+      "grad_norm": 4.375,
+      "learning_rate": 0.00010509164229744089,
+      "loss": 4.4307,
+      "step": 281500
+    },
+    {
+      "epoch": 2.4238119834710745,
+      "grad_norm": 4.53125,
+      "learning_rate": 0.00010506458521414415,
+      "loss": 4.2277,
+      "step": 281550
+    },
+    {
+      "epoch": 2.4242424242424243,
+      "grad_norm": 4.28125,
+      "learning_rate": 0.00010503752775911773,
+      "loss": 4.6191,
+      "step": 281600
+    },
+    {
+      "epoch": 2.424672865013774,
+      "grad_norm": 3.125,
+      "learning_rate": 0.00010501046993434765,
+      "loss": 4.3869,
+      "step": 281650
+    },
+    {
+      "epoch": 2.425103305785124,
+      "grad_norm": 6.75,
+      "learning_rate": 0.00010498341174181983,
+      "loss": 4.1685,
+      "step": 281700
+    },
+    {
+      "epoch": 2.4255337465564737,
+      "grad_norm": 1.8359375,
+      "learning_rate": 0.00010495635318352029,
+      "loss": 4.2421,
+      "step": 281750
+    },
+    {
+      "epoch": 2.425964187327824,
+      "grad_norm": 1.9765625,
+      "learning_rate": 0.00010492929426143512,
+      "loss": 4.3515,
+      "step": 281800
+    },
+    {
+      "epoch": 2.4263946280991737,
+      "grad_norm": 2.15625,
+      "learning_rate": 0.00010490223497755036,
+      "loss": 4.0464,
+      "step": 281850
+    },
+    {
+      "epoch": 2.4268250688705235,
+      "grad_norm": 3.96875,
+      "learning_rate": 0.00010487517533385207,
+      "loss": 4.2715,
+      "step": 281900
+    },
+    {
+      "epoch": 2.4272555096418733,
+      "grad_norm": 2.171875,
+      "learning_rate": 0.00010484811533232644,
+      "loss": 4.6647,
+      "step": 281950
+    },
+    {
+      "epoch": 2.427685950413223,
+      "grad_norm": 2.515625,
+      "learning_rate": 0.00010482105497495955,
+      "loss": 4.4087,
+      "step": 282000
+    },
+    {
+      "epoch": 2.427685950413223,
+      "eval_loss": 5.012827396392822,
+      "eval_runtime": 25.3249,
+      "eval_samples_per_second": 25.272,
+      "eval_steps_per_second": 12.636,
+      "eval_tts_loss": 7.408275696474385,
+      "step": 282000
+    },
+    {
+      "epoch": 2.428116391184573,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.00010479399426373761,
+      "loss": 4.7418,
+      "step": 282050
+    },
+    {
+      "epoch": 2.4285468319559227,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.00010476693320064678,
+      "loss": 4.465,
+      "step": 282100
+    },
+    {
+      "epoch": 2.4289772727272725,
+      "grad_norm": 4.625,
+      "learning_rate": 0.00010473987178767334,
+      "loss": 4.4881,
+      "step": 282150
+    },
+    {
+      "epoch": 2.4294077134986227,
+      "grad_norm": 1.90625,
+      "learning_rate": 0.00010471281002680351,
+      "loss": 4.5577,
+      "step": 282200
+    },
+    {
+      "epoch": 2.4298381542699725,
+      "grad_norm": 3.125,
+      "learning_rate": 0.00010468574792002353,
+      "loss": 4.5803,
+      "step": 282250
+    },
+    {
+      "epoch": 2.4302685950413223,
+      "grad_norm": 3.375,
+      "learning_rate": 0.00010465868546931976,
+      "loss": 4.1768,
+      "step": 282300
+    },
+    {
+      "epoch": 2.430699035812672,
+      "grad_norm": 3.0625,
+      "learning_rate": 0.0001046316226766785,
+      "loss": 4.6854,
+      "step": 282350
+    },
+    {
+      "epoch": 2.431129476584022,
+      "grad_norm": 3.578125,
+      "learning_rate": 0.00010460455954408608,
+      "loss": 4.343,
+      "step": 282400
+    },
+    {
+      "epoch": 2.431559917355372,
+      "grad_norm": 3.34375,
+      "learning_rate": 0.00010457749607352892,
+      "loss": 4.566,
+      "step": 282450
+    },
+    {
+      "epoch": 2.431990358126722,
+      "grad_norm": 2.453125,
+      "learning_rate": 0.00010455043226699339,
+      "loss": 4.4499,
+      "step": 282500
+    },
+    {
+      "epoch": 2.4324207988980717,
+      "grad_norm": 3.21875,
+      "learning_rate": 0.00010452336812646594,
+      "loss": 4.4572,
+      "step": 282550
+    },
+    {
+      "epoch": 2.4328512396694215,
+      "grad_norm": 2.078125,
+      "learning_rate": 0.00010449630365393297,
+      "loss": 4.1014,
+      "step": 282600
+    },
+    {
+      "epoch": 2.4332816804407713,
+      "grad_norm": 1.90625,
+      "learning_rate": 0.00010446923885138101,
+      "loss": 4.2052,
+      "step": 282650
+    },
+    {
+      "epoch": 2.433712121212121,
+      "grad_norm": 4.15625,
+      "learning_rate": 0.00010444217372079652,
+      "loss": 4.1167,
+      "step": 282700
+    },
+    {
+      "epoch": 2.434142561983471,
+      "grad_norm": 2.9375,
+      "learning_rate": 0.00010441510826416603,
+      "loss": 4.4092,
+      "step": 282750
+    },
+    {
+      "epoch": 2.4345730027548207,
+      "grad_norm": 3.28125,
+      "learning_rate": 0.00010438804248347609,
+      "loss": 4.0695,
+      "step": 282800
+    },
+    {
+      "epoch": 2.435003443526171,
+      "grad_norm": 1.9140625,
+      "learning_rate": 0.00010436097638071331,
+      "loss": 4.3029,
+      "step": 282850
+    },
+    {
+      "epoch": 2.4354338842975207,
+      "grad_norm": 0.82421875,
+      "learning_rate": 0.0001043339099578642,
+      "loss": 4.471,
+      "step": 282900
+    },
+    {
+      "epoch": 2.4358643250688705,
+      "grad_norm": 6.96875,
+      "learning_rate": 0.00010430684321691547,
+      "loss": 4.39,
+      "step": 282950
+    },
+    {
+      "epoch": 2.4362947658402203,
+      "grad_norm": 2.484375,
+      "learning_rate": 0.0001042797761598537,
+      "loss": 4.2399,
+      "step": 283000
+    },
+    {
+      "epoch": 2.43672520661157,
+      "grad_norm": 3.15625,
+      "learning_rate": 0.00010425270878866553,
+      "loss": 4.3552,
+      "step": 283050
+    },
+    {
+      "epoch": 2.4371556473829203,
+      "grad_norm": 3.0,
+      "learning_rate": 0.00010422564110533772,
+      "loss": 4.3942,
+      "step": 283100
+    },
+    {
+      "epoch": 2.43758608815427,
+      "grad_norm": 1.5390625,
+      "learning_rate": 0.00010419857311185693,
+      "loss": 3.9753,
+      "step": 283150
+    },
+    {
+      "epoch": 2.43801652892562,
+      "grad_norm": 3.34375,
+      "learning_rate": 0.00010417150481020994,
+      "loss": 4.5873,
+      "step": 283200
+    },
+    {
+      "epoch": 2.4384469696969697,
+      "grad_norm": 3.640625,
+      "learning_rate": 0.00010414443620238343,
+      "loss": 4.5218,
+      "step": 283250
+    },
+    {
+      "epoch": 2.4388774104683195,
+      "grad_norm": 4.8125,
+      "learning_rate": 0.00010411736729036421,
+      "loss": 4.5451,
+      "step": 283300
+    },
+    {
+      "epoch": 2.4393078512396693,
+      "grad_norm": 4.3125,
+      "learning_rate": 0.00010409029807613914,
+      "loss": 4.564,
+      "step": 283350
+    },
+    {
+      "epoch": 2.439738292011019,
+      "grad_norm": 3.9375,
+      "learning_rate": 0.00010406322856169497,
+      "loss": 4.5764,
+      "step": 283400
+    },
+    {
+      "epoch": 2.440168732782369,
+      "grad_norm": 3.09375,
+      "learning_rate": 0.00010403615874901857,
+      "loss": 4.0948,
+      "step": 283450
+    },
+    {
+      "epoch": 2.440599173553719,
+      "grad_norm": 2.46875,
+      "learning_rate": 0.00010400908864009677,
+      "loss": 4.5149,
+      "step": 283500
+    },
+    {
+      "epoch": 2.441029614325069,
+      "grad_norm": 3.21875,
+      "learning_rate": 0.00010398201823691653,
+      "loss": 4.5937,
+      "step": 283550
+    },
+    {
+      "epoch": 2.4414600550964187,
+      "grad_norm": 4.53125,
+      "learning_rate": 0.00010395494754146471,
+      "loss": 4.6073,
+      "step": 283600
+    },
+    {
+      "epoch": 2.4418904958677685,
+      "grad_norm": 8.25,
+      "learning_rate": 0.00010392787655572823,
+      "loss": 4.2603,
+      "step": 283650
+    },
+    {
+      "epoch": 2.4423209366391183,
+      "grad_norm": 3.03125,
+      "learning_rate": 0.00010390080528169407,
+      "loss": 4.4448,
+      "step": 283700
+    },
+    {
+      "epoch": 2.4427513774104685,
+      "grad_norm": 2.546875,
+      "learning_rate": 0.00010387373372134923,
+      "loss": 4.5935,
+      "step": 283750
+    },
+    {
+      "epoch": 2.4431818181818183,
+      "grad_norm": 2.9375,
+      "learning_rate": 0.00010384666187668061,
+      "loss": 4.577,
+      "step": 283800
+    },
+    {
+      "epoch": 2.443612258953168,
+      "grad_norm": 4.125,
+      "learning_rate": 0.00010381958974967534,
+      "loss": 4.351,
+      "step": 283850
+    },
+    {
+      "epoch": 2.444042699724518,
+      "grad_norm": 1.9140625,
+      "learning_rate": 0.00010379251734232036,
+      "loss": 3.7904,
+      "step": 283900
+    },
+    {
+      "epoch": 2.4444731404958677,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.00010376544465660278,
+      "loss": 4.6604,
+      "step": 283950
+    },
+    {
+      "epoch": 2.4449035812672175,
+      "grad_norm": 3.921875,
+      "learning_rate": 0.00010373837169450966,
+      "loss": 4.3474,
+      "step": 284000
+    },
+    {
+      "epoch": 2.4453340220385673,
+      "grad_norm": 2.984375,
+      "learning_rate": 0.00010371129845802812,
+      "loss": 4.0781,
+      "step": 284050
+    },
+    {
+      "epoch": 2.445764462809917,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.00010368422494914525,
+      "loss": 4.3142,
+      "step": 284100
+    },
+    {
+      "epoch": 2.4461949035812673,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.00010365715116984818,
+      "loss": 4.3244,
+      "step": 284150
+    },
+    {
+      "epoch": 2.446625344352617,
+      "grad_norm": 2.8125,
+      "learning_rate": 0.0001036300771221241,
+      "loss": 4.0259,
+      "step": 284200
+    },
+    {
+      "epoch": 2.447055785123967,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.00010360300280796018,
+      "loss": 4.2041,
+      "step": 284250
+    },
+    {
+      "epoch": 2.4474862258953167,
+      "grad_norm": 1.5078125,
+      "learning_rate": 0.00010357592822934356,
+      "loss": 4.0192,
+      "step": 284300
+    },
+    {
+      "epoch": 2.4479166666666665,
+      "grad_norm": 1.9140625,
+      "learning_rate": 0.00010354885338826155,
+      "loss": 4.3531,
+      "step": 284350
+    },
+    {
+      "epoch": 2.4483471074380168,
+      "grad_norm": 3.4375,
+      "learning_rate": 0.00010352177828670131,
+      "loss": 4.2375,
+      "step": 284400
+    },
+    {
+      "epoch": 2.4487775482093666,
+      "grad_norm": 5.15625,
+      "learning_rate": 0.00010349470292665011,
+      "loss": 4.535,
+      "step": 284450
+    },
+    {
+      "epoch": 2.4492079889807163,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00010346762731009527,
+      "loss": 4.5689,
+      "step": 284500
+    },
+    {
+      "epoch": 2.449638429752066,
+      "grad_norm": 3.078125,
+      "learning_rate": 0.00010344055143902403,
+      "loss": 4.5238,
+      "step": 284550
+    },
+    {
+      "epoch": 2.450068870523416,
+      "grad_norm": 3.296875,
+      "learning_rate": 0.00010341347531542372,
+      "loss": 4.5314,
+      "step": 284600
+    },
+    {
+      "epoch": 2.4504993112947657,
+      "grad_norm": 3.5,
+      "learning_rate": 0.00010338639894128166,
+      "loss": 4.9696,
+      "step": 284650
+    },
+    {
+      "epoch": 2.4509297520661155,
+      "grad_norm": 3.4375,
+      "learning_rate": 0.0001033593223185852,
+      "loss": 4.3965,
+      "step": 284700
+    },
+    {
+      "epoch": 2.4513601928374658,
+      "grad_norm": 2.0,
+      "learning_rate": 0.00010333224544932174,
+      "loss": 4.423,
+      "step": 284750
+    },
+    {
+      "epoch": 2.4517906336088156,
+      "grad_norm": 3.15625,
+      "learning_rate": 0.00010330516833547859,
+      "loss": 4.2458,
+      "step": 284800
+    },
+    {
+      "epoch": 2.4522210743801653,
+      "grad_norm": 4.5,
+      "learning_rate": 0.0001032780909790432,
+      "loss": 4.3486,
+      "step": 284850
+    },
+    {
+      "epoch": 2.452651515151515,
+      "grad_norm": 3.171875,
+      "learning_rate": 0.00010325101338200303,
+      "loss": 4.2645,
+      "step": 284900
+    },
+    {
+      "epoch": 2.453081955922865,
+      "grad_norm": 2.78125,
+      "learning_rate": 0.00010322393554634541,
+      "loss": 4.4039,
+      "step": 284950
+    },
+    {
+      "epoch": 2.4535123966942147,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.0001031968574740579,
+      "loss": 3.922,
+      "step": 285000
+    },
+    {
+      "epoch": 2.4535123966942147,
+      "eval_loss": 5.01358699798584,
+      "eval_runtime": 24.04,
+      "eval_samples_per_second": 26.622,
+      "eval_steps_per_second": 13.311,
+      "eval_tts_loss": 7.434466893334985,
+      "step": 285000
+    },
+    {
+      "epoch": 2.453942837465565,
+      "grad_norm": 4.71875,
+      "learning_rate": 0.00010316977916712793,
+      "loss": 4.2276,
+      "step": 285050
+    },
+    {
+      "epoch": 2.4543732782369148,
+      "grad_norm": 2.75,
+      "learning_rate": 0.00010314270062754297,
+      "loss": 4.484,
+      "step": 285100
+    },
+    {
+      "epoch": 2.4548037190082646,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.00010311562185729056,
+      "loss": 4.2807,
+      "step": 285150
+    },
+    {
+      "epoch": 2.4552341597796143,
+      "grad_norm": 3.109375,
+      "learning_rate": 0.00010308854285835819,
+      "loss": 4.2822,
+      "step": 285200
+    },
+    {
+      "epoch": 2.455664600550964,
+      "grad_norm": 4.71875,
+      "learning_rate": 0.00010306146363273346,
+      "loss": 4.5957,
+      "step": 285250
+    },
+    {
+      "epoch": 2.456095041322314,
+      "grad_norm": 3.296875,
+      "learning_rate": 0.00010303438418240383,
+      "loss": 4.2958,
+      "step": 285300
+    },
+    {
+      "epoch": 2.4565254820936637,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.00010300730450935694,
+      "loss": 4.3225,
+      "step": 285350
+    },
+    {
+      "epoch": 2.456955922865014,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00010298022461558041,
+      "loss": 4.4848,
+      "step": 285400
+    },
+    {
+      "epoch": 2.4573863636363638,
+      "grad_norm": 3.015625,
+      "learning_rate": 0.00010295314450306177,
+      "loss": 4.4985,
+      "step": 285450
+    },
+    {
+      "epoch": 2.4578168044077136,
+      "grad_norm": 3.25,
+      "learning_rate": 0.00010292606417378869,
+      "loss": 4.2136,
+      "step": 285500
+    },
+    {
+      "epoch": 2.4582472451790633,
+      "grad_norm": 5.59375,
+      "learning_rate": 0.0001028989836297488,
+      "loss": 4.3001,
+      "step": 285550
+    },
+    {
+      "epoch": 2.458677685950413,
+      "grad_norm": 3.859375,
+      "learning_rate": 0.00010287190287292977,
+      "loss": 4.4967,
+      "step": 285600
+    },
+    {
+      "epoch": 2.459108126721763,
+      "grad_norm": 3.5625,
+      "learning_rate": 0.00010284482190531921,
+      "loss": 4.2818,
+      "step": 285650
+    },
+    {
+      "epoch": 2.459538567493113,
+      "grad_norm": 3.359375,
+      "learning_rate": 0.00010281774072890484,
+      "loss": 4.3762,
+      "step": 285700
+    },
+    {
+      "epoch": 2.459969008264463,
+      "grad_norm": 3.34375,
+      "learning_rate": 0.00010279065934567438,
+      "loss": 4.5424,
+      "step": 285750
+    },
+    {
+      "epoch": 2.4603994490358128,
+      "grad_norm": 3.421875,
+      "learning_rate": 0.00010276357775761553,
+      "loss": 3.999,
+      "step": 285800
+    },
+    {
+      "epoch": 2.4608298898071626,
+      "grad_norm": 4.3125,
+      "learning_rate": 0.00010273649596671602,
+      "loss": 4.1942,
+      "step": 285850
+    },
+    {
+      "epoch": 2.4612603305785123,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.00010270941397496358,
+      "loss": 4.6793,
+      "step": 285900
+    },
+    {
+      "epoch": 2.461690771349862,
+      "grad_norm": 2.984375,
+      "learning_rate": 0.00010268233178434601,
+      "loss": 4.1216,
+      "step": 285950
+    },
+    {
+      "epoch": 2.462121212121212,
+      "grad_norm": 2.0625,
+      "learning_rate": 0.00010265524939685103,
+      "loss": 4.3523,
+      "step": 286000
+    },
+    {
+      "epoch": 2.462551652892562,
+      "grad_norm": 1.6171875,
+      "learning_rate": 0.00010262816681446648,
+      "loss": 4.3751,
+      "step": 286050
+    },
+    {
+      "epoch": 2.462982093663912,
+      "grad_norm": 3.265625,
+      "learning_rate": 0.00010260108403918012,
+      "loss": 4.0416,
+      "step": 286100
+    },
+    {
+      "epoch": 2.4634125344352618,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.00010257400107297981,
+      "loss": 4.2597,
+      "step": 286150
+    },
+    {
+      "epoch": 2.4638429752066116,
+      "grad_norm": 3.0625,
+      "learning_rate": 0.00010254691791785332,
+      "loss": 4.2485,
+      "step": 286200
+    },
+    {
+      "epoch": 2.4642734159779613,
+      "grad_norm": 2.1875,
+      "learning_rate": 0.00010251983457578858,
+      "loss": 4.167,
+      "step": 286250
+    },
+    {
+      "epoch": 2.464703856749311,
+      "grad_norm": 2.25,
+      "learning_rate": 0.00010249275104877338,
+      "loss": 4.576,
+      "step": 286300
+    },
+    {
+      "epoch": 2.4651342975206614,
+      "grad_norm": 3.21875,
+      "learning_rate": 0.00010246566733879561,
+      "loss": 4.4953,
+      "step": 286350
+    },
+    {
+      "epoch": 2.465564738292011,
+      "grad_norm": 3.34375,
+      "learning_rate": 0.00010243858344784316,
+      "loss": 4.4542,
+      "step": 286400
+    },
+    {
+      "epoch": 2.465995179063361,
+      "grad_norm": 1.7109375,
+      "learning_rate": 0.00010241149937790397,
+      "loss": 4.1143,
+      "step": 286450
+    },
+    {
+      "epoch": 2.4664256198347108,
+      "grad_norm": 3.140625,
+      "learning_rate": 0.00010238441513096589,
+      "loss": 4.3711,
+      "step": 286500
+    },
+    {
+      "epoch": 2.4668560606060606,
+      "grad_norm": 3.40625,
+      "learning_rate": 0.00010235733070901685,
+      "loss": 4.3277,
+      "step": 286550
+    },
+    {
+      "epoch": 2.4672865013774103,
+      "grad_norm": 2.859375,
+      "learning_rate": 0.00010233024611404485,
+      "loss": 4.3605,
+      "step": 286600
+    },
+    {
+      "epoch": 2.46771694214876,
+      "grad_norm": 3.859375,
+      "learning_rate": 0.00010230316134803774,
+      "loss": 4.157,
+      "step": 286650
+    },
+    {
+      "epoch": 2.4681473829201104,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.0001022760764129836,
+      "loss": 4.446,
+      "step": 286700
+    },
+    {
+      "epoch": 2.46857782369146,
+      "grad_norm": 2.890625,
+      "learning_rate": 0.0001022489913108703,
+      "loss": 4.4489,
+      "step": 286750
+    },
+    {
+      "epoch": 2.46900826446281,
+      "grad_norm": 2.71875,
+      "learning_rate": 0.00010222190604368593,
+      "loss": 3.9839,
+      "step": 286800
+    },
+    {
+      "epoch": 2.4694387052341598,
+      "grad_norm": 2.359375,
+      "learning_rate": 0.00010219482061341841,
+      "loss": 4.1401,
+      "step": 286850
+    },
+    {
+      "epoch": 2.4698691460055096,
+      "grad_norm": 3.03125,
+      "learning_rate": 0.00010216773502205574,
+      "loss": 3.8893,
+      "step": 286900
+    },
+    {
+      "epoch": 2.4702995867768593,
+      "grad_norm": 3.9375,
+      "learning_rate": 0.00010214064927158606,
+      "loss": 4.4247,
+      "step": 286950
+    },
+    {
+      "epoch": 2.4707300275482096,
+      "grad_norm": 3.6875,
+      "learning_rate": 0.00010211356336399729,
+      "loss": 4.2582,
+      "step": 287000
+    },
+    {
+      "epoch": 2.4711604683195594,
+      "grad_norm": 4.5,
+      "learning_rate": 0.00010208647730127752,
+      "loss": 4.2929,
+      "step": 287050
+    },
+    {
+      "epoch": 2.471590909090909,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.00010205939108541478,
+      "loss": 4.0117,
+      "step": 287100
+    },
+    {
+      "epoch": 2.472021349862259,
+      "grad_norm": 0.828125,
+      "learning_rate": 0.0001020323047183972,
+      "loss": 4.225,
+      "step": 287150
+    },
+    {
+      "epoch": 2.4724517906336088,
+      "grad_norm": 5.03125,
+      "learning_rate": 0.00010200521820221283,
+      "loss": 4.5542,
+      "step": 287200
+    },
+    {
+      "epoch": 2.4728822314049586,
+      "grad_norm": 4.1875,
+      "learning_rate": 0.00010197813153884972,
+      "loss": 4.7844,
+      "step": 287250
+    },
+    {
+      "epoch": 2.4733126721763083,
+      "grad_norm": 6.03125,
+      "learning_rate": 0.00010195104473029605,
+      "loss": 4.5003,
+      "step": 287300
+    },
+    {
+      "epoch": 2.4737431129476586,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.00010192395777853988,
+      "loss": 4.0351,
+      "step": 287350
+    },
+    {
+      "epoch": 2.4741735537190084,
+      "grad_norm": 3.640625,
+      "learning_rate": 0.00010189687068556933,
+      "loss": 4.3536,
+      "step": 287400
+    },
+    {
+      "epoch": 2.474603994490358,
+      "grad_norm": 2.90625,
+      "learning_rate": 0.00010186978345337257,
+      "loss": 4.382,
+      "step": 287450
+    },
+    {
+      "epoch": 2.475034435261708,
+      "grad_norm": 2.84375,
+      "learning_rate": 0.00010184269608393773,
+      "loss": 4.3086,
+      "step": 287500
+    },
+    {
+      "epoch": 2.4754648760330578,
+      "grad_norm": 4.28125,
+      "learning_rate": 0.00010181560857925294,
+      "loss": 4.0329,
+      "step": 287550
+    },
+    {
+      "epoch": 2.4758953168044076,
+      "grad_norm": 2.953125,
+      "learning_rate": 0.00010178852094130642,
+      "loss": 4.5042,
+      "step": 287600
+    },
+    {
+      "epoch": 2.476325757575758,
+      "grad_norm": 1.234375,
+      "learning_rate": 0.00010176143317208631,
+      "loss": 4.2629,
+      "step": 287650
+    },
+    {
+      "epoch": 2.4767561983471076,
+      "grad_norm": 3.609375,
+      "learning_rate": 0.00010173434527358079,
+      "loss": 4.5112,
+      "step": 287700
+    },
+    {
+      "epoch": 2.4771866391184574,
+      "grad_norm": 3.28125,
+      "learning_rate": 0.00010170725724777803,
+      "loss": 4.0819,
+      "step": 287750
+    },
+    {
+      "epoch": 2.477617079889807,
+      "grad_norm": 3.5625,
+      "learning_rate": 0.00010168016909666629,
+      "loss": 4.3617,
+      "step": 287800
+    },
+    {
+      "epoch": 2.478047520661157,
+      "grad_norm": 4.21875,
+      "learning_rate": 0.00010165308082223375,
+      "loss": 3.8755,
+      "step": 287850
+    },
+    {
+      "epoch": 2.4784779614325068,
+      "grad_norm": 1.2109375,
+      "learning_rate": 0.00010162599242646862,
+      "loss": 4.414,
+      "step": 287900
+    },
+    {
+      "epoch": 2.4789084022038566,
+      "grad_norm": 3.40625,
+      "learning_rate": 0.00010159890391135916,
+      "loss": 4.2458,
+      "step": 287950
+    },
+    {
+      "epoch": 2.479338842975207,
+      "grad_norm": 2.53125,
+      "learning_rate": 0.00010157181527889358,
+      "loss": 4.603,
+      "step": 288000
+    },
+    {
+      "epoch": 2.479338842975207,
+      "eval_loss": 5.010517597198486,
+      "eval_runtime": 24.4818,
+      "eval_samples_per_second": 26.142,
+      "eval_steps_per_second": 13.071,
+      "eval_tts_loss": 7.395733432660736,
+      "step": 288000
+    },
+    {
+      "epoch": 2.4797692837465566,
+      "grad_norm": 3.1875,
+      "learning_rate": 0.00010154472653106016,
+      "loss": 4.504,
+      "step": 288050
+    },
+    {
+      "epoch": 2.4801997245179064,
+      "grad_norm": 3.671875,
+      "learning_rate": 0.00010151763766984713,
+      "loss": 4.1512,
+      "step": 288100
+    },
+    {
+      "epoch": 2.480630165289256,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.00010149054869724274,
+      "loss": 4.0913,
+      "step": 288150
+    },
+    {
+      "epoch": 2.481060606060606,
+      "grad_norm": 2.609375,
+      "learning_rate": 0.00010146345961523532,
+      "loss": 4.299,
+      "step": 288200
+    },
+    {
+      "epoch": 2.4814910468319558,
+      "grad_norm": 3.125,
+      "learning_rate": 0.00010143637042581307,
+      "loss": 4.2602,
+      "step": 288250
+    },
+    {
+      "epoch": 2.481921487603306,
+      "grad_norm": 3.078125,
+      "learning_rate": 0.00010140928113096435,
+      "loss": 4.3977,
+      "step": 288300
+    },
+    {
+      "epoch": 2.482351928374656,
+      "grad_norm": 3.234375,
+      "learning_rate": 0.00010138219173267742,
+      "loss": 4.3319,
+      "step": 288350
+    },
+    {
+      "epoch": 2.4827823691460056,
+      "grad_norm": 3.890625,
+      "learning_rate": 0.00010135510223294059,
+      "loss": 3.8356,
+      "step": 288400
+    },
+    {
+      "epoch": 2.4832128099173554,
+      "grad_norm": 1.9921875,
+      "learning_rate": 0.00010132801263374215,
+      "loss": 4.0408,
+      "step": 288450
+    },
+    {
+      "epoch": 2.483643250688705,
+      "grad_norm": 4.3125,
+      "learning_rate": 0.00010130092293707047,
+      "loss": 4.2576,
+      "step": 288500
+    },
+    {
+      "epoch": 2.484073691460055,
+      "grad_norm": 2.25,
+      "learning_rate": 0.00010127383314491384,
+      "loss": 4.3203,
+      "step": 288550
+    },
+    {
+      "epoch": 2.4845041322314048,
+      "grad_norm": 3.078125,
+      "learning_rate": 0.00010124674325926057,
+      "loss": 4.4197,
+      "step": 288600
+    },
+    {
+      "epoch": 2.484934573002755,
+      "grad_norm": 2.40625,
+      "learning_rate": 0.00010121965328209905,
+      "loss": 4.369,
+      "step": 288650
+    },
+    {
+      "epoch": 2.485365013774105,
+      "grad_norm": 2.109375,
+      "learning_rate": 0.00010119256321541761,
+      "loss": 4.5727,
+      "step": 288700
+    },
+    {
+      "epoch": 2.4857954545454546,
+      "grad_norm": 2.828125,
+      "learning_rate": 0.00010116547306120461,
+      "loss": 4.3041,
+      "step": 288750
+    },
+    {
+      "epoch": 2.4862258953168044,
+      "grad_norm": 5.8125,
+      "learning_rate": 0.00010113838282144834,
+      "loss": 4.2564,
+      "step": 288800
+    },
+    {
+      "epoch": 2.486656336088154,
+      "grad_norm": 4.25,
+      "learning_rate": 0.00010111129249813728,
+      "loss": 4.7618,
+      "step": 288850
+    },
+    {
+      "epoch": 2.487086776859504,
+      "grad_norm": 2.28125,
+      "learning_rate": 0.00010108420209325971,
+      "loss": 4.2519,
+      "step": 288900
+    },
+    {
+      "epoch": 2.487517217630854,
+      "grad_norm": 6.375,
+      "learning_rate": 0.00010105711160880402,
+      "loss": 4.4035,
+      "step": 288950
+    },
+    {
+      "epoch": 2.487947658402204,
+      "grad_norm": 2.734375,
+      "learning_rate": 0.00010103002104675866,
+      "loss": 4.3746,
+      "step": 289000
+    },
+    {
+      "epoch": 2.488378099173554,
+      "grad_norm": 3.625,
+      "learning_rate": 0.00010100293040911196,
+      "loss": 4.0388,
+      "step": 289050
+    },
+    {
+      "epoch": 2.4888085399449036,
+      "grad_norm": 3.90625,
+      "learning_rate": 0.00010097583969785231,
+      "loss": 4.3886,
+      "step": 289100
+    },
+    {
+      "epoch": 2.4892389807162534,
+      "grad_norm": 3.75,
+      "learning_rate": 0.00010094874891496812,
+      "loss": 4.766,
+      "step": 289150
+    },
+    {
+      "epoch": 2.489669421487603,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00010092165806244783,
+      "loss": 4.5697,
+      "step": 289200
+    },
+    {
+      "epoch": 2.490099862258953,
+      "grad_norm": 2.59375,
+      "learning_rate": 0.00010089456714227981,
+      "loss": 4.4505,
+      "step": 289250
+    },
+    {
+      "epoch": 2.490530303030303,
+      "grad_norm": 2.703125,
+      "learning_rate": 0.00010086747615645245,
+      "loss": 4.6242,
+      "step": 289300
+    },
+    {
+      "epoch": 2.490960743801653,
+      "grad_norm": 1.7734375,
+      "learning_rate": 0.00010084038510695424,
+      "loss": 4.5188,
+      "step": 289350
+    },
+    {
+      "epoch": 2.491391184573003,
+      "grad_norm": 2.640625,
+      "learning_rate": 0.00010081329399577357,
+      "loss": 4.3866,
+      "step": 289400
+    },
+    {
+      "epoch": 2.4918216253443526,
+      "grad_norm": 2.34375,
+      "learning_rate": 0.00010078620282489885,
+      "loss": 4.3092,
+      "step": 289450
+    },
+    {
+      "epoch": 2.4922520661157024,
+      "grad_norm": 3.0625,
+      "learning_rate": 0.00010075911159631857,
+      "loss": 4.5881,
+      "step": 289500
+    },
+    {
+      "epoch": 2.492682506887052,
+      "grad_norm": 2.09375,
+      "learning_rate": 0.00010073202031202112,
+      "loss": 4.678,
+      "step": 289550
+    },
+    {
+      "epoch": 2.4931129476584024,
+      "grad_norm": 3.453125,
+      "learning_rate": 0.00010070492897399493,
+      "loss": 4.1268,
+      "step": 289600
+    },
+    {
+      "epoch": 2.493543388429752,
+      "grad_norm": 3.34375,
+      "learning_rate": 0.00010067783758422851,
+      "loss": 4.1757,
+      "step": 289650
+    },
+    {
+      "epoch": 2.493973829201102,
+      "grad_norm": 3.328125,
+      "learning_rate": 0.00010065074614471023,
+      "loss": 4.5447,
+      "step": 289700
+    },
+    {
+      "epoch": 2.494404269972452,
+      "grad_norm": 1.875,
+      "learning_rate": 0.00010062365465742862,
+      "loss": 4.5332,
+      "step": 289750
+    },
+    {
+      "epoch": 2.4948347107438016,
+      "grad_norm": 2.203125,
+      "learning_rate": 0.00010059656312437207,
+      "loss": 4.2415,
+      "step": 289800
+    },
+    {
+      "epoch": 2.4952651515151514,
+      "grad_norm": 3.640625,
+      "learning_rate": 0.0001005694715475291,
+      "loss": 4.4303,
+      "step": 289850
+    },
+    {
+      "epoch": 2.495695592286501,
+      "grad_norm": 2.984375,
+      "learning_rate": 0.00010054237992888811,
+      "loss": 4.4299,
+      "step": 289900
+    },
+    {
+      "epoch": 2.4961260330578514,
+      "grad_norm": 4.21875,
+      "learning_rate": 0.00010051528827043761,
+      "loss": 4.5791,
+      "step": 289950
+    },
+    {
+      "epoch": 2.496556473829201,
+      "grad_norm": 7.125,
+      "learning_rate": 0.00010048819657416606,
+      "loss": 4.5008,
+      "step": 290000
+    },
+    {
+      "epoch": 2.496986914600551,
+      "grad_norm": 3.90625,
+      "learning_rate": 0.00010046110484206195,
+      "loss": 4.5097,
+      "step": 290050
+    },
+    {
+      "epoch": 2.497417355371901,
+      "grad_norm": 3.171875,
+      "learning_rate": 0.00010043401307611371,
+      "loss": 4.4785,
+      "step": 290100
+    },
+    {
+      "epoch": 2.4978477961432506,
+      "grad_norm": 1.5625,
+      "learning_rate": 0.00010040692127830987,
+      "loss": 4.389,
+      "step": 290150
+    },
+    {
+      "epoch": 2.4982782369146004,
+      "grad_norm": 1.796875,
+      "learning_rate": 0.00010037982945063887,
+      "loss": 4.2936,
+      "step": 290200
+    },
+    {
+      "epoch": 2.4987086776859506,
+      "grad_norm": 4.1875,
+      "learning_rate": 0.00010035273759508919,
+      "loss": 4.1188,
+      "step": 290250
+    },
+    {
+      "epoch": 2.4991391184573004,
+      "grad_norm": 3.328125,
+      "learning_rate": 0.00010032564571364935,
+      "loss": 4.3657,
+      "step": 290300
+    },
+    {
+      "epoch": 2.49956955922865,
+      "grad_norm": 4.4375,
+      "learning_rate": 0.00010029855380830779,
+      "loss": 4.3694,
+      "step": 290350
+    },
+    {
+      "epoch": 2.5,
+      "grad_norm": 3.328125,
+      "learning_rate": 0.00010027146188105307,
+      "loss": 4.3072,
+      "step": 290400
+    },
+    {
+      "epoch": 2.50043044077135,
+      "grad_norm": 2.015625,
+      "learning_rate": 0.00010024436993387361,
+      "loss": 4.0094,
+      "step": 290450
+    },
+    {
+      "epoch": 2.5008608815426996,
+      "grad_norm": 2.234375,
+      "learning_rate": 0.00010021727796875791,
+      "loss": 4.2795,
+      "step": 290500
+    },
+    {
+      "epoch": 2.5012913223140494,
+      "grad_norm": 2.578125,
+      "learning_rate": 0.0001001901859876945,
+      "loss": 4.1298,
+      "step": 290550
+    },
+    {
+      "epoch": 2.501721763085399,
+      "grad_norm": 4.0,
+      "learning_rate": 0.00010016309399267181,
+      "loss": 4.3885,
+      "step": 290600
+    },
+    {
+      "epoch": 2.5021522038567494,
+      "grad_norm": 3.9375,
+      "learning_rate": 0.0001001360019856784,
+      "loss": 4.7,
+      "step": 290650
+    },
+    {
+      "epoch": 2.502582644628099,
+      "grad_norm": 6.0,
+      "learning_rate": 0.00010010890996870272,
+      "loss": 4.2198,
+      "step": 290700
+    },
+    {
+      "epoch": 2.503013085399449,
+      "grad_norm": 4.0625,
+      "learning_rate": 0.0001000818179437333,
+      "loss": 4.289,
+      "step": 290750
+    },
+    {
+      "epoch": 2.503443526170799,
+      "grad_norm": 3.296875,
+      "learning_rate": 0.00010005472591275862,
+      "loss": 4.1972,
+      "step": 290800
+    },
+    {
+      "epoch": 2.503873966942149,
+      "grad_norm": 2.046875,
+      "learning_rate": 0.00010002763387776715,
+      "loss": 4.49,
+      "step": 290850
+    },
+    {
+      "epoch": 2.504304407713499,
+      "grad_norm": 2.796875,
+      "learning_rate": 0.00010000054184074746,
+      "loss": 4.5324,
+      "step": 290900
+    },
+    {
+      "epoch": 2.5047348484848486,
+      "grad_norm": 4.625,
+      "learning_rate": 9.997344980368795e-05,
+      "loss": 4.7044,
+      "step": 290950
+    },
+    {
+      "epoch": 2.5051652892561984,
+      "grad_norm": 6.125,
+      "learning_rate": 9.994635776857717e-05,
+      "loss": 4.3919,
+      "step": 291000
+    },
+    {
+      "epoch": 2.5051652892561984,
+      "eval_loss": 5.009791851043701,
+      "eval_runtime": 24.0994,
+      "eval_samples_per_second": 26.557,
+      "eval_steps_per_second": 13.278,
+      "eval_tts_loss": 7.433074333890757,
+      "step": 291000
+    },
+    {
+      "epoch": 2.505595730027548,
+      "grad_norm": 1.6640625,
+      "learning_rate": 9.991926573740364e-05,
+      "loss": 4.2164,
+      "step": 291050
+    },
+    {
+      "epoch": 2.506026170798898,
+      "grad_norm": 3.640625,
+      "learning_rate": 9.989217371215585e-05,
+      "loss": 4.427,
+      "step": 291100
+    },
+    {
+      "epoch": 2.506456611570248,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.986508169482223e-05,
+      "loss": 4.3297,
+      "step": 291150
+    },
+    {
+      "epoch": 2.5068870523415976,
+      "grad_norm": 4.8125,
+      "learning_rate": 9.983798968739132e-05,
+      "loss": 4.1809,
+      "step": 291200
+    },
+    {
+      "epoch": 2.5073174931129474,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.981089769185166e-05,
+      "loss": 4.3658,
+      "step": 291250
+    },
+    {
+      "epoch": 2.5077479338842976,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.978380571019167e-05,
+      "loss": 4.3938,
+      "step": 291300
+    },
+    {
+      "epoch": 2.5081783746556474,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.975671374439993e-05,
+      "loss": 4.0513,
+      "step": 291350
+    },
+    {
+      "epoch": 2.508608815426997,
+      "grad_norm": 4.59375,
+      "learning_rate": 9.972962179646481e-05,
+      "loss": 4.4905,
+      "step": 291400
+    },
+    {
+      "epoch": 2.509039256198347,
+      "grad_norm": 4.1875,
+      "learning_rate": 9.97025298683749e-05,
+      "loss": 4.2303,
+      "step": 291450
+    },
+    {
+      "epoch": 2.5094696969696972,
+      "grad_norm": 5.15625,
+      "learning_rate": 9.967543796211865e-05,
+      "loss": 4.0584,
+      "step": 291500
+    },
+    {
+      "epoch": 2.509900137741047,
+      "grad_norm": 2.484375,
+      "learning_rate": 9.964834607968457e-05,
+      "loss": 4.0834,
+      "step": 291550
+    },
+    {
+      "epoch": 2.510330578512397,
+      "grad_norm": 2.265625,
+      "learning_rate": 9.96212542230611e-05,
+      "loss": 4.53,
+      "step": 291600
+    },
+    {
+      "epoch": 2.5107610192837466,
+      "grad_norm": 2.390625,
+      "learning_rate": 9.95941623942368e-05,
+      "loss": 4.3153,
+      "step": 291650
+    },
+    {
+      "epoch": 2.5111914600550964,
+      "grad_norm": 2.453125,
+      "learning_rate": 9.956707059520007e-05,
+      "loss": 3.8703,
+      "step": 291700
+    },
+    {
+      "epoch": 2.511621900826446,
+      "grad_norm": 4.09375,
+      "learning_rate": 9.953997882793943e-05,
+      "loss": 4.2936,
+      "step": 291750
+    },
+    {
+      "epoch": 2.512052341597796,
+      "grad_norm": 2.28125,
+      "learning_rate": 9.951288709444336e-05,
+      "loss": 4.3772,
+      "step": 291800
+    },
+    {
+      "epoch": 2.512482782369146,
+      "grad_norm": 1.75,
+      "learning_rate": 9.948579539670033e-05,
+      "loss": 4.2879,
+      "step": 291850
+    },
+    {
+      "epoch": 2.512913223140496,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.945870373669883e-05,
+      "loss": 4.5433,
+      "step": 291900
+    },
+    {
+      "epoch": 2.513343663911846,
+      "grad_norm": 3.984375,
+      "learning_rate": 9.94316121164273e-05,
+      "loss": 4.6699,
+      "step": 291950
+    },
+    {
+      "epoch": 2.5137741046831956,
+      "grad_norm": 3.609375,
+      "learning_rate": 9.94045205378742e-05,
+      "loss": 4.537,
+      "step": 292000
+    },
+    {
+      "epoch": 2.5142045454545454,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.937742900302801e-05,
+      "loss": 4.537,
+      "step": 292050
+    },
+    {
+      "epoch": 2.514634986225895,
+      "grad_norm": 4.5625,
+      "learning_rate": 9.935033751387721e-05,
+      "loss": 4.3589,
+      "step": 292100
+    },
+    {
+      "epoch": 2.5150654269972454,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.932324607241028e-05,
+      "loss": 4.3449,
+      "step": 292150
+    },
+    {
+      "epoch": 2.5154958677685952,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.929615468061556e-05,
+      "loss": 4.543,
+      "step": 292200
+    },
+    {
+      "epoch": 2.515926308539945,
+      "grad_norm": 3.671875,
+      "learning_rate": 9.926906334048158e-05,
+      "loss": 4.1635,
+      "step": 292250
+    },
+    {
+      "epoch": 2.516356749311295,
+      "grad_norm": 2.25,
+      "learning_rate": 9.924197205399682e-05,
+      "loss": 4.2854,
+      "step": 292300
+    },
+    {
+      "epoch": 2.5167871900826446,
+      "grad_norm": 3.5,
+      "learning_rate": 9.921488082314964e-05,
+      "loss": 4.3496,
+      "step": 292350
+    },
+    {
+      "epoch": 2.5172176308539944,
+      "grad_norm": 3.453125,
+      "learning_rate": 9.918778964992856e-05,
+      "loss": 4.4551,
+      "step": 292400
+    },
+    {
+      "epoch": 2.517648071625344,
+      "grad_norm": 1.75,
+      "learning_rate": 9.916069853632195e-05,
+      "loss": 4.4931,
+      "step": 292450
+    },
+    {
+      "epoch": 2.518078512396694,
+      "grad_norm": 2.375,
+      "learning_rate": 9.913360748431825e-05,
+      "loss": 4.0472,
+      "step": 292500
+    },
+    {
+      "epoch": 2.5185089531680442,
+      "grad_norm": 4.21875,
+      "learning_rate": 9.910651649590588e-05,
+      "loss": 3.9426,
+      "step": 292550
+    },
+    {
+      "epoch": 2.518939393939394,
+      "grad_norm": 2.328125,
+      "learning_rate": 9.907942557307331e-05,
+      "loss": 4.8473,
+      "step": 292600
+    },
+    {
+      "epoch": 2.519369834710744,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.905233471780894e-05,
+      "loss": 4.7175,
+      "step": 292650
+    },
+    {
+      "epoch": 2.5198002754820936,
+      "grad_norm": 3.53125,
+      "learning_rate": 9.902524393210113e-05,
+      "loss": 4.1179,
+      "step": 292700
+    },
+    {
+      "epoch": 2.5202307162534434,
+      "grad_norm": 5.1875,
+      "learning_rate": 9.899815321793831e-05,
+      "loss": 4.4922,
+      "step": 292750
+    },
+    {
+      "epoch": 2.5206611570247937,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.897106257730892e-05,
+      "loss": 4.6808,
+      "step": 292800
+    },
+    {
+      "epoch": 2.5210915977961434,
+      "grad_norm": 2.015625,
+      "learning_rate": 9.89439720122013e-05,
+      "loss": 4.5483,
+      "step": 292850
+    },
+    {
+      "epoch": 2.5215220385674932,
+      "grad_norm": 4.40625,
+      "learning_rate": 9.891688152460388e-05,
+      "loss": 4.3343,
+      "step": 292900
+    },
+    {
+      "epoch": 2.521952479338843,
+      "grad_norm": 5.96875,
+      "learning_rate": 9.888979111650501e-05,
+      "loss": 4.1144,
+      "step": 292950
+    },
+    {
+      "epoch": 2.522382920110193,
+      "grad_norm": 4.4375,
+      "learning_rate": 9.886270078989309e-05,
+      "loss": 4.2403,
+      "step": 293000
+    },
+    {
+      "epoch": 2.5228133608815426,
+      "grad_norm": 2.140625,
+      "learning_rate": 9.883561054675647e-05,
+      "loss": 4.4512,
+      "step": 293050
+    },
+    {
+      "epoch": 2.5232438016528924,
+      "grad_norm": 2.203125,
+      "learning_rate": 9.880852038908356e-05,
+      "loss": 4.4038,
+      "step": 293100
+    },
+    {
+      "epoch": 2.523674242424242,
+      "grad_norm": 2.390625,
+      "learning_rate": 9.87814303188627e-05,
+      "loss": 4.5217,
+      "step": 293150
+    },
+    {
+      "epoch": 2.5241046831955924,
+      "grad_norm": 1.8359375,
+      "learning_rate": 9.875434033808219e-05,
+      "loss": 4.3784,
+      "step": 293200
+    },
+    {
+      "epoch": 2.5245351239669422,
+      "grad_norm": 4.1875,
+      "learning_rate": 9.872725044873042e-05,
+      "loss": 4.6214,
+      "step": 293250
+    },
+    {
+      "epoch": 2.524965564738292,
+      "grad_norm": 3.046875,
+      "learning_rate": 9.870016065279575e-05,
+      "loss": 4.4362,
+      "step": 293300
+    },
+    {
+      "epoch": 2.525396005509642,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.867307095226648e-05,
+      "loss": 4.4533,
+      "step": 293350
+    },
+    {
+      "epoch": 2.5258264462809916,
+      "grad_norm": 4.34375,
+      "learning_rate": 9.864598134913093e-05,
+      "loss": 4.121,
+      "step": 293400
+    },
+    {
+      "epoch": 2.526256887052342,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.86188918453775e-05,
+      "loss": 4.3346,
+      "step": 293450
+    },
+    {
+      "epoch": 2.5266873278236917,
+      "grad_norm": 2.28125,
+      "learning_rate": 9.85918024429944e-05,
+      "loss": 4.4135,
+      "step": 293500
+    },
+    {
+      "epoch": 2.5271177685950414,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.856471314396994e-05,
+      "loss": 4.0899,
+      "step": 293550
+    },
+    {
+      "epoch": 2.5275482093663912,
+      "grad_norm": 4.75,
+      "learning_rate": 9.853762395029247e-05,
+      "loss": 4.2385,
+      "step": 293600
+    },
+    {
+      "epoch": 2.527978650137741,
+      "grad_norm": 4.0625,
+      "learning_rate": 9.851053486395027e-05,
+      "loss": 4.1959,
+      "step": 293650
+    },
+    {
+      "epoch": 2.528409090909091,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.848344588693161e-05,
+      "loss": 4.6997,
+      "step": 293700
+    },
+    {
+      "epoch": 2.5288395316804406,
+      "grad_norm": 1.9921875,
+      "learning_rate": 9.845635702122474e-05,
+      "loss": 4.1244,
+      "step": 293750
+    },
+    {
+      "epoch": 2.5292699724517904,
+      "grad_norm": 3.828125,
+      "learning_rate": 9.842926826881796e-05,
+      "loss": 4.0847,
+      "step": 293800
+    },
+    {
+      "epoch": 2.5297004132231407,
+      "grad_norm": 4.625,
+      "learning_rate": 9.840217963169947e-05,
+      "loss": 4.4527,
+      "step": 293850
+    },
+    {
+      "epoch": 2.5301308539944904,
+      "grad_norm": 2.28125,
+      "learning_rate": 9.837509111185758e-05,
+      "loss": 4.1335,
+      "step": 293900
+    },
+    {
+      "epoch": 2.5305612947658402,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.834800271128054e-05,
+      "loss": 4.1885,
+      "step": 293950
+    },
+    {
+      "epoch": 2.53099173553719,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.832091443195653e-05,
+      "loss": 4.0911,
+      "step": 294000
+    },
+    {
+      "epoch": 2.53099173553719,
+      "eval_loss": 5.008902072906494,
+      "eval_runtime": 24.6594,
+      "eval_samples_per_second": 25.954,
+      "eval_steps_per_second": 12.977,
+      "eval_tts_loss": 7.441221903996289,
+      "step": 294000
+    },
+    {
+      "epoch": 2.53142217630854,
+      "grad_norm": 3.5,
+      "learning_rate": 9.829382627587376e-05,
+      "loss": 4.6011,
+      "step": 294050
+    },
+    {
+      "epoch": 2.53185261707989,
+      "grad_norm": 1.9375,
+      "learning_rate": 9.826673824502046e-05,
+      "loss": 4.3629,
+      "step": 294100
+    },
+    {
+      "epoch": 2.53228305785124,
+      "grad_norm": 3.984375,
+      "learning_rate": 9.823965034138488e-05,
+      "loss": 4.542,
+      "step": 294150
+    },
+    {
+      "epoch": 2.5327134986225897,
+      "grad_norm": 2.359375,
+      "learning_rate": 9.82125625669552e-05,
+      "loss": 4.335,
+      "step": 294200
+    },
+    {
+      "epoch": 2.5331439393939394,
+      "grad_norm": 1.359375,
+      "learning_rate": 9.818547492371953e-05,
+      "loss": 4.2868,
+      "step": 294250
+    },
+    {
+      "epoch": 2.5335743801652892,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.815838741366611e-05,
+      "loss": 4.1282,
+      "step": 294300
+    },
+    {
+      "epoch": 2.534004820936639,
+      "grad_norm": 0.73828125,
+      "learning_rate": 9.81313000387831e-05,
+      "loss": 4.3212,
+      "step": 294350
+    },
+    {
+      "epoch": 2.534435261707989,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.810421280105863e-05,
+      "loss": 4.2713,
+      "step": 294400
+    },
+    {
+      "epoch": 2.5348657024793386,
+      "grad_norm": 2.53125,
+      "learning_rate": 9.80771257024809e-05,
+      "loss": 4.413,
+      "step": 294450
+    },
+    {
+      "epoch": 2.535296143250689,
+      "grad_norm": 4.71875,
+      "learning_rate": 9.805003874503796e-05,
+      "loss": 4.48,
+      "step": 294500
+    },
+    {
+      "epoch": 2.5357265840220387,
+      "grad_norm": 2.75,
+      "learning_rate": 9.802295193071803e-05,
+      "loss": 4.4144,
+      "step": 294550
+    },
+    {
+      "epoch": 2.5361570247933884,
+      "grad_norm": 2.390625,
+      "learning_rate": 9.799586526150913e-05,
+      "loss": 4.4056,
+      "step": 294600
+    },
+    {
+      "epoch": 2.5365874655647382,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.796877873939943e-05,
+      "loss": 4.5464,
+      "step": 294650
+    },
+    {
+      "epoch": 2.537017906336088,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.794169236637702e-05,
+      "loss": 4.303,
+      "step": 294700
+    },
+    {
+      "epoch": 2.5374483471074383,
+      "grad_norm": 2.15625,
+      "learning_rate": 9.791460614442992e-05,
+      "loss": 4.0937,
+      "step": 294750
+    },
+    {
+      "epoch": 2.537878787878788,
+      "grad_norm": 2.5,
+      "learning_rate": 9.788752007554625e-05,
+      "loss": 4.4437,
+      "step": 294800
+    },
+    {
+      "epoch": 2.538309228650138,
+      "grad_norm": 4.0625,
+      "learning_rate": 9.786043416171407e-05,
+      "loss": 4.3343,
+      "step": 294850
+    },
+    {
+      "epoch": 2.5387396694214877,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.78333484049214e-05,
+      "loss": 5.0317,
+      "step": 294900
+    },
+    {
+      "epoch": 2.5391701101928374,
+      "grad_norm": 2.875,
+      "learning_rate": 9.780626280715635e-05,
+      "loss": 4.1116,
+      "step": 294950
+    },
+    {
+      "epoch": 2.5396005509641872,
+      "grad_norm": 4.65625,
+      "learning_rate": 9.777917737040684e-05,
+      "loss": 4.694,
+      "step": 295000
+    },
+    {
+      "epoch": 2.540030991735537,
+      "grad_norm": 2.453125,
+      "learning_rate": 9.775209209666095e-05,
+      "loss": 4.398,
+      "step": 295050
+    },
+    {
+      "epoch": 2.540461432506887,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.772500698790665e-05,
+      "loss": 4.5284,
+      "step": 295100
+    },
+    {
+      "epoch": 2.540891873278237,
+      "grad_norm": 5.0,
+      "learning_rate": 9.769792204613196e-05,
+      "loss": 4.2916,
+      "step": 295150
+    },
+    {
+      "epoch": 2.541322314049587,
+      "grad_norm": 0.984375,
+      "learning_rate": 9.767083727332483e-05,
+      "loss": 4.2949,
+      "step": 295200
+    },
+    {
+      "epoch": 2.5417527548209367,
+      "grad_norm": 3.0,
+      "learning_rate": 9.764375267147326e-05,
+      "loss": 4.4463,
+      "step": 295250
+    },
+    {
+      "epoch": 2.5421831955922864,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.761666824256512e-05,
+      "loss": 4.1976,
+      "step": 295300
+    },
+    {
+      "epoch": 2.5426136363636362,
+      "grad_norm": 4.59375,
+      "learning_rate": 9.758958398858845e-05,
+      "loss": 4.4993,
+      "step": 295350
+    },
+    {
+      "epoch": 2.5430440771349865,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.756249991153109e-05,
+      "loss": 4.1564,
+      "step": 295400
+    },
+    {
+      "epoch": 2.5434745179063363,
+      "grad_norm": 2.28125,
+      "learning_rate": 9.7535416013381e-05,
+      "loss": 4.7065,
+      "step": 295450
+    },
+    {
+      "epoch": 2.543904958677686,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.750833229612611e-05,
+      "loss": 4.4891,
+      "step": 295500
+    },
+    {
+      "epoch": 2.544335399449036,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.748124876175426e-05,
+      "loss": 4.3254,
+      "step": 295550
+    },
+    {
+      "epoch": 2.5447658402203857,
+      "grad_norm": 1.6015625,
+      "learning_rate": 9.745416541225329e-05,
+      "loss": 4.2103,
+      "step": 295600
+    },
+    {
+      "epoch": 2.5451962809917354,
+      "grad_norm": 6.65625,
+      "learning_rate": 9.742708224961112e-05,
+      "loss": 4.2909,
+      "step": 295650
+    },
+    {
+      "epoch": 2.5456267217630852,
+      "grad_norm": 2.375,
+      "learning_rate": 9.739999927581558e-05,
+      "loss": 4.0355,
+      "step": 295700
+    },
+    {
+      "epoch": 2.546057162534435,
+      "grad_norm": 2.5,
+      "learning_rate": 9.737291649285455e-05,
+      "loss": 4.1197,
+      "step": 295750
+    },
+    {
+      "epoch": 2.5464876033057853,
+      "grad_norm": 4.0625,
+      "learning_rate": 9.734583390271573e-05,
+      "loss": 4.5149,
+      "step": 295800
+    },
+    {
+      "epoch": 2.546918044077135,
+      "grad_norm": 3.6875,
+      "learning_rate": 9.731875150738699e-05,
+      "loss": 4.6415,
+      "step": 295850
+    },
+    {
+      "epoch": 2.547348484848485,
+      "grad_norm": 4.3125,
+      "learning_rate": 9.729166930885616e-05,
+      "loss": 4.3265,
+      "step": 295900
+    },
+    {
+      "epoch": 2.5477789256198347,
+      "grad_norm": 3.78125,
+      "learning_rate": 9.726458730911093e-05,
+      "loss": 4.6339,
+      "step": 295950
+    },
+    {
+      "epoch": 2.5482093663911844,
+      "grad_norm": 3.703125,
+      "learning_rate": 9.723750551013916e-05,
+      "loss": 4.1391,
+      "step": 296000
+    },
+    {
+      "epoch": 2.5486398071625347,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.721042391392852e-05,
+      "loss": 4.2668,
+      "step": 296050
+    },
+    {
+      "epoch": 2.5490702479338845,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.718334252246676e-05,
+      "loss": 4.5348,
+      "step": 296100
+    },
+    {
+      "epoch": 2.5495006887052343,
+      "grad_norm": 4.875,
+      "learning_rate": 9.715626133774158e-05,
+      "loss": 4.3282,
+      "step": 296150
+    },
+    {
+      "epoch": 2.549931129476584,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.712918036174072e-05,
+      "loss": 4.2539,
+      "step": 296200
+    },
+    {
+      "epoch": 2.550361570247934,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.710209959645187e-05,
+      "loss": 4.851,
+      "step": 296250
+    },
+    {
+      "epoch": 2.5507920110192837,
+      "grad_norm": 2.234375,
+      "learning_rate": 9.707501904386264e-05,
+      "loss": 4.5381,
+      "step": 296300
+    },
+    {
+      "epoch": 2.5512224517906334,
+      "grad_norm": 2.171875,
+      "learning_rate": 9.704793870596071e-05,
+      "loss": 4.3863,
+      "step": 296350
+    },
+    {
+      "epoch": 2.5516528925619832,
+      "grad_norm": 3.546875,
+      "learning_rate": 9.702085858473375e-05,
+      "loss": 4.4474,
+      "step": 296400
+    },
+    {
+      "epoch": 2.5520833333333335,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.699377868216932e-05,
+      "loss": 4.3004,
+      "step": 296450
+    },
+    {
+      "epoch": 2.5525137741046833,
+      "grad_norm": 2.375,
+      "learning_rate": 9.696669900025513e-05,
+      "loss": 4.4782,
+      "step": 296500
+    },
+    {
+      "epoch": 2.552944214876033,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.693961954097866e-05,
+      "loss": 4.3658,
+      "step": 296550
+    },
+    {
+      "epoch": 2.553374655647383,
+      "grad_norm": 4.0,
+      "learning_rate": 9.691254030632756e-05,
+      "loss": 4.2612,
+      "step": 296600
+    },
+    {
+      "epoch": 2.5538050964187327,
+      "grad_norm": 5.59375,
+      "learning_rate": 9.688546129828933e-05,
+      "loss": 4.4855,
+      "step": 296650
+    },
+    {
+      "epoch": 2.554235537190083,
+      "grad_norm": 2.578125,
+      "learning_rate": 9.685838251885157e-05,
+      "loss": 4.4261,
+      "step": 296700
+    },
+    {
+      "epoch": 2.5546659779614327,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.683130397000178e-05,
+      "loss": 4.4219,
+      "step": 296750
+    },
+    {
+      "epoch": 2.5550964187327825,
+      "grad_norm": 3.328125,
+      "learning_rate": 9.680422565372744e-05,
+      "loss": 4.3975,
+      "step": 296800
+    },
+    {
+      "epoch": 2.5555268595041323,
+      "grad_norm": 2.65625,
+      "learning_rate": 9.677714757201603e-05,
+      "loss": 4.1262,
+      "step": 296850
+    },
+    {
+      "epoch": 2.555957300275482,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.67500697268551e-05,
+      "loss": 4.4234,
+      "step": 296900
+    },
+    {
+      "epoch": 2.556387741046832,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.672299212023202e-05,
+      "loss": 4.505,
+      "step": 296950
+    },
+    {
+      "epoch": 2.5568181818181817,
+      "grad_norm": 2.359375,
+      "learning_rate": 9.669591475413429e-05,
+      "loss": 4.6743,
+      "step": 297000
+    },
+    {
+      "epoch": 2.5568181818181817,
+      "eval_loss": 5.007340431213379,
+      "eval_runtime": 24.2976,
+      "eval_samples_per_second": 26.34,
+      "eval_steps_per_second": 13.17,
+      "eval_tts_loss": 7.371025063424796,
+      "step": 297000
+    },
+    {
+      "epoch": 2.5572486225895315,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.666883763054933e-05,
+      "loss": 4.1647,
+      "step": 297050
+    },
+    {
+      "epoch": 2.5576790633608817,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.664176075146452e-05,
+      "loss": 4.2352,
+      "step": 297100
+    },
+    {
+      "epoch": 2.5581095041322315,
+      "grad_norm": 3.671875,
+      "learning_rate": 9.661468411886721e-05,
+      "loss": 4.6513,
+      "step": 297150
+    },
+    {
+      "epoch": 2.5585399449035813,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.65876077347448e-05,
+      "loss": 4.4155,
+      "step": 297200
+    },
+    {
+      "epoch": 2.558970385674931,
+      "grad_norm": 2.0,
+      "learning_rate": 9.656053160108466e-05,
+      "loss": 4.3145,
+      "step": 297250
+    },
+    {
+      "epoch": 2.559400826446281,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.653345571987413e-05,
+      "loss": 4.5716,
+      "step": 297300
+    },
+    {
+      "epoch": 2.559831267217631,
+      "grad_norm": 2.3125,
+      "learning_rate": 9.650638009310044e-05,
+      "loss": 3.7966,
+      "step": 297350
+    },
+    {
+      "epoch": 2.560261707988981,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.647930472275097e-05,
+      "loss": 4.9684,
+      "step": 297400
+    },
+    {
+      "epoch": 2.5606921487603307,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.645222961081292e-05,
+      "loss": 4.0982,
+      "step": 297450
+    },
+    {
+      "epoch": 2.5611225895316805,
+      "grad_norm": 4.0625,
+      "learning_rate": 9.64251547592736e-05,
+      "loss": 4.3799,
+      "step": 297500
+    },
+    {
+      "epoch": 2.5615530303030303,
+      "grad_norm": 3.765625,
+      "learning_rate": 9.639808017012027e-05,
+      "loss": 4.2548,
+      "step": 297550
+    },
+    {
+      "epoch": 2.56198347107438,
+      "grad_norm": 3.5625,
+      "learning_rate": 9.637100584534008e-05,
+      "loss": 4.3646,
+      "step": 297600
+    },
+    {
+      "epoch": 2.56241391184573,
+      "grad_norm": 3.328125,
+      "learning_rate": 9.634393178692023e-05,
+      "loss": 4.1925,
+      "step": 297650
+    },
+    {
+      "epoch": 2.5628443526170797,
+      "grad_norm": 2.453125,
+      "learning_rate": 9.631685799684792e-05,
+      "loss": 4.6156,
+      "step": 297700
+    },
+    {
+      "epoch": 2.56327479338843,
+      "grad_norm": 3.25,
+      "learning_rate": 9.628978447711032e-05,
+      "loss": 4.0818,
+      "step": 297750
+    },
+    {
+      "epoch": 2.5637052341597797,
+      "grad_norm": 2.53125,
+      "learning_rate": 9.62627112296946e-05,
+      "loss": 4.1033,
+      "step": 297800
+    },
+    {
+      "epoch": 2.5641356749311295,
+      "grad_norm": 1.40625,
+      "learning_rate": 9.623563825658778e-05,
+      "loss": 4.2805,
+      "step": 297850
+    },
+    {
+      "epoch": 2.5645661157024793,
+      "grad_norm": 1.6875,
+      "learning_rate": 9.620856555977701e-05,
+      "loss": 4.1623,
+      "step": 297900
+    },
+    {
+      "epoch": 2.564996556473829,
+      "grad_norm": 4.46875,
+      "learning_rate": 9.61814931412494e-05,
+      "loss": 4.4439,
+      "step": 297950
+    },
+    {
+      "epoch": 2.5654269972451793,
+      "grad_norm": 3.640625,
+      "learning_rate": 9.615442100299195e-05,
+      "loss": 4.342,
+      "step": 298000
+    },
+    {
+      "epoch": 2.565857438016529,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.612734914699178e-05,
+      "loss": 4.2367,
+      "step": 298050
+    },
+    {
+      "epoch": 2.566287878787879,
+      "grad_norm": 2.3125,
+      "learning_rate": 9.610027757523579e-05,
+      "loss": 4.1732,
+      "step": 298100
+    },
+    {
+      "epoch": 2.5667183195592287,
+      "grad_norm": 4.40625,
+      "learning_rate": 9.607320628971106e-05,
+      "loss": 4.0481,
+      "step": 298150
+    },
+    {
+      "epoch": 2.5671487603305785,
+      "grad_norm": 1.5625,
+      "learning_rate": 9.604613529240453e-05,
+      "loss": 4.5346,
+      "step": 298200
+    },
+    {
+      "epoch": 2.5675792011019283,
+      "grad_norm": 6.0,
+      "learning_rate": 9.601906458530316e-05,
+      "loss": 4.3702,
+      "step": 298250
+    },
+    {
+      "epoch": 2.568009641873278,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.599199417039393e-05,
+      "loss": 4.2452,
+      "step": 298300
+    },
+    {
+      "epoch": 2.568440082644628,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.596492404966365e-05,
+      "loss": 4.4884,
+      "step": 298350
+    },
+    {
+      "epoch": 2.568870523415978,
+      "grad_norm": 6.28125,
+      "learning_rate": 9.593785422509927e-05,
+      "loss": 4.8544,
+      "step": 298400
+    },
+    {
+      "epoch": 2.569300964187328,
+      "grad_norm": 2.34375,
+      "learning_rate": 9.591078469868767e-05,
+      "loss": 4.5118,
+      "step": 298450
+    },
+    {
+      "epoch": 2.5697314049586777,
+      "grad_norm": 3.859375,
+      "learning_rate": 9.588371547241565e-05,
+      "loss": 4.3172,
+      "step": 298500
+    },
+    {
+      "epoch": 2.5701618457300275,
+      "grad_norm": 1.2734375,
+      "learning_rate": 9.585664654827011e-05,
+      "loss": 4.4204,
+      "step": 298550
+    },
+    {
+      "epoch": 2.5705922865013773,
+      "grad_norm": 2.390625,
+      "learning_rate": 9.582957792823773e-05,
+      "loss": 4.5699,
+      "step": 298600
+    },
+    {
+      "epoch": 2.5710227272727275,
+      "grad_norm": 5.53125,
+      "learning_rate": 9.58025096143054e-05,
+      "loss": 4.1956,
+      "step": 298650
+    },
+    {
+      "epoch": 2.5714531680440773,
+      "grad_norm": 1.6796875,
+      "learning_rate": 9.577544160845981e-05,
+      "loss": 4.45,
+      "step": 298700
+    },
+    {
+      "epoch": 2.571883608815427,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.574837391268774e-05,
+      "loss": 4.5122,
+      "step": 298750
+    },
+    {
+      "epoch": 2.572314049586777,
+      "grad_norm": 1.3515625,
+      "learning_rate": 9.572130652897585e-05,
+      "loss": 4.0651,
+      "step": 298800
+    },
+    {
+      "epoch": 2.5727444903581267,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.56942394593109e-05,
+      "loss": 4.397,
+      "step": 298850
+    },
+    {
+      "epoch": 2.5731749311294765,
+      "grad_norm": 4.96875,
+      "learning_rate": 9.566717270567946e-05,
+      "loss": 4.3147,
+      "step": 298900
+    },
+    {
+      "epoch": 2.5736053719008263,
+      "grad_norm": 2.625,
+      "learning_rate": 9.564010627006825e-05,
+      "loss": 4.6492,
+      "step": 298950
+    },
+    {
+      "epoch": 2.574035812672176,
+      "grad_norm": 4.21875,
+      "learning_rate": 9.561304015446384e-05,
+      "loss": 4.2182,
+      "step": 299000
+    },
+    {
+      "epoch": 2.5744662534435263,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.558597436085285e-05,
+      "loss": 4.0775,
+      "step": 299050
+    },
+    {
+      "epoch": 2.574896694214876,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.555890889122188e-05,
+      "loss": 4.2896,
+      "step": 299100
+    },
+    {
+      "epoch": 2.575327134986226,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.553184374755741e-05,
+      "loss": 4.6974,
+      "step": 299150
+    },
+    {
+      "epoch": 2.5757575757575757,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.550477893184598e-05,
+      "loss": 4.0553,
+      "step": 299200
+    },
+    {
+      "epoch": 2.5761880165289255,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.547771444607411e-05,
+      "loss": 4.1128,
+      "step": 299250
+    },
+    {
+      "epoch": 2.5766184573002757,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.54506502922283e-05,
+      "loss": 4.7973,
+      "step": 299300
+    },
+    {
+      "epoch": 2.5770488980716255,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.542358647229497e-05,
+      "loss": 4.7141,
+      "step": 299350
+    },
+    {
+      "epoch": 2.5774793388429753,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.539652298826052e-05,
+      "loss": 4.482,
+      "step": 299400
+    },
+    {
+      "epoch": 2.577909779614325,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.536945984211136e-05,
+      "loss": 4.3513,
+      "step": 299450
+    },
+    {
+      "epoch": 2.578340220385675,
+      "grad_norm": 3.6875,
+      "learning_rate": 9.534239703583393e-05,
+      "loss": 4.7105,
+      "step": 299500
+    },
+    {
+      "epoch": 2.5787706611570247,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.531533457141451e-05,
+      "loss": 4.2576,
+      "step": 299550
+    },
+    {
+      "epoch": 2.5792011019283745,
+      "grad_norm": 3.0,
+      "learning_rate": 9.528827245083948e-05,
+      "loss": 5.111,
+      "step": 299600
+    },
+    {
+      "epoch": 2.5796315426997243,
+      "grad_norm": 4.3125,
+      "learning_rate": 9.526121067609511e-05,
+      "loss": 4.373,
+      "step": 299650
+    },
+    {
+      "epoch": 2.5800619834710745,
+      "grad_norm": 0.640625,
+      "learning_rate": 9.523414924916765e-05,
+      "loss": 4.4143,
+      "step": 299700
+    },
+    {
+      "epoch": 2.5804924242424243,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.520708817204338e-05,
+      "loss": 4.4131,
+      "step": 299750
+    },
+    {
+      "epoch": 2.580922865013774,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.518002744670855e-05,
+      "loss": 4.2544,
+      "step": 299800
+    },
+    {
+      "epoch": 2.581353305785124,
+      "grad_norm": 3.015625,
+      "learning_rate": 9.515296707514936e-05,
+      "loss": 4.5132,
+      "step": 299850
+    },
+    {
+      "epoch": 2.5817837465564737,
+      "grad_norm": 2.515625,
+      "learning_rate": 9.51259070593519e-05,
+      "loss": 4.3841,
+      "step": 299900
+    },
+    {
+      "epoch": 2.582214187327824,
+      "grad_norm": 5.25,
+      "learning_rate": 9.50988474013024e-05,
+      "loss": 4.3082,
+      "step": 299950
+    },
+    {
+      "epoch": 2.5826446280991737,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.507178810298696e-05,
+      "loss": 4.4722,
+      "step": 300000
+    },
+    {
+      "epoch": 2.5826446280991737,
+      "eval_loss": 5.005288124084473,
+      "eval_runtime": 24.7806,
+      "eval_samples_per_second": 25.827,
+      "eval_steps_per_second": 12.913,
+      "eval_tts_loss": 7.403621951830839,
+      "step": 300000
+    },
+    {
+      "epoch": 2.5830750688705235,
+      "grad_norm": 2.546875,
+      "learning_rate": 9.504472916639166e-05,
+      "loss": 4.026,
+      "step": 300050
+    },
+    {
+      "epoch": 2.5835055096418733,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.50176705935026e-05,
+      "loss": 4.4608,
+      "step": 300100
+    },
+    {
+      "epoch": 2.583935950413223,
+      "grad_norm": 2.65625,
+      "learning_rate": 9.499061238630578e-05,
+      "loss": 4.0954,
+      "step": 300150
+    },
+    {
+      "epoch": 2.584366391184573,
+      "grad_norm": 2.109375,
+      "learning_rate": 9.496355454678724e-05,
+      "loss": 4.4024,
+      "step": 300200
+    },
+    {
+      "epoch": 2.5847968319559227,
+      "grad_norm": 1.8046875,
+      "learning_rate": 9.493649707693295e-05,
+      "loss": 4.1422,
+      "step": 300250
+    },
+    {
+      "epoch": 2.5852272727272725,
+      "grad_norm": 1.7734375,
+      "learning_rate": 9.490943997872887e-05,
+      "loss": 4.2486,
+      "step": 300300
+    },
+    {
+      "epoch": 2.5856577134986227,
+      "grad_norm": 4.03125,
+      "learning_rate": 9.4882383254161e-05,
+      "loss": 4.6009,
+      "step": 300350
+    },
+    {
+      "epoch": 2.5860881542699725,
+      "grad_norm": 0.9921875,
+      "learning_rate": 9.485532690521514e-05,
+      "loss": 4.3778,
+      "step": 300400
+    },
+    {
+      "epoch": 2.5865185950413223,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.48282709338772e-05,
+      "loss": 4.1779,
+      "step": 300450
+    },
+    {
+      "epoch": 2.586949035812672,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.480121534213307e-05,
+      "loss": 4.3436,
+      "step": 300500
+    },
+    {
+      "epoch": 2.587379476584022,
+      "grad_norm": 1.7890625,
+      "learning_rate": 9.477416013196854e-05,
+      "loss": 4.2219,
+      "step": 300550
+    },
+    {
+      "epoch": 2.587809917355372,
+      "grad_norm": 3.71875,
+      "learning_rate": 9.474710530536941e-05,
+      "loss": 4.4922,
+      "step": 300600
+    },
+    {
+      "epoch": 2.588240358126722,
+      "grad_norm": 4.34375,
+      "learning_rate": 9.472005086432149e-05,
+      "loss": 4.5593,
+      "step": 300650
+    },
+    {
+      "epoch": 2.5886707988980717,
+      "grad_norm": 4.28125,
+      "learning_rate": 9.469299681081046e-05,
+      "loss": 4.2956,
+      "step": 300700
+    },
+    {
+      "epoch": 2.5891012396694215,
+      "grad_norm": 5.96875,
+      "learning_rate": 9.466594314682201e-05,
+      "loss": 4.4582,
+      "step": 300750
+    },
+    {
+      "epoch": 2.5895316804407713,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.463888987434187e-05,
+      "loss": 4.2782,
+      "step": 300800
+    },
+    {
+      "epoch": 2.589962121212121,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.461183699535568e-05,
+      "loss": 4.3627,
+      "step": 300850
+    },
+    {
+      "epoch": 2.590392561983471,
+      "grad_norm": 6.59375,
+      "learning_rate": 9.458478451184911e-05,
+      "loss": 4.5938,
+      "step": 300900
+    },
+    {
+      "epoch": 2.5908230027548207,
+      "grad_norm": 1.7734375,
+      "learning_rate": 9.455773242580765e-05,
+      "loss": 4.3129,
+      "step": 300950
+    },
+    {
+      "epoch": 2.591253443526171,
+      "grad_norm": 1.7890625,
+      "learning_rate": 9.453068073921694e-05,
+      "loss": 4.263,
+      "step": 301000
+    },
+    {
+      "epoch": 2.5916838842975207,
+      "grad_norm": 3.375,
+      "learning_rate": 9.450362945406247e-05,
+      "loss": 4.3501,
+      "step": 301050
+    },
+    {
+      "epoch": 2.5921143250688705,
+      "grad_norm": 4.75,
+      "learning_rate": 9.447657857232977e-05,
+      "loss": 4.1093,
+      "step": 301100
+    },
+    {
+      "epoch": 2.5925447658402203,
+      "grad_norm": 5.125,
+      "learning_rate": 9.444952809600438e-05,
+      "loss": 4.3862,
+      "step": 301150
+    },
+    {
+      "epoch": 2.59297520661157,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.442247802707166e-05,
+      "loss": 4.3417,
+      "step": 301200
+    },
+    {
+      "epoch": 2.5934056473829203,
+      "grad_norm": 2.4375,
+      "learning_rate": 9.439542836751703e-05,
+      "loss": 4.2905,
+      "step": 301250
+    },
+    {
+      "epoch": 2.59383608815427,
+      "grad_norm": 4.34375,
+      "learning_rate": 9.436837911932589e-05,
+      "loss": 3.9661,
+      "step": 301300
+    },
+    {
+      "epoch": 2.59426652892562,
+      "grad_norm": 2.46875,
+      "learning_rate": 9.434133028448362e-05,
+      "loss": 4.647,
+      "step": 301350
+    },
+    {
+      "epoch": 2.5946969696969697,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.431428186497558e-05,
+      "loss": 4.1314,
+      "step": 301400
+    },
+    {
+      "epoch": 2.5951274104683195,
+      "grad_norm": 1.9609375,
+      "learning_rate": 9.428723386278695e-05,
+      "loss": 4.648,
+      "step": 301450
+    },
+    {
+      "epoch": 2.5955578512396693,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.426018627990308e-05,
+      "loss": 4.2566,
+      "step": 301500
+    },
+    {
+      "epoch": 2.595988292011019,
+      "grad_norm": 5.5625,
+      "learning_rate": 9.42331391183092e-05,
+      "loss": 4.388,
+      "step": 301550
+    },
+    {
+      "epoch": 2.596418732782369,
+      "grad_norm": 4.40625,
+      "learning_rate": 9.420609237999048e-05,
+      "loss": 4.2848,
+      "step": 301600
+    },
+    {
+      "epoch": 2.596849173553719,
+      "grad_norm": 4.75,
+      "learning_rate": 9.417904606693216e-05,
+      "loss": 4.2478,
+      "step": 301650
+    },
+    {
+      "epoch": 2.597279614325069,
+      "grad_norm": 4.375,
+      "learning_rate": 9.41520001811193e-05,
+      "loss": 4.457,
+      "step": 301700
+    },
+    {
+      "epoch": 2.5977100550964187,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.412495472453705e-05,
+      "loss": 4.2519,
+      "step": 301750
+    },
+    {
+      "epoch": 2.5981404958677685,
+      "grad_norm": 5.8125,
+      "learning_rate": 9.409790969917048e-05,
+      "loss": 4.4381,
+      "step": 301800
+    },
+    {
+      "epoch": 2.5985709366391183,
+      "grad_norm": 4.1875,
+      "learning_rate": 9.407086510700464e-05,
+      "loss": 4.5783,
+      "step": 301850
+    },
+    {
+      "epoch": 2.5990013774104685,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.404382095002457e-05,
+      "loss": 4.3697,
+      "step": 301900
+    },
+    {
+      "epoch": 2.5994318181818183,
+      "grad_norm": 3.453125,
+      "learning_rate": 9.40167772302152e-05,
+      "loss": 4.6052,
+      "step": 301950
+    },
+    {
+      "epoch": 2.599862258953168,
+      "grad_norm": 3.125,
+      "learning_rate": 9.398973394956148e-05,
+      "loss": 4.3887,
+      "step": 302000
+    },
+    {
+      "epoch": 2.600292699724518,
+      "grad_norm": 3.875,
+      "learning_rate": 9.39626911100484e-05,
+      "loss": 4.4233,
+      "step": 302050
+    },
+    {
+      "epoch": 2.6007231404958677,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.393564871366078e-05,
+      "loss": 4.5385,
+      "step": 302100
+    },
+    {
+      "epoch": 2.6011535812672175,
+      "grad_norm": 2.21875,
+      "learning_rate": 9.390860676238354e-05,
+      "loss": 3.8624,
+      "step": 302150
+    },
+    {
+      "epoch": 2.6015840220385673,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.388156525820142e-05,
+      "loss": 4.312,
+      "step": 302200
+    },
+    {
+      "epoch": 2.602014462809917,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.385452420309925e-05,
+      "loss": 4.5539,
+      "step": 302250
+    },
+    {
+      "epoch": 2.6024449035812673,
+      "grad_norm": 1.3515625,
+      "learning_rate": 9.382748359906176e-05,
+      "loss": 4.2204,
+      "step": 302300
+    },
+    {
+      "epoch": 2.602875344352617,
+      "grad_norm": 6.21875,
+      "learning_rate": 9.38004434480737e-05,
+      "loss": 4.3204,
+      "step": 302350
+    },
+    {
+      "epoch": 2.603305785123967,
+      "grad_norm": 1.21875,
+      "learning_rate": 9.37734037521198e-05,
+      "loss": 4.338,
+      "step": 302400
+    },
+    {
+      "epoch": 2.6037362258953167,
+      "grad_norm": 3.671875,
+      "learning_rate": 9.374636451318463e-05,
+      "loss": 4.3437,
+      "step": 302450
+    },
+    {
+      "epoch": 2.6041666666666665,
+      "grad_norm": 5.09375,
+      "learning_rate": 9.371932573325285e-05,
+      "loss": 4.466,
+      "step": 302500
+    },
+    {
+      "epoch": 2.6045971074380168,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.369228741430906e-05,
+      "loss": 4.1902,
+      "step": 302550
+    },
+    {
+      "epoch": 2.6050275482093666,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.366524955833779e-05,
+      "loss": 4.3701,
+      "step": 302600
+    },
+    {
+      "epoch": 2.6054579889807163,
+      "grad_norm": 4.6875,
+      "learning_rate": 9.363821216732356e-05,
+      "loss": 4.3691,
+      "step": 302650
+    },
+    {
+      "epoch": 2.605888429752066,
+      "grad_norm": 1.7421875,
+      "learning_rate": 9.361117524325091e-05,
+      "loss": 4.4504,
+      "step": 302700
+    },
+    {
+      "epoch": 2.606318870523416,
+      "grad_norm": 2.75,
+      "learning_rate": 9.358413878810424e-05,
+      "loss": 4.3391,
+      "step": 302750
+    },
+    {
+      "epoch": 2.6067493112947657,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.355710280386796e-05,
+      "loss": 4.5579,
+      "step": 302800
+    },
+    {
+      "epoch": 2.6071797520661155,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.353006729252646e-05,
+      "loss": 4.1653,
+      "step": 302850
+    },
+    {
+      "epoch": 2.6076101928374653,
+      "grad_norm": 7.125,
+      "learning_rate": 9.350303225606413e-05,
+      "loss": 4.2967,
+      "step": 302900
+    },
+    {
+      "epoch": 2.6080406336088156,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.347599769646526e-05,
+      "loss": 4.4816,
+      "step": 302950
+    },
+    {
+      "epoch": 2.6084710743801653,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.344896361571411e-05,
+      "loss": 4.0751,
+      "step": 303000
+    },
+    {
+      "epoch": 2.6084710743801653,
+      "eval_loss": 5.003540515899658,
+      "eval_runtime": 24.2401,
+      "eval_samples_per_second": 26.403,
+      "eval_steps_per_second": 13.201,
+      "eval_tts_loss": 7.406452009501407,
+      "step": 303000
+    },
+    {
+      "epoch": 2.608901515151515,
+      "grad_norm": 2.515625,
+      "learning_rate": 9.342193001579491e-05,
+      "loss": 3.9798,
+      "step": 303050
+    },
+    {
+      "epoch": 2.609331955922865,
+      "grad_norm": 2.453125,
+      "learning_rate": 9.339489689869192e-05,
+      "loss": 4.2675,
+      "step": 303100
+    },
+    {
+      "epoch": 2.6097623966942147,
+      "grad_norm": 3.78125,
+      "learning_rate": 9.336786426638927e-05,
+      "loss": 4.3256,
+      "step": 303150
+    },
+    {
+      "epoch": 2.610192837465565,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.334083212087115e-05,
+      "loss": 4.5578,
+      "step": 303200
+    },
+    {
+      "epoch": 2.6106232782369148,
+      "grad_norm": 1.9375,
+      "learning_rate": 9.331380046412161e-05,
+      "loss": 4.2694,
+      "step": 303250
+    },
+    {
+      "epoch": 2.6110537190082646,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.32867692981247e-05,
+      "loss": 3.8954,
+      "step": 303300
+    },
+    {
+      "epoch": 2.6114841597796143,
+      "grad_norm": 4.0625,
+      "learning_rate": 9.325973862486449e-05,
+      "loss": 4.3279,
+      "step": 303350
+    },
+    {
+      "epoch": 2.611914600550964,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.323270844632498e-05,
+      "loss": 4.3109,
+      "step": 303400
+    },
+    {
+      "epoch": 2.612345041322314,
+      "grad_norm": 1.46875,
+      "learning_rate": 9.320567876449013e-05,
+      "loss": 4.4989,
+      "step": 303450
+    },
+    {
+      "epoch": 2.6127754820936637,
+      "grad_norm": 4.5625,
+      "learning_rate": 9.31786495813438e-05,
+      "loss": 4.5941,
+      "step": 303500
+    },
+    {
+      "epoch": 2.6132059228650135,
+      "grad_norm": 1.78125,
+      "learning_rate": 9.31516208988699e-05,
+      "loss": 4.352,
+      "step": 303550
+    },
+    {
+      "epoch": 2.6136363636363638,
+      "grad_norm": 4.6875,
+      "learning_rate": 9.312459271905234e-05,
+      "loss": 4.4323,
+      "step": 303600
+    },
+    {
+      "epoch": 2.6140668044077136,
+      "grad_norm": 1.5859375,
+      "learning_rate": 9.309756504387484e-05,
+      "loss": 4.2064,
+      "step": 303650
+    },
+    {
+      "epoch": 2.6144972451790633,
+      "grad_norm": 4.09375,
+      "learning_rate": 9.307053787532127e-05,
+      "loss": 4.3542,
+      "step": 303700
+    },
+    {
+      "epoch": 2.614927685950413,
+      "grad_norm": 2.875,
+      "learning_rate": 9.304351121537526e-05,
+      "loss": 4.3965,
+      "step": 303750
+    },
+    {
+      "epoch": 2.615358126721763,
+      "grad_norm": 4.15625,
+      "learning_rate": 9.301648506602057e-05,
+      "loss": 4.4847,
+      "step": 303800
+    },
+    {
+      "epoch": 2.615788567493113,
+      "grad_norm": 3.90625,
+      "learning_rate": 9.298945942924085e-05,
+      "loss": 3.9876,
+      "step": 303850
+    },
+    {
+      "epoch": 2.616219008264463,
+      "grad_norm": 2.796875,
+      "learning_rate": 9.296243430701975e-05,
+      "loss": 4.4824,
+      "step": 303900
+    },
+    {
+      "epoch": 2.6166494490358128,
+      "grad_norm": 4.875,
+      "learning_rate": 9.293540970134081e-05,
+      "loss": 4.5415,
+      "step": 303950
+    },
+    {
+      "epoch": 2.6170798898071626,
+      "grad_norm": 4.5625,
+      "learning_rate": 9.29083856141876e-05,
+      "loss": 4.3902,
+      "step": 304000
+    },
+    {
+      "epoch": 2.6175103305785123,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.288136204754362e-05,
+      "loss": 4.1748,
+      "step": 304050
+    },
+    {
+      "epoch": 2.617940771349862,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.285433900339235e-05,
+      "loss": 4.463,
+      "step": 304100
+    },
+    {
+      "epoch": 2.618371212121212,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.282731648371721e-05,
+      "loss": 4.2591,
+      "step": 304150
+    },
+    {
+      "epoch": 2.6188016528925617,
+      "grad_norm": 4.84375,
+      "learning_rate": 9.280029449050167e-05,
+      "loss": 4.427,
+      "step": 304200
+    },
+    {
+      "epoch": 2.619232093663912,
+      "grad_norm": 2.359375,
+      "learning_rate": 9.277327302572895e-05,
+      "loss": 4.069,
+      "step": 304250
+    },
+    {
+      "epoch": 2.6196625344352618,
+      "grad_norm": 2.546875,
+      "learning_rate": 9.274625209138249e-05,
+      "loss": 4.5566,
+      "step": 304300
+    },
+    {
+      "epoch": 2.6200929752066116,
+      "grad_norm": 0.765625,
+      "learning_rate": 9.271923168944548e-05,
+      "loss": 4.4964,
+      "step": 304350
+    },
+    {
+      "epoch": 2.6205234159779613,
+      "grad_norm": 5.65625,
+      "learning_rate": 9.269221182190122e-05,
+      "loss": 4.4107,
+      "step": 304400
+    },
+    {
+      "epoch": 2.620953856749311,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.266519249073289e-05,
+      "loss": 4.243,
+      "step": 304450
+    },
+    {
+      "epoch": 2.6213842975206614,
+      "grad_norm": 4.84375,
+      "learning_rate": 9.263817369792368e-05,
+      "loss": 4.6178,
+      "step": 304500
+    },
+    {
+      "epoch": 2.621814738292011,
+      "grad_norm": 3.625,
+      "learning_rate": 9.261115544545663e-05,
+      "loss": 4.0694,
+      "step": 304550
+    },
+    {
+      "epoch": 2.622245179063361,
+      "grad_norm": 1.2109375,
+      "learning_rate": 9.25841377353149e-05,
+      "loss": 4.5586,
+      "step": 304600
+    },
+    {
+      "epoch": 2.6226756198347108,
+      "grad_norm": 2.1875,
+      "learning_rate": 9.25571205694815e-05,
+      "loss": 4.3658,
+      "step": 304650
+    },
+    {
+      "epoch": 2.6231060606060606,
+      "grad_norm": 4.03125,
+      "learning_rate": 9.25301039499394e-05,
+      "loss": 4.7799,
+      "step": 304700
+    },
+    {
+      "epoch": 2.6235365013774103,
+      "grad_norm": 2.046875,
+      "learning_rate": 9.250308787867167e-05,
+      "loss": 4.1392,
+      "step": 304750
+    },
+    {
+      "epoch": 2.62396694214876,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.247607235766112e-05,
+      "loss": 3.9677,
+      "step": 304800
+    },
+    {
+      "epoch": 2.62439738292011,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.244905738889067e-05,
+      "loss": 4.5434,
+      "step": 304850
+    },
+    {
+      "epoch": 2.62482782369146,
+      "grad_norm": 3.5,
+      "learning_rate": 9.242204297434315e-05,
+      "loss": 4.2262,
+      "step": 304900
+    },
+    {
+      "epoch": 2.62525826446281,
+      "grad_norm": 4.5,
+      "learning_rate": 9.23950291160014e-05,
+      "loss": 4.6274,
+      "step": 304950
+    },
+    {
+      "epoch": 2.6256887052341598,
+      "grad_norm": 4.25,
+      "learning_rate": 9.236801581584815e-05,
+      "loss": 3.9821,
+      "step": 305000
+    },
+    {
+      "epoch": 2.6261191460055096,
+      "grad_norm": 1.953125,
+      "learning_rate": 9.234100307586609e-05,
+      "loss": 4.1777,
+      "step": 305050
+    },
+    {
+      "epoch": 2.6265495867768593,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.231399089803793e-05,
+      "loss": 4.6839,
+      "step": 305100
+    },
+    {
+      "epoch": 2.6269800275482096,
+      "grad_norm": 2.609375,
+      "learning_rate": 9.228697928434633e-05,
+      "loss": 4.225,
+      "step": 305150
+    },
+    {
+      "epoch": 2.6274104683195594,
+      "grad_norm": 5.375,
+      "learning_rate": 9.225996823677381e-05,
+      "loss": 4.1882,
+      "step": 305200
+    },
+    {
+      "epoch": 2.627840909090909,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.223295775730303e-05,
+      "loss": 4.1015,
+      "step": 305250
+    },
+    {
+      "epoch": 2.628271349862259,
+      "grad_norm": 2.453125,
+      "learning_rate": 9.220594784791638e-05,
+      "loss": 4.3895,
+      "step": 305300
+    },
+    {
+      "epoch": 2.6287017906336088,
+      "grad_norm": 2.484375,
+      "learning_rate": 9.217893851059642e-05,
+      "loss": 4.2169,
+      "step": 305350
+    },
+    {
+      "epoch": 2.6291322314049586,
+      "grad_norm": 2.828125,
+      "learning_rate": 9.215192974732552e-05,
+      "loss": 4.5952,
+      "step": 305400
+    },
+    {
+      "epoch": 2.6295626721763083,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.212492156008614e-05,
+      "loss": 4.5745,
+      "step": 305450
+    },
+    {
+      "epoch": 2.629993112947658,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.209791395086056e-05,
+      "loss": 4.4447,
+      "step": 305500
+    },
+    {
+      "epoch": 2.6304235537190084,
+      "grad_norm": 3.453125,
+      "learning_rate": 9.207090692163106e-05,
+      "loss": 4.5392,
+      "step": 305550
+    },
+    {
+      "epoch": 2.630853994490358,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.204390047437994e-05,
+      "loss": 4.5285,
+      "step": 305600
+    },
+    {
+      "epoch": 2.631284435261708,
+      "grad_norm": 1.671875,
+      "learning_rate": 9.201689461108943e-05,
+      "loss": 4.3483,
+      "step": 305650
+    },
+    {
+      "epoch": 2.6317148760330578,
+      "grad_norm": 4.84375,
+      "learning_rate": 9.198988933374164e-05,
+      "loss": 4.5677,
+      "step": 305700
+    },
+    {
+      "epoch": 2.6321453168044076,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.19628846443188e-05,
+      "loss": 3.7258,
+      "step": 305750
+    },
+    {
+      "epoch": 2.632575757575758,
+      "grad_norm": 4.53125,
+      "learning_rate": 9.193588054480289e-05,
+      "loss": 4.6498,
+      "step": 305800
+    },
+    {
+      "epoch": 2.6330061983471076,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.190887703717601e-05,
+      "loss": 4.3877,
+      "step": 305850
+    },
+    {
+      "epoch": 2.6334366391184574,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.188187412342013e-05,
+      "loss": 4.3732,
+      "step": 305900
+    },
+    {
+      "epoch": 2.633867079889807,
+      "grad_norm": 2.875,
+      "learning_rate": 9.185487180551723e-05,
+      "loss": 4.2059,
+      "step": 305950
+    },
+    {
+      "epoch": 2.634297520661157,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.182787008544925e-05,
+      "loss": 4.5497,
+      "step": 306000
+    },
+    {
+      "epoch": 2.634297520661157,
+      "eval_loss": 5.004040718078613,
+      "eval_runtime": 24.771,
+      "eval_samples_per_second": 25.837,
+      "eval_steps_per_second": 12.918,
+      "eval_tts_loss": 7.402762199822888,
+      "step": 306000
+    },
+    {
+      "epoch": 2.6347279614325068,
+      "grad_norm": 1.875,
+      "learning_rate": 9.1800868965198e-05,
+      "loss": 3.8846,
+      "step": 306050
+    },
+    {
+      "epoch": 2.6351584022038566,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.17738684467453e-05,
+      "loss": 4.1791,
+      "step": 306100
+    },
+    {
+      "epoch": 2.6355888429752063,
+      "grad_norm": 4.75,
+      "learning_rate": 9.174686853207299e-05,
+      "loss": 4.5738,
+      "step": 306150
+    },
+    {
+      "epoch": 2.6360192837465566,
+      "grad_norm": 1.71875,
+      "learning_rate": 9.171986922316275e-05,
+      "loss": 4.4295,
+      "step": 306200
+    },
+    {
+      "epoch": 2.6364497245179064,
+      "grad_norm": 6.9375,
+      "learning_rate": 9.169287052199629e-05,
+      "loss": 4.3924,
+      "step": 306250
+    },
+    {
+      "epoch": 2.636880165289256,
+      "grad_norm": 1.1640625,
+      "learning_rate": 9.166587243055531e-05,
+      "loss": 4.0633,
+      "step": 306300
+    },
+    {
+      "epoch": 2.637310606060606,
+      "grad_norm": 3.625,
+      "learning_rate": 9.163887495082133e-05,
+      "loss": 4.5509,
+      "step": 306350
+    },
+    {
+      "epoch": 2.6377410468319558,
+      "grad_norm": 1.453125,
+      "learning_rate": 9.161187808477593e-05,
+      "loss": 4.4327,
+      "step": 306400
+    },
+    {
+      "epoch": 2.638171487603306,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.158488183440061e-05,
+      "loss": 4.0974,
+      "step": 306450
+    },
+    {
+      "epoch": 2.638601928374656,
+      "grad_norm": 3.125,
+      "learning_rate": 9.15578862016769e-05,
+      "loss": 4.3478,
+      "step": 306500
+    },
+    {
+      "epoch": 2.6390323691460056,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.15308911885862e-05,
+      "loss": 4.5721,
+      "step": 306550
+    },
+    {
+      "epoch": 2.6394628099173554,
+      "grad_norm": 4.78125,
+      "learning_rate": 9.150389679710981e-05,
+      "loss": 4.5783,
+      "step": 306600
+    },
+    {
+      "epoch": 2.639893250688705,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.147690302922913e-05,
+      "loss": 4.3638,
+      "step": 306650
+    },
+    {
+      "epoch": 2.640323691460055,
+      "grad_norm": 3.6875,
+      "learning_rate": 9.144990988692543e-05,
+      "loss": 4.6775,
+      "step": 306700
+    },
+    {
+      "epoch": 2.6407541322314048,
+      "grad_norm": 3.09375,
+      "learning_rate": 9.142291737217995e-05,
+      "loss": 4.2685,
+      "step": 306750
+    },
+    {
+      "epoch": 2.6411845730027546,
+      "grad_norm": 4.84375,
+      "learning_rate": 9.139592548697391e-05,
+      "loss": 4.3172,
+      "step": 306800
+    },
+    {
+      "epoch": 2.641615013774105,
+      "grad_norm": 3.96875,
+      "learning_rate": 9.136893423328841e-05,
+      "loss": 4.1565,
+      "step": 306850
+    },
+    {
+      "epoch": 2.6420454545454546,
+      "grad_norm": 5.125,
+      "learning_rate": 9.134194361310454e-05,
+      "loss": 4.5736,
+      "step": 306900
+    },
+    {
+      "epoch": 2.6424758953168044,
+      "grad_norm": 3.59375,
+      "learning_rate": 9.131495362840339e-05,
+      "loss": 4.1663,
+      "step": 306950
+    },
+    {
+      "epoch": 2.642906336088154,
+      "grad_norm": 3.375,
+      "learning_rate": 9.128796428116597e-05,
+      "loss": 4.481,
+      "step": 307000
+    },
+    {
+      "epoch": 2.643336776859504,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.126097557337326e-05,
+      "loss": 4.3028,
+      "step": 307050
+    },
+    {
+      "epoch": 2.643767217630854,
+      "grad_norm": 3.90625,
+      "learning_rate": 9.12339875070061e-05,
+      "loss": 4.3013,
+      "step": 307100
+    },
+    {
+      "epoch": 2.644197658402204,
+      "grad_norm": 4.125,
+      "learning_rate": 9.120700008404537e-05,
+      "loss": 4.3392,
+      "step": 307150
+    },
+    {
+      "epoch": 2.644628099173554,
+      "grad_norm": 2.09375,
+      "learning_rate": 9.118001330647195e-05,
+      "loss": 4.4566,
+      "step": 307200
+    },
+    {
+      "epoch": 2.6450585399449036,
+      "grad_norm": 3.0,
+      "learning_rate": 9.115302717626657e-05,
+      "loss": 4.1723,
+      "step": 307250
+    },
+    {
+      "epoch": 2.6454889807162534,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.112604169540999e-05,
+      "loss": 4.3946,
+      "step": 307300
+    },
+    {
+      "epoch": 2.645919421487603,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.109905686588281e-05,
+      "loss": 4.4125,
+      "step": 307350
+    },
+    {
+      "epoch": 2.646349862258953,
+      "grad_norm": 2.078125,
+      "learning_rate": 9.107207268966573e-05,
+      "loss": 4.5274,
+      "step": 307400
+    },
+    {
+      "epoch": 2.6467803030303028,
+      "grad_norm": 3.203125,
+      "learning_rate": 9.10450891687393e-05,
+      "loss": 4.5369,
+      "step": 307450
+    },
+    {
+      "epoch": 2.647210743801653,
+      "grad_norm": 0.94921875,
+      "learning_rate": 9.101810630508407e-05,
+      "loss": 4.0234,
+      "step": 307500
+    },
+    {
+      "epoch": 2.647641184573003,
+      "grad_norm": 4.15625,
+      "learning_rate": 9.09911241006805e-05,
+      "loss": 4.0093,
+      "step": 307550
+    },
+    {
+      "epoch": 2.6480716253443526,
+      "grad_norm": 3.59375,
+      "learning_rate": 9.096414255750904e-05,
+      "loss": 4.5366,
+      "step": 307600
+    },
+    {
+      "epoch": 2.6485020661157024,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.093716167755006e-05,
+      "loss": 4.8575,
+      "step": 307650
+    },
+    {
+      "epoch": 2.6489325068870526,
+      "grad_norm": 5.125,
+      "learning_rate": 9.091018146278394e-05,
+      "loss": 4.5015,
+      "step": 307700
+    },
+    {
+      "epoch": 2.6493629476584024,
+      "grad_norm": 3.875,
+      "learning_rate": 9.088320191519092e-05,
+      "loss": 4.2045,
+      "step": 307750
+    },
+    {
+      "epoch": 2.649793388429752,
+      "grad_norm": 2.625,
+      "learning_rate": 9.08562230367513e-05,
+      "loss": 4.1992,
+      "step": 307800
+    },
+    {
+      "epoch": 2.650223829201102,
+      "grad_norm": 4.75,
+      "learning_rate": 9.082924482944519e-05,
+      "loss": 4.6196,
+      "step": 307850
+    },
+    {
+      "epoch": 2.650654269972452,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.08022672952528e-05,
+      "loss": 4.1372,
+      "step": 307900
+    },
+    {
+      "epoch": 2.6510847107438016,
+      "grad_norm": 2.671875,
+      "learning_rate": 9.077529043615421e-05,
+      "loss": 4.1683,
+      "step": 307950
+    },
+    {
+      "epoch": 2.6515151515151514,
+      "grad_norm": 1.953125,
+      "learning_rate": 9.074831425412943e-05,
+      "loss": 4.5398,
+      "step": 308000
+    },
+    {
+      "epoch": 2.651945592286501,
+      "grad_norm": 2.6875,
+      "learning_rate": 9.072133875115849e-05,
+      "loss": 4.5371,
+      "step": 308050
+    },
+    {
+      "epoch": 2.652376033057851,
+      "grad_norm": 4.03125,
+      "learning_rate": 9.069436392922135e-05,
+      "loss": 4.1481,
+      "step": 308100
+    },
+    {
+      "epoch": 2.652806473829201,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.066738979029784e-05,
+      "loss": 4.3286,
+      "step": 308150
+    },
+    {
+      "epoch": 2.653236914600551,
+      "grad_norm": 5.125,
+      "learning_rate": 9.064041633636783e-05,
+      "loss": 4.01,
+      "step": 308200
+    },
+    {
+      "epoch": 2.653667355371901,
+      "grad_norm": 2.8125,
+      "learning_rate": 9.061344356941116e-05,
+      "loss": 4.4116,
+      "step": 308250
+    },
+    {
+      "epoch": 2.6540977961432506,
+      "grad_norm": 1.8359375,
+      "learning_rate": 9.05864714914075e-05,
+      "loss": 4.0769,
+      "step": 308300
+    },
+    {
+      "epoch": 2.654528236914601,
+      "grad_norm": 1.625,
+      "learning_rate": 9.055950010433663e-05,
+      "loss": 4.2289,
+      "step": 308350
+    },
+    {
+      "epoch": 2.6549586776859506,
+      "grad_norm": 3.5,
+      "learning_rate": 9.053252941017811e-05,
+      "loss": 4.2436,
+      "step": 308400
+    },
+    {
+      "epoch": 2.6553891184573004,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.050555941091156e-05,
+      "loss": 4.5338,
+      "step": 308450
+    },
+    {
+      "epoch": 2.65581955922865,
+      "grad_norm": 2.328125,
+      "learning_rate": 9.04785901085165e-05,
+      "loss": 3.9787,
+      "step": 308500
+    },
+    {
+      "epoch": 2.65625,
+      "grad_norm": 2.359375,
+      "learning_rate": 9.045162150497248e-05,
+      "loss": 4.3154,
+      "step": 308550
+    },
+    {
+      "epoch": 2.65668044077135,
+      "grad_norm": 2.578125,
+      "learning_rate": 9.04246536022589e-05,
+      "loss": 4.209,
+      "step": 308600
+    },
+    {
+      "epoch": 2.6571108815426996,
+      "grad_norm": 5.15625,
+      "learning_rate": 9.03976864023551e-05,
+      "loss": 4.3636,
+      "step": 308650
+    },
+    {
+      "epoch": 2.6575413223140494,
+      "grad_norm": 2.421875,
+      "learning_rate": 9.037071990724049e-05,
+      "loss": 4.2426,
+      "step": 308700
+    },
+    {
+      "epoch": 2.657971763085399,
+      "grad_norm": 5.0625,
+      "learning_rate": 9.034375411889431e-05,
+      "loss": 4.1477,
+      "step": 308750
+    },
+    {
+      "epoch": 2.6584022038567494,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.03167890392958e-05,
+      "loss": 4.0072,
+      "step": 308800
+    },
+    {
+      "epoch": 2.658832644628099,
+      "grad_norm": 2.9375,
+      "learning_rate": 9.028982467042417e-05,
+      "loss": 4.3343,
+      "step": 308850
+    },
+    {
+      "epoch": 2.659263085399449,
+      "grad_norm": 1.96875,
+      "learning_rate": 9.026286101425846e-05,
+      "loss": 4.3219,
+      "step": 308900
+    },
+    {
+      "epoch": 2.659693526170799,
+      "grad_norm": 2.46875,
+      "learning_rate": 9.023589807277785e-05,
+      "loss": 4.4663,
+      "step": 308950
+    },
+    {
+      "epoch": 2.660123966942149,
+      "grad_norm": 3.59375,
+      "learning_rate": 9.020893584796129e-05,
+      "loss": 4.3882,
+      "step": 309000
+    },
+    {
+      "epoch": 2.660123966942149,
+      "eval_loss": 5.002894401550293,
+      "eval_runtime": 25.5644,
+      "eval_samples_per_second": 25.035,
+      "eval_steps_per_second": 12.517,
+      "eval_tts_loss": 7.4085076570409605,
+      "step": 309000
+    },
+    {
+      "epoch": 2.660554407713499,
+      "grad_norm": 4.4375,
+      "learning_rate": 9.018197434178779e-05,
+      "loss": 4.4659,
+      "step": 309050
+    },
+    {
+      "epoch": 2.6609848484848486,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.015501355623627e-05,
+      "loss": 4.2047,
+      "step": 309100
+    },
+    {
+      "epoch": 2.6614152892561984,
+      "grad_norm": 4.96875,
+      "learning_rate": 9.012805349328553e-05,
+      "loss": 4.5465,
+      "step": 309150
+    },
+    {
+      "epoch": 2.661845730027548,
+      "grad_norm": 2.40625,
+      "learning_rate": 9.010109415491441e-05,
+      "loss": 4.293,
+      "step": 309200
+    },
+    {
+      "epoch": 2.662276170798898,
+      "grad_norm": 4.125,
+      "learning_rate": 9.007413554310172e-05,
+      "loss": 4.6997,
+      "step": 309250
+    },
+    {
+      "epoch": 2.662706611570248,
+      "grad_norm": 1.6953125,
+      "learning_rate": 9.00471776598261e-05,
+      "loss": 4.3926,
+      "step": 309300
+    },
+    {
+      "epoch": 2.6631370523415976,
+      "grad_norm": 2.0625,
+      "learning_rate": 9.002022050706627e-05,
+      "loss": 4.2365,
+      "step": 309350
+    },
+    {
+      "epoch": 2.6635674931129474,
+      "grad_norm": 1.15625,
+      "learning_rate": 8.999326408680073e-05,
+      "loss": 4.0975,
+      "step": 309400
+    },
+    {
+      "epoch": 2.6639979338842976,
+      "grad_norm": 1.9921875,
+      "learning_rate": 8.99663084010081e-05,
+      "loss": 4.5361,
+      "step": 309450
+    },
+    {
+      "epoch": 2.6644283746556474,
+      "grad_norm": 2.390625,
+      "learning_rate": 8.993935345166683e-05,
+      "loss": 4.0661,
+      "step": 309500
+    },
+    {
+      "epoch": 2.664858815426997,
+      "grad_norm": 4.25,
+      "learning_rate": 8.991239924075538e-05,
+      "loss": 4.163,
+      "step": 309550
+    },
+    {
+      "epoch": 2.665289256198347,
+      "grad_norm": 2.6875,
+      "learning_rate": 8.988544577025214e-05,
+      "loss": 4.4862,
+      "step": 309600
+    },
+    {
+      "epoch": 2.6657196969696972,
+      "grad_norm": 2.65625,
+      "learning_rate": 8.985849304213542e-05,
+      "loss": 4.3983,
+      "step": 309650
+    },
+    {
+      "epoch": 2.666150137741047,
+      "grad_norm": 3.265625,
+      "learning_rate": 8.983154105838345e-05,
+      "loss": 4.2226,
+      "step": 309700
+    },
+    {
+      "epoch": 2.666580578512397,
+      "grad_norm": 3.78125,
+      "learning_rate": 8.980458982097453e-05,
+      "loss": 4.4132,
+      "step": 309750
+    },
+    {
+      "epoch": 2.6670110192837466,
+      "grad_norm": 2.78125,
+      "learning_rate": 8.977763933188674e-05,
+      "loss": 4.3468,
+      "step": 309800
+    },
+    {
+      "epoch": 2.6674414600550964,
+      "grad_norm": 2.84375,
+      "learning_rate": 8.975068959309824e-05,
+      "loss": 4.3531,
+      "step": 309850
+    },
+    {
+      "epoch": 2.667871900826446,
+      "grad_norm": 3.515625,
+      "learning_rate": 8.972374060658712e-05,
+      "loss": 4.5711,
+      "step": 309900
+    },
+    {
+      "epoch": 2.668302341597796,
+      "grad_norm": 3.28125,
+      "learning_rate": 8.96967923743313e-05,
+      "loss": 3.7813,
+      "step": 309950
+    },
+    {
+      "epoch": 2.668732782369146,
+      "grad_norm": 1.4765625,
+      "learning_rate": 8.966984489830872e-05,
+      "loss": 4.4286,
+      "step": 310000
+    },
+    {
+      "epoch": 2.669163223140496,
+      "grad_norm": 3.90625,
+      "learning_rate": 8.964289818049731e-05,
+      "loss": 4.2551,
+      "step": 310050
+    },
+    {
+      "epoch": 2.669593663911846,
+      "grad_norm": 2.5625,
+      "learning_rate": 8.96159522228749e-05,
+      "loss": 4.4058,
+      "step": 310100
+    },
+    {
+      "epoch": 2.6700241046831956,
+      "grad_norm": 2.578125,
+      "learning_rate": 8.958900702741931e-05,
+      "loss": 4.2769,
+      "step": 310150
+    },
+    {
+      "epoch": 2.6704545454545454,
+      "grad_norm": 3.078125,
+      "learning_rate": 8.956206259610813e-05,
+      "loss": 4.1848,
+      "step": 310200
+    },
+    {
+      "epoch": 2.670884986225895,
+      "grad_norm": 4.1875,
+      "learning_rate": 8.95351189309191e-05,
+      "loss": 4.6105,
+      "step": 310250
+    },
+    {
+      "epoch": 2.6713154269972454,
+      "grad_norm": 2.203125,
+      "learning_rate": 8.950817603382986e-05,
+      "loss": 4.3688,
+      "step": 310300
+    },
+    {
+      "epoch": 2.6717458677685952,
+      "grad_norm": 2.6875,
+      "learning_rate": 8.94812339068179e-05,
+      "loss": 4.1359,
+      "step": 310350
+    },
+    {
+      "epoch": 2.672176308539945,
+      "grad_norm": 3.359375,
+      "learning_rate": 8.945429255186078e-05,
+      "loss": 4.2189,
+      "step": 310400
+    },
+    {
+      "epoch": 2.672606749311295,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.942735197093587e-05,
+      "loss": 4.7217,
+      "step": 310450
+    },
+    {
+      "epoch": 2.6730371900826446,
+      "grad_norm": 2.265625,
+      "learning_rate": 8.940041216602057e-05,
+      "loss": 4.417,
+      "step": 310500
+    },
+    {
+      "epoch": 2.6734676308539944,
+      "grad_norm": 2.828125,
+      "learning_rate": 8.93734731390922e-05,
+      "loss": 4.0018,
+      "step": 310550
+    },
+    {
+      "epoch": 2.673898071625344,
+      "grad_norm": 2.21875,
+      "learning_rate": 8.934653489212805e-05,
+      "loss": 4.4157,
+      "step": 310600
+    },
+    {
+      "epoch": 2.674328512396694,
+      "grad_norm": 2.5,
+      "learning_rate": 8.931959742710535e-05,
+      "loss": 4.2251,
+      "step": 310650
+    },
+    {
+      "epoch": 2.6747589531680442,
+      "grad_norm": 6.15625,
+      "learning_rate": 8.929266074600118e-05,
+      "loss": 4.4576,
+      "step": 310700
+    },
+    {
+      "epoch": 2.675189393939394,
+      "grad_norm": 2.578125,
+      "learning_rate": 8.926572485079268e-05,
+      "loss": 4.0139,
+      "step": 310750
+    },
+    {
+      "epoch": 2.675619834710744,
+      "grad_norm": 2.15625,
+      "learning_rate": 8.923878974345688e-05,
+      "loss": 4.3964,
+      "step": 310800
+    },
+    {
+      "epoch": 2.6760502754820936,
+      "grad_norm": 1.0078125,
+      "learning_rate": 8.921185542597076e-05,
+      "loss": 4.3175,
+      "step": 310850
+    },
+    {
+      "epoch": 2.6764807162534434,
+      "grad_norm": 1.3515625,
+      "learning_rate": 8.918492190031126e-05,
+      "loss": 4.1622,
+      "step": 310900
+    },
+    {
+      "epoch": 2.6769111570247937,
+      "grad_norm": 2.6875,
+      "learning_rate": 8.915798916845521e-05,
+      "loss": 4.2837,
+      "step": 310950
+    },
+    {
+      "epoch": 2.6773415977961434,
+      "grad_norm": 1.5234375,
+      "learning_rate": 8.913105723237942e-05,
+      "loss": 4.3683,
+      "step": 311000
+    },
+    {
+      "epoch": 2.6777720385674932,
+      "grad_norm": 1.4453125,
+      "learning_rate": 8.910412609406064e-05,
+      "loss": 3.8637,
+      "step": 311050
+    },
+    {
+      "epoch": 2.678202479338843,
+      "grad_norm": 4.28125,
+      "learning_rate": 8.907719575547558e-05,
+      "loss": 4.5455,
+      "step": 311100
+    },
+    {
+      "epoch": 2.678632920110193,
+      "grad_norm": 3.8125,
+      "learning_rate": 8.905026621860086e-05,
+      "loss": 4.5472,
+      "step": 311150
+    },
+    {
+      "epoch": 2.6790633608815426,
+      "grad_norm": 2.046875,
+      "learning_rate": 8.902333748541303e-05,
+      "loss": 4.1928,
+      "step": 311200
+    },
+    {
+      "epoch": 2.6794938016528924,
+      "grad_norm": 3.140625,
+      "learning_rate": 8.899640955788858e-05,
+      "loss": 4.2625,
+      "step": 311250
+    },
+    {
+      "epoch": 2.679924242424242,
+      "grad_norm": 4.125,
+      "learning_rate": 8.896948243800402e-05,
+      "loss": 4.4184,
+      "step": 311300
+    },
+    {
+      "epoch": 2.6803546831955924,
+      "grad_norm": 3.671875,
+      "learning_rate": 8.89425561277357e-05,
+      "loss": 4.6541,
+      "step": 311350
+    },
+    {
+      "epoch": 2.6807851239669422,
+      "grad_norm": 3.484375,
+      "learning_rate": 8.891563062906002e-05,
+      "loss": 4.4166,
+      "step": 311400
+    },
+    {
+      "epoch": 2.681215564738292,
+      "grad_norm": 1.015625,
+      "learning_rate": 8.888870594395315e-05,
+      "loss": 4.1524,
+      "step": 311450
+    },
+    {
+      "epoch": 2.681646005509642,
+      "grad_norm": 5.625,
+      "learning_rate": 8.886178207439139e-05,
+      "loss": 4.4343,
+      "step": 311500
+    },
+    {
+      "epoch": 2.6820764462809916,
+      "grad_norm": 3.109375,
+      "learning_rate": 8.883485902235082e-05,
+      "loss": 4.2866,
+      "step": 311550
+    },
+    {
+      "epoch": 2.682506887052342,
+      "grad_norm": 4.25,
+      "learning_rate": 8.88079367898076e-05,
+      "loss": 4.5307,
+      "step": 311600
+    },
+    {
+      "epoch": 2.6829373278236917,
+      "grad_norm": 2.515625,
+      "learning_rate": 8.878101537873774e-05,
+      "loss": 4.6749,
+      "step": 311650
+    },
+    {
+      "epoch": 2.6833677685950414,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.875409479111725e-05,
+      "loss": 4.3074,
+      "step": 311700
+    },
+    {
+      "epoch": 2.6837982093663912,
+      "grad_norm": 2.5,
+      "learning_rate": 8.872717502892194e-05,
+      "loss": 4.3442,
+      "step": 311750
+    },
+    {
+      "epoch": 2.684228650137741,
+      "grad_norm": 3.40625,
+      "learning_rate": 8.870025609412776e-05,
+      "loss": 4.1657,
+      "step": 311800
+    },
+    {
+      "epoch": 2.684659090909091,
+      "grad_norm": 4.125,
+      "learning_rate": 8.867333798871047e-05,
+      "loss": 4.5745,
+      "step": 311850
+    },
+    {
+      "epoch": 2.6850895316804406,
+      "grad_norm": 2.875,
+      "learning_rate": 8.86464207146458e-05,
+      "loss": 4.3193,
+      "step": 311900
+    },
+    {
+      "epoch": 2.6855199724517904,
+      "grad_norm": 2.40625,
+      "learning_rate": 8.861950427390944e-05,
+      "loss": 4.3696,
+      "step": 311950
+    },
+    {
+      "epoch": 2.6859504132231407,
+      "grad_norm": 3.21875,
+      "learning_rate": 8.859258866847697e-05,
+      "loss": 4.2616,
+      "step": 312000
+    },
+    {
+      "epoch": 2.6859504132231407,
+      "eval_loss": 5.001897811889648,
+      "eval_runtime": 24.5929,
+      "eval_samples_per_second": 26.024,
+      "eval_steps_per_second": 13.012,
+      "eval_tts_loss": 7.426488727864566,
+      "step": 312000
+    },
+    {
+      "epoch": 2.6863808539944904,
+      "grad_norm": 1.1875,
+      "learning_rate": 8.856567390032393e-05,
+      "loss": 4.5049,
+      "step": 312050
+    },
+    {
+      "epoch": 2.6868112947658402,
+      "grad_norm": 6.40625,
+      "learning_rate": 8.853875997142582e-05,
+      "loss": 4.7234,
+      "step": 312100
+    },
+    {
+      "epoch": 2.68724173553719,
+      "grad_norm": 4.8125,
+      "learning_rate": 8.851184688375809e-05,
+      "loss": 4.2044,
+      "step": 312150
+    },
+    {
+      "epoch": 2.68767217630854,
+      "grad_norm": 4.0,
+      "learning_rate": 8.848493463929611e-05,
+      "loss": 4.3651,
+      "step": 312200
+    },
+    {
+      "epoch": 2.68810261707989,
+      "grad_norm": 5.34375,
+      "learning_rate": 8.84580232400151e-05,
+      "loss": 4.4014,
+      "step": 312250
+    },
+    {
+      "epoch": 2.68853305785124,
+      "grad_norm": 1.75,
+      "learning_rate": 8.843111268789034e-05,
+      "loss": 4.4415,
+      "step": 312300
+    },
+    {
+      "epoch": 2.6889634986225897,
+      "grad_norm": 4.25,
+      "learning_rate": 8.840420298489707e-05,
+      "loss": 4.3472,
+      "step": 312350
+    },
+    {
+      "epoch": 2.6893939393939394,
+      "grad_norm": 2.84375,
+      "learning_rate": 8.83772941330103e-05,
+      "loss": 4.4325,
+      "step": 312400
+    },
+    {
+      "epoch": 2.6898243801652892,
+      "grad_norm": 2.546875,
+      "learning_rate": 8.83503861342052e-05,
+      "loss": 4.6666,
+      "step": 312450
+    },
+    {
+      "epoch": 2.690254820936639,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.832347899045663e-05,
+      "loss": 4.1108,
+      "step": 312500
+    },
+    {
+      "epoch": 2.690685261707989,
+      "grad_norm": 2.25,
+      "learning_rate": 8.829657270373963e-05,
+      "loss": 4.1628,
+      "step": 312550
+    },
+    {
+      "epoch": 2.6911157024793386,
+      "grad_norm": 3.875,
+      "learning_rate": 8.826966727602895e-05,
+      "loss": 4.2323,
+      "step": 312600
+    },
+    {
+      "epoch": 2.691546143250689,
+      "grad_norm": 2.953125,
+      "learning_rate": 8.824276270929951e-05,
+      "loss": 4.3412,
+      "step": 312650
+    },
+    {
+      "epoch": 2.6919765840220387,
+      "grad_norm": 2.890625,
+      "learning_rate": 8.821585900552601e-05,
+      "loss": 4.4091,
+      "step": 312700
+    },
+    {
+      "epoch": 2.6924070247933884,
+      "grad_norm": 4.6875,
+      "learning_rate": 8.818895616668304e-05,
+      "loss": 4.3031,
+      "step": 312750
+    },
+    {
+      "epoch": 2.6928374655647382,
+      "grad_norm": 4.375,
+      "learning_rate": 8.816205419474528e-05,
+      "loss": 4.247,
+      "step": 312800
+    },
+    {
+      "epoch": 2.693267906336088,
+      "grad_norm": 2.71875,
+      "learning_rate": 8.81351530916873e-05,
+      "loss": 4.325,
+      "step": 312850
+    },
+    {
+      "epoch": 2.6936983471074383,
+      "grad_norm": 4.9375,
+      "learning_rate": 8.810825285948355e-05,
+      "loss": 4.1087,
+      "step": 312900
+    },
+    {
+      "epoch": 2.694128787878788,
+      "grad_norm": 4.75,
+      "learning_rate": 8.808135350010847e-05,
+      "loss": 4.6989,
+      "step": 312950
+    },
+    {
+      "epoch": 2.694559228650138,
+      "grad_norm": 3.671875,
+      "learning_rate": 8.805445501553638e-05,
+      "loss": 4.1459,
+      "step": 313000
+    },
+    {
+      "epoch": 2.6949896694214877,
+      "grad_norm": 2.703125,
+      "learning_rate": 8.802755740774159e-05,
+      "loss": 4.3638,
+      "step": 313050
+    },
+    {
+      "epoch": 2.6954201101928374,
+      "grad_norm": 1.703125,
+      "learning_rate": 8.800066067869831e-05,
+      "loss": 4.018,
+      "step": 313100
+    },
+    {
+      "epoch": 2.6958505509641872,
+      "grad_norm": 2.78125,
+      "learning_rate": 8.797376483038073e-05,
+      "loss": 4.4241,
+      "step": 313150
+    },
+    {
+      "epoch": 2.696280991735537,
+      "grad_norm": 1.9609375,
+      "learning_rate": 8.794686986476297e-05,
+      "loss": 4.2152,
+      "step": 313200
+    },
+    {
+      "epoch": 2.696711432506887,
+      "grad_norm": 2.015625,
+      "learning_rate": 8.7919975783819e-05,
+      "loss": 4.6765,
+      "step": 313250
+    },
+    {
+      "epoch": 2.697141873278237,
+      "grad_norm": 3.796875,
+      "learning_rate": 8.789308258952282e-05,
+      "loss": 4.2798,
+      "step": 313300
+    },
+    {
+      "epoch": 2.697572314049587,
+      "grad_norm": 2.421875,
+      "learning_rate": 8.786619028384833e-05,
+      "loss": 4.7068,
+      "step": 313350
+    },
+    {
+      "epoch": 2.6980027548209367,
+      "grad_norm": 3.015625,
+      "learning_rate": 8.783929886876936e-05,
+      "loss": 4.0879,
+      "step": 313400
+    },
+    {
+      "epoch": 2.6984331955922864,
+      "grad_norm": 3.28125,
+      "learning_rate": 8.781240834625968e-05,
+      "loss": 4.5351,
+      "step": 313450
+    },
+    {
+      "epoch": 2.6988636363636362,
+      "grad_norm": 2.515625,
+      "learning_rate": 8.778551871829305e-05,
+      "loss": 4.3604,
+      "step": 313500
+    },
+    {
+      "epoch": 2.6992940771349865,
+      "grad_norm": 3.984375,
+      "learning_rate": 8.775862998684304e-05,
+      "loss": 4.4568,
+      "step": 313550
+    },
+    {
+      "epoch": 2.6997245179063363,
+      "grad_norm": 3.015625,
+      "learning_rate": 8.773174215388323e-05,
+      "loss": 4.458,
+      "step": 313600
+    },
+    {
+      "epoch": 2.700154958677686,
+      "grad_norm": 2.625,
+      "learning_rate": 8.770485522138716e-05,
+      "loss": 4.5919,
+      "step": 313650
+    },
+    {
+      "epoch": 2.700585399449036,
+      "grad_norm": 2.546875,
+      "learning_rate": 8.767796919132827e-05,
+      "loss": 4.2773,
+      "step": 313700
+    },
+    {
+      "epoch": 2.7010158402203857,
+      "grad_norm": 2.734375,
+      "learning_rate": 8.765108406567997e-05,
+      "loss": 4.3466,
+      "step": 313750
+    },
+    {
+      "epoch": 2.7014462809917354,
+      "grad_norm": 2.59375,
+      "learning_rate": 8.762419984641547e-05,
+      "loss": 4.325,
+      "step": 313800
+    },
+    {
+      "epoch": 2.7018767217630852,
+      "grad_norm": 2.078125,
+      "learning_rate": 8.759731653550807e-05,
+      "loss": 4.3278,
+      "step": 313850
+    },
+    {
+      "epoch": 2.702307162534435,
+      "grad_norm": 3.046875,
+      "learning_rate": 8.757043413493098e-05,
+      "loss": 4.4759,
+      "step": 313900
+    },
+    {
+      "epoch": 2.7027376033057853,
+      "grad_norm": 2.9375,
+      "learning_rate": 8.754355264665726e-05,
+      "loss": 4.7913,
+      "step": 313950
+    },
+    {
+      "epoch": 2.703168044077135,
+      "grad_norm": 4.8125,
+      "learning_rate": 8.751667207266004e-05,
+      "loss": 3.9926,
+      "step": 314000
+    },
+    {
+      "epoch": 2.703598484848485,
+      "grad_norm": 1.8671875,
+      "learning_rate": 8.748979241491216e-05,
+      "loss": 4.0226,
+      "step": 314050
+    },
+    {
+      "epoch": 2.7040289256198347,
+      "grad_norm": 2.875,
+      "learning_rate": 8.746291367538663e-05,
+      "loss": 4.6195,
+      "step": 314100
+    },
+    {
+      "epoch": 2.7044593663911844,
+      "grad_norm": 3.28125,
+      "learning_rate": 8.743603585605624e-05,
+      "loss": 4.275,
+      "step": 314150
+    },
+    {
+      "epoch": 2.7048898071625347,
+      "grad_norm": 3.25,
+      "learning_rate": 8.740915895889381e-05,
+      "loss": 4.3307,
+      "step": 314200
+    },
+    {
+      "epoch": 2.7053202479338845,
+      "grad_norm": 2.34375,
+      "learning_rate": 8.738228298587204e-05,
+      "loss": 4.1049,
+      "step": 314250
+    },
+    {
+      "epoch": 2.7057506887052343,
+      "grad_norm": 6.09375,
+      "learning_rate": 8.735540793896351e-05,
+      "loss": 4.5113,
+      "step": 314300
+    },
+    {
+      "epoch": 2.706181129476584,
+      "grad_norm": 1.7421875,
+      "learning_rate": 8.732853382014083e-05,
+      "loss": 4.1478,
+      "step": 314350
+    },
+    {
+      "epoch": 2.706611570247934,
+      "grad_norm": 2.796875,
+      "learning_rate": 8.730166063137653e-05,
+      "loss": 4.1365,
+      "step": 314400
+    },
+    {
+      "epoch": 2.7070420110192837,
+      "grad_norm": 2.359375,
+      "learning_rate": 8.7274788374643e-05,
+      "loss": 4.4434,
+      "step": 314450
+    },
+    {
+      "epoch": 2.7074724517906334,
+      "grad_norm": 3.609375,
+      "learning_rate": 8.724791705191267e-05,
+      "loss": 4.1965,
+      "step": 314500
+    },
+    {
+      "epoch": 2.7079028925619832,
+      "grad_norm": 4.65625,
+      "learning_rate": 8.722104666515774e-05,
+      "loss": 4.0563,
+      "step": 314550
+    },
+    {
+      "epoch": 2.7083333333333335,
+      "grad_norm": 4.53125,
+      "learning_rate": 8.71941772163505e-05,
+      "loss": 4.0379,
+      "step": 314600
+    },
+    {
+      "epoch": 2.7087637741046833,
+      "grad_norm": 3.0625,
+      "learning_rate": 8.71673087074631e-05,
+      "loss": 4.3457,
+      "step": 314650
+    },
+    {
+      "epoch": 2.709194214876033,
+      "grad_norm": 3.65625,
+      "learning_rate": 8.714044114046764e-05,
+      "loss": 4.409,
+      "step": 314700
+    },
+    {
+      "epoch": 2.709624655647383,
+      "grad_norm": 2.328125,
+      "learning_rate": 8.711357451733613e-05,
+      "loss": 4.4346,
+      "step": 314750
+    },
+    {
+      "epoch": 2.7100550964187327,
+      "grad_norm": 2.140625,
+      "learning_rate": 8.708670884004054e-05,
+      "loss": 3.9265,
+      "step": 314800
+    },
+    {
+      "epoch": 2.710485537190083,
+      "grad_norm": 3.4375,
+      "learning_rate": 8.70598441105527e-05,
+      "loss": 4.1622,
+      "step": 314850
+    },
+    {
+      "epoch": 2.7109159779614327,
+      "grad_norm": 2.640625,
+      "learning_rate": 8.70329803308445e-05,
+      "loss": 4.7735,
+      "step": 314900
+    },
+    {
+      "epoch": 2.7113464187327825,
+      "grad_norm": 3.21875,
+      "learning_rate": 8.70061175028876e-05,
+      "loss": 4.4975,
+      "step": 314950
+    },
+    {
+      "epoch": 2.7117768595041323,
+      "grad_norm": 2.15625,
+      "learning_rate": 8.697925562865378e-05,
+      "loss": 4.4245,
+      "step": 315000
+    },
+    {
+      "epoch": 2.7117768595041323,
+      "eval_loss": 5.002341270446777,
+      "eval_runtime": 24.7637,
+      "eval_samples_per_second": 25.844,
+      "eval_steps_per_second": 12.922,
+      "eval_tts_loss": 7.452000815631489,
+      "step": 315000
+    },
+    {
+      "epoch": 2.712207300275482,
+      "grad_norm": 3.125,
+      "learning_rate": 8.695239471011453e-05,
+      "loss": 4.445,
+      "step": 315050
+    },
+    {
+      "epoch": 2.712637741046832,
+      "grad_norm": 3.59375,
+      "learning_rate": 8.692553474924145e-05,
+      "loss": 4.0807,
+      "step": 315100
+    },
+    {
+      "epoch": 2.7130681818181817,
+      "grad_norm": 1.640625,
+      "learning_rate": 8.689867574800598e-05,
+      "loss": 4.406,
+      "step": 315150
+    },
+    {
+      "epoch": 2.7134986225895315,
+      "grad_norm": 1.609375,
+      "learning_rate": 8.687181770837953e-05,
+      "loss": 4.2764,
+      "step": 315200
+    },
+    {
+      "epoch": 2.7139290633608817,
+      "grad_norm": 1.8515625,
+      "learning_rate": 8.684496063233344e-05,
+      "loss": 4.3596,
+      "step": 315250
+    },
+    {
+      "epoch": 2.7143595041322315,
+      "grad_norm": 3.859375,
+      "learning_rate": 8.681810452183892e-05,
+      "loss": 4.4398,
+      "step": 315300
+    },
+    {
+      "epoch": 2.7147899449035813,
+      "grad_norm": 3.484375,
+      "learning_rate": 8.679124937886714e-05,
+      "loss": 3.9907,
+      "step": 315350
+    },
+    {
+      "epoch": 2.715220385674931,
+      "grad_norm": 6.03125,
+      "learning_rate": 8.676439520538923e-05,
+      "loss": 4.0684,
+      "step": 315400
+    },
+    {
+      "epoch": 2.715650826446281,
+      "grad_norm": 4.1875,
+      "learning_rate": 8.673754200337628e-05,
+      "loss": 4.5319,
+      "step": 315450
+    },
+    {
+      "epoch": 2.716081267217631,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.671068977479918e-05,
+      "loss": 4.2654,
+      "step": 315500
+    },
+    {
+      "epoch": 2.716511707988981,
+      "grad_norm": 3.15625,
+      "learning_rate": 8.668383852162889e-05,
+      "loss": 4.3361,
+      "step": 315550
+    },
+    {
+      "epoch": 2.7169421487603307,
+      "grad_norm": 5.25,
+      "learning_rate": 8.665698824583619e-05,
+      "loss": 4.5006,
+      "step": 315600
+    },
+    {
+      "epoch": 2.7173725895316805,
+      "grad_norm": 2.578125,
+      "learning_rate": 8.663013894939183e-05,
+      "loss": 4.5736,
+      "step": 315650
+    },
+    {
+      "epoch": 2.7178030303030303,
+      "grad_norm": 2.0625,
+      "learning_rate": 8.66032906342665e-05,
+      "loss": 4.2864,
+      "step": 315700
+    },
+    {
+      "epoch": 2.71823347107438,
+      "grad_norm": 4.34375,
+      "learning_rate": 8.657644330243082e-05,
+      "loss": 4.1847,
+      "step": 315750
+    },
+    {
+      "epoch": 2.71866391184573,
+      "grad_norm": 2.1875,
+      "learning_rate": 8.654959695585537e-05,
+      "loss": 4.2824,
+      "step": 315800
+    },
+    {
+      "epoch": 2.7190943526170797,
+      "grad_norm": 3.609375,
+      "learning_rate": 8.65227515965105e-05,
+      "loss": 4.395,
+      "step": 315850
+    },
+    {
+      "epoch": 2.71952479338843,
+      "grad_norm": 2.546875,
+      "learning_rate": 8.649590722636667e-05,
+      "loss": 4.3194,
+      "step": 315900
+    },
+    {
+      "epoch": 2.7199552341597797,
+      "grad_norm": 3.0625,
+      "learning_rate": 8.646906384739421e-05,
+      "loss": 4.2922,
+      "step": 315950
+    },
+    {
+      "epoch": 2.7203856749311295,
+      "grad_norm": 2.765625,
+      "learning_rate": 8.644222146156335e-05,
+      "loss": 4.3868,
+      "step": 316000
+    },
+    {
+      "epoch": 2.7208161157024793,
+      "grad_norm": 4.875,
+      "learning_rate": 8.641538007084428e-05,
+      "loss": 4.2453,
+      "step": 316050
+    },
+    {
+      "epoch": 2.721246556473829,
+      "grad_norm": 4.28125,
+      "learning_rate": 8.638853967720706e-05,
+      "loss": 4.1028,
+      "step": 316100
+    },
+    {
+      "epoch": 2.7216769972451793,
+      "grad_norm": 2.28125,
+      "learning_rate": 8.636170028262176e-05,
+      "loss": 4.0804,
+      "step": 316150
+    },
+    {
+      "epoch": 2.722107438016529,
+      "grad_norm": 3.015625,
+      "learning_rate": 8.63348618890583e-05,
+      "loss": 4.1789,
+      "step": 316200
+    },
+    {
+      "epoch": 2.722537878787879,
+      "grad_norm": 3.234375,
+      "learning_rate": 8.630802449848659e-05,
+      "loss": 4.144,
+      "step": 316250
+    },
+    {
+      "epoch": 2.7229683195592287,
+      "grad_norm": 2.453125,
+      "learning_rate": 8.628118811287644e-05,
+      "loss": 4.2576,
+      "step": 316300
+    },
+    {
+      "epoch": 2.7233987603305785,
+      "grad_norm": 4.53125,
+      "learning_rate": 8.625435273419753e-05,
+      "loss": 4.4293,
+      "step": 316350
+    },
+    {
+      "epoch": 2.7238292011019283,
+      "grad_norm": 2.15625,
+      "learning_rate": 8.622751836441956e-05,
+      "loss": 4.2431,
+      "step": 316400
+    },
+    {
+      "epoch": 2.724259641873278,
+      "grad_norm": 2.25,
+      "learning_rate": 8.620068500551212e-05,
+      "loss": 4.1925,
+      "step": 316450
+    },
+    {
+      "epoch": 2.724690082644628,
+      "grad_norm": 4.125,
+      "learning_rate": 8.617385265944468e-05,
+      "loss": 4.2578,
+      "step": 316500
+    },
+    {
+      "epoch": 2.725120523415978,
+      "grad_norm": 2.890625,
+      "learning_rate": 8.614702132818678e-05,
+      "loss": 4.1409,
+      "step": 316550
+    },
+    {
+      "epoch": 2.725550964187328,
+      "grad_norm": 2.984375,
+      "learning_rate": 8.612019101370764e-05,
+      "loss": 4.4209,
+      "step": 316600
+    },
+    {
+      "epoch": 2.7259814049586777,
+      "grad_norm": 3.546875,
+      "learning_rate": 8.609336171797665e-05,
+      "loss": 4.2466,
+      "step": 316650
+    },
+    {
+      "epoch": 2.7264118457300275,
+      "grad_norm": 2.296875,
+      "learning_rate": 8.606653344296296e-05,
+      "loss": 4.5445,
+      "step": 316700
+    },
+    {
+      "epoch": 2.7268422865013773,
+      "grad_norm": 1.25,
+      "learning_rate": 8.603970619063575e-05,
+      "loss": 3.8468,
+      "step": 316750
+    },
+    {
+      "epoch": 2.7272727272727275,
+      "grad_norm": 2.46875,
+      "learning_rate": 8.60128799629641e-05,
+      "loss": 4.4241,
+      "step": 316800
+    },
+    {
+      "epoch": 2.7277031680440773,
+      "grad_norm": 0.69140625,
+      "learning_rate": 8.598605476191695e-05,
+      "loss": 4.2696,
+      "step": 316850
+    },
+    {
+      "epoch": 2.728133608815427,
+      "grad_norm": 4.75,
+      "learning_rate": 8.595923058946321e-05,
+      "loss": 4.4268,
+      "step": 316900
+    },
+    {
+      "epoch": 2.728564049586777,
+      "grad_norm": 4.4375,
+      "learning_rate": 8.593240744757175e-05,
+      "loss": 4.2739,
+      "step": 316950
+    },
+    {
+      "epoch": 2.7289944903581267,
+      "grad_norm": 3.90625,
+      "learning_rate": 8.590558533821129e-05,
+      "loss": 4.3778,
+      "step": 317000
+    },
+    {
+      "epoch": 2.7294249311294765,
+      "grad_norm": 2.359375,
+      "learning_rate": 8.58787642633506e-05,
+      "loss": 4.3922,
+      "step": 317050
+    },
+    {
+      "epoch": 2.7298553719008263,
+      "grad_norm": 2.84375,
+      "learning_rate": 8.585194422495817e-05,
+      "loss": 4.4721,
+      "step": 317100
+    },
+    {
+      "epoch": 2.730285812672176,
+      "grad_norm": 4.03125,
+      "learning_rate": 8.582512522500263e-05,
+      "loss": 4.6698,
+      "step": 317150
+    },
+    {
+      "epoch": 2.7307162534435263,
+      "grad_norm": 3.484375,
+      "learning_rate": 8.579830726545237e-05,
+      "loss": 4.3732,
+      "step": 317200
+    },
+    {
+      "epoch": 2.731146694214876,
+      "grad_norm": 1.921875,
+      "learning_rate": 8.577149034827581e-05,
+      "loss": 4.2507,
+      "step": 317250
+    },
+    {
+      "epoch": 2.731577134986226,
+      "grad_norm": 1.0234375,
+      "learning_rate": 8.574467447544126e-05,
+      "loss": 4.0641,
+      "step": 317300
+    },
+    {
+      "epoch": 2.7320075757575757,
+      "grad_norm": 1.6875,
+      "learning_rate": 8.571785964891696e-05,
+      "loss": 4.2794,
+      "step": 317350
+    },
+    {
+      "epoch": 2.7324380165289255,
+      "grad_norm": 3.609375,
+      "learning_rate": 8.569104587067097e-05,
+      "loss": 4.4496,
+      "step": 317400
+    },
+    {
+      "epoch": 2.7328684573002757,
+      "grad_norm": 4.15625,
+      "learning_rate": 8.566423314267144e-05,
+      "loss": 4.6087,
+      "step": 317450
+    },
+    {
+      "epoch": 2.7332988980716255,
+      "grad_norm": 2.296875,
+      "learning_rate": 8.563742146688638e-05,
+      "loss": 4.4325,
+      "step": 317500
+    },
+    {
+      "epoch": 2.7337293388429753,
+      "grad_norm": 6.09375,
+      "learning_rate": 8.561061084528366e-05,
+      "loss": 4.1216,
+      "step": 317550
+    },
+    {
+      "epoch": 2.734159779614325,
+      "grad_norm": 4.3125,
+      "learning_rate": 8.558380127983119e-05,
+      "loss": 4.425,
+      "step": 317600
+    },
+    {
+      "epoch": 2.734590220385675,
+      "grad_norm": 4.0,
+      "learning_rate": 8.555699277249664e-05,
+      "loss": 4.1315,
+      "step": 317650
+    },
+    {
+      "epoch": 2.7350206611570247,
+      "grad_norm": 4.6875,
+      "learning_rate": 8.553018532524776e-05,
+      "loss": 4.515,
+      "step": 317700
+    },
+    {
+      "epoch": 2.7354511019283745,
+      "grad_norm": 2.796875,
+      "learning_rate": 8.550337894005213e-05,
+      "loss": 4.1332,
+      "step": 317750
+    },
+    {
+      "epoch": 2.7358815426997243,
+      "grad_norm": 4.21875,
+      "learning_rate": 8.547657361887731e-05,
+      "loss": 4.2553,
+      "step": 317800
+    },
+    {
+      "epoch": 2.7363119834710745,
+      "grad_norm": 4.0,
+      "learning_rate": 8.544976936369079e-05,
+      "loss": 4.5011,
+      "step": 317850
+    },
+    {
+      "epoch": 2.7367424242424243,
+      "grad_norm": 4.34375,
+      "learning_rate": 8.542296617645981e-05,
+      "loss": 4.4846,
+      "step": 317900
+    },
+    {
+      "epoch": 2.737172865013774,
+      "grad_norm": 5.75,
+      "learning_rate": 8.539616405915176e-05,
+      "loss": 4.4545,
+      "step": 317950
+    },
+    {
+      "epoch": 2.737603305785124,
+      "grad_norm": 2.28125,
+      "learning_rate": 8.536936301373389e-05,
+      "loss": 4.4221,
+      "step": 318000
+    },
+    {
+      "epoch": 2.737603305785124,
+      "eval_loss": 4.999110221862793,
+      "eval_runtime": 25.5297,
+      "eval_samples_per_second": 25.069,
+      "eval_steps_per_second": 12.534,
+      "eval_tts_loss": 7.4143139091922885,
+      "step": 318000
+    },
+    {
+      "epoch": 2.7380337465564737,
+      "grad_norm": 4.25,
+      "learning_rate": 8.534256304217325e-05,
+      "loss": 4.3131,
+      "step": 318050
+    },
+    {
+      "epoch": 2.738464187327824,
+      "grad_norm": 2.25,
+      "learning_rate": 8.531576414643701e-05,
+      "loss": 4.1048,
+      "step": 318100
+    },
+    {
+      "epoch": 2.7388946280991737,
+      "grad_norm": 1.7265625,
+      "learning_rate": 8.528896632849202e-05,
+      "loss": 4.3546,
+      "step": 318150
+    },
+    {
+      "epoch": 2.7393250688705235,
+      "grad_norm": 3.234375,
+      "learning_rate": 8.526216959030528e-05,
+      "loss": 4.2491,
+      "step": 318200
+    },
+    {
+      "epoch": 2.7397555096418733,
+      "grad_norm": 3.9375,
+      "learning_rate": 8.523537393384357e-05,
+      "loss": 4.1753,
+      "step": 318250
+    },
+    {
+      "epoch": 2.740185950413223,
+      "grad_norm": 2.46875,
+      "learning_rate": 8.520857936107367e-05,
+      "loss": 4.1779,
+      "step": 318300
+    },
+    {
+      "epoch": 2.740616391184573,
+      "grad_norm": 2.125,
+      "learning_rate": 8.518178587396222e-05,
+      "loss": 4.3107,
+      "step": 318350
+    },
+    {
+      "epoch": 2.7410468319559227,
+      "grad_norm": 1.9140625,
+      "learning_rate": 8.51549934744758e-05,
+      "loss": 4.4846,
+      "step": 318400
+    },
+    {
+      "epoch": 2.7414772727272725,
+      "grad_norm": 3.265625,
+      "learning_rate": 8.512820216458088e-05,
+      "loss": 4.4076,
+      "step": 318450
+    },
+    {
+      "epoch": 2.7419077134986227,
+      "grad_norm": 3.453125,
+      "learning_rate": 8.510141194624397e-05,
+      "loss": 4.3227,
+      "step": 318500
+    },
+    {
+      "epoch": 2.7423381542699725,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.507462282143134e-05,
+      "loss": 4.4398,
+      "step": 318550
+    },
+    {
+      "epoch": 2.7427685950413223,
+      "grad_norm": 2.9375,
+      "learning_rate": 8.504783479210931e-05,
+      "loss": 4.3605,
+      "step": 318600
+    },
+    {
+      "epoch": 2.743199035812672,
+      "grad_norm": 4.28125,
+      "learning_rate": 8.502104786024402e-05,
+      "loss": 4.2704,
+      "step": 318650
+    },
+    {
+      "epoch": 2.743629476584022,
+      "grad_norm": 3.71875,
+      "learning_rate": 8.49942620278016e-05,
+      "loss": 4.5545,
+      "step": 318700
+    },
+    {
+      "epoch": 2.744059917355372,
+      "grad_norm": 3.5,
+      "learning_rate": 8.496747729674804e-05,
+      "loss": 4.6623,
+      "step": 318750
+    },
+    {
+      "epoch": 2.744490358126722,
+      "grad_norm": 2.78125,
+      "learning_rate": 8.494069366904928e-05,
+      "loss": 4.2208,
+      "step": 318800
+    },
+    {
+      "epoch": 2.7449207988980717,
+      "grad_norm": 1.859375,
+      "learning_rate": 8.491391114667127e-05,
+      "loss": 3.9635,
+      "step": 318850
+    },
+    {
+      "epoch": 2.7453512396694215,
+      "grad_norm": 4.40625,
+      "learning_rate": 8.48871297315797e-05,
+      "loss": 3.867,
+      "step": 318900
+    },
+    {
+      "epoch": 2.7457816804407713,
+      "grad_norm": 6.3125,
+      "learning_rate": 8.486034942574028e-05,
+      "loss": 4.4222,
+      "step": 318950
+    },
+    {
+      "epoch": 2.746212121212121,
+      "grad_norm": 3.53125,
+      "learning_rate": 8.483357023111862e-05,
+      "loss": 4.3181,
+      "step": 319000
+    },
+    {
+      "epoch": 2.746642561983471,
+      "grad_norm": 1.7265625,
+      "learning_rate": 8.48067921496803e-05,
+      "loss": 4.6035,
+      "step": 319050
+    },
+    {
+      "epoch": 2.7470730027548207,
+      "grad_norm": 2.578125,
+      "learning_rate": 8.478001518339074e-05,
+      "loss": 4.1172,
+      "step": 319100
+    },
+    {
+      "epoch": 2.747503443526171,
+      "grad_norm": 3.234375,
+      "learning_rate": 8.475323933421534e-05,
+      "loss": 3.7132,
+      "step": 319150
+    },
+    {
+      "epoch": 2.7479338842975207,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.472646460411937e-05,
+      "loss": 3.9813,
+      "step": 319200
+    },
+    {
+      "epoch": 2.7483643250688705,
+      "grad_norm": 3.28125,
+      "learning_rate": 8.469969099506799e-05,
+      "loss": 4.2913,
+      "step": 319250
+    },
+    {
+      "epoch": 2.7487947658402203,
+      "grad_norm": 2.84375,
+      "learning_rate": 8.467291850902638e-05,
+      "loss": 4.1881,
+      "step": 319300
+    },
+    {
+      "epoch": 2.74922520661157,
+      "grad_norm": 1.1484375,
+      "learning_rate": 8.46461471479596e-05,
+      "loss": 3.9789,
+      "step": 319350
+    },
+    {
+      "epoch": 2.7496556473829203,
+      "grad_norm": 2.40625,
+      "learning_rate": 8.461937691383261e-05,
+      "loss": 4.2677,
+      "step": 319400
+    },
+    {
+      "epoch": 2.75008608815427,
+      "grad_norm": 3.1875,
+      "learning_rate": 8.459260780861021e-05,
+      "loss": 4.3883,
+      "step": 319450
+    },
+    {
+      "epoch": 2.75051652892562,
+      "grad_norm": 4.1875,
+      "learning_rate": 8.456583983425726e-05,
+      "loss": 4.1065,
+      "step": 319500
+    },
+    {
+      "epoch": 2.7509469696969697,
+      "grad_norm": 2.875,
+      "learning_rate": 8.453907299273847e-05,
+      "loss": 4.1442,
+      "step": 319550
+    },
+    {
+      "epoch": 2.7513774104683195,
+      "grad_norm": 2.28125,
+      "learning_rate": 8.451230728601843e-05,
+      "loss": 4.2141,
+      "step": 319600
+    },
+    {
+      "epoch": 2.7518078512396693,
+      "grad_norm": 2.96875,
+      "learning_rate": 8.448554271606176e-05,
+      "loss": 4.4881,
+      "step": 319650
+    },
+    {
+      "epoch": 2.752238292011019,
+      "grad_norm": 3.0,
+      "learning_rate": 8.445877928483283e-05,
+      "loss": 4.2699,
+      "step": 319700
+    },
+    {
+      "epoch": 2.752668732782369,
+      "grad_norm": 5.09375,
+      "learning_rate": 8.443201699429609e-05,
+      "loss": 4.3307,
+      "step": 319750
+    },
+    {
+      "epoch": 2.753099173553719,
+      "grad_norm": 3.28125,
+      "learning_rate": 8.440525584641579e-05,
+      "loss": 4.5304,
+      "step": 319800
+    },
+    {
+      "epoch": 2.753529614325069,
+      "grad_norm": 1.5390625,
+      "learning_rate": 8.437849584315616e-05,
+      "loss": 4.2557,
+      "step": 319850
+    },
+    {
+      "epoch": 2.7539600550964187,
+      "grad_norm": 5.40625,
+      "learning_rate": 8.435173698648136e-05,
+      "loss": 3.9056,
+      "step": 319900
+    },
+    {
+      "epoch": 2.7543904958677685,
+      "grad_norm": 1.453125,
+      "learning_rate": 8.432497927835536e-05,
+      "loss": 3.999,
+      "step": 319950
+    },
+    {
+      "epoch": 2.7548209366391183,
+      "grad_norm": 3.0625,
+      "learning_rate": 8.429822272074214e-05,
+      "loss": 4.0828,
+      "step": 320000
+    },
+    {
+      "epoch": 2.7552513774104685,
+      "grad_norm": 2.09375,
+      "learning_rate": 8.427146731560564e-05,
+      "loss": 4.4632,
+      "step": 320050
+    },
+    {
+      "epoch": 2.7556818181818183,
+      "grad_norm": 3.546875,
+      "learning_rate": 8.424471306490956e-05,
+      "loss": 4.6003,
+      "step": 320100
+    },
+    {
+      "epoch": 2.756112258953168,
+      "grad_norm": 3.4375,
+      "learning_rate": 8.421795997061768e-05,
+      "loss": 4.1435,
+      "step": 320150
+    },
+    {
+      "epoch": 2.756542699724518,
+      "grad_norm": 3.453125,
+      "learning_rate": 8.419120803469354e-05,
+      "loss": 4.5449,
+      "step": 320200
+    },
+    {
+      "epoch": 2.7569731404958677,
+      "grad_norm": 1.6953125,
+      "learning_rate": 8.416445725910074e-05,
+      "loss": 4.2561,
+      "step": 320250
+    },
+    {
+      "epoch": 2.7574035812672175,
+      "grad_norm": 3.65625,
+      "learning_rate": 8.413770764580267e-05,
+      "loss": 4.5275,
+      "step": 320300
+    },
+    {
+      "epoch": 2.7578340220385673,
+      "grad_norm": 3.640625,
+      "learning_rate": 8.411095919676275e-05,
+      "loss": 3.8868,
+      "step": 320350
+    },
+    {
+      "epoch": 2.758264462809917,
+      "grad_norm": 3.328125,
+      "learning_rate": 8.408421191394428e-05,
+      "loss": 4.6279,
+      "step": 320400
+    },
+    {
+      "epoch": 2.7586949035812673,
+      "grad_norm": 4.71875,
+      "learning_rate": 8.405746579931039e-05,
+      "loss": 4.6871,
+      "step": 320450
+    },
+    {
+      "epoch": 2.759125344352617,
+      "grad_norm": 4.125,
+      "learning_rate": 8.403072085482418e-05,
+      "loss": 4.4073,
+      "step": 320500
+    },
+    {
+      "epoch": 2.759555785123967,
+      "grad_norm": 4.3125,
+      "learning_rate": 8.400397708244872e-05,
+      "loss": 3.9658,
+      "step": 320550
+    },
+    {
+      "epoch": 2.7599862258953167,
+      "grad_norm": 4.5625,
+      "learning_rate": 8.397723448414692e-05,
+      "loss": 4.3502,
+      "step": 320600
+    },
+    {
+      "epoch": 2.7604166666666665,
+      "grad_norm": 6.15625,
+      "learning_rate": 8.395049306188167e-05,
+      "loss": 4.7147,
+      "step": 320650
+    },
+    {
+      "epoch": 2.7608471074380168,
+      "grad_norm": 2.8125,
+      "learning_rate": 8.392375281761564e-05,
+      "loss": 4.5776,
+      "step": 320700
+    },
+    {
+      "epoch": 2.7612775482093666,
+      "grad_norm": 3.984375,
+      "learning_rate": 8.38970137533116e-05,
+      "loss": 4.6361,
+      "step": 320750
+    },
+    {
+      "epoch": 2.7617079889807163,
+      "grad_norm": 2.96875,
+      "learning_rate": 8.387027587093208e-05,
+      "loss": 4.465,
+      "step": 320800
+    },
+    {
+      "epoch": 2.762138429752066,
+      "grad_norm": 3.8125,
+      "learning_rate": 8.384353917243961e-05,
+      "loss": 4.4795,
+      "step": 320850
+    },
+    {
+      "epoch": 2.762568870523416,
+      "grad_norm": 3.84375,
+      "learning_rate": 8.381680365979663e-05,
+      "loss": 4.6065,
+      "step": 320900
+    },
+    {
+      "epoch": 2.7629993112947657,
+      "grad_norm": 2.328125,
+      "learning_rate": 8.379006933496544e-05,
+      "loss": 4.1891,
+      "step": 320950
+    },
+    {
+      "epoch": 2.7634297520661155,
+      "grad_norm": 3.21875,
+      "learning_rate": 8.376333619990828e-05,
+      "loss": 4.2837,
+      "step": 321000
+    },
+    {
+      "epoch": 2.7634297520661155,
+      "eval_loss": 4.997475624084473,
+      "eval_runtime": 25.4314,
+      "eval_samples_per_second": 25.166,
+      "eval_steps_per_second": 12.583,
+      "eval_tts_loss": 7.414691866568385,
+      "step": 321000
+    },
+    {
+      "epoch": 2.7638601928374653,
+      "grad_norm": 2.3125,
+      "learning_rate": 8.373660425658728e-05,
+      "loss": 4.1129,
+      "step": 321050
+    },
+    {
+      "epoch": 2.7642906336088156,
+      "grad_norm": 3.703125,
+      "learning_rate": 8.370987350696456e-05,
+      "loss": 3.9958,
+      "step": 321100
+    },
+    {
+      "epoch": 2.7647210743801653,
+      "grad_norm": 1.984375,
+      "learning_rate": 8.368314395300206e-05,
+      "loss": 4.2864,
+      "step": 321150
+    },
+    {
+      "epoch": 2.765151515151515,
+      "grad_norm": 2.96875,
+      "learning_rate": 8.365641559666173e-05,
+      "loss": 4.3726,
+      "step": 321200
+    },
+    {
+      "epoch": 2.765581955922865,
+      "grad_norm": 3.96875,
+      "learning_rate": 8.362968843990528e-05,
+      "loss": 4.388,
+      "step": 321250
+    },
+    {
+      "epoch": 2.7660123966942147,
+      "grad_norm": 2.921875,
+      "learning_rate": 8.360296248469451e-05,
+      "loss": 4.1838,
+      "step": 321300
+    },
+    {
+      "epoch": 2.766442837465565,
+      "grad_norm": 2.625,
+      "learning_rate": 8.3576237732991e-05,
+      "loss": 4.2318,
+      "step": 321350
+    },
+    {
+      "epoch": 2.7668732782369148,
+      "grad_norm": 3.21875,
+      "learning_rate": 8.35495141867563e-05,
+      "loss": 4.6858,
+      "step": 321400
+    },
+    {
+      "epoch": 2.7673037190082646,
+      "grad_norm": 3.796875,
+      "learning_rate": 8.35227918479519e-05,
+      "loss": 3.9743,
+      "step": 321450
+    },
+    {
+      "epoch": 2.7677341597796143,
+      "grad_norm": 2.53125,
+      "learning_rate": 8.349607071853909e-05,
+      "loss": 3.9725,
+      "step": 321500
+    },
+    {
+      "epoch": 2.768164600550964,
+      "grad_norm": 2.828125,
+      "learning_rate": 8.346935080047917e-05,
+      "loss": 4.2781,
+      "step": 321550
+    },
+    {
+      "epoch": 2.768595041322314,
+      "grad_norm": 4.375,
+      "learning_rate": 8.344263209573336e-05,
+      "loss": 4.5143,
+      "step": 321600
+    },
+    {
+      "epoch": 2.7690254820936637,
+      "grad_norm": 2.703125,
+      "learning_rate": 8.34159146062627e-05,
+      "loss": 4.4649,
+      "step": 321650
+    },
+    {
+      "epoch": 2.7694559228650135,
+      "grad_norm": 3.921875,
+      "learning_rate": 8.338919833402829e-05,
+      "loss": 4.2749,
+      "step": 321700
+    },
+    {
+      "epoch": 2.7698863636363638,
+      "grad_norm": 5.0,
+      "learning_rate": 8.336248328099092e-05,
+      "loss": 4.374,
+      "step": 321750
+    },
+    {
+      "epoch": 2.7703168044077136,
+      "grad_norm": 2.796875,
+      "learning_rate": 8.333576944911151e-05,
+      "loss": 4.0905,
+      "step": 321800
+    },
+    {
+      "epoch": 2.7707472451790633,
+      "grad_norm": 3.015625,
+      "learning_rate": 8.330905684035076e-05,
+      "loss": 4.5605,
+      "step": 321850
+    },
+    {
+      "epoch": 2.771177685950413,
+      "grad_norm": 1.3515625,
+      "learning_rate": 8.328234545666931e-05,
+      "loss": 4.5177,
+      "step": 321900
+    },
+    {
+      "epoch": 2.771608126721763,
+      "grad_norm": 4.21875,
+      "learning_rate": 8.325563530002778e-05,
+      "loss": 4.6202,
+      "step": 321950
+    },
+    {
+      "epoch": 2.772038567493113,
+      "grad_norm": 4.125,
+      "learning_rate": 8.322892637238657e-05,
+      "loss": 4.573,
+      "step": 322000
+    },
+    {
+      "epoch": 2.772469008264463,
+      "grad_norm": 2.453125,
+      "learning_rate": 8.320221867570605e-05,
+      "loss": 4.1262,
+      "step": 322050
+    },
+    {
+      "epoch": 2.7728994490358128,
+      "grad_norm": 4.4375,
+      "learning_rate": 8.317551221194657e-05,
+      "loss": 4.2955,
+      "step": 322100
+    },
+    {
+      "epoch": 2.7733298898071626,
+      "grad_norm": 3.09375,
+      "learning_rate": 8.314880698306828e-05,
+      "loss": 4.0413,
+      "step": 322150
+    },
+    {
+      "epoch": 2.7737603305785123,
+      "grad_norm": 2.875,
+      "learning_rate": 8.312210299103133e-05,
+      "loss": 4.5331,
+      "step": 322200
+    },
+    {
+      "epoch": 2.774190771349862,
+      "grad_norm": 2.203125,
+      "learning_rate": 8.309540023779567e-05,
+      "loss": 4.3642,
+      "step": 322250
+    },
+    {
+      "epoch": 2.774621212121212,
+      "grad_norm": 3.1875,
+      "learning_rate": 8.306869872532129e-05,
+      "loss": 4.3883,
+      "step": 322300
+    },
+    {
+      "epoch": 2.7750516528925617,
+      "grad_norm": 4.84375,
+      "learning_rate": 8.304199845556797e-05,
+      "loss": 4.3788,
+      "step": 322350
+    },
+    {
+      "epoch": 2.775482093663912,
+      "grad_norm": 4.625,
+      "learning_rate": 8.301529943049547e-05,
+      "loss": 4.217,
+      "step": 322400
+    },
+    {
+      "epoch": 2.7759125344352618,
+      "grad_norm": 2.03125,
+      "learning_rate": 8.29886016520635e-05,
+      "loss": 4.4452,
+      "step": 322450
+    },
+    {
+      "epoch": 2.7763429752066116,
+      "grad_norm": 5.09375,
+      "learning_rate": 8.296190512223154e-05,
+      "loss": 3.9561,
+      "step": 322500
+    },
+    {
+      "epoch": 2.7767734159779613,
+      "grad_norm": 2.828125,
+      "learning_rate": 8.293520984295907e-05,
+      "loss": 4.6054,
+      "step": 322550
+    },
+    {
+      "epoch": 2.777203856749311,
+      "grad_norm": 1.8046875,
+      "learning_rate": 8.290851581620548e-05,
+      "loss": 4.2483,
+      "step": 322600
+    },
+    {
+      "epoch": 2.7776342975206614,
+      "grad_norm": 3.328125,
+      "learning_rate": 8.288182304393007e-05,
+      "loss": 4.6909,
+      "step": 322650
+    },
+    {
+      "epoch": 2.778064738292011,
+      "grad_norm": 4.4375,
+      "learning_rate": 8.2855131528092e-05,
+      "loss": 4.0004,
+      "step": 322700
+    },
+    {
+      "epoch": 2.778495179063361,
+      "grad_norm": 5.03125,
+      "learning_rate": 8.282844127065043e-05,
+      "loss": 4.2721,
+      "step": 322750
+    },
+    {
+      "epoch": 2.7789256198347108,
+      "grad_norm": 2.671875,
+      "learning_rate": 8.280175227356432e-05,
+      "loss": 4.1614,
+      "step": 322800
+    },
+    {
+      "epoch": 2.7793560606060606,
+      "grad_norm": 2.765625,
+      "learning_rate": 8.277506453879254e-05,
+      "loss": 4.5671,
+      "step": 322850
+    },
+    {
+      "epoch": 2.7797865013774103,
+      "grad_norm": 1.8046875,
+      "learning_rate": 8.274837806829398e-05,
+      "loss": 4.1722,
+      "step": 322900
+    },
+    {
+      "epoch": 2.78021694214876,
+      "grad_norm": 4.09375,
+      "learning_rate": 8.272169286402738e-05,
+      "loss": 4.592,
+      "step": 322950
+    },
+    {
+      "epoch": 2.78064738292011,
+      "grad_norm": 3.390625,
+      "learning_rate": 8.269500892795136e-05,
+      "loss": 4.2389,
+      "step": 323000
+    },
+    {
+      "epoch": 2.78107782369146,
+      "grad_norm": 3.265625,
+      "learning_rate": 8.266832626202441e-05,
+      "loss": 4.3429,
+      "step": 323050
+    },
+    {
+      "epoch": 2.78150826446281,
+      "grad_norm": 7.0,
+      "learning_rate": 8.264164486820502e-05,
+      "loss": 4.503,
+      "step": 323100
+    },
+    {
+      "epoch": 2.7819387052341598,
+      "grad_norm": 4.03125,
+      "learning_rate": 8.261496474845156e-05,
+      "loss": 4.3183,
+      "step": 323150
+    },
+    {
+      "epoch": 2.7823691460055096,
+      "grad_norm": 1.8125,
+      "learning_rate": 8.258828590472229e-05,
+      "loss": 4.2086,
+      "step": 323200
+    },
+    {
+      "epoch": 2.7827995867768593,
+      "grad_norm": 2.65625,
+      "learning_rate": 8.25616083389754e-05,
+      "loss": 4.4939,
+      "step": 323250
+    },
+    {
+      "epoch": 2.7832300275482096,
+      "grad_norm": 4.96875,
+      "learning_rate": 8.253493205316889e-05,
+      "loss": 4.1386,
+      "step": 323300
+    },
+    {
+      "epoch": 2.7836604683195594,
+      "grad_norm": 2.0,
+      "learning_rate": 8.250825704926081e-05,
+      "loss": 3.9069,
+      "step": 323350
+    },
+    {
+      "epoch": 2.784090909090909,
+      "grad_norm": 2.75,
+      "learning_rate": 8.248158332920902e-05,
+      "loss": 4.3003,
+      "step": 323400
+    },
+    {
+      "epoch": 2.784521349862259,
+      "grad_norm": 2.03125,
+      "learning_rate": 8.245491089497133e-05,
+      "loss": 4.5793,
+      "step": 323450
+    },
+    {
+      "epoch": 2.7849517906336088,
+      "grad_norm": 3.984375,
+      "learning_rate": 8.242823974850544e-05,
+      "loss": 4.0986,
+      "step": 323500
+    },
+    {
+      "epoch": 2.7853822314049586,
+      "grad_norm": 3.359375,
+      "learning_rate": 8.240156989176893e-05,
+      "loss": 4.2117,
+      "step": 323550
+    },
+    {
+      "epoch": 2.7858126721763083,
+      "grad_norm": 1.7421875,
+      "learning_rate": 8.237490132671932e-05,
+      "loss": 4.8066,
+      "step": 323600
+    },
+    {
+      "epoch": 2.786243112947658,
+      "grad_norm": 2.28125,
+      "learning_rate": 8.234823405531405e-05,
+      "loss": 4.4939,
+      "step": 323650
+    },
+    {
+      "epoch": 2.7866735537190084,
+      "grad_norm": 2.453125,
+      "learning_rate": 8.232156807951041e-05,
+      "loss": 4.3521,
+      "step": 323700
+    },
+    {
+      "epoch": 2.787103994490358,
+      "grad_norm": 1.1015625,
+      "learning_rate": 8.229490340126567e-05,
+      "loss": 4.1612,
+      "step": 323750
+    },
+    {
+      "epoch": 2.787534435261708,
+      "grad_norm": 1.2265625,
+      "learning_rate": 8.226824002253689e-05,
+      "loss": 3.9199,
+      "step": 323800
+    },
+    {
+      "epoch": 2.7879648760330578,
+      "grad_norm": 3.234375,
+      "learning_rate": 8.224157794528115e-05,
+      "loss": 4.5263,
+      "step": 323850
+    },
+    {
+      "epoch": 2.7883953168044076,
+      "grad_norm": 2.625,
+      "learning_rate": 8.221491717145537e-05,
+      "loss": 4.17,
+      "step": 323900
+    },
+    {
+      "epoch": 2.788825757575758,
+      "grad_norm": 2.203125,
+      "learning_rate": 8.218825770301639e-05,
+      "loss": 4.3201,
+      "step": 323950
+    },
+    {
+      "epoch": 2.7892561983471076,
+      "grad_norm": 4.21875,
+      "learning_rate": 8.216159954192103e-05,
+      "loss": 4.525,
+      "step": 324000
+    },
+    {
+      "epoch": 2.7892561983471076,
+      "eval_loss": 4.999444484710693,
+      "eval_runtime": 24.0622,
+      "eval_samples_per_second": 26.598,
+      "eval_steps_per_second": 13.299,
+      "eval_tts_loss": 7.436658738592688,
+      "step": 324000
+    },
+    {
+      "epoch": 2.7896866391184574,
+      "grad_norm": 3.40625,
+      "learning_rate": 8.213494269012585e-05,
+      "loss": 4.4447,
+      "step": 324050
+    },
+    {
+      "epoch": 2.790117079889807,
+      "grad_norm": 1.421875,
+      "learning_rate": 8.210828714958743e-05,
+      "loss": 4.5229,
+      "step": 324100
+    },
+    {
+      "epoch": 2.790547520661157,
+      "grad_norm": 1.9375,
+      "learning_rate": 8.208163292226225e-05,
+      "loss": 4.4509,
+      "step": 324150
+    },
+    {
+      "epoch": 2.7909779614325068,
+      "grad_norm": 3.15625,
+      "learning_rate": 8.205498001010665e-05,
+      "loss": 3.9698,
+      "step": 324200
+    },
+    {
+      "epoch": 2.7914084022038566,
+      "grad_norm": 3.71875,
+      "learning_rate": 8.202832841507693e-05,
+      "loss": 4.2365,
+      "step": 324250
+    },
+    {
+      "epoch": 2.7918388429752063,
+      "grad_norm": 3.390625,
+      "learning_rate": 8.200167813912922e-05,
+      "loss": 4.1722,
+      "step": 324300
+    },
+    {
+      "epoch": 2.7922692837465566,
+      "grad_norm": 3.703125,
+      "learning_rate": 8.197502918421962e-05,
+      "loss": 4.169,
+      "step": 324350
+    },
+    {
+      "epoch": 2.7926997245179064,
+      "grad_norm": 4.15625,
+      "learning_rate": 8.194838155230407e-05,
+      "loss": 4.5398,
+      "step": 324400
+    },
+    {
+      "epoch": 2.793130165289256,
+      "grad_norm": 2.671875,
+      "learning_rate": 8.192173524533848e-05,
+      "loss": 4.2018,
+      "step": 324450
+    },
+    {
+      "epoch": 2.793560606060606,
+      "grad_norm": 2.578125,
+      "learning_rate": 8.189509026527865e-05,
+      "loss": 3.9646,
+      "step": 324500
+    },
+    {
+      "epoch": 2.7939910468319558,
+      "grad_norm": 3.234375,
+      "learning_rate": 8.186844661408027e-05,
+      "loss": 4.381,
+      "step": 324550
+    },
+    {
+      "epoch": 2.794421487603306,
+      "grad_norm": 3.328125,
+      "learning_rate": 8.184180429369883e-05,
+      "loss": 4.3156,
+      "step": 324600
+    },
+    {
+      "epoch": 2.794851928374656,
+      "grad_norm": 1.7734375,
+      "learning_rate": 8.18151633060899e-05,
+      "loss": 4.1109,
+      "step": 324650
+    },
+    {
+      "epoch": 2.7952823691460056,
+      "grad_norm": 3.578125,
+      "learning_rate": 8.178852365320889e-05,
+      "loss": 4.4717,
+      "step": 324700
+    },
+    {
+      "epoch": 2.7957128099173554,
+      "grad_norm": 2.0625,
+      "learning_rate": 8.176188533701102e-05,
+      "loss": 4.3858,
+      "step": 324750
+    },
+    {
+      "epoch": 2.796143250688705,
+      "grad_norm": 3.421875,
+      "learning_rate": 8.173524835945159e-05,
+      "loss": 4.4907,
+      "step": 324800
+    },
+    {
+      "epoch": 2.796573691460055,
+      "grad_norm": 3.6875,
+      "learning_rate": 8.170861272248557e-05,
+      "loss": 4.3155,
+      "step": 324850
+    },
+    {
+      "epoch": 2.7970041322314048,
+      "grad_norm": 3.75,
+      "learning_rate": 8.168197842806805e-05,
+      "loss": 4.5615,
+      "step": 324900
+    },
+    {
+      "epoch": 2.7974345730027546,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.165534547815387e-05,
+      "loss": 4.0864,
+      "step": 324950
+    },
+    {
+      "epoch": 2.797865013774105,
+      "grad_norm": 2.96875,
+      "learning_rate": 8.16287138746979e-05,
+      "loss": 4.136,
+      "step": 325000
+    },
+    {
+      "epoch": 2.7982954545454546,
+      "grad_norm": 5.9375,
+      "learning_rate": 8.160208361965483e-05,
+      "loss": 4.412,
+      "step": 325050
+    },
+    {
+      "epoch": 2.7987258953168044,
+      "grad_norm": 1.6171875,
+      "learning_rate": 8.157545471497919e-05,
+      "loss": 4.2311,
+      "step": 325100
+    },
+    {
+      "epoch": 2.799156336088154,
+      "grad_norm": 2.109375,
+      "learning_rate": 8.154882716262552e-05,
+      "loss": 4.1473,
+      "step": 325150
+    },
+    {
+      "epoch": 2.799586776859504,
+      "grad_norm": 3.21875,
+      "learning_rate": 8.152220096454826e-05,
+      "loss": 3.4352,
+      "step": 325200
+    },
+    {
+      "epoch": 2.800017217630854,
+      "grad_norm": 3.78125,
+      "learning_rate": 8.149557612270169e-05,
+      "loss": 3.9129,
+      "step": 325250
+    },
+    {
+      "epoch": 2.800447658402204,
+      "grad_norm": 0.6484375,
+      "learning_rate": 8.146895263904004e-05,
+      "loss": 4.5044,
+      "step": 325300
+    },
+    {
+      "epoch": 2.800878099173554,
+      "grad_norm": 4.0625,
+      "learning_rate": 8.144233051551738e-05,
+      "loss": 4.4709,
+      "step": 325350
+    },
+    {
+      "epoch": 2.8013085399449036,
+      "grad_norm": 3.234375,
+      "learning_rate": 8.141570975408773e-05,
+      "loss": 4.1304,
+      "step": 325400
+    },
+    {
+      "epoch": 2.8017389807162534,
+      "grad_norm": 4.09375,
+      "learning_rate": 8.138909035670499e-05,
+      "loss": 4.0429,
+      "step": 325450
+    },
+    {
+      "epoch": 2.802169421487603,
+      "grad_norm": 4.0,
+      "learning_rate": 8.136247232532298e-05,
+      "loss": 4.4678,
+      "step": 325500
+    },
+    {
+      "epoch": 2.802599862258953,
+      "grad_norm": 4.25,
+      "learning_rate": 8.133585566189542e-05,
+      "loss": 4.4711,
+      "step": 325550
+    },
+    {
+      "epoch": 2.8030303030303028,
+      "grad_norm": 4.96875,
+      "learning_rate": 8.13092403683759e-05,
+      "loss": 4.771,
+      "step": 325600
+    },
+    {
+      "epoch": 2.803460743801653,
+      "grad_norm": 0.89453125,
+      "learning_rate": 8.128262644671788e-05,
+      "loss": 4.1548,
+      "step": 325650
+    },
+    {
+      "epoch": 2.803891184573003,
+      "grad_norm": 2.125,
+      "learning_rate": 8.125601389887484e-05,
+      "loss": 4.255,
+      "step": 325700
+    },
+    {
+      "epoch": 2.8043216253443526,
+      "grad_norm": 3.609375,
+      "learning_rate": 8.122940272680005e-05,
+      "loss": 4.3555,
+      "step": 325750
+    },
+    {
+      "epoch": 2.8047520661157024,
+      "grad_norm": 4.25,
+      "learning_rate": 8.120279293244673e-05,
+      "loss": 4.3582,
+      "step": 325800
+    },
+    {
+      "epoch": 2.8051825068870526,
+      "grad_norm": 2.296875,
+      "learning_rate": 8.117618451776792e-05,
+      "loss": 4.4223,
+      "step": 325850
+    },
+    {
+      "epoch": 2.8056129476584024,
+      "grad_norm": 3.171875,
+      "learning_rate": 8.11495774847167e-05,
+      "loss": 4.4934,
+      "step": 325900
+    },
+    {
+      "epoch": 2.806043388429752,
+      "grad_norm": 4.71875,
+      "learning_rate": 8.11229718352459e-05,
+      "loss": 4.4963,
+      "step": 325950
+    },
+    {
+      "epoch": 2.806473829201102,
+      "grad_norm": 2.40625,
+      "learning_rate": 8.109636757130838e-05,
+      "loss": 4.4024,
+      "step": 326000
+    },
+    {
+      "epoch": 2.806904269972452,
+      "grad_norm": 4.53125,
+      "learning_rate": 8.106976469485683e-05,
+      "loss": 4.5773,
+      "step": 326050
+    },
+    {
+      "epoch": 2.8073347107438016,
+      "grad_norm": 4.25,
+      "learning_rate": 8.104316320784381e-05,
+      "loss": 4.3996,
+      "step": 326100
+    },
+    {
+      "epoch": 2.8077651515151514,
+      "grad_norm": 3.484375,
+      "learning_rate": 8.10165631122218e-05,
+      "loss": 4.6319,
+      "step": 326150
+    },
+    {
+      "epoch": 2.808195592286501,
+      "grad_norm": 3.765625,
+      "learning_rate": 8.09899644099432e-05,
+      "loss": 4.1598,
+      "step": 326200
+    },
+    {
+      "epoch": 2.808626033057851,
+      "grad_norm": 3.90625,
+      "learning_rate": 8.096336710296035e-05,
+      "loss": 4.3487,
+      "step": 326250
+    },
+    {
+      "epoch": 2.809056473829201,
+      "grad_norm": 1.34375,
+      "learning_rate": 8.093677119322535e-05,
+      "loss": 4.4717,
+      "step": 326300
+    },
+    {
+      "epoch": 2.809486914600551,
+      "grad_norm": 3.5,
+      "learning_rate": 8.09101766826904e-05,
+      "loss": 4.333,
+      "step": 326350
+    },
+    {
+      "epoch": 2.809917355371901,
+      "grad_norm": 2.03125,
+      "learning_rate": 8.088358357330739e-05,
+      "loss": 4.4636,
+      "step": 326400
+    },
+    {
+      "epoch": 2.8103477961432506,
+      "grad_norm": 1.5859375,
+      "learning_rate": 8.085699186702818e-05,
+      "loss": 4.2457,
+      "step": 326450
+    },
+    {
+      "epoch": 2.810778236914601,
+      "grad_norm": 3.9375,
+      "learning_rate": 8.083040156580457e-05,
+      "loss": 4.3138,
+      "step": 326500
+    },
+    {
+      "epoch": 2.8112086776859506,
+      "grad_norm": 3.671875,
+      "learning_rate": 8.080381267158827e-05,
+      "loss": 4.4918,
+      "step": 326550
+    },
+    {
+      "epoch": 2.8116391184573004,
+      "grad_norm": 3.65625,
+      "learning_rate": 8.077722518633085e-05,
+      "loss": 4.5321,
+      "step": 326600
+    },
+    {
+      "epoch": 2.81206955922865,
+      "grad_norm": 7.03125,
+      "learning_rate": 8.07506391119837e-05,
+      "loss": 4.3943,
+      "step": 326650
+    },
+    {
+      "epoch": 2.8125,
+      "grad_norm": 1.8984375,
+      "learning_rate": 8.072405445049821e-05,
+      "loss": 4.6043,
+      "step": 326700
+    },
+    {
+      "epoch": 2.81293044077135,
+      "grad_norm": 2.984375,
+      "learning_rate": 8.069747120382568e-05,
+      "loss": 4.4846,
+      "step": 326750
+    },
+    {
+      "epoch": 2.8133608815426996,
+      "grad_norm": 4.96875,
+      "learning_rate": 8.06708893739172e-05,
+      "loss": 4.2954,
+      "step": 326800
+    },
+    {
+      "epoch": 2.8137913223140494,
+      "grad_norm": 3.46875,
+      "learning_rate": 8.064430896272392e-05,
+      "loss": 4.4399,
+      "step": 326850
+    },
+    {
+      "epoch": 2.814221763085399,
+      "grad_norm": 2.859375,
+      "learning_rate": 8.061772997219665e-05,
+      "loss": 4.3473,
+      "step": 326900
+    },
+    {
+      "epoch": 2.8146522038567494,
+      "grad_norm": 3.625,
+      "learning_rate": 8.059115240428634e-05,
+      "loss": 4.1814,
+      "step": 326950
+    },
+    {
+      "epoch": 2.815082644628099,
+      "grad_norm": 3.671875,
+      "learning_rate": 8.056457626094364e-05,
+      "loss": 4.4147,
+      "step": 327000
+    },
+    {
+      "epoch": 2.815082644628099,
+      "eval_loss": 4.998778343200684,
+      "eval_runtime": 24.2422,
+      "eval_samples_per_second": 26.4,
+      "eval_steps_per_second": 13.2,
+      "eval_tts_loss": 7.420693159245713,
+      "step": 327000
+    },
+    {
+      "epoch": 2.815513085399449,
+      "grad_norm": 3.6875,
+      "learning_rate": 8.053800154411925e-05,
+      "loss": 4.2546,
+      "step": 327050
+    },
+    {
+      "epoch": 2.815943526170799,
+      "grad_norm": 3.296875,
+      "learning_rate": 8.051142825576369e-05,
+      "loss": 4.4716,
+      "step": 327100
+    },
+    {
+      "epoch": 2.816373966942149,
+      "grad_norm": 2.125,
+      "learning_rate": 8.048485639782735e-05,
+      "loss": 4.811,
+      "step": 327150
+    },
+    {
+      "epoch": 2.816804407713499,
+      "grad_norm": 5.0625,
+      "learning_rate": 8.045828597226056e-05,
+      "loss": 4.6568,
+      "step": 327200
+    },
+    {
+      "epoch": 2.8172348484848486,
+      "grad_norm": 5.28125,
+      "learning_rate": 8.043171698101354e-05,
+      "loss": 4.1752,
+      "step": 327250
+    },
+    {
+      "epoch": 2.8176652892561984,
+      "grad_norm": 3.859375,
+      "learning_rate": 8.040514942603636e-05,
+      "loss": 4.2854,
+      "step": 327300
+    },
+    {
+      "epoch": 2.818095730027548,
+      "grad_norm": 2.484375,
+      "learning_rate": 8.037858330927912e-05,
+      "loss": 4.3418,
+      "step": 327350
+    },
+    {
+      "epoch": 2.818526170798898,
+      "grad_norm": 1.2109375,
+      "learning_rate": 8.03520186326916e-05,
+      "loss": 4.0692,
+      "step": 327400
+    },
+    {
+      "epoch": 2.818956611570248,
+      "grad_norm": 3.25,
+      "learning_rate": 8.032545539822365e-05,
+      "loss": 4.7863,
+      "step": 327450
+    },
+    {
+      "epoch": 2.8193870523415976,
+      "grad_norm": 2.46875,
+      "learning_rate": 8.029889360782493e-05,
+      "loss": 4.0983,
+      "step": 327500
+    },
+    {
+      "epoch": 2.8198174931129474,
+      "grad_norm": 5.5625,
+      "learning_rate": 8.027233326344502e-05,
+      "loss": 4.908,
+      "step": 327550
+    },
+    {
+      "epoch": 2.8202479338842976,
+      "grad_norm": 3.34375,
+      "learning_rate": 8.024577436703346e-05,
+      "loss": 4.3652,
+      "step": 327600
+    },
+    {
+      "epoch": 2.8206783746556474,
+      "grad_norm": 2.0625,
+      "learning_rate": 8.021921692053953e-05,
+      "loss": 4.3827,
+      "step": 327650
+    },
+    {
+      "epoch": 2.821108815426997,
+      "grad_norm": 5.0625,
+      "learning_rate": 8.01926609259125e-05,
+      "loss": 4.5317,
+      "step": 327700
+    },
+    {
+      "epoch": 2.821539256198347,
+      "grad_norm": 3.140625,
+      "learning_rate": 8.016610638510154e-05,
+      "loss": 4.5239,
+      "step": 327750
+    },
+    {
+      "epoch": 2.8219696969696972,
+      "grad_norm": 1.890625,
+      "learning_rate": 8.013955330005571e-05,
+      "loss": 4.1893,
+      "step": 327800
+    },
+    {
+      "epoch": 2.822400137741047,
+      "grad_norm": 4.0,
+      "learning_rate": 8.011300167272398e-05,
+      "loss": 4.3599,
+      "step": 327850
+    },
+    {
+      "epoch": 2.822830578512397,
+      "grad_norm": 3.03125,
+      "learning_rate": 8.008645150505507e-05,
+      "loss": 3.8317,
+      "step": 327900
+    },
+    {
+      "epoch": 2.8232610192837466,
+      "grad_norm": 2.546875,
+      "learning_rate": 8.005990279899782e-05,
+      "loss": 4.2058,
+      "step": 327950
+    },
+    {
+      "epoch": 2.8236914600550964,
+      "grad_norm": 2.875,
+      "learning_rate": 8.003335555650077e-05,
+      "loss": 4.5639,
+      "step": 328000
+    },
+    {
+      "epoch": 2.824121900826446,
+      "grad_norm": 2.53125,
+      "learning_rate": 8.000680977951245e-05,
+      "loss": 4.4624,
+      "step": 328050
+    },
+    {
+      "epoch": 2.824552341597796,
+      "grad_norm": 7.5625,
+      "learning_rate": 7.998026546998134e-05,
+      "loss": 4.2624,
+      "step": 328100
+    },
+    {
+      "epoch": 2.824982782369146,
+      "grad_norm": 2.265625,
+      "learning_rate": 7.995372262985564e-05,
+      "loss": 4.0531,
+      "step": 328150
+    },
+    {
+      "epoch": 2.825413223140496,
+      "grad_norm": 1.0,
+      "learning_rate": 7.992718126108354e-05,
+      "loss": 3.8754,
+      "step": 328200
+    },
+    {
+      "epoch": 2.825843663911846,
+      "grad_norm": 3.078125,
+      "learning_rate": 7.990064136561317e-05,
+      "loss": 4.5383,
+      "step": 328250
+    },
+    {
+      "epoch": 2.8262741046831956,
+      "grad_norm": 3.0,
+      "learning_rate": 7.987410294539249e-05,
+      "loss": 4.5454,
+      "step": 328300
+    },
+    {
+      "epoch": 2.8267045454545454,
+      "grad_norm": 2.96875,
+      "learning_rate": 7.984756600236932e-05,
+      "loss": 4.6176,
+      "step": 328350
+    },
+    {
+      "epoch": 2.827134986225895,
+      "grad_norm": 3.46875,
+      "learning_rate": 7.98210305384915e-05,
+      "loss": 4.4,
+      "step": 328400
+    },
+    {
+      "epoch": 2.8275654269972454,
+      "grad_norm": 3.140625,
+      "learning_rate": 7.979449655570656e-05,
+      "loss": 4.4922,
+      "step": 328450
+    },
+    {
+      "epoch": 2.8279958677685952,
+      "grad_norm": 7.5625,
+      "learning_rate": 7.976796405596215e-05,
+      "loss": 4.6449,
+      "step": 328500
+    },
+    {
+      "epoch": 2.828426308539945,
+      "grad_norm": 3.65625,
+      "learning_rate": 7.974143304120562e-05,
+      "loss": 4.5647,
+      "step": 328550
+    },
+    {
+      "epoch": 2.828856749311295,
+      "grad_norm": 2.203125,
+      "learning_rate": 7.971490351338432e-05,
+      "loss": 4.4015,
+      "step": 328600
+    },
+    {
+      "epoch": 2.8292871900826446,
+      "grad_norm": 3.28125,
+      "learning_rate": 7.96883754744455e-05,
+      "loss": 4.1724,
+      "step": 328650
+    },
+    {
+      "epoch": 2.8297176308539944,
+      "grad_norm": 2.046875,
+      "learning_rate": 7.966184892633616e-05,
+      "loss": 4.0353,
+      "step": 328700
+    },
+    {
+      "epoch": 2.830148071625344,
+      "grad_norm": 3.171875,
+      "learning_rate": 7.963532387100336e-05,
+      "loss": 4.069,
+      "step": 328750
+    },
+    {
+      "epoch": 2.830578512396694,
+      "grad_norm": 1.921875,
+      "learning_rate": 7.960880031039399e-05,
+      "loss": 4.1332,
+      "step": 328800
+    },
+    {
+      "epoch": 2.8310089531680442,
+      "grad_norm": 3.21875,
+      "learning_rate": 7.958227824645478e-05,
+      "loss": 4.1883,
+      "step": 328850
+    },
+    {
+      "epoch": 2.831439393939394,
+      "grad_norm": 2.625,
+      "learning_rate": 7.955575768113245e-05,
+      "loss": 4.0097,
+      "step": 328900
+    },
+    {
+      "epoch": 2.831869834710744,
+      "grad_norm": 4.6875,
+      "learning_rate": 7.95292386163735e-05,
+      "loss": 4.3936,
+      "step": 328950
+    },
+    {
+      "epoch": 2.8323002754820936,
+      "grad_norm": 3.234375,
+      "learning_rate": 7.950272105412441e-05,
+      "loss": 4.3752,
+      "step": 329000
+    },
+    {
+      "epoch": 2.8327307162534434,
+      "grad_norm": 3.96875,
+      "learning_rate": 7.947620499633146e-05,
+      "loss": 4.2651,
+      "step": 329050
+    },
+    {
+      "epoch": 2.8331611570247937,
+      "grad_norm": 3.5,
+      "learning_rate": 7.94496904449409e-05,
+      "loss": 4.508,
+      "step": 329100
+    },
+    {
+      "epoch": 2.8335915977961434,
+      "grad_norm": 2.453125,
+      "learning_rate": 7.94231774018989e-05,
+      "loss": 4.187,
+      "step": 329150
+    },
+    {
+      "epoch": 2.8340220385674932,
+      "grad_norm": 1.875,
+      "learning_rate": 7.939666586915139e-05,
+      "loss": 4.3722,
+      "step": 329200
+    },
+    {
+      "epoch": 2.834452479338843,
+      "grad_norm": 3.453125,
+      "learning_rate": 7.937015584864425e-05,
+      "loss": 4.2752,
+      "step": 329250
+    },
+    {
+      "epoch": 2.834882920110193,
+      "grad_norm": 4.5625,
+      "learning_rate": 7.93436473423233e-05,
+      "loss": 4.2452,
+      "step": 329300
+    },
+    {
+      "epoch": 2.8353133608815426,
+      "grad_norm": 3.1875,
+      "learning_rate": 7.931714035213418e-05,
+      "loss": 4.1229,
+      "step": 329350
+    },
+    {
+      "epoch": 2.8357438016528924,
+      "grad_norm": 3.46875,
+      "learning_rate": 7.929063488002249e-05,
+      "loss": 4.189,
+      "step": 329400
+    },
+    {
+      "epoch": 2.836174242424242,
+      "grad_norm": 2.453125,
+      "learning_rate": 7.926413092793361e-05,
+      "loss": 4.4492,
+      "step": 329450
+    },
+    {
+      "epoch": 2.8366046831955924,
+      "grad_norm": 8.375,
+      "learning_rate": 7.923762849781292e-05,
+      "loss": 4.411,
+      "step": 329500
+    },
+    {
+      "epoch": 2.8370351239669422,
+      "grad_norm": 3.234375,
+      "learning_rate": 7.92111275916056e-05,
+      "loss": 4.5448,
+      "step": 329550
+    },
+    {
+      "epoch": 2.837465564738292,
+      "grad_norm": 4.09375,
+      "learning_rate": 7.918462821125681e-05,
+      "loss": 4.248,
+      "step": 329600
+    },
+    {
+      "epoch": 2.837896005509642,
+      "grad_norm": 2.75,
+      "learning_rate": 7.915813035871154e-05,
+      "loss": 4.1487,
+      "step": 329650
+    },
+    {
+      "epoch": 2.8383264462809916,
+      "grad_norm": 3.5,
+      "learning_rate": 7.913163403591464e-05,
+      "loss": 4.2946,
+      "step": 329700
+    },
+    {
+      "epoch": 2.838756887052342,
+      "grad_norm": 3.078125,
+      "learning_rate": 7.91051392448109e-05,
+      "loss": 4.5097,
+      "step": 329750
+    },
+    {
+      "epoch": 2.8391873278236917,
+      "grad_norm": 3.171875,
+      "learning_rate": 7.907864598734497e-05,
+      "loss": 4.3428,
+      "step": 329800
+    },
+    {
+      "epoch": 2.8396177685950414,
+      "grad_norm": 4.09375,
+      "learning_rate": 7.905215426546144e-05,
+      "loss": 4.3565,
+      "step": 329850
+    },
+    {
+      "epoch": 2.8400482093663912,
+      "grad_norm": 3.4375,
+      "learning_rate": 7.902566408110472e-05,
+      "loss": 4.2442,
+      "step": 329900
+    },
+    {
+      "epoch": 2.840478650137741,
+      "grad_norm": 3.53125,
+      "learning_rate": 7.899917543621911e-05,
+      "loss": 3.9229,
+      "step": 329950
+    },
+    {
+      "epoch": 2.840909090909091,
+      "grad_norm": 3.640625,
+      "learning_rate": 7.897268833274885e-05,
+      "loss": 4.2746,
+      "step": 330000
+    },
+    {
+      "epoch": 2.840909090909091,
+      "eval_loss": 4.996027946472168,
+      "eval_runtime": 24.5372,
+      "eval_samples_per_second": 26.083,
+      "eval_steps_per_second": 13.041,
+      "eval_tts_loss": 7.445830441324012,
+      "step": 330000
+    },
+    {
+      "epoch": 2.8413395316804406,
+      "grad_norm": 2.015625,
+      "learning_rate": 7.8946202772638e-05,
+      "loss": 4.1043,
+      "step": 330050
+    },
+    {
+      "epoch": 2.8417699724517904,
+      "grad_norm": 2.8125,
+      "learning_rate": 7.891971875783056e-05,
+      "loss": 4.4002,
+      "step": 330100
+    },
+    {
+      "epoch": 2.8422004132231407,
+      "grad_norm": 1.9140625,
+      "learning_rate": 7.889323629027045e-05,
+      "loss": 4.2002,
+      "step": 330150
+    },
+    {
+      "epoch": 2.8426308539944904,
+      "grad_norm": 3.046875,
+      "learning_rate": 7.88667553719014e-05,
+      "loss": 4.1802,
+      "step": 330200
+    },
+    {
+      "epoch": 2.8430612947658402,
+      "grad_norm": 3.4375,
+      "learning_rate": 7.884027600466698e-05,
+      "loss": 4.3297,
+      "step": 330250
+    },
+    {
+      "epoch": 2.84349173553719,
+      "grad_norm": 3.828125,
+      "learning_rate": 7.881379819051078e-05,
+      "loss": 4.3222,
+      "step": 330300
+    },
+    {
+      "epoch": 2.84392217630854,
+      "grad_norm": 4.25,
+      "learning_rate": 7.878732193137624e-05,
+      "loss": 4.2271,
+      "step": 330350
+    },
+    {
+      "epoch": 2.84435261707989,
+      "grad_norm": 3.90625,
+      "learning_rate": 7.87608472292066e-05,
+      "loss": 4.4208,
+      "step": 330400
+    },
+    {
+      "epoch": 2.84478305785124,
+      "grad_norm": 3.921875,
+      "learning_rate": 7.873437408594512e-05,
+      "loss": 4.3129,
+      "step": 330450
+    },
+    {
+      "epoch": 2.8452134986225897,
+      "grad_norm": 3.6875,
+      "learning_rate": 7.870790250353479e-05,
+      "loss": 3.6925,
+      "step": 330500
+    },
+    {
+      "epoch": 2.8456439393939394,
+      "grad_norm": 2.140625,
+      "learning_rate": 7.868143248391862e-05,
+      "loss": 4.0339,
+      "step": 330550
+    },
+    {
+      "epoch": 2.8460743801652892,
+      "grad_norm": 5.15625,
+      "learning_rate": 7.865496402903943e-05,
+      "loss": 4.3239,
+      "step": 330600
+    },
+    {
+      "epoch": 2.846504820936639,
+      "grad_norm": 2.453125,
+      "learning_rate": 7.862849714083995e-05,
+      "loss": 4.1416,
+      "step": 330650
+    },
+    {
+      "epoch": 2.846935261707989,
+      "grad_norm": 3.40625,
+      "learning_rate": 7.860203182126282e-05,
+      "loss": 4.4766,
+      "step": 330700
+    },
+    {
+      "epoch": 2.8473657024793386,
+      "grad_norm": 1.9765625,
+      "learning_rate": 7.857556807225051e-05,
+      "loss": 4.2611,
+      "step": 330750
+    },
+    {
+      "epoch": 2.847796143250689,
+      "grad_norm": 3.4375,
+      "learning_rate": 7.854910589574538e-05,
+      "loss": 4.422,
+      "step": 330800
+    },
+    {
+      "epoch": 2.8482265840220387,
+      "grad_norm": 2.765625,
+      "learning_rate": 7.852264529368976e-05,
+      "loss": 4.1691,
+      "step": 330850
+    },
+    {
+      "epoch": 2.8486570247933884,
+      "grad_norm": 1.578125,
+      "learning_rate": 7.849618626802573e-05,
+      "loss": 4.2793,
+      "step": 330900
+    },
+    {
+      "epoch": 2.8490874655647382,
+      "grad_norm": 4.1875,
+      "learning_rate": 7.846972882069542e-05,
+      "loss": 4.3258,
+      "step": 330950
+    },
+    {
+      "epoch": 2.849517906336088,
+      "grad_norm": 3.125,
+      "learning_rate": 7.844327295364063e-05,
+      "loss": 4.4516,
+      "step": 331000
+    },
+    {
+      "epoch": 2.8499483471074383,
+      "grad_norm": 3.546875,
+      "learning_rate": 7.841681866880324e-05,
+      "loss": 4.3668,
+      "step": 331050
+    },
+    {
+      "epoch": 2.850378787878788,
+      "grad_norm": 3.3125,
+      "learning_rate": 7.839036596812492e-05,
+      "loss": 4.2713,
+      "step": 331100
+    },
+    {
+      "epoch": 2.850809228650138,
+      "grad_norm": 7.0625,
+      "learning_rate": 7.836391485354724e-05,
+      "loss": 4.019,
+      "step": 331150
+    },
+    {
+      "epoch": 2.8512396694214877,
+      "grad_norm": 4.21875,
+      "learning_rate": 7.833746532701171e-05,
+      "loss": 4.4856,
+      "step": 331200
+    },
+    {
+      "epoch": 2.8516701101928374,
+      "grad_norm": 3.515625,
+      "learning_rate": 7.831101739045959e-05,
+      "loss": 3.9797,
+      "step": 331250
+    },
+    {
+      "epoch": 2.8521005509641872,
+      "grad_norm": 5.96875,
+      "learning_rate": 7.828457104583209e-05,
+      "loss": 4.4868,
+      "step": 331300
+    },
+    {
+      "epoch": 2.852530991735537,
+      "grad_norm": 2.046875,
+      "learning_rate": 7.825812629507037e-05,
+      "loss": 4.5017,
+      "step": 331350
+    },
+    {
+      "epoch": 2.852961432506887,
+      "grad_norm": 3.125,
+      "learning_rate": 7.823168314011542e-05,
+      "loss": 4.5807,
+      "step": 331400
+    },
+    {
+      "epoch": 2.853391873278237,
+      "grad_norm": 2.59375,
+      "learning_rate": 7.820524158290811e-05,
+      "loss": 4.3126,
+      "step": 331450
+    },
+    {
+      "epoch": 2.853822314049587,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.817880162538912e-05,
+      "loss": 4.4338,
+      "step": 331500
+    },
+    {
+      "epoch": 2.8542527548209367,
+      "grad_norm": 1.34375,
+      "learning_rate": 7.815236326949919e-05,
+      "loss": 4.0954,
+      "step": 331550
+    },
+    {
+      "epoch": 2.8546831955922864,
+      "grad_norm": 4.25,
+      "learning_rate": 7.812592651717876e-05,
+      "loss": 4.3199,
+      "step": 331600
+    },
+    {
+      "epoch": 2.8551136363636362,
+      "grad_norm": 2.328125,
+      "learning_rate": 7.809949137036825e-05,
+      "loss": 4.1058,
+      "step": 331650
+    },
+    {
+      "epoch": 2.8555440771349865,
+      "grad_norm": 1.90625,
+      "learning_rate": 7.8073057831008e-05,
+      "loss": 4.442,
+      "step": 331700
+    },
+    {
+      "epoch": 2.8559745179063363,
+      "grad_norm": 1.34375,
+      "learning_rate": 7.80466259010381e-05,
+      "loss": 4.6728,
+      "step": 331750
+    },
+    {
+      "epoch": 2.856404958677686,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.802019558239861e-05,
+      "loss": 3.9543,
+      "step": 331800
+    },
+    {
+      "epoch": 2.856835399449036,
+      "grad_norm": 2.53125,
+      "learning_rate": 7.799376687702949e-05,
+      "loss": 4.1872,
+      "step": 331850
+    },
+    {
+      "epoch": 2.8572658402203857,
+      "grad_norm": 4.71875,
+      "learning_rate": 7.796733978687054e-05,
+      "loss": 4.2554,
+      "step": 331900
+    },
+    {
+      "epoch": 2.8576962809917354,
+      "grad_norm": 3.40625,
+      "learning_rate": 7.794091431386141e-05,
+      "loss": 4.0566,
+      "step": 331950
+    },
+    {
+      "epoch": 2.8581267217630852,
+      "grad_norm": 2.390625,
+      "learning_rate": 7.791449045994177e-05,
+      "loss": 4.1505,
+      "step": 332000
+    },
+    {
+      "epoch": 2.858557162534435,
+      "grad_norm": 2.90625,
+      "learning_rate": 7.788806822705097e-05,
+      "loss": 4.206,
+      "step": 332050
+    },
+    {
+      "epoch": 2.8589876033057853,
+      "grad_norm": 2.421875,
+      "learning_rate": 7.78616476171284e-05,
+      "loss": 4.3296,
+      "step": 332100
+    },
+    {
+      "epoch": 2.859418044077135,
+      "grad_norm": 3.4375,
+      "learning_rate": 7.783522863211324e-05,
+      "loss": 4.6101,
+      "step": 332150
+    },
+    {
+      "epoch": 2.859848484848485,
+      "grad_norm": 5.28125,
+      "learning_rate": 7.780881127394464e-05,
+      "loss": 4.1777,
+      "step": 332200
+    },
+    {
+      "epoch": 2.8602789256198347,
+      "grad_norm": 2.484375,
+      "learning_rate": 7.778239554456156e-05,
+      "loss": 4.3672,
+      "step": 332250
+    },
+    {
+      "epoch": 2.8607093663911844,
+      "grad_norm": 3.28125,
+      "learning_rate": 7.775598144590281e-05,
+      "loss": 4.2635,
+      "step": 332300
+    },
+    {
+      "epoch": 2.8611398071625347,
+      "grad_norm": 3.515625,
+      "learning_rate": 7.772956897990715e-05,
+      "loss": 4.4765,
+      "step": 332350
+    },
+    {
+      "epoch": 2.8615702479338845,
+      "grad_norm": 3.8125,
+      "learning_rate": 7.770315814851324e-05,
+      "loss": 4.271,
+      "step": 332400
+    },
+    {
+      "epoch": 2.8620006887052343,
+      "grad_norm": 2.890625,
+      "learning_rate": 7.767674895365953e-05,
+      "loss": 4.1485,
+      "step": 332450
+    },
+    {
+      "epoch": 2.862431129476584,
+      "grad_norm": 3.0625,
+      "learning_rate": 7.765034139728446e-05,
+      "loss": 4.219,
+      "step": 332500
+    },
+    {
+      "epoch": 2.862861570247934,
+      "grad_norm": 2.734375,
+      "learning_rate": 7.762393548132619e-05,
+      "loss": 4.3381,
+      "step": 332550
+    },
+    {
+      "epoch": 2.8632920110192837,
+      "grad_norm": 4.625,
+      "learning_rate": 7.759753120772295e-05,
+      "loss": 4.2912,
+      "step": 332600
+    },
+    {
+      "epoch": 2.8637224517906334,
+      "grad_norm": 3.3125,
+      "learning_rate": 7.75711285784127e-05,
+      "loss": 4.348,
+      "step": 332650
+    },
+    {
+      "epoch": 2.8641528925619832,
+      "grad_norm": 2.34375,
+      "learning_rate": 7.754472759533334e-05,
+      "loss": 4.2607,
+      "step": 332700
+    },
+    {
+      "epoch": 2.8645833333333335,
+      "grad_norm": 3.546875,
+      "learning_rate": 7.751832826042271e-05,
+      "loss": 3.9201,
+      "step": 332750
+    },
+    {
+      "epoch": 2.8650137741046833,
+      "grad_norm": 3.296875,
+      "learning_rate": 7.749193057561839e-05,
+      "loss": 4.6225,
+      "step": 332800
+    },
+    {
+      "epoch": 2.865444214876033,
+      "grad_norm": 2.421875,
+      "learning_rate": 7.746553454285793e-05,
+      "loss": 4.6142,
+      "step": 332850
+    },
+    {
+      "epoch": 2.865874655647383,
+      "grad_norm": 4.28125,
+      "learning_rate": 7.743914016407875e-05,
+      "loss": 4.2471,
+      "step": 332900
+    },
+    {
+      "epoch": 2.8663050964187327,
+      "grad_norm": 2.0,
+      "learning_rate": 7.741274744121814e-05,
+      "loss": 4.7241,
+      "step": 332950
+    },
+    {
+      "epoch": 2.866735537190083,
+      "grad_norm": 2.4375,
+      "learning_rate": 7.73863563762133e-05,
+      "loss": 3.8558,
+      "step": 333000
+    },
+    {
+      "epoch": 2.866735537190083,
+      "eval_loss": 4.995693206787109,
+      "eval_runtime": 24.1386,
+      "eval_samples_per_second": 26.514,
+      "eval_steps_per_second": 13.257,
+      "eval_tts_loss": 7.401987731875897,
+      "step": 333000
+    },
+    {
+      "epoch": 2.8671659779614327,
+      "grad_norm": 2.140625,
+      "learning_rate": 7.735996697100121e-05,
+      "loss": 4.163,
+      "step": 333050
+    },
+    {
+      "epoch": 2.8675964187327825,
+      "grad_norm": 2.484375,
+      "learning_rate": 7.733357922751885e-05,
+      "loss": 4.6496,
+      "step": 333100
+    },
+    {
+      "epoch": 2.8680268595041323,
+      "grad_norm": 3.03125,
+      "learning_rate": 7.730719314770299e-05,
+      "loss": 4.5372,
+      "step": 333150
+    },
+    {
+      "epoch": 2.868457300275482,
+      "grad_norm": 4.5625,
+      "learning_rate": 7.728080873349032e-05,
+      "loss": 4.3275,
+      "step": 333200
+    },
+    {
+      "epoch": 2.868887741046832,
+      "grad_norm": 3.046875,
+      "learning_rate": 7.725442598681746e-05,
+      "loss": 4.3856,
+      "step": 333250
+    },
+    {
+      "epoch": 2.8693181818181817,
+      "grad_norm": 7.375,
+      "learning_rate": 7.722804490962078e-05,
+      "loss": 4.099,
+      "step": 333300
+    },
+    {
+      "epoch": 2.8697486225895315,
+      "grad_norm": 2.78125,
+      "learning_rate": 7.720166550383658e-05,
+      "loss": 4.1062,
+      "step": 333350
+    },
+    {
+      "epoch": 2.8701790633608817,
+      "grad_norm": 2.90625,
+      "learning_rate": 7.717528777140107e-05,
+      "loss": 4.503,
+      "step": 333400
+    },
+    {
+      "epoch": 2.8706095041322315,
+      "grad_norm": 4.65625,
+      "learning_rate": 7.714891171425036e-05,
+      "loss": 4.3076,
+      "step": 333450
+    },
+    {
+      "epoch": 2.8710399449035813,
+      "grad_norm": 4.3125,
+      "learning_rate": 7.712253733432039e-05,
+      "loss": 4.5337,
+      "step": 333500
+    },
+    {
+      "epoch": 2.871470385674931,
+      "grad_norm": 3.40625,
+      "learning_rate": 7.70961646335469e-05,
+      "loss": 4.3035,
+      "step": 333550
+    },
+    {
+      "epoch": 2.871900826446281,
+      "grad_norm": 3.5625,
+      "learning_rate": 7.706979361386565e-05,
+      "loss": 4.2263,
+      "step": 333600
+    },
+    {
+      "epoch": 2.872331267217631,
+      "grad_norm": 1.234375,
+      "learning_rate": 7.704342427721225e-05,
+      "loss": 4.1651,
+      "step": 333650
+    },
+    {
+      "epoch": 2.872761707988981,
+      "grad_norm": 2.953125,
+      "learning_rate": 7.701705662552209e-05,
+      "loss": 4.5041,
+      "step": 333700
+    },
+    {
+      "epoch": 2.8731921487603307,
+      "grad_norm": 3.78125,
+      "learning_rate": 7.699069066073053e-05,
+      "loss": 4.4416,
+      "step": 333750
+    },
+    {
+      "epoch": 2.8736225895316805,
+      "grad_norm": 2.140625,
+      "learning_rate": 7.696432638477281e-05,
+      "loss": 4.2717,
+      "step": 333800
+    },
+    {
+      "epoch": 2.8740530303030303,
+      "grad_norm": 4.59375,
+      "learning_rate": 7.693796379958391e-05,
+      "loss": 4.3637,
+      "step": 333850
+    },
+    {
+      "epoch": 2.87448347107438,
+      "grad_norm": 2.625,
+      "learning_rate": 7.691160290709886e-05,
+      "loss": 4.1766,
+      "step": 333900
+    },
+    {
+      "epoch": 2.87491391184573,
+      "grad_norm": 3.40625,
+      "learning_rate": 7.68852437092525e-05,
+      "loss": 4.4558,
+      "step": 333950
+    },
+    {
+      "epoch": 2.8753443526170797,
+      "grad_norm": 2.90625,
+      "learning_rate": 7.685888620797949e-05,
+      "loss": 4.2098,
+      "step": 334000
+    },
+    {
+      "epoch": 2.87577479338843,
+      "grad_norm": 2.140625,
+      "learning_rate": 7.683253040521451e-05,
+      "loss": 4.1785,
+      "step": 334050
+    },
+    {
+      "epoch": 2.8762052341597797,
+      "grad_norm": 4.40625,
+      "learning_rate": 7.680617630289189e-05,
+      "loss": 4.3321,
+      "step": 334100
+    },
+    {
+      "epoch": 2.8766356749311295,
+      "grad_norm": 2.65625,
+      "learning_rate": 7.677982390294605e-05,
+      "loss": 4.3175,
+      "step": 334150
+    },
+    {
+      "epoch": 2.8770661157024793,
+      "grad_norm": 2.96875,
+      "learning_rate": 7.675347320731116e-05,
+      "loss": 4.594,
+      "step": 334200
+    },
+    {
+      "epoch": 2.877496556473829,
+      "grad_norm": 2.5,
+      "learning_rate": 7.672712421792133e-05,
+      "loss": 4.4218,
+      "step": 334250
+    },
+    {
+      "epoch": 2.8779269972451793,
+      "grad_norm": 4.5,
+      "learning_rate": 7.670077693671052e-05,
+      "loss": 4.1439,
+      "step": 334300
+    },
+    {
+      "epoch": 2.878357438016529,
+      "grad_norm": 2.578125,
+      "learning_rate": 7.667443136561255e-05,
+      "loss": 4.44,
+      "step": 334350
+    },
+    {
+      "epoch": 2.878787878787879,
+      "grad_norm": 2.984375,
+      "learning_rate": 7.664808750656111e-05,
+      "loss": 4.25,
+      "step": 334400
+    },
+    {
+      "epoch": 2.8792183195592287,
+      "grad_norm": 2.515625,
+      "learning_rate": 7.662174536148982e-05,
+      "loss": 4.7782,
+      "step": 334450
+    },
+    {
+      "epoch": 2.8796487603305785,
+      "grad_norm": 2.203125,
+      "learning_rate": 7.65954049323321e-05,
+      "loss": 3.9941,
+      "step": 334500
+    },
+    {
+      "epoch": 2.8800792011019283,
+      "grad_norm": 2.0,
+      "learning_rate": 7.656906622102136e-05,
+      "loss": 3.7434,
+      "step": 334550
+    },
+    {
+      "epoch": 2.880509641873278,
+      "grad_norm": 2.40625,
+      "learning_rate": 7.654272922949069e-05,
+      "loss": 4.2841,
+      "step": 334600
+    },
+    {
+      "epoch": 2.880940082644628,
+      "grad_norm": 3.265625,
+      "learning_rate": 7.651639395967324e-05,
+      "loss": 4.3047,
+      "step": 334650
+    },
+    {
+      "epoch": 2.881370523415978,
+      "grad_norm": 3.921875,
+      "learning_rate": 7.649006041350194e-05,
+      "loss": 4.0306,
+      "step": 334700
+    },
+    {
+      "epoch": 2.881800964187328,
+      "grad_norm": 3.203125,
+      "learning_rate": 7.646372859290963e-05,
+      "loss": 4.4538,
+      "step": 334750
+    },
+    {
+      "epoch": 2.8822314049586777,
+      "grad_norm": 4.46875,
+      "learning_rate": 7.643739849982903e-05,
+      "loss": 4.3576,
+      "step": 334800
+    },
+    {
+      "epoch": 2.8826618457300275,
+      "grad_norm": 2.84375,
+      "learning_rate": 7.641107013619265e-05,
+      "loss": 4.5172,
+      "step": 334850
+    },
+    {
+      "epoch": 2.8830922865013773,
+      "grad_norm": 3.25,
+      "learning_rate": 7.638474350393296e-05,
+      "loss": 3.9946,
+      "step": 334900
+    },
+    {
+      "epoch": 2.8835227272727275,
+      "grad_norm": 4.96875,
+      "learning_rate": 7.635841860498229e-05,
+      "loss": 4.3149,
+      "step": 334950
+    },
+    {
+      "epoch": 2.8839531680440773,
+      "grad_norm": 2.78125,
+      "learning_rate": 7.633209544127286e-05,
+      "loss": 4.4252,
+      "step": 335000
+    },
+    {
+      "epoch": 2.884383608815427,
+      "grad_norm": 2.859375,
+      "learning_rate": 7.630577401473671e-05,
+      "loss": 4.2135,
+      "step": 335050
+    },
+    {
+      "epoch": 2.884814049586777,
+      "grad_norm": 3.03125,
+      "learning_rate": 7.627945432730572e-05,
+      "loss": 4.4764,
+      "step": 335100
+    },
+    {
+      "epoch": 2.8852444903581267,
+      "grad_norm": 3.609375,
+      "learning_rate": 7.625313638091179e-05,
+      "loss": 4.3044,
+      "step": 335150
+    },
+    {
+      "epoch": 2.8856749311294765,
+      "grad_norm": 4.6875,
+      "learning_rate": 7.622682017748651e-05,
+      "loss": 3.9791,
+      "step": 335200
+    },
+    {
+      "epoch": 2.8861053719008263,
+      "grad_norm": 1.859375,
+      "learning_rate": 7.62005057189615e-05,
+      "loss": 4.4244,
+      "step": 335250
+    },
+    {
+      "epoch": 2.886535812672176,
+      "grad_norm": 3.625,
+      "learning_rate": 7.617419300726819e-05,
+      "loss": 4.4314,
+      "step": 335300
+    },
+    {
+      "epoch": 2.8869662534435263,
+      "grad_norm": 3.34375,
+      "learning_rate": 7.614788204433784e-05,
+      "loss": 4.3055,
+      "step": 335350
+    },
+    {
+      "epoch": 2.887396694214876,
+      "grad_norm": 4.5,
+      "learning_rate": 7.612157283210159e-05,
+      "loss": 4.2885,
+      "step": 335400
+    },
+    {
+      "epoch": 2.887827134986226,
+      "grad_norm": 3.671875,
+      "learning_rate": 7.609526537249054e-05,
+      "loss": 4.5623,
+      "step": 335450
+    },
+    {
+      "epoch": 2.8882575757575757,
+      "grad_norm": 0.6484375,
+      "learning_rate": 7.60689596674356e-05,
+      "loss": 4.2664,
+      "step": 335500
+    },
+    {
+      "epoch": 2.8886880165289255,
+      "grad_norm": 2.84375,
+      "learning_rate": 7.60426557188675e-05,
+      "loss": 4.4225,
+      "step": 335550
+    },
+    {
+      "epoch": 2.8891184573002757,
+      "grad_norm": 3.421875,
+      "learning_rate": 7.601635352871696e-05,
+      "loss": 4.1482,
+      "step": 335600
+    },
+    {
+      "epoch": 2.8895488980716255,
+      "grad_norm": 3.484375,
+      "learning_rate": 7.599005309891445e-05,
+      "loss": 4.4473,
+      "step": 335650
+    },
+    {
+      "epoch": 2.8899793388429753,
+      "grad_norm": 2.75,
+      "learning_rate": 7.596375443139038e-05,
+      "loss": 4.4659,
+      "step": 335700
+    },
+    {
+      "epoch": 2.890409779614325,
+      "grad_norm": 1.5234375,
+      "learning_rate": 7.593745752807501e-05,
+      "loss": 4.1006,
+      "step": 335750
+    },
+    {
+      "epoch": 2.890840220385675,
+      "grad_norm": 2.3125,
+      "learning_rate": 7.591116239089851e-05,
+      "loss": 4.3912,
+      "step": 335800
+    },
+    {
+      "epoch": 2.8912706611570247,
+      "grad_norm": 3.546875,
+      "learning_rate": 7.588486902179088e-05,
+      "loss": 4.4834,
+      "step": 335850
+    },
+    {
+      "epoch": 2.8917011019283745,
+      "grad_norm": 2.765625,
+      "learning_rate": 7.585857742268194e-05,
+      "loss": 4.587,
+      "step": 335900
+    },
+    {
+      "epoch": 2.8921315426997243,
+      "grad_norm": 2.921875,
+      "learning_rate": 7.583228759550148e-05,
+      "loss": 4.0826,
+      "step": 335950
+    },
+    {
+      "epoch": 2.8925619834710745,
+      "grad_norm": 3.828125,
+      "learning_rate": 7.580599954217912e-05,
+      "loss": 4.5829,
+      "step": 336000
+    },
+    {
+      "epoch": 2.8925619834710745,
+      "eval_loss": 4.996245861053467,
+      "eval_runtime": 24.6638,
+      "eval_samples_per_second": 25.949,
+      "eval_steps_per_second": 12.974,
+      "eval_tts_loss": 7.448037665631593,
+      "step": 336000
+    },
+    {
+      "epoch": 2.8929924242424243,
+      "grad_norm": 1.4609375,
+      "learning_rate": 7.577971326464432e-05,
+      "loss": 4.3964,
+      "step": 336050
+    },
+    {
+      "epoch": 2.893422865013774,
+      "grad_norm": 2.03125,
+      "learning_rate": 7.57534287648265e-05,
+      "loss": 4.2171,
+      "step": 336100
+    },
+    {
+      "epoch": 2.893853305785124,
+      "grad_norm": 5.03125,
+      "learning_rate": 7.57271460446548e-05,
+      "loss": 4.2809,
+      "step": 336150
+    },
+    {
+      "epoch": 2.8942837465564737,
+      "grad_norm": 1.65625,
+      "learning_rate": 7.570086510605839e-05,
+      "loss": 4.2913,
+      "step": 336200
+    },
+    {
+      "epoch": 2.894714187327824,
+      "grad_norm": 5.25,
+      "learning_rate": 7.567458595096615e-05,
+      "loss": 4.3189,
+      "step": 336250
+    },
+    {
+      "epoch": 2.8951446280991737,
+      "grad_norm": 1.578125,
+      "learning_rate": 7.564830858130698e-05,
+      "loss": 4.2536,
+      "step": 336300
+    },
+    {
+      "epoch": 2.8955750688705235,
+      "grad_norm": 3.359375,
+      "learning_rate": 7.56220329990096e-05,
+      "loss": 4.2402,
+      "step": 336350
+    },
+    {
+      "epoch": 2.8960055096418733,
+      "grad_norm": 4.21875,
+      "learning_rate": 7.559575920600255e-05,
+      "loss": 4.3815,
+      "step": 336400
+    },
+    {
+      "epoch": 2.896435950413223,
+      "grad_norm": 1.25,
+      "learning_rate": 7.556948720421422e-05,
+      "loss": 4.6188,
+      "step": 336450
+    },
+    {
+      "epoch": 2.896866391184573,
+      "grad_norm": 3.40625,
+      "learning_rate": 7.554321699557299e-05,
+      "loss": 4.3958,
+      "step": 336500
+    },
+    {
+      "epoch": 2.8972968319559227,
+      "grad_norm": 1.9375,
+      "learning_rate": 7.551694858200701e-05,
+      "loss": 4.5251,
+      "step": 336550
+    },
+    {
+      "epoch": 2.8977272727272725,
+      "grad_norm": 2.6875,
+      "learning_rate": 7.549068196544436e-05,
+      "loss": 4.4589,
+      "step": 336600
+    },
+    {
+      "epoch": 2.8981577134986227,
+      "grad_norm": 1.6171875,
+      "learning_rate": 7.546441714781287e-05,
+      "loss": 4.2934,
+      "step": 336650
+    },
+    {
+      "epoch": 2.8985881542699725,
+      "grad_norm": 4.3125,
+      "learning_rate": 7.54381541310404e-05,
+      "loss": 4.5141,
+      "step": 336700
+    },
+    {
+      "epoch": 2.8990185950413223,
+      "grad_norm": 3.375,
+      "learning_rate": 7.541189291705455e-05,
+      "loss": 4.6853,
+      "step": 336750
+    },
+    {
+      "epoch": 2.899449035812672,
+      "grad_norm": 2.390625,
+      "learning_rate": 7.538563350778285e-05,
+      "loss": 4.0215,
+      "step": 336800
+    },
+    {
+      "epoch": 2.899879476584022,
+      "grad_norm": 3.171875,
+      "learning_rate": 7.535937590515274e-05,
+      "loss": 4.3871,
+      "step": 336850
+    },
+    {
+      "epoch": 2.900309917355372,
+      "grad_norm": 2.46875,
+      "learning_rate": 7.533312011109141e-05,
+      "loss": 3.9731,
+      "step": 336900
+    },
+    {
+      "epoch": 2.900740358126722,
+      "grad_norm": 1.859375,
+      "learning_rate": 7.530686612752596e-05,
+      "loss": 4.3554,
+      "step": 336950
+    },
+    {
+      "epoch": 2.9011707988980717,
+      "grad_norm": 1.203125,
+      "learning_rate": 7.52806139563834e-05,
+      "loss": 4.1233,
+      "step": 337000
+    },
+    {
+      "epoch": 2.9016012396694215,
+      "grad_norm": 3.46875,
+      "learning_rate": 7.525436359959062e-05,
+      "loss": 4.4772,
+      "step": 337050
+    },
+    {
+      "epoch": 2.9020316804407713,
+      "grad_norm": 4.375,
+      "learning_rate": 7.522811505907433e-05,
+      "loss": 4.1946,
+      "step": 337100
+    },
+    {
+      "epoch": 2.902462121212121,
+      "grad_norm": 4.1875,
+      "learning_rate": 7.520186833676106e-05,
+      "loss": 4.287,
+      "step": 337150
+    },
+    {
+      "epoch": 2.902892561983471,
+      "grad_norm": 3.046875,
+      "learning_rate": 7.517562343457729e-05,
+      "loss": 3.7122,
+      "step": 337200
+    },
+    {
+      "epoch": 2.9033230027548207,
+      "grad_norm": 5.75,
+      "learning_rate": 7.514938035444939e-05,
+      "loss": 4.0659,
+      "step": 337250
+    },
+    {
+      "epoch": 2.903753443526171,
+      "grad_norm": 2.59375,
+      "learning_rate": 7.512313909830346e-05,
+      "loss": 4.3358,
+      "step": 337300
+    },
+    {
+      "epoch": 2.9041838842975207,
+      "grad_norm": 3.984375,
+      "learning_rate": 7.509689966806562e-05,
+      "loss": 4.3521,
+      "step": 337350
+    },
+    {
+      "epoch": 2.9046143250688705,
+      "grad_norm": 3.140625,
+      "learning_rate": 7.50706620656618e-05,
+      "loss": 4.6674,
+      "step": 337400
+    },
+    {
+      "epoch": 2.9050447658402203,
+      "grad_norm": 2.4375,
+      "learning_rate": 7.504442629301769e-05,
+      "loss": 4.2609,
+      "step": 337450
+    },
+    {
+      "epoch": 2.90547520661157,
+      "grad_norm": 1.8671875,
+      "learning_rate": 7.5018192352059e-05,
+      "loss": 4.6744,
+      "step": 337500
+    },
+    {
+      "epoch": 2.9059056473829203,
+      "grad_norm": 5.375,
+      "learning_rate": 7.499196024471127e-05,
+      "loss": 4.5588,
+      "step": 337550
+    },
+    {
+      "epoch": 2.90633608815427,
+      "grad_norm": 3.484375,
+      "learning_rate": 7.496572997289983e-05,
+      "loss": 4.3246,
+      "step": 337600
+    },
+    {
+      "epoch": 2.90676652892562,
+      "grad_norm": 1.765625,
+      "learning_rate": 7.493950153854998e-05,
+      "loss": 4.2777,
+      "step": 337650
+    },
+    {
+      "epoch": 2.9071969696969697,
+      "grad_norm": 3.125,
+      "learning_rate": 7.491327494358677e-05,
+      "loss": 4.1976,
+      "step": 337700
+    },
+    {
+      "epoch": 2.9076274104683195,
+      "grad_norm": 3.234375,
+      "learning_rate": 7.488705018993522e-05,
+      "loss": 4.2998,
+      "step": 337750
+    },
+    {
+      "epoch": 2.9080578512396693,
+      "grad_norm": 2.34375,
+      "learning_rate": 7.486082727952013e-05,
+      "loss": 4.076,
+      "step": 337800
+    },
+    {
+      "epoch": 2.908488292011019,
+      "grad_norm": 2.453125,
+      "learning_rate": 7.483460621426624e-05,
+      "loss": 4.347,
+      "step": 337850
+    },
+    {
+      "epoch": 2.908918732782369,
+      "grad_norm": 3.96875,
+      "learning_rate": 7.480838699609812e-05,
+      "loss": 4.5208,
+      "step": 337900
+    },
+    {
+      "epoch": 2.909349173553719,
+      "grad_norm": 3.15625,
+      "learning_rate": 7.47821696269402e-05,
+      "loss": 4.348,
+      "step": 337950
+    },
+    {
+      "epoch": 2.909779614325069,
+      "grad_norm": 2.84375,
+      "learning_rate": 7.475595410871674e-05,
+      "loss": 4.4903,
+      "step": 338000
+    },
+    {
+      "epoch": 2.9102100550964187,
+      "grad_norm": 4.09375,
+      "learning_rate": 7.472974044335196e-05,
+      "loss": 4.4813,
+      "step": 338050
+    },
+    {
+      "epoch": 2.9106404958677685,
+      "grad_norm": 1.359375,
+      "learning_rate": 7.470352863276983e-05,
+      "loss": 4.2015,
+      "step": 338100
+    },
+    {
+      "epoch": 2.9110709366391183,
+      "grad_norm": 1.4453125,
+      "learning_rate": 7.467731867889434e-05,
+      "loss": 4.3932,
+      "step": 338150
+    },
+    {
+      "epoch": 2.9115013774104685,
+      "grad_norm": 2.375,
+      "learning_rate": 7.46511105836491e-05,
+      "loss": 4.1417,
+      "step": 338200
+    },
+    {
+      "epoch": 2.9119318181818183,
+      "grad_norm": 4.78125,
+      "learning_rate": 7.462490434895786e-05,
+      "loss": 3.9601,
+      "step": 338250
+    },
+    {
+      "epoch": 2.912362258953168,
+      "grad_norm": 1.109375,
+      "learning_rate": 7.4598699976744e-05,
+      "loss": 4.3112,
+      "step": 338300
+    },
+    {
+      "epoch": 2.912792699724518,
+      "grad_norm": 2.265625,
+      "learning_rate": 7.457249746893091e-05,
+      "loss": 4.291,
+      "step": 338350
+    },
+    {
+      "epoch": 2.9132231404958677,
+      "grad_norm": 2.5625,
+      "learning_rate": 7.454629682744186e-05,
+      "loss": 4.316,
+      "step": 338400
+    },
+    {
+      "epoch": 2.9136535812672175,
+      "grad_norm": 3.171875,
+      "learning_rate": 7.452009805419981e-05,
+      "loss": 4.4949,
+      "step": 338450
+    },
+    {
+      "epoch": 2.9140840220385673,
+      "grad_norm": 3.0625,
+      "learning_rate": 7.449390115112774e-05,
+      "loss": 4.0673,
+      "step": 338500
+    },
+    {
+      "epoch": 2.914514462809917,
+      "grad_norm": 2.96875,
+      "learning_rate": 7.446770612014841e-05,
+      "loss": 4.5295,
+      "step": 338550
+    },
+    {
+      "epoch": 2.9149449035812673,
+      "grad_norm": 1.9296875,
+      "learning_rate": 7.444151296318458e-05,
+      "loss": 4.0253,
+      "step": 338600
+    },
+    {
+      "epoch": 2.915375344352617,
+      "grad_norm": 7.375,
+      "learning_rate": 7.441532168215872e-05,
+      "loss": 4.2479,
+      "step": 338650
+    },
+    {
+      "epoch": 2.915805785123967,
+      "grad_norm": 5.34375,
+      "learning_rate": 7.438913227899316e-05,
+      "loss": 4.3128,
+      "step": 338700
+    },
+    {
+      "epoch": 2.9162362258953167,
+      "grad_norm": 3.921875,
+      "learning_rate": 7.43629447556102e-05,
+      "loss": 4.5874,
+      "step": 338750
+    },
+    {
+      "epoch": 2.9166666666666665,
+      "grad_norm": 2.625,
+      "learning_rate": 7.433675911393193e-05,
+      "loss": 4.0499,
+      "step": 338800
+    },
+    {
+      "epoch": 2.9170971074380168,
+      "grad_norm": 2.03125,
+      "learning_rate": 7.43105753558803e-05,
+      "loss": 4.124,
+      "step": 338850
+    },
+    {
+      "epoch": 2.9175275482093666,
+      "grad_norm": 4.5625,
+      "learning_rate": 7.428439348337721e-05,
+      "loss": 4.111,
+      "step": 338900
+    },
+    {
+      "epoch": 2.9179579889807163,
+      "grad_norm": 5.6875,
+      "learning_rate": 7.425821349834431e-05,
+      "loss": 4.3855,
+      "step": 338950
+    },
+    {
+      "epoch": 2.918388429752066,
+      "grad_norm": 2.90625,
+      "learning_rate": 7.423203540270311e-05,
+      "loss": 4.3686,
+      "step": 339000
+    },
+    {
+      "epoch": 2.918388429752066,
+      "eval_loss": 4.994027614593506,
+      "eval_runtime": 23.9056,
+      "eval_samples_per_second": 26.772,
+      "eval_steps_per_second": 13.386,
+      "eval_tts_loss": 7.426867337462221,
+      "step": 339000
+    },
+    {
+      "epoch": 2.918818870523416,
+      "grad_norm": 5.1875,
+      "learning_rate": 7.42058591983751e-05,
+      "loss": 4.6089,
+      "step": 339050
+    },
+    {
+      "epoch": 2.9192493112947657,
+      "grad_norm": 3.078125,
+      "learning_rate": 7.417968488728153e-05,
+      "loss": 4.0212,
+      "step": 339100
+    },
+    {
+      "epoch": 2.9196797520661155,
+      "grad_norm": 3.28125,
+      "learning_rate": 7.415351247134351e-05,
+      "loss": 4.0897,
+      "step": 339150
+    },
+    {
+      "epoch": 2.9201101928374653,
+      "grad_norm": 3.625,
+      "learning_rate": 7.412734195248212e-05,
+      "loss": 4.6838,
+      "step": 339200
+    },
+    {
+      "epoch": 2.9205406336088156,
+      "grad_norm": 1.75,
+      "learning_rate": 7.410117333261812e-05,
+      "loss": 4.1019,
+      "step": 339250
+    },
+    {
+      "epoch": 2.9209710743801653,
+      "grad_norm": 4.6875,
+      "learning_rate": 7.407500661367228e-05,
+      "loss": 4.2213,
+      "step": 339300
+    },
+    {
+      "epoch": 2.921401515151515,
+      "grad_norm": 4.28125,
+      "learning_rate": 7.404884179756517e-05,
+      "loss": 4.1974,
+      "step": 339350
+    },
+    {
+      "epoch": 2.921831955922865,
+      "grad_norm": 1.46875,
+      "learning_rate": 7.402267888621725e-05,
+      "loss": 4.2541,
+      "step": 339400
+    },
+    {
+      "epoch": 2.9222623966942147,
+      "grad_norm": 2.46875,
+      "learning_rate": 7.399651788154882e-05,
+      "loss": 4.234,
+      "step": 339450
+    },
+    {
+      "epoch": 2.922692837465565,
+      "grad_norm": 2.15625,
+      "learning_rate": 7.397035878548002e-05,
+      "loss": 4.2452,
+      "step": 339500
+    },
+    {
+      "epoch": 2.9231232782369148,
+      "grad_norm": 2.890625,
+      "learning_rate": 7.394420159993085e-05,
+      "loss": 4.6126,
+      "step": 339550
+    },
+    {
+      "epoch": 2.9235537190082646,
+      "grad_norm": 3.59375,
+      "learning_rate": 7.391804632682126e-05,
+      "loss": 4.3926,
+      "step": 339600
+    },
+    {
+      "epoch": 2.9239841597796143,
+      "grad_norm": 1.8515625,
+      "learning_rate": 7.389189296807093e-05,
+      "loss": 4.395,
+      "step": 339650
+    },
+    {
+      "epoch": 2.924414600550964,
+      "grad_norm": 3.09375,
+      "learning_rate": 7.386574152559952e-05,
+      "loss": 4.0477,
+      "step": 339700
+    },
+    {
+      "epoch": 2.924845041322314,
+      "grad_norm": 2.765625,
+      "learning_rate": 7.38395920013264e-05,
+      "loss": 4.1257,
+      "step": 339750
+    },
+    {
+      "epoch": 2.9252754820936637,
+      "grad_norm": 2.6875,
+      "learning_rate": 7.381344439717098e-05,
+      "loss": 4.0482,
+      "step": 339800
+    },
+    {
+      "epoch": 2.9257059228650135,
+      "grad_norm": 4.8125,
+      "learning_rate": 7.378729871505237e-05,
+      "loss": 4.4351,
+      "step": 339850
+    },
+    {
+      "epoch": 2.9261363636363638,
+      "grad_norm": 4.25,
+      "learning_rate": 7.376115495688965e-05,
+      "loss": 4.7657,
+      "step": 339900
+    },
+    {
+      "epoch": 2.9265668044077136,
+      "grad_norm": 2.9375,
+      "learning_rate": 7.373501312460173e-05,
+      "loss": 4.4889,
+      "step": 339950
+    },
+    {
+      "epoch": 2.9269972451790633,
+      "grad_norm": 3.265625,
+      "learning_rate": 7.370887322010731e-05,
+      "loss": 4.2453,
+      "step": 340000
+    },
+    {
+      "epoch": 2.927427685950413,
+      "grad_norm": 3.34375,
+      "learning_rate": 7.368273524532502e-05,
+      "loss": 4.4345,
+      "step": 340050
+    },
+    {
+      "epoch": 2.927858126721763,
+      "grad_norm": 3.640625,
+      "learning_rate": 7.365659920217335e-05,
+      "loss": 4.2598,
+      "step": 340100
+    },
+    {
+      "epoch": 2.928288567493113,
+      "grad_norm": 1.6484375,
+      "learning_rate": 7.363046509257061e-05,
+      "loss": 3.9422,
+      "step": 340150
+    },
+    {
+      "epoch": 2.928719008264463,
+      "grad_norm": 5.5,
+      "learning_rate": 7.360433291843503e-05,
+      "loss": 4.3143,
+      "step": 340200
+    },
+    {
+      "epoch": 2.9291494490358128,
+      "grad_norm": 3.578125,
+      "learning_rate": 7.357820268168458e-05,
+      "loss": 4.3546,
+      "step": 340250
+    },
+    {
+      "epoch": 2.9295798898071626,
+      "grad_norm": 3.515625,
+      "learning_rate": 7.355207438423723e-05,
+      "loss": 4.3261,
+      "step": 340300
+    },
+    {
+      "epoch": 2.9300103305785123,
+      "grad_norm": 3.90625,
+      "learning_rate": 7.35259480280107e-05,
+      "loss": 4.3995,
+      "step": 340350
+    },
+    {
+      "epoch": 2.930440771349862,
+      "grad_norm": 1.8125,
+      "learning_rate": 7.349982361492262e-05,
+      "loss": 4.3694,
+      "step": 340400
+    },
+    {
+      "epoch": 2.930871212121212,
+      "grad_norm": 6.625,
+      "learning_rate": 7.347370114689051e-05,
+      "loss": 4.585,
+      "step": 340450
+    },
+    {
+      "epoch": 2.9313016528925617,
+      "grad_norm": 3.5625,
+      "learning_rate": 7.344758062583165e-05,
+      "loss": 4.4143,
+      "step": 340500
+    },
+    {
+      "epoch": 2.931732093663912,
+      "grad_norm": 2.375,
+      "learning_rate": 7.342146205366321e-05,
+      "loss": 4.4472,
+      "step": 340550
+    },
+    {
+      "epoch": 2.9321625344352618,
+      "grad_norm": 3.734375,
+      "learning_rate": 7.339534543230228e-05,
+      "loss": 4.6417,
+      "step": 340600
+    },
+    {
+      "epoch": 2.9325929752066116,
+      "grad_norm": 4.1875,
+      "learning_rate": 7.336923076366577e-05,
+      "loss": 4.1988,
+      "step": 340650
+    },
+    {
+      "epoch": 2.9330234159779613,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.334311804967044e-05,
+      "loss": 4.5914,
+      "step": 340700
+    },
+    {
+      "epoch": 2.933453856749311,
+      "grad_norm": 4.625,
+      "learning_rate": 7.331700729223285e-05,
+      "loss": 4.2424,
+      "step": 340750
+    },
+    {
+      "epoch": 2.9338842975206614,
+      "grad_norm": 3.796875,
+      "learning_rate": 7.329089849326952e-05,
+      "loss": 4.4046,
+      "step": 340800
+    },
+    {
+      "epoch": 2.934314738292011,
+      "grad_norm": 3.390625,
+      "learning_rate": 7.32647916546968e-05,
+      "loss": 4.1883,
+      "step": 340850
+    },
+    {
+      "epoch": 2.934745179063361,
+      "grad_norm": 3.78125,
+      "learning_rate": 7.323868677843082e-05,
+      "loss": 4.3866,
+      "step": 340900
+    },
+    {
+      "epoch": 2.9351756198347108,
+      "grad_norm": 2.78125,
+      "learning_rate": 7.32125838663877e-05,
+      "loss": 4.2655,
+      "step": 340950
+    },
+    {
+      "epoch": 2.9356060606060606,
+      "grad_norm": 4.03125,
+      "learning_rate": 7.318648292048327e-05,
+      "loss": 4.1353,
+      "step": 341000
+    },
+    {
+      "epoch": 2.9360365013774103,
+      "grad_norm": 4.6875,
+      "learning_rate": 7.316038394263327e-05,
+      "loss": 4.0012,
+      "step": 341050
+    },
+    {
+      "epoch": 2.93646694214876,
+      "grad_norm": 4.34375,
+      "learning_rate": 7.313428693475336e-05,
+      "loss": 4.4765,
+      "step": 341100
+    },
+    {
+      "epoch": 2.93689738292011,
+      "grad_norm": 4.4375,
+      "learning_rate": 7.310819189875899e-05,
+      "loss": 4.46,
+      "step": 341150
+    },
+    {
+      "epoch": 2.93732782369146,
+      "grad_norm": 1.7578125,
+      "learning_rate": 7.308209883656547e-05,
+      "loss": 4.4946,
+      "step": 341200
+    },
+    {
+      "epoch": 2.93775826446281,
+      "grad_norm": 1.5546875,
+      "learning_rate": 7.3056007750088e-05,
+      "loss": 4.5813,
+      "step": 341250
+    },
+    {
+      "epoch": 2.9381887052341598,
+      "grad_norm": 3.328125,
+      "learning_rate": 7.302991864124158e-05,
+      "loss": 4.4975,
+      "step": 341300
+    },
+    {
+      "epoch": 2.9386191460055096,
+      "grad_norm": 1.109375,
+      "learning_rate": 7.30038315119411e-05,
+      "loss": 4.2773,
+      "step": 341350
+    },
+    {
+      "epoch": 2.9390495867768593,
+      "grad_norm": 3.875,
+      "learning_rate": 7.297774636410129e-05,
+      "loss": 4.102,
+      "step": 341400
+    },
+    {
+      "epoch": 2.9394800275482096,
+      "grad_norm": 4.5625,
+      "learning_rate": 7.295166319963676e-05,
+      "loss": 4.309,
+      "step": 341450
+    },
+    {
+      "epoch": 2.9399104683195594,
+      "grad_norm": 2.59375,
+      "learning_rate": 7.292558202046197e-05,
+      "loss": 4.0754,
+      "step": 341500
+    },
+    {
+      "epoch": 2.940340909090909,
+      "grad_norm": 6.59375,
+      "learning_rate": 7.289950282849123e-05,
+      "loss": 4.4291,
+      "step": 341550
+    },
+    {
+      "epoch": 2.940771349862259,
+      "grad_norm": 3.75,
+      "learning_rate": 7.287342562563863e-05,
+      "loss": 4.4011,
+      "step": 341600
+    },
+    {
+      "epoch": 2.9412017906336088,
+      "grad_norm": 6.21875,
+      "learning_rate": 7.284735041381824e-05,
+      "loss": 4.5579,
+      "step": 341650
+    },
+    {
+      "epoch": 2.9416322314049586,
+      "grad_norm": 4.875,
+      "learning_rate": 7.28212771949439e-05,
+      "loss": 4.4128,
+      "step": 341700
+    },
+    {
+      "epoch": 2.9420626721763083,
+      "grad_norm": 2.609375,
+      "learning_rate": 7.279520597092938e-05,
+      "loss": 4.4299,
+      "step": 341750
+    },
+    {
+      "epoch": 2.942493112947658,
+      "grad_norm": 4.6875,
+      "learning_rate": 7.276913674368817e-05,
+      "loss": 4.401,
+      "step": 341800
+    },
+    {
+      "epoch": 2.9429235537190084,
+      "grad_norm": 1.953125,
+      "learning_rate": 7.274306951513374e-05,
+      "loss": 4.4946,
+      "step": 341850
+    },
+    {
+      "epoch": 2.943353994490358,
+      "grad_norm": 3.4375,
+      "learning_rate": 7.271700428717936e-05,
+      "loss": 4.6358,
+      "step": 341900
+    },
+    {
+      "epoch": 2.943784435261708,
+      "grad_norm": 3.875,
+      "learning_rate": 7.269094106173815e-05,
+      "loss": 4.2043,
+      "step": 341950
+    },
+    {
+      "epoch": 2.9442148760330578,
+      "grad_norm": 6.28125,
+      "learning_rate": 7.266487984072315e-05,
+      "loss": 4.0789,
+      "step": 342000
+    },
+    {
+      "epoch": 2.9442148760330578,
+      "eval_loss": 4.994072914123535,
+      "eval_runtime": 23.8197,
+      "eval_samples_per_second": 26.869,
+      "eval_steps_per_second": 13.434,
+      "eval_tts_loss": 7.446218288815809,
+      "step": 342000
+    },
+    {
+      "epoch": 2.9446453168044076,
+      "grad_norm": 3.1875,
+      "learning_rate": 7.263882062604714e-05,
+      "loss": 4.4023,
+      "step": 342050
+    },
+    {
+      "epoch": 2.945075757575758,
+      "grad_norm": 2.71875,
+      "learning_rate": 7.261276341962282e-05,
+      "loss": 4.4711,
+      "step": 342100
+    },
+    {
+      "epoch": 2.9455061983471076,
+      "grad_norm": 4.34375,
+      "learning_rate": 7.258670822336273e-05,
+      "loss": 4.3143,
+      "step": 342150
+    },
+    {
+      "epoch": 2.9459366391184574,
+      "grad_norm": 2.828125,
+      "learning_rate": 7.256065503917928e-05,
+      "loss": 4.4861,
+      "step": 342200
+    },
+    {
+      "epoch": 2.946367079889807,
+      "grad_norm": 8.0625,
+      "learning_rate": 7.253460386898476e-05,
+      "loss": 4.478,
+      "step": 342250
+    },
+    {
+      "epoch": 2.946797520661157,
+      "grad_norm": 4.0,
+      "learning_rate": 7.250855471469115e-05,
+      "loss": 4.476,
+      "step": 342300
+    },
+    {
+      "epoch": 2.9472279614325068,
+      "grad_norm": 1.515625,
+      "learning_rate": 7.248250757821053e-05,
+      "loss": 4.4263,
+      "step": 342350
+    },
+    {
+      "epoch": 2.9476584022038566,
+      "grad_norm": 2.5625,
+      "learning_rate": 7.245646246145461e-05,
+      "loss": 4.5628,
+      "step": 342400
+    },
+    {
+      "epoch": 2.9480888429752063,
+      "grad_norm": 1.9765625,
+      "learning_rate": 7.243041936633508e-05,
+      "loss": 4.1868,
+      "step": 342450
+    },
+    {
+      "epoch": 2.9485192837465566,
+      "grad_norm": 3.5625,
+      "learning_rate": 7.24043782947635e-05,
+      "loss": 4.2413,
+      "step": 342500
+    },
+    {
+      "epoch": 2.9489497245179064,
+      "grad_norm": 4.34375,
+      "learning_rate": 7.237833924865116e-05,
+      "loss": 4.2355,
+      "step": 342550
+    },
+    {
+      "epoch": 2.949380165289256,
+      "grad_norm": 4.5,
+      "learning_rate": 7.235230222990926e-05,
+      "loss": 4.2631,
+      "step": 342600
+    },
+    {
+      "epoch": 2.949810606060606,
+      "grad_norm": 3.671875,
+      "learning_rate": 7.23262672404489e-05,
+      "loss": 3.7034,
+      "step": 342650
+    },
+    {
+      "epoch": 2.9502410468319558,
+      "grad_norm": 2.796875,
+      "learning_rate": 7.2300234282181e-05,
+      "loss": 4.3717,
+      "step": 342700
+    },
+    {
+      "epoch": 2.950671487603306,
+      "grad_norm": 1.9140625,
+      "learning_rate": 7.227420335701632e-05,
+      "loss": 4.2489,
+      "step": 342750
+    },
+    {
+      "epoch": 2.951101928374656,
+      "grad_norm": 2.703125,
+      "learning_rate": 7.224817446686542e-05,
+      "loss": 4.4818,
+      "step": 342800
+    },
+    {
+      "epoch": 2.9515323691460056,
+      "grad_norm": 2.46875,
+      "learning_rate": 7.222214761363882e-05,
+      "loss": 4.1592,
+      "step": 342850
+    },
+    {
+      "epoch": 2.9519628099173554,
+      "grad_norm": 3.765625,
+      "learning_rate": 7.219612279924683e-05,
+      "loss": 4.5082,
+      "step": 342900
+    },
+    {
+      "epoch": 2.952393250688705,
+      "grad_norm": 1.703125,
+      "learning_rate": 7.217010002559959e-05,
+      "loss": 4.38,
+      "step": 342950
+    },
+    {
+      "epoch": 2.952823691460055,
+      "grad_norm": 2.0,
+      "learning_rate": 7.214407929460714e-05,
+      "loss": 4.2772,
+      "step": 343000
+    },
+    {
+      "epoch": 2.9532541322314048,
+      "grad_norm": 4.125,
+      "learning_rate": 7.211806060817934e-05,
+      "loss": 4.2779,
+      "step": 343050
+    },
+    {
+      "epoch": 2.9536845730027546,
+      "grad_norm": 3.234375,
+      "learning_rate": 7.209204396822591e-05,
+      "loss": 4.096,
+      "step": 343100
+    },
+    {
+      "epoch": 2.954115013774105,
+      "grad_norm": 3.890625,
+      "learning_rate": 7.206602937665638e-05,
+      "loss": 4.2168,
+      "step": 343150
+    },
+    {
+      "epoch": 2.9545454545454546,
+      "grad_norm": 2.625,
+      "learning_rate": 7.20400168353802e-05,
+      "loss": 4.2189,
+      "step": 343200
+    },
+    {
+      "epoch": 2.9549758953168044,
+      "grad_norm": 4.15625,
+      "learning_rate": 7.201400634630662e-05,
+      "loss": 4.3709,
+      "step": 343250
+    },
+    {
+      "epoch": 2.955406336088154,
+      "grad_norm": 3.90625,
+      "learning_rate": 7.198799791134481e-05,
+      "loss": 4.6831,
+      "step": 343300
+    },
+    {
+      "epoch": 2.955836776859504,
+      "grad_norm": 4.0625,
+      "learning_rate": 7.196199153240362e-05,
+      "loss": 4.3318,
+      "step": 343350
+    },
+    {
+      "epoch": 2.956267217630854,
+      "grad_norm": 2.3125,
+      "learning_rate": 7.193598721139196e-05,
+      "loss": 4.4729,
+      "step": 343400
+    },
+    {
+      "epoch": 2.956697658402204,
+      "grad_norm": 3.015625,
+      "learning_rate": 7.190998495021845e-05,
+      "loss": 4.5371,
+      "step": 343450
+    },
+    {
+      "epoch": 2.957128099173554,
+      "grad_norm": 2.109375,
+      "learning_rate": 7.188398475079159e-05,
+      "loss": 4.4545,
+      "step": 343500
+    },
+    {
+      "epoch": 2.9575585399449036,
+      "grad_norm": 4.0625,
+      "learning_rate": 7.185798661501981e-05,
+      "loss": 4.5769,
+      "step": 343550
+    },
+    {
+      "epoch": 2.9579889807162534,
+      "grad_norm": 4.5,
+      "learning_rate": 7.183199054481122e-05,
+      "loss": 4.2668,
+      "step": 343600
+    },
+    {
+      "epoch": 2.958419421487603,
+      "grad_norm": 4.09375,
+      "learning_rate": 7.180599654207391e-05,
+      "loss": 4.313,
+      "step": 343650
+    },
+    {
+      "epoch": 2.958849862258953,
+      "grad_norm": 2.390625,
+      "learning_rate": 7.178000460871583e-05,
+      "loss": 4.3008,
+      "step": 343700
+    },
+    {
+      "epoch": 2.9592803030303028,
+      "grad_norm": 3.375,
+      "learning_rate": 7.175401474664467e-05,
+      "loss": 4.3496,
+      "step": 343750
+    },
+    {
+      "epoch": 2.959710743801653,
+      "grad_norm": 4.59375,
+      "learning_rate": 7.172802695776808e-05,
+      "loss": 4.4027,
+      "step": 343800
+    },
+    {
+      "epoch": 2.960141184573003,
+      "grad_norm": 3.03125,
+      "learning_rate": 7.170204124399345e-05,
+      "loss": 4.5014,
+      "step": 343850
+    },
+    {
+      "epoch": 2.9605716253443526,
+      "grad_norm": 3.25,
+      "learning_rate": 7.167605760722813e-05,
+      "loss": 4.5292,
+      "step": 343900
+    },
+    {
+      "epoch": 2.9610020661157024,
+      "grad_norm": 3.46875,
+      "learning_rate": 7.165007604937922e-05,
+      "loss": 4.5672,
+      "step": 343950
+    },
+    {
+      "epoch": 2.9614325068870526,
+      "grad_norm": 2.3125,
+      "learning_rate": 7.162409657235371e-05,
+      "loss": 4.3536,
+      "step": 344000
+    },
+    {
+      "epoch": 2.9618629476584024,
+      "grad_norm": 2.546875,
+      "learning_rate": 7.159811917805854e-05,
+      "loss": 4.3316,
+      "step": 344050
+    },
+    {
+      "epoch": 2.962293388429752,
+      "grad_norm": 1.71875,
+      "learning_rate": 7.157214386840026e-05,
+      "loss": 4.0619,
+      "step": 344100
+    },
+    {
+      "epoch": 2.962723829201102,
+      "grad_norm": 3.75,
+      "learning_rate": 7.154617064528545e-05,
+      "loss": 4.8132,
+      "step": 344150
+    },
+    {
+      "epoch": 2.963154269972452,
+      "grad_norm": 3.65625,
+      "learning_rate": 7.15201995106205e-05,
+      "loss": 4.3088,
+      "step": 344200
+    },
+    {
+      "epoch": 2.9635847107438016,
+      "grad_norm": 3.4375,
+      "learning_rate": 7.149423046631163e-05,
+      "loss": 4.1888,
+      "step": 344250
+    },
+    {
+      "epoch": 2.9640151515151514,
+      "grad_norm": 1.6796875,
+      "learning_rate": 7.146826351426496e-05,
+      "loss": 4.3703,
+      "step": 344300
+    },
+    {
+      "epoch": 2.964445592286501,
+      "grad_norm": 3.71875,
+      "learning_rate": 7.14422986563863e-05,
+      "loss": 4.1878,
+      "step": 344350
+    },
+    {
+      "epoch": 2.964876033057851,
+      "grad_norm": 3.140625,
+      "learning_rate": 7.141633589458147e-05,
+      "loss": 4.5171,
+      "step": 344400
+    },
+    {
+      "epoch": 2.965306473829201,
+      "grad_norm": 3.21875,
+      "learning_rate": 7.139037523075614e-05,
+      "loss": 4.3684,
+      "step": 344450
+    },
+    {
+      "epoch": 2.965736914600551,
+      "grad_norm": 3.328125,
+      "learning_rate": 7.136441666681566e-05,
+      "loss": 4.2267,
+      "step": 344500
+    },
+    {
+      "epoch": 2.966167355371901,
+      "grad_norm": 3.234375,
+      "learning_rate": 7.133846020466543e-05,
+      "loss": 4.395,
+      "step": 344550
+    },
+    {
+      "epoch": 2.9665977961432506,
+      "grad_norm": 3.28125,
+      "learning_rate": 7.131250584621054e-05,
+      "loss": 4.2683,
+      "step": 344600
+    },
+    {
+      "epoch": 2.967028236914601,
+      "grad_norm": 2.796875,
+      "learning_rate": 7.128655359335599e-05,
+      "loss": 4.5993,
+      "step": 344650
+    },
+    {
+      "epoch": 2.9674586776859506,
+      "grad_norm": 3.21875,
+      "learning_rate": 7.12606034480066e-05,
+      "loss": 4.2759,
+      "step": 344700
+    },
+    {
+      "epoch": 2.9678891184573004,
+      "grad_norm": 2.890625,
+      "learning_rate": 7.123465541206714e-05,
+      "loss": 3.9885,
+      "step": 344750
+    },
+    {
+      "epoch": 2.96831955922865,
+      "grad_norm": 2.59375,
+      "learning_rate": 7.120870948744202e-05,
+      "loss": 4.4197,
+      "step": 344800
+    },
+    {
+      "epoch": 2.96875,
+      "grad_norm": 2.984375,
+      "learning_rate": 7.118276567603574e-05,
+      "loss": 4.2508,
+      "step": 344850
+    },
+    {
+      "epoch": 2.96918044077135,
+      "grad_norm": 3.609375,
+      "learning_rate": 7.115682397975242e-05,
+      "loss": 4.3649,
+      "step": 344900
+    },
+    {
+      "epoch": 2.9696108815426996,
+      "grad_norm": 3.25,
+      "learning_rate": 7.113088440049617e-05,
+      "loss": 4.3533,
+      "step": 344950
+    },
+    {
+      "epoch": 2.9700413223140494,
+      "grad_norm": 2.703125,
+      "learning_rate": 7.110494694017089e-05,
+      "loss": 4.5324,
+      "step": 345000
+    },
+    {
+      "epoch": 2.9700413223140494,
+      "eval_loss": 4.992117404937744,
+      "eval_runtime": 24.5137,
+      "eval_samples_per_second": 26.108,
+      "eval_steps_per_second": 13.054,
+      "eval_tts_loss": 7.4279264540100876,
+      "step": 345000
+    },
+    {
+      "epoch": 2.970471763085399,
+      "grad_norm": 3.640625,
+      "learning_rate": 7.107901160068032e-05,
+      "loss": 4.0649,
+      "step": 345050
+    },
+    {
+      "epoch": 2.9709022038567494,
+      "grad_norm": 3.703125,
+      "learning_rate": 7.105307838392812e-05,
+      "loss": 4.1369,
+      "step": 345100
+    },
+    {
+      "epoch": 2.971332644628099,
+      "grad_norm": 4.5625,
+      "learning_rate": 7.102714729181767e-05,
+      "loss": 4.4068,
+      "step": 345150
+    },
+    {
+      "epoch": 2.971763085399449,
+      "grad_norm": 1.96875,
+      "learning_rate": 7.100121832625224e-05,
+      "loss": 4.3105,
+      "step": 345200
+    },
+    {
+      "epoch": 2.972193526170799,
+      "grad_norm": 1.5,
+      "learning_rate": 7.097529148913503e-05,
+      "loss": 4.1333,
+      "step": 345250
+    },
+    {
+      "epoch": 2.972623966942149,
+      "grad_norm": 2.71875,
+      "learning_rate": 7.094936678236894e-05,
+      "loss": 4.4415,
+      "step": 345300
+    },
+    {
+      "epoch": 2.973054407713499,
+      "grad_norm": 3.328125,
+      "learning_rate": 7.092344420785686e-05,
+      "loss": 4.3307,
+      "step": 345350
+    },
+    {
+      "epoch": 2.9734848484848486,
+      "grad_norm": 3.65625,
+      "learning_rate": 7.089752376750138e-05,
+      "loss": 4.5526,
+      "step": 345400
+    },
+    {
+      "epoch": 2.9739152892561984,
+      "grad_norm": 3.0,
+      "learning_rate": 7.087160546320506e-05,
+      "loss": 4.5478,
+      "step": 345450
+    },
+    {
+      "epoch": 2.974345730027548,
+      "grad_norm": 4.6875,
+      "learning_rate": 7.084568929687022e-05,
+      "loss": 3.944,
+      "step": 345500
+    },
+    {
+      "epoch": 2.974776170798898,
+      "grad_norm": 1.765625,
+      "learning_rate": 7.081977527039905e-05,
+      "loss": 4.3547,
+      "step": 345550
+    },
+    {
+      "epoch": 2.975206611570248,
+      "grad_norm": 3.3125,
+      "learning_rate": 7.079386338569363e-05,
+      "loss": 4.3263,
+      "step": 345600
+    },
+    {
+      "epoch": 2.9756370523415976,
+      "grad_norm": 2.171875,
+      "learning_rate": 7.076795364465578e-05,
+      "loss": 4.5249,
+      "step": 345650
+    },
+    {
+      "epoch": 2.9760674931129474,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.074204604918722e-05,
+      "loss": 4.5851,
+      "step": 345700
+    },
+    {
+      "epoch": 2.9764979338842976,
+      "grad_norm": 2.9375,
+      "learning_rate": 7.071614060118954e-05,
+      "loss": 4.3908,
+      "step": 345750
+    },
+    {
+      "epoch": 2.9769283746556474,
+      "grad_norm": 3.90625,
+      "learning_rate": 7.069023730256413e-05,
+      "loss": 4.6488,
+      "step": 345800
+    },
+    {
+      "epoch": 2.977358815426997,
+      "grad_norm": 5.625,
+      "learning_rate": 7.066433615521228e-05,
+      "loss": 4.5108,
+      "step": 345850
+    },
+    {
+      "epoch": 2.977789256198347,
+      "grad_norm": 4.40625,
+      "learning_rate": 7.0638437161035e-05,
+      "loss": 4.0877,
+      "step": 345900
+    },
+    {
+      "epoch": 2.9782196969696972,
+      "grad_norm": 4.1875,
+      "learning_rate": 7.061254032193327e-05,
+      "loss": 4.3139,
+      "step": 345950
+    },
+    {
+      "epoch": 2.978650137741047,
+      "grad_norm": 2.6875,
+      "learning_rate": 7.058664563980783e-05,
+      "loss": 4.5791,
+      "step": 346000
+    },
+    {
+      "epoch": 2.979080578512397,
+      "grad_norm": 2.5,
+      "learning_rate": 7.056075311655932e-05,
+      "loss": 4.2769,
+      "step": 346050
+    },
+    {
+      "epoch": 2.9795110192837466,
+      "grad_norm": 1.875,
+      "learning_rate": 7.053486275408822e-05,
+      "loss": 4.3123,
+      "step": 346100
+    },
+    {
+      "epoch": 2.9799414600550964,
+      "grad_norm": 3.015625,
+      "learning_rate": 7.050897455429479e-05,
+      "loss": 4.254,
+      "step": 346150
+    },
+    {
+      "epoch": 2.980371900826446,
+      "grad_norm": 3.359375,
+      "learning_rate": 7.048308851907916e-05,
+      "loss": 4.4651,
+      "step": 346200
+    },
+    {
+      "epoch": 2.980802341597796,
+      "grad_norm": 3.359375,
+      "learning_rate": 7.045720465034132e-05,
+      "loss": 4.1671,
+      "step": 346250
+    },
+    {
+      "epoch": 2.981232782369146,
+      "grad_norm": 2.734375,
+      "learning_rate": 7.043132294998112e-05,
+      "loss": 4.3572,
+      "step": 346300
+    },
+    {
+      "epoch": 2.981663223140496,
+      "grad_norm": 5.53125,
+      "learning_rate": 7.040544341989821e-05,
+      "loss": 4.4563,
+      "step": 346350
+    },
+    {
+      "epoch": 2.982093663911846,
+      "grad_norm": 4.0,
+      "learning_rate": 7.037956606199202e-05,
+      "loss": 4.4577,
+      "step": 346400
+    },
+    {
+      "epoch": 2.9825241046831956,
+      "grad_norm": 2.515625,
+      "learning_rate": 7.0353690878162e-05,
+      "loss": 4.0317,
+      "step": 346450
+    },
+    {
+      "epoch": 2.9829545454545454,
+      "grad_norm": 2.5,
+      "learning_rate": 7.032781787030727e-05,
+      "loss": 4.3282,
+      "step": 346500
+    },
+    {
+      "epoch": 2.983384986225895,
+      "grad_norm": 3.171875,
+      "learning_rate": 7.030194704032685e-05,
+      "loss": 4.5129,
+      "step": 346550
+    },
+    {
+      "epoch": 2.9838154269972454,
+      "grad_norm": 3.171875,
+      "learning_rate": 7.027607839011963e-05,
+      "loss": 4.4157,
+      "step": 346600
+    },
+    {
+      "epoch": 2.9842458677685952,
+      "grad_norm": 1.8203125,
+      "learning_rate": 7.025021192158433e-05,
+      "loss": 4.1739,
+      "step": 346650
+    },
+    {
+      "epoch": 2.984676308539945,
+      "grad_norm": 3.09375,
+      "learning_rate": 7.022434763661945e-05,
+      "loss": 4.338,
+      "step": 346700
+    },
+    {
+      "epoch": 2.985106749311295,
+      "grad_norm": 3.40625,
+      "learning_rate": 7.019848553712337e-05,
+      "loss": 4.2847,
+      "step": 346750
+    },
+    {
+      "epoch": 2.9855371900826446,
+      "grad_norm": 5.0,
+      "learning_rate": 7.017262562499436e-05,
+      "loss": 3.9216,
+      "step": 346800
+    },
+    {
+      "epoch": 2.9859676308539944,
+      "grad_norm": 3.71875,
+      "learning_rate": 7.014676790213042e-05,
+      "loss": 4.4951,
+      "step": 346850
+    },
+    {
+      "epoch": 2.986398071625344,
+      "grad_norm": 4.3125,
+      "learning_rate": 7.012091237042955e-05,
+      "loss": 4.6992,
+      "step": 346900
+    },
+    {
+      "epoch": 2.986828512396694,
+      "grad_norm": 1.625,
+      "learning_rate": 7.009505903178936e-05,
+      "loss": 4.6149,
+      "step": 346950
+    },
+    {
+      "epoch": 2.9872589531680442,
+      "grad_norm": 3.8125,
+      "learning_rate": 7.006920788810752e-05,
+      "loss": 4.3064,
+      "step": 347000
+    },
+    {
+      "epoch": 2.987689393939394,
+      "grad_norm": 2.140625,
+      "learning_rate": 7.00433589412814e-05,
+      "loss": 4.2437,
+      "step": 347050
+    },
+    {
+      "epoch": 2.988119834710744,
+      "grad_norm": 3.640625,
+      "learning_rate": 7.001751219320829e-05,
+      "loss": 4.0124,
+      "step": 347100
+    },
+    {
+      "epoch": 2.9885502754820936,
+      "grad_norm": 2.640625,
+      "learning_rate": 6.999166764578532e-05,
+      "loss": 4.1106,
+      "step": 347150
+    },
+    {
+      "epoch": 2.9889807162534434,
+      "grad_norm": 1.6484375,
+      "learning_rate": 6.996582530090936e-05,
+      "loss": 3.7507,
+      "step": 347200
+    },
+    {
+      "epoch": 2.9894111570247937,
+      "grad_norm": 3.03125,
+      "learning_rate": 6.993998516047717e-05,
+      "loss": 4.8152,
+      "step": 347250
+    },
+    {
+      "epoch": 2.9898415977961434,
+      "grad_norm": 2.8125,
+      "learning_rate": 6.99141472263854e-05,
+      "loss": 4.5709,
+      "step": 347300
+    },
+    {
+      "epoch": 2.9902720385674932,
+      "grad_norm": 2.265625,
+      "learning_rate": 6.988831150053051e-05,
+      "loss": 4.0828,
+      "step": 347350
+    },
+    {
+      "epoch": 2.990702479338843,
+      "grad_norm": 3.390625,
+      "learning_rate": 6.98624779848088e-05,
+      "loss": 4.6084,
+      "step": 347400
+    },
+    {
+      "epoch": 2.991132920110193,
+      "grad_norm": 4.5,
+      "learning_rate": 6.983664668111632e-05,
+      "loss": 4.2643,
+      "step": 347450
+    },
+    {
+      "epoch": 2.9915633608815426,
+      "grad_norm": 3.5,
+      "learning_rate": 6.981081759134907e-05,
+      "loss": 4.4713,
+      "step": 347500
+    },
+    {
+      "epoch": 2.9919938016528924,
+      "grad_norm": 3.265625,
+      "learning_rate": 6.978499071740286e-05,
+      "loss": 4.2272,
+      "step": 347550
+    },
+    {
+      "epoch": 2.992424242424242,
+      "grad_norm": 3.921875,
+      "learning_rate": 6.975916606117332e-05,
+      "loss": 4.6481,
+      "step": 347600
+    },
+    {
+      "epoch": 2.9928546831955924,
+      "grad_norm": 2.828125,
+      "learning_rate": 6.973334362455597e-05,
+      "loss": 4.2598,
+      "step": 347650
+    },
+    {
+      "epoch": 2.9932851239669422,
+      "grad_norm": 3.671875,
+      "learning_rate": 6.970752340944604e-05,
+      "loss": 3.8855,
+      "step": 347700
+    },
+    {
+      "epoch": 2.993715564738292,
+      "grad_norm": 3.421875,
+      "learning_rate": 6.968170541773871e-05,
+      "loss": 4.099,
+      "step": 347750
+    },
+    {
+      "epoch": 2.994146005509642,
+      "grad_norm": 2.84375,
+      "learning_rate": 6.965588965132896e-05,
+      "loss": 4.5199,
+      "step": 347800
+    },
+    {
+      "epoch": 2.9945764462809916,
+      "grad_norm": 1.8125,
+      "learning_rate": 6.963007611211166e-05,
+      "loss": 4.1152,
+      "step": 347850
+    },
+    {
+      "epoch": 2.995006887052342,
+      "grad_norm": 4.125,
+      "learning_rate": 6.960426480198143e-05,
+      "loss": 3.9402,
+      "step": 347900
+    },
+    {
+      "epoch": 2.9954373278236917,
+      "grad_norm": 3.515625,
+      "learning_rate": 6.957845572283275e-05,
+      "loss": 4.5469,
+      "step": 347950
+    },
+    {
+      "epoch": 2.9958677685950414,
+      "grad_norm": 2.84375,
+      "learning_rate": 6.955264887655995e-05,
+      "loss": 4.2333,
+      "step": 348000
+    },
+    {
+      "epoch": 2.9958677685950414,
+      "eval_loss": 4.991213321685791,
+      "eval_runtime": 24.0364,
+      "eval_samples_per_second": 26.626,
+      "eval_steps_per_second": 13.313,
+      "eval_tts_loss": 7.409507157671467,
+      "step": 348000
+    },
+    {
+      "epoch": 2.9962982093663912,
+      "grad_norm": 1.171875,
+      "learning_rate": 6.952684426505724e-05,
+      "loss": 4.1471,
+      "step": 348050
+    },
+    {
+      "epoch": 2.996728650137741,
+      "grad_norm": 3.25,
+      "learning_rate": 6.950104189021856e-05,
+      "loss": 3.7776,
+      "step": 348100
+    },
+    {
+      "epoch": 2.997159090909091,
+      "grad_norm": 2.796875,
+      "learning_rate": 6.947524175393785e-05,
+      "loss": 4.6694,
+      "step": 348150
+    },
+    {
+      "epoch": 2.9975895316804406,
+      "grad_norm": 3.828125,
+      "learning_rate": 6.944944385810868e-05,
+      "loss": 4.5849,
+      "step": 348200
+    },
+    {
+      "epoch": 2.9980199724517904,
+      "grad_norm": 4.09375,
+      "learning_rate": 6.94236482046246e-05,
+      "loss": 4.2124,
+      "step": 348250
+    },
+    {
+      "epoch": 2.9984504132231407,
+      "grad_norm": 3.8125,
+      "learning_rate": 6.939785479537893e-05,
+      "loss": 4.3135,
+      "step": 348300
+    },
+    {
+      "epoch": 2.9988808539944904,
+      "grad_norm": 3.015625,
+      "learning_rate": 6.93720636322649e-05,
+      "loss": 4.493,
+      "step": 348350
+    },
+    {
+      "epoch": 2.9993112947658402,
+      "grad_norm": 3.90625,
+      "learning_rate": 6.934627471717549e-05,
+      "loss": 4.5723,
+      "step": 348400
+    },
+    {
+      "epoch": 2.99974173553719,
+      "grad_norm": 5.875,
+      "learning_rate": 6.93204880520036e-05,
+      "loss": 4.3788,
+      "step": 348450
+    },
+    {
+      "epoch": 3.00017217630854,
+      "grad_norm": 1.3671875,
+      "learning_rate": 6.929470363864185e-05,
+      "loss": 4.0193,
+      "step": 348500
+    },
+    {
+      "epoch": 3.0006026170798896,
+      "grad_norm": 3.109375,
+      "learning_rate": 6.92689214789828e-05,
+      "loss": 4.0593,
+      "step": 348550
+    },
+    {
+      "epoch": 3.00103305785124,
+      "grad_norm": 4.5625,
+      "learning_rate": 6.924314157491877e-05,
+      "loss": 4.3015,
+      "step": 348600
+    },
+    {
+      "epoch": 3.0014634986225897,
+      "grad_norm": 4.28125,
+      "learning_rate": 6.921736392834197e-05,
+      "loss": 4.4952,
+      "step": 348650
+    },
+    {
+      "epoch": 3.0018939393939394,
+      "grad_norm": 2.734375,
+      "learning_rate": 6.919158854114447e-05,
+      "loss": 4.2368,
+      "step": 348700
+    },
+    {
+      "epoch": 3.0023243801652892,
+      "grad_norm": 3.765625,
+      "learning_rate": 6.916581541521807e-05,
+      "loss": 4.4221,
+      "step": 348750
+    },
+    {
+      "epoch": 3.002754820936639,
+      "grad_norm": 2.9375,
+      "learning_rate": 6.914004455245446e-05,
+      "loss": 3.8055,
+      "step": 348800
+    },
+    {
+      "epoch": 3.003185261707989,
+      "grad_norm": 3.546875,
+      "learning_rate": 6.91142759547452e-05,
+      "loss": 4.6014,
+      "step": 348850
+    },
+    {
+      "epoch": 3.0036157024793386,
+      "grad_norm": 2.515625,
+      "learning_rate": 6.908850962398162e-05,
+      "loss": 4.2205,
+      "step": 348900
+    },
+    {
+      "epoch": 3.004046143250689,
+      "grad_norm": 2.203125,
+      "learning_rate": 6.906274556205495e-05,
+      "loss": 4.2182,
+      "step": 348950
+    },
+    {
+      "epoch": 3.0044765840220387,
+      "grad_norm": 5.28125,
+      "learning_rate": 6.903698377085614e-05,
+      "loss": 4.0105,
+      "step": 349000
+    },
+    {
+      "epoch": 3.0049070247933884,
+      "grad_norm": 5.53125,
+      "learning_rate": 6.901122425227615e-05,
+      "loss": 4.2359,
+      "step": 349050
+    },
+    {
+      "epoch": 3.0053374655647382,
+      "grad_norm": 3.640625,
+      "learning_rate": 6.89854670082056e-05,
+      "loss": 4.2774,
+      "step": 349100
+    },
+    {
+      "epoch": 3.005767906336088,
+      "grad_norm": 1.9140625,
+      "learning_rate": 6.895971204053504e-05,
+      "loss": 4.3043,
+      "step": 349150
+    },
+    {
+      "epoch": 3.006198347107438,
+      "grad_norm": 2.5625,
+      "learning_rate": 6.893395935115487e-05,
+      "loss": 4.2211,
+      "step": 349200
+    },
+    {
+      "epoch": 3.006628787878788,
+      "grad_norm": 1.6640625,
+      "learning_rate": 6.890820894195522e-05,
+      "loss": 4.2544,
+      "step": 349250
+    },
+    {
+      "epoch": 3.007059228650138,
+      "grad_norm": 2.359375,
+      "learning_rate": 6.888246081482613e-05,
+      "loss": 4.342,
+      "step": 349300
+    },
+    {
+      "epoch": 3.0074896694214877,
+      "grad_norm": 2.921875,
+      "learning_rate": 6.885671497165745e-05,
+      "loss": 4.3731,
+      "step": 349350
+    },
+    {
+      "epoch": 3.0079201101928374,
+      "grad_norm": 4.21875,
+      "learning_rate": 6.883097141433892e-05,
+      "loss": 4.0846,
+      "step": 349400
+    },
+    {
+      "epoch": 3.0083505509641872,
+      "grad_norm": 4.40625,
+      "learning_rate": 6.880523014476004e-05,
+      "loss": 4.2551,
+      "step": 349450
+    },
+    {
+      "epoch": 3.008780991735537,
+      "grad_norm": 2.671875,
+      "learning_rate": 6.87794911648101e-05,
+      "loss": 4.831,
+      "step": 349500
+    },
+    {
+      "epoch": 3.009211432506887,
+      "grad_norm": 2.125,
+      "learning_rate": 6.875375447637836e-05,
+      "loss": 4.7646,
+      "step": 349550
+    },
+    {
+      "epoch": 3.009641873278237,
+      "grad_norm": 1.671875,
+      "learning_rate": 6.87280200813538e-05,
+      "loss": 3.9587,
+      "step": 349600
+    },
+    {
+      "epoch": 3.010072314049587,
+      "grad_norm": 2.515625,
+      "learning_rate": 6.870228798162528e-05,
+      "loss": 4.1138,
+      "step": 349650
+    },
+    {
+      "epoch": 3.0105027548209367,
+      "grad_norm": 2.015625,
+      "learning_rate": 6.867655817908152e-05,
+      "loss": 4.4146,
+      "step": 349700
+    },
+    {
+      "epoch": 3.0109331955922864,
+      "grad_norm": 3.5,
+      "learning_rate": 6.865083067561098e-05,
+      "loss": 4.4591,
+      "step": 349750
+    },
+    {
+      "epoch": 3.0113636363636362,
+      "grad_norm": 4.28125,
+      "learning_rate": 6.862510547310199e-05,
+      "loss": 4.1988,
+      "step": 349800
+    },
+    {
+      "epoch": 3.011794077134986,
+      "grad_norm": 3.828125,
+      "learning_rate": 6.859938257344277e-05,
+      "loss": 4.5588,
+      "step": 349850
+    },
+    {
+      "epoch": 3.0122245179063363,
+      "grad_norm": 2.671875,
+      "learning_rate": 6.857366197852132e-05,
+      "loss": 4.3576,
+      "step": 349900
+    },
+    {
+      "epoch": 3.012654958677686,
+      "grad_norm": 3.109375,
+      "learning_rate": 6.854794369022548e-05,
+      "loss": 4.2082,
+      "step": 349950
+    },
+    {
+      "epoch": 3.013085399449036,
+      "grad_norm": 1.625,
+      "learning_rate": 6.852222771044287e-05,
+      "loss": 4.2093,
+      "step": 350000
+    },
+    {
+      "epoch": 3.0135158402203857,
+      "grad_norm": 3.140625,
+      "learning_rate": 6.849651404106102e-05,
+      "loss": 4.2408,
+      "step": 350050
+    },
+    {
+      "epoch": 3.0139462809917354,
+      "grad_norm": 4.375,
+      "learning_rate": 6.847080268396728e-05,
+      "loss": 4.0275,
+      "step": 350100
+    },
+    {
+      "epoch": 3.0143767217630852,
+      "grad_norm": 4.375,
+      "learning_rate": 6.844509364104876e-05,
+      "loss": 4.0008,
+      "step": 350150
+    },
+    {
+      "epoch": 3.0148071625344355,
+      "grad_norm": 4.59375,
+      "learning_rate": 6.84193869141925e-05,
+      "loss": 4.1721,
+      "step": 350200
+    },
+    {
+      "epoch": 3.0152376033057853,
+      "grad_norm": 1.859375,
+      "learning_rate": 6.839368250528529e-05,
+      "loss": 4.5617,
+      "step": 350250
+    },
+    {
+      "epoch": 3.015668044077135,
+      "grad_norm": 5.4375,
+      "learning_rate": 6.836798041621377e-05,
+      "loss": 4.3456,
+      "step": 350300
+    },
+    {
+      "epoch": 3.016098484848485,
+      "grad_norm": 3.515625,
+      "learning_rate": 6.834228064886442e-05,
+      "loss": 4.0823,
+      "step": 350350
+    },
+    {
+      "epoch": 3.0165289256198347,
+      "grad_norm": 1.9453125,
+      "learning_rate": 6.831658320512357e-05,
+      "loss": 3.9468,
+      "step": 350400
+    },
+    {
+      "epoch": 3.0169593663911844,
+      "grad_norm": 3.09375,
+      "learning_rate": 6.829088808687734e-05,
+      "loss": 4.402,
+      "step": 350450
+    },
+    {
+      "epoch": 3.0173898071625342,
+      "grad_norm": 4.46875,
+      "learning_rate": 6.826519529601174e-05,
+      "loss": 4.019,
+      "step": 350500
+    },
+    {
+      "epoch": 3.0178202479338845,
+      "grad_norm": 2.71875,
+      "learning_rate": 6.823950483441247e-05,
+      "loss": 4.2083,
+      "step": 350550
+    },
+    {
+      "epoch": 3.0182506887052343,
+      "grad_norm": 5.03125,
+      "learning_rate": 6.821381670396524e-05,
+      "loss": 4.0946,
+      "step": 350600
+    },
+    {
+      "epoch": 3.018681129476584,
+      "grad_norm": 1.6171875,
+      "learning_rate": 6.818813090655544e-05,
+      "loss": 4.2547,
+      "step": 350650
+    },
+    {
+      "epoch": 3.019111570247934,
+      "grad_norm": 3.703125,
+      "learning_rate": 6.816244744406841e-05,
+      "loss": 4.1576,
+      "step": 350700
+    },
+    {
+      "epoch": 3.0195420110192837,
+      "grad_norm": 3.640625,
+      "learning_rate": 6.813676631838926e-05,
+      "loss": 4.4224,
+      "step": 350750
+    },
+    {
+      "epoch": 3.0199724517906334,
+      "grad_norm": 2.890625,
+      "learning_rate": 6.81110875314029e-05,
+      "loss": 4.1079,
+      "step": 350800
+    },
+    {
+      "epoch": 3.0204028925619837,
+      "grad_norm": 3.4375,
+      "learning_rate": 6.808541108499407e-05,
+      "loss": 4.2391,
+      "step": 350850
+    },
+    {
+      "epoch": 3.0208333333333335,
+      "grad_norm": 4.28125,
+      "learning_rate": 6.805973698104742e-05,
+      "loss": 4.4265,
+      "step": 350900
+    },
+    {
+      "epoch": 3.0212637741046833,
+      "grad_norm": 1.4609375,
+      "learning_rate": 6.803406522144735e-05,
+      "loss": 4.4175,
+      "step": 350950
+    },
+    {
+      "epoch": 3.021694214876033,
+      "grad_norm": 3.5,
+      "learning_rate": 6.800839580807816e-05,
+      "loss": 4.0121,
+      "step": 351000
+    },
+    {
+      "epoch": 3.021694214876033,
+      "eval_loss": 4.993182182312012,
+      "eval_runtime": 25.3417,
+      "eval_samples_per_second": 25.255,
+      "eval_steps_per_second": 12.627,
+      "eval_tts_loss": 7.477537839047033,
+      "step": 351000
+    },
+    {
+      "epoch": 3.022124655647383,
+      "grad_norm": 2.8125,
+      "learning_rate": 6.798272874282383e-05,
+      "loss": 4.2912,
+      "step": 351050
+    },
+    {
+      "epoch": 3.0225550964187327,
+      "grad_norm": 3.4375,
+      "learning_rate": 6.795706402756835e-05,
+      "loss": 4.4703,
+      "step": 351100
+    },
+    {
+      "epoch": 3.0229855371900825,
+      "grad_norm": 5.34375,
+      "learning_rate": 6.79314016641954e-05,
+      "loss": 4.2438,
+      "step": 351150
+    },
+    {
+      "epoch": 3.0234159779614327,
+      "grad_norm": 3.140625,
+      "learning_rate": 6.790574165458858e-05,
+      "loss": 4.2153,
+      "step": 351200
+    },
+    {
+      "epoch": 3.0238464187327825,
+      "grad_norm": 2.265625,
+      "learning_rate": 6.78800840006313e-05,
+      "loss": 4.5285,
+      "step": 351250
+    },
+    {
+      "epoch": 3.0242768595041323,
+      "grad_norm": 2.15625,
+      "learning_rate": 6.785442870420674e-05,
+      "loss": 4.4665,
+      "step": 351300
+    },
+    {
+      "epoch": 3.024707300275482,
+      "grad_norm": 4.0625,
+      "learning_rate": 6.782877576719793e-05,
+      "loss": 4.1483,
+      "step": 351350
+    },
+    {
+      "epoch": 3.025137741046832,
+      "grad_norm": 2.1875,
+      "learning_rate": 6.780312519148775e-05,
+      "loss": 4.2245,
+      "step": 351400
+    },
+    {
+      "epoch": 3.0255681818181817,
+      "grad_norm": 1.765625,
+      "learning_rate": 6.777747697895893e-05,
+      "loss": 4.2463,
+      "step": 351450
+    },
+    {
+      "epoch": 3.025998622589532,
+      "grad_norm": 2.125,
+      "learning_rate": 6.775183113149399e-05,
+      "loss": 4.3165,
+      "step": 351500
+    },
+    {
+      "epoch": 3.0264290633608817,
+      "grad_norm": 2.53125,
+      "learning_rate": 6.772618765097522e-05,
+      "loss": 4.3179,
+      "step": 351550
+    },
+    {
+      "epoch": 3.0268595041322315,
+      "grad_norm": 2.953125,
+      "learning_rate": 6.770054653928485e-05,
+      "loss": 4.4555,
+      "step": 351600
+    },
+    {
+      "epoch": 3.0272899449035813,
+      "grad_norm": 3.1875,
+      "learning_rate": 6.767490779830487e-05,
+      "loss": 4.261,
+      "step": 351650
+    },
+    {
+      "epoch": 3.027720385674931,
+      "grad_norm": 3.703125,
+      "learning_rate": 6.76492714299171e-05,
+      "loss": 4.2654,
+      "step": 351700
+    },
+    {
+      "epoch": 3.028150826446281,
+      "grad_norm": 4.40625,
+      "learning_rate": 6.762363743600325e-05,
+      "loss": 4.1454,
+      "step": 351750
+    },
+    {
+      "epoch": 3.0285812672176307,
+      "grad_norm": 2.96875,
+      "learning_rate": 6.759800581844474e-05,
+      "loss": 4.2149,
+      "step": 351800
+    },
+    {
+      "epoch": 3.029011707988981,
+      "grad_norm": 3.65625,
+      "learning_rate": 6.757237657912283e-05,
+      "loss": 4.4111,
+      "step": 351850
+    },
+    {
+      "epoch": 3.0294421487603307,
+      "grad_norm": 3.484375,
+      "learning_rate": 6.754674971991875e-05,
+      "loss": 4.2162,
+      "step": 351900
+    },
+    {
+      "epoch": 3.0298725895316805,
+      "grad_norm": 2.84375,
+      "learning_rate": 6.752112524271341e-05,
+      "loss": 4.3516,
+      "step": 351950
+    },
+    {
+      "epoch": 3.0303030303030303,
+      "grad_norm": 5.28125,
+      "learning_rate": 6.749550314938758e-05,
+      "loss": 4.2341,
+      "step": 352000
+    },
+    {
+      "epoch": 3.03073347107438,
+      "grad_norm": 3.3125,
+      "learning_rate": 6.746988344182192e-05,
+      "loss": 4.4311,
+      "step": 352050
+    },
+    {
+      "epoch": 3.03116391184573,
+      "grad_norm": 3.171875,
+      "learning_rate": 6.74442661218968e-05,
+      "loss": 4.5223,
+      "step": 352100
+    },
+    {
+      "epoch": 3.03159435261708,
+      "grad_norm": 2.765625,
+      "learning_rate": 6.741865119149252e-05,
+      "loss": 4.2839,
+      "step": 352150
+    },
+    {
+      "epoch": 3.03202479338843,
+      "grad_norm": 2.4375,
+      "learning_rate": 6.739303865248911e-05,
+      "loss": 4.2244,
+      "step": 352200
+    },
+    {
+      "epoch": 3.0324552341597797,
+      "grad_norm": 4.40625,
+      "learning_rate": 6.736742850676652e-05,
+      "loss": 4.4552,
+      "step": 352250
+    },
+    {
+      "epoch": 3.0328856749311295,
+      "grad_norm": 2.984375,
+      "learning_rate": 6.73418207562045e-05,
+      "loss": 4.2598,
+      "step": 352300
+    },
+    {
+      "epoch": 3.0333161157024793,
+      "grad_norm": 4.65625,
+      "learning_rate": 6.731621540268257e-05,
+      "loss": 4.6048,
+      "step": 352350
+    },
+    {
+      "epoch": 3.033746556473829,
+      "grad_norm": 2.96875,
+      "learning_rate": 6.729061244808006e-05,
+      "loss": 4.5114,
+      "step": 352400
+    },
+    {
+      "epoch": 3.034176997245179,
+      "grad_norm": 3.65625,
+      "learning_rate": 6.726501189427624e-05,
+      "loss": 4.3409,
+      "step": 352450
+    },
+    {
+      "epoch": 3.034607438016529,
+      "grad_norm": 2.140625,
+      "learning_rate": 6.72394137431501e-05,
+      "loss": 4.1391,
+      "step": 352500
+    },
+    {
+      "epoch": 3.035037878787879,
+      "grad_norm": 1.9921875,
+      "learning_rate": 6.721381799658057e-05,
+      "loss": 4.074,
+      "step": 352550
+    },
+    {
+      "epoch": 3.0354683195592287,
+      "grad_norm": 1.96875,
+      "learning_rate": 6.718822465644619e-05,
+      "loss": 4.5309,
+      "step": 352600
+    },
+    {
+      "epoch": 3.0358987603305785,
+      "grad_norm": 2.546875,
+      "learning_rate": 6.716263372462554e-05,
+      "loss": 4.1559,
+      "step": 352650
+    },
+    {
+      "epoch": 3.0363292011019283,
+      "grad_norm": 1.859375,
+      "learning_rate": 6.713704520299694e-05,
+      "loss": 4.3863,
+      "step": 352700
+    },
+    {
+      "epoch": 3.036759641873278,
+      "grad_norm": 3.21875,
+      "learning_rate": 6.711145909343848e-05,
+      "loss": 4.4605,
+      "step": 352750
+    },
+    {
+      "epoch": 3.0371900826446283,
+      "grad_norm": 3.5,
+      "learning_rate": 6.708587539782821e-05,
+      "loss": 4.263,
+      "step": 352800
+    },
+    {
+      "epoch": 3.037620523415978,
+      "grad_norm": 4.9375,
+      "learning_rate": 6.706029411804386e-05,
+      "loss": 4.2533,
+      "step": 352850
+    },
+    {
+      "epoch": 3.038050964187328,
+      "grad_norm": 4.15625,
+      "learning_rate": 6.703471525596301e-05,
+      "loss": 4.565,
+      "step": 352900
+    },
+    {
+      "epoch": 3.0384814049586777,
+      "grad_norm": 3.078125,
+      "learning_rate": 6.700913881346316e-05,
+      "loss": 4.2158,
+      "step": 352950
+    },
+    {
+      "epoch": 3.0389118457300275,
+      "grad_norm": 4.71875,
+      "learning_rate": 6.698356479242154e-05,
+      "loss": 4.629,
+      "step": 353000
+    },
+    {
+      "epoch": 3.0393422865013773,
+      "grad_norm": 2.65625,
+      "learning_rate": 6.695799319471525e-05,
+      "loss": 4.5001,
+      "step": 353050
+    },
+    {
+      "epoch": 3.039772727272727,
+      "grad_norm": 4.3125,
+      "learning_rate": 6.693242402222113e-05,
+      "loss": 4.2241,
+      "step": 353100
+    },
+    {
+      "epoch": 3.0402031680440773,
+      "grad_norm": 3.796875,
+      "learning_rate": 6.690685727681594e-05,
+      "loss": 4.0977,
+      "step": 353150
+    },
+    {
+      "epoch": 3.040633608815427,
+      "grad_norm": 2.125,
+      "learning_rate": 6.688129296037624e-05,
+      "loss": 4.4082,
+      "step": 353200
+    },
+    {
+      "epoch": 3.041064049586777,
+      "grad_norm": 4.0,
+      "learning_rate": 6.685573107477836e-05,
+      "loss": 4.0613,
+      "step": 353250
+    },
+    {
+      "epoch": 3.0414944903581267,
+      "grad_norm": 4.21875,
+      "learning_rate": 6.683017162189855e-05,
+      "loss": 4.1511,
+      "step": 353300
+    },
+    {
+      "epoch": 3.0419249311294765,
+      "grad_norm": 5.03125,
+      "learning_rate": 6.680461460361277e-05,
+      "loss": 4.8417,
+      "step": 353350
+    },
+    {
+      "epoch": 3.0423553719008263,
+      "grad_norm": 1.9375,
+      "learning_rate": 6.677906002179681e-05,
+      "loss": 4.3596,
+      "step": 353400
+    },
+    {
+      "epoch": 3.0427858126721765,
+      "grad_norm": 3.171875,
+      "learning_rate": 6.675350787832639e-05,
+      "loss": 4.5467,
+      "step": 353450
+    },
+    {
+      "epoch": 3.0432162534435263,
+      "grad_norm": 3.328125,
+      "learning_rate": 6.672795817507697e-05,
+      "loss": 4.5393,
+      "step": 353500
+    },
+    {
+      "epoch": 3.043646694214876,
+      "grad_norm": 2.828125,
+      "learning_rate": 6.670241091392386e-05,
+      "loss": 4.4852,
+      "step": 353550
+    },
+    {
+      "epoch": 3.044077134986226,
+      "grad_norm": 3.609375,
+      "learning_rate": 6.66768660967421e-05,
+      "loss": 4.1685,
+      "step": 353600
+    },
+    {
+      "epoch": 3.0445075757575757,
+      "grad_norm": 4.3125,
+      "learning_rate": 6.665132372540668e-05,
+      "loss": 4.2069,
+      "step": 353650
+    },
+    {
+      "epoch": 3.0449380165289255,
+      "grad_norm": 4.71875,
+      "learning_rate": 6.662578380179236e-05,
+      "loss": 4.3321,
+      "step": 353700
+    },
+    {
+      "epoch": 3.0453684573002753,
+      "grad_norm": 1.8828125,
+      "learning_rate": 6.660024632777368e-05,
+      "loss": 4.2612,
+      "step": 353750
+    },
+    {
+      "epoch": 3.0457988980716255,
+      "grad_norm": 1.640625,
+      "learning_rate": 6.657471130522508e-05,
+      "loss": 3.9872,
+      "step": 353800
+    },
+    {
+      "epoch": 3.0462293388429753,
+      "grad_norm": 3.40625,
+      "learning_rate": 6.654917873602075e-05,
+      "loss": 4.615,
+      "step": 353850
+    },
+    {
+      "epoch": 3.046659779614325,
+      "grad_norm": 1.453125,
+      "learning_rate": 6.652364862203474e-05,
+      "loss": 4.3082,
+      "step": 353900
+    },
+    {
+      "epoch": 3.047090220385675,
+      "grad_norm": 3.875,
+      "learning_rate": 6.649812096514087e-05,
+      "loss": 4.5583,
+      "step": 353950
+    },
+    {
+      "epoch": 3.0475206611570247,
+      "grad_norm": 4.5,
+      "learning_rate": 6.647259576721288e-05,
+      "loss": 4.2475,
+      "step": 354000
+    },
+    {
+      "epoch": 3.0475206611570247,
+      "eval_loss": 4.994631767272949,
+      "eval_runtime": 24.3067,
+      "eval_samples_per_second": 26.33,
+      "eval_steps_per_second": 13.165,
+      "eval_tts_loss": 7.493267624532595,
+      "step": 354000
+    },
+    {
+      "epoch": 3.0479511019283745,
+      "grad_norm": 5.375,
+      "learning_rate": 6.644707303012418e-05,
+      "loss": 3.9716,
+      "step": 354050
+    },
+    {
+      "epoch": 3.0483815426997247,
+      "grad_norm": 3.375,
+      "learning_rate": 6.642155275574818e-05,
+      "loss": 4.414,
+      "step": 354100
+    },
+    {
+      "epoch": 3.0488119834710745,
+      "grad_norm": 3.4375,
+      "learning_rate": 6.639603494595789e-05,
+      "loss": 4.5595,
+      "step": 354150
+    },
+    {
+      "epoch": 3.0492424242424243,
+      "grad_norm": 3.453125,
+      "learning_rate": 6.637051960262639e-05,
+      "loss": 4.3779,
+      "step": 354200
+    },
+    {
+      "epoch": 3.049672865013774,
+      "grad_norm": 3.984375,
+      "learning_rate": 6.634500672762637e-05,
+      "loss": 4.3522,
+      "step": 354250
+    },
+    {
+      "epoch": 3.050103305785124,
+      "grad_norm": 6.65625,
+      "learning_rate": 6.631949632283044e-05,
+      "loss": 4.3906,
+      "step": 354300
+    },
+    {
+      "epoch": 3.0505337465564737,
+      "grad_norm": 2.921875,
+      "learning_rate": 6.629398839011105e-05,
+      "loss": 4.3624,
+      "step": 354350
+    },
+    {
+      "epoch": 3.0509641873278235,
+      "grad_norm": 3.265625,
+      "learning_rate": 6.626848293134038e-05,
+      "loss": 4.305,
+      "step": 354400
+    },
+    {
+      "epoch": 3.0513946280991737,
+      "grad_norm": 2.46875,
+      "learning_rate": 6.624297994839047e-05,
+      "loss": 3.9894,
+      "step": 354450
+    },
+    {
+      "epoch": 3.0518250688705235,
+      "grad_norm": 3.203125,
+      "learning_rate": 6.621747944313319e-05,
+      "loss": 4.1997,
+      "step": 354500
+    },
+    {
+      "epoch": 3.0522555096418733,
+      "grad_norm": 4.4375,
+      "learning_rate": 6.619198141744027e-05,
+      "loss": 4.1079,
+      "step": 354550
+    },
+    {
+      "epoch": 3.052685950413223,
+      "grad_norm": 3.203125,
+      "learning_rate": 6.616648587318317e-05,
+      "loss": 4.6535,
+      "step": 354600
+    },
+    {
+      "epoch": 3.053116391184573,
+      "grad_norm": 1.1328125,
+      "learning_rate": 6.61409928122332e-05,
+      "loss": 3.9644,
+      "step": 354650
+    },
+    {
+      "epoch": 3.0535468319559227,
+      "grad_norm": 3.96875,
+      "learning_rate": 6.61155022364615e-05,
+      "loss": 4.3938,
+      "step": 354700
+    },
+    {
+      "epoch": 3.053977272727273,
+      "grad_norm": 3.65625,
+      "learning_rate": 6.609001414773904e-05,
+      "loss": 4.1628,
+      "step": 354750
+    },
+    {
+      "epoch": 3.0544077134986227,
+      "grad_norm": 3.40625,
+      "learning_rate": 6.606452854793656e-05,
+      "loss": 4.5047,
+      "step": 354800
+    },
+    {
+      "epoch": 3.0548381542699725,
+      "grad_norm": 2.71875,
+      "learning_rate": 6.603904543892471e-05,
+      "loss": 4.3843,
+      "step": 354850
+    },
+    {
+      "epoch": 3.0552685950413223,
+      "grad_norm": 4.34375,
+      "learning_rate": 6.601356482257386e-05,
+      "loss": 4.1097,
+      "step": 354900
+    },
+    {
+      "epoch": 3.055699035812672,
+      "grad_norm": 3.78125,
+      "learning_rate": 6.598808670075418e-05,
+      "loss": 4.438,
+      "step": 354950
+    },
+    {
+      "epoch": 3.056129476584022,
+      "grad_norm": 2.65625,
+      "learning_rate": 6.596261107533578e-05,
+      "loss": 4.269,
+      "step": 355000
+    },
+    {
+      "epoch": 3.0565599173553717,
+      "grad_norm": 2.65625,
+      "learning_rate": 6.593713794818849e-05,
+      "loss": 4.0392,
+      "step": 355050
+    },
+    {
+      "epoch": 3.056990358126722,
+      "grad_norm": 3.84375,
+      "learning_rate": 6.591166732118204e-05,
+      "loss": 4.1408,
+      "step": 355100
+    },
+    {
+      "epoch": 3.0574207988980717,
+      "grad_norm": 3.484375,
+      "learning_rate": 6.588619919618579e-05,
+      "loss": 4.3555,
+      "step": 355150
+    },
+    {
+      "epoch": 3.0578512396694215,
+      "grad_norm": 3.6875,
+      "learning_rate": 6.586073357506915e-05,
+      "loss": 4.3225,
+      "step": 355200
+    },
+    {
+      "epoch": 3.0582816804407713,
+      "grad_norm": 3.625,
+      "learning_rate": 6.583527045970122e-05,
+      "loss": 4.3289,
+      "step": 355250
+    },
+    {
+      "epoch": 3.058712121212121,
+      "grad_norm": 3.625,
+      "learning_rate": 6.580980985195091e-05,
+      "loss": 4.3943,
+      "step": 355300
+    },
+    {
+      "epoch": 3.059142561983471,
+      "grad_norm": 5.46875,
+      "learning_rate": 6.578435175368704e-05,
+      "loss": 4.1192,
+      "step": 355350
+    },
+    {
+      "epoch": 3.059573002754821,
+      "grad_norm": 3.921875,
+      "learning_rate": 6.57588961667781e-05,
+      "loss": 4.5199,
+      "step": 355400
+    },
+    {
+      "epoch": 3.060003443526171,
+      "grad_norm": 4.4375,
+      "learning_rate": 6.57334430930925e-05,
+      "loss": 4.5455,
+      "step": 355450
+    },
+    {
+      "epoch": 3.0604338842975207,
+      "grad_norm": 2.875,
+      "learning_rate": 6.570799253449846e-05,
+      "loss": 4.2386,
+      "step": 355500
+    },
+    {
+      "epoch": 3.0608643250688705,
+      "grad_norm": 1.7265625,
+      "learning_rate": 6.568254449286399e-05,
+      "loss": 4.1244,
+      "step": 355550
+    },
+    {
+      "epoch": 3.0612947658402203,
+      "grad_norm": 4.1875,
+      "learning_rate": 6.565709897005696e-05,
+      "loss": 4.0862,
+      "step": 355600
+    },
+    {
+      "epoch": 3.06172520661157,
+      "grad_norm": 4.375,
+      "learning_rate": 6.563165596794492e-05,
+      "loss": 4.3905,
+      "step": 355650
+    },
+    {
+      "epoch": 3.06215564738292,
+      "grad_norm": 4.65625,
+      "learning_rate": 6.560621548839538e-05,
+      "loss": 4.3767,
+      "step": 355700
+    },
+    {
+      "epoch": 3.06258608815427,
+      "grad_norm": 3.3125,
+      "learning_rate": 6.558077753327566e-05,
+      "loss": 4.4376,
+      "step": 355750
+    },
+    {
+      "epoch": 3.06301652892562,
+      "grad_norm": 7.09375,
+      "learning_rate": 6.555534210445278e-05,
+      "loss": 4.1551,
+      "step": 355800
+    },
+    {
+      "epoch": 3.0634469696969697,
+      "grad_norm": 1.6484375,
+      "learning_rate": 6.552990920379369e-05,
+      "loss": 4.0698,
+      "step": 355850
+    },
+    {
+      "epoch": 3.0638774104683195,
+      "grad_norm": 9.1875,
+      "learning_rate": 6.550447883316512e-05,
+      "loss": 4.166,
+      "step": 355900
+    },
+    {
+      "epoch": 3.0643078512396693,
+      "grad_norm": 6.0,
+      "learning_rate": 6.547905099443358e-05,
+      "loss": 4.5483,
+      "step": 355950
+    },
+    {
+      "epoch": 3.064738292011019,
+      "grad_norm": 2.84375,
+      "learning_rate": 6.545362568946542e-05,
+      "loss": 4.5924,
+      "step": 356000
+    },
+    {
+      "epoch": 3.0651687327823693,
+      "grad_norm": 3.09375,
+      "learning_rate": 6.542820292012681e-05,
+      "loss": 4.1123,
+      "step": 356050
+    },
+    {
+      "epoch": 3.065599173553719,
+      "grad_norm": 3.171875,
+      "learning_rate": 6.54027826882837e-05,
+      "loss": 4.089,
+      "step": 356100
+    },
+    {
+      "epoch": 3.066029614325069,
+      "grad_norm": 1.671875,
+      "learning_rate": 6.537736499580194e-05,
+      "loss": 4.47,
+      "step": 356150
+    },
+    {
+      "epoch": 3.0664600550964187,
+      "grad_norm": 3.5625,
+      "learning_rate": 6.535194984454707e-05,
+      "loss": 4.4509,
+      "step": 356200
+    },
+    {
+      "epoch": 3.0668904958677685,
+      "grad_norm": 3.4375,
+      "learning_rate": 6.532653723638453e-05,
+      "loss": 4.2035,
+      "step": 356250
+    },
+    {
+      "epoch": 3.0673209366391183,
+      "grad_norm": 4.40625,
+      "learning_rate": 6.530112717317956e-05,
+      "loss": 4.0482,
+      "step": 356300
+    },
+    {
+      "epoch": 3.067751377410468,
+      "grad_norm": 3.71875,
+      "learning_rate": 6.52757196567972e-05,
+      "loss": 4.4188,
+      "step": 356350
+    },
+    {
+      "epoch": 3.0681818181818183,
+      "grad_norm": 2.6875,
+      "learning_rate": 6.525031468910232e-05,
+      "loss": 4.5306,
+      "step": 356400
+    },
+    {
+      "epoch": 3.068612258953168,
+      "grad_norm": 5.40625,
+      "learning_rate": 6.522491227195956e-05,
+      "loss": 4.6338,
+      "step": 356450
+    },
+    {
+      "epoch": 3.069042699724518,
+      "grad_norm": 2.546875,
+      "learning_rate": 6.519951240723344e-05,
+      "loss": 4.0925,
+      "step": 356500
+    },
+    {
+      "epoch": 3.0694731404958677,
+      "grad_norm": 3.828125,
+      "learning_rate": 6.517411509678819e-05,
+      "loss": 4.1743,
+      "step": 356550
+    },
+    {
+      "epoch": 3.0699035812672175,
+      "grad_norm": 3.890625,
+      "learning_rate": 6.5148720342488e-05,
+      "loss": 4.6664,
+      "step": 356600
+    },
+    {
+      "epoch": 3.0703340220385673,
+      "grad_norm": 3.203125,
+      "learning_rate": 6.512332814619677e-05,
+      "loss": 4.5723,
+      "step": 356650
+    },
+    {
+      "epoch": 3.0707644628099175,
+      "grad_norm": 3.546875,
+      "learning_rate": 6.509793850977817e-05,
+      "loss": 4.1521,
+      "step": 356700
+    },
+    {
+      "epoch": 3.0711949035812673,
+      "grad_norm": 4.53125,
+      "learning_rate": 6.507255143509579e-05,
+      "loss": 4.2396,
+      "step": 356750
+    },
+    {
+      "epoch": 3.071625344352617,
+      "grad_norm": 3.359375,
+      "learning_rate": 6.504716692401301e-05,
+      "loss": 4.3324,
+      "step": 356800
+    },
+    {
+      "epoch": 3.072055785123967,
+      "grad_norm": 2.0,
+      "learning_rate": 6.502178497839296e-05,
+      "loss": 4.1093,
+      "step": 356850
+    },
+    {
+      "epoch": 3.0724862258953167,
+      "grad_norm": 4.0,
+      "learning_rate": 6.499640560009867e-05,
+      "loss": 4.0299,
+      "step": 356900
+    },
+    {
+      "epoch": 3.0729166666666665,
+      "grad_norm": 2.84375,
+      "learning_rate": 6.497102879099288e-05,
+      "loss": 4.2207,
+      "step": 356950
+    },
+    {
+      "epoch": 3.0733471074380168,
+      "grad_norm": 3.53125,
+      "learning_rate": 6.494565455293817e-05,
+      "loss": 4.0947,
+      "step": 357000
+    },
+    {
+      "epoch": 3.0733471074380168,
+      "eval_loss": 4.9925127029418945,
+      "eval_runtime": 24.1044,
+      "eval_samples_per_second": 26.551,
+      "eval_steps_per_second": 13.276,
+      "eval_tts_loss": 7.451498809537497,
+      "step": 357000
+    },
+    {
+      "epoch": 3.0737775482093666,
+      "grad_norm": 3.046875,
+      "learning_rate": 6.492028288779702e-05,
+      "loss": 4.3632,
+      "step": 357050
+    },
+    {
+      "epoch": 3.0742079889807163,
+      "grad_norm": 2.78125,
+      "learning_rate": 6.489491379743165e-05,
+      "loss": 4.4784,
+      "step": 357100
+    },
+    {
+      "epoch": 3.074638429752066,
+      "grad_norm": 2.8125,
+      "learning_rate": 6.486954728370408e-05,
+      "loss": 4.1263,
+      "step": 357150
+    },
+    {
+      "epoch": 3.075068870523416,
+      "grad_norm": 5.40625,
+      "learning_rate": 6.484418334847612e-05,
+      "loss": 4.5602,
+      "step": 357200
+    },
+    {
+      "epoch": 3.0754993112947657,
+      "grad_norm": 3.171875,
+      "learning_rate": 6.481882199360947e-05,
+      "loss": 4.2463,
+      "step": 357250
+    },
+    {
+      "epoch": 3.0759297520661155,
+      "grad_norm": 3.4375,
+      "learning_rate": 6.47934632209656e-05,
+      "loss": 4.3203,
+      "step": 357300
+    },
+    {
+      "epoch": 3.0763601928374658,
+      "grad_norm": 2.921875,
+      "learning_rate": 6.476810703240577e-05,
+      "loss": 4.1856,
+      "step": 357350
+    },
+    {
+      "epoch": 3.0767906336088156,
+      "grad_norm": 3.828125,
+      "learning_rate": 6.474275342979112e-05,
+      "loss": 4.1917,
+      "step": 357400
+    },
+    {
+      "epoch": 3.0772210743801653,
+      "grad_norm": 3.296875,
+      "learning_rate": 6.471740241498245e-05,
+      "loss": 4.1496,
+      "step": 357450
+    },
+    {
+      "epoch": 3.077651515151515,
+      "grad_norm": 2.046875,
+      "learning_rate": 6.469205398984058e-05,
+      "loss": 4.2282,
+      "step": 357500
+    },
+    {
+      "epoch": 3.078081955922865,
+      "grad_norm": 3.234375,
+      "learning_rate": 6.466670815622595e-05,
+      "loss": 4.3859,
+      "step": 357550
+    },
+    {
+      "epoch": 3.0785123966942147,
+      "grad_norm": 6.125,
+      "learning_rate": 6.464136491599893e-05,
+      "loss": 4.2796,
+      "step": 357600
+    },
+    {
+      "epoch": 3.078942837465565,
+      "grad_norm": 2.140625,
+      "learning_rate": 6.461602427101963e-05,
+      "loss": 4.5376,
+      "step": 357650
+    },
+    {
+      "epoch": 3.0793732782369148,
+      "grad_norm": 4.6875,
+      "learning_rate": 6.459068622314808e-05,
+      "loss": 4.2966,
+      "step": 357700
+    },
+    {
+      "epoch": 3.0798037190082646,
+      "grad_norm": 1.3515625,
+      "learning_rate": 6.456535077424391e-05,
+      "loss": 4.4985,
+      "step": 357750
+    },
+    {
+      "epoch": 3.0802341597796143,
+      "grad_norm": 3.09375,
+      "learning_rate": 6.454001792616679e-05,
+      "loss": 4.3831,
+      "step": 357800
+    },
+    {
+      "epoch": 3.080664600550964,
+      "grad_norm": 3.15625,
+      "learning_rate": 6.451468768077602e-05,
+      "loss": 4.3182,
+      "step": 357850
+    },
+    {
+      "epoch": 3.081095041322314,
+      "grad_norm": 4.34375,
+      "learning_rate": 6.448936003993085e-05,
+      "loss": 4.103,
+      "step": 357900
+    },
+    {
+      "epoch": 3.0815254820936637,
+      "grad_norm": 3.28125,
+      "learning_rate": 6.446403500549027e-05,
+      "loss": 4.2688,
+      "step": 357950
+    },
+    {
+      "epoch": 3.081955922865014,
+      "grad_norm": 2.03125,
+      "learning_rate": 6.443871257931307e-05,
+      "loss": 4.353,
+      "step": 358000
+    },
+    {
+      "epoch": 3.0823863636363638,
+      "grad_norm": 2.765625,
+      "learning_rate": 6.441339276325782e-05,
+      "loss": 4.4004,
+      "step": 358050
+    },
+    {
+      "epoch": 3.0828168044077136,
+      "grad_norm": 4.34375,
+      "learning_rate": 6.438807555918296e-05,
+      "loss": 4.3556,
+      "step": 358100
+    },
+    {
+      "epoch": 3.0832472451790633,
+      "grad_norm": 4.53125,
+      "learning_rate": 6.436276096894678e-05,
+      "loss": 4.1774,
+      "step": 358150
+    },
+    {
+      "epoch": 3.083677685950413,
+      "grad_norm": 2.359375,
+      "learning_rate": 6.433744899440726e-05,
+      "loss": 4.448,
+      "step": 358200
+    },
+    {
+      "epoch": 3.084108126721763,
+      "grad_norm": 2.0625,
+      "learning_rate": 6.431213963742223e-05,
+      "loss": 4.0852,
+      "step": 358250
+    },
+    {
+      "epoch": 3.084538567493113,
+      "grad_norm": 2.578125,
+      "learning_rate": 6.428683289984938e-05,
+      "loss": 4.1548,
+      "step": 358300
+    },
+    {
+      "epoch": 3.084969008264463,
+      "grad_norm": 3.59375,
+      "learning_rate": 6.426152878354614e-05,
+      "loss": 4.3454,
+      "step": 358350
+    },
+    {
+      "epoch": 3.0853994490358128,
+      "grad_norm": 5.28125,
+      "learning_rate": 6.423622729036979e-05,
+      "loss": 4.1213,
+      "step": 358400
+    },
+    {
+      "epoch": 3.0858298898071626,
+      "grad_norm": 2.734375,
+      "learning_rate": 6.421092842217746e-05,
+      "loss": 4.2708,
+      "step": 358450
+    },
+    {
+      "epoch": 3.0862603305785123,
+      "grad_norm": 4.71875,
+      "learning_rate": 6.418563218082597e-05,
+      "loss": 4.1605,
+      "step": 358500
+    },
+    {
+      "epoch": 3.086690771349862,
+      "grad_norm": 3.203125,
+      "learning_rate": 6.416033856817199e-05,
+      "loss": 4.3796,
+      "step": 358550
+    },
+    {
+      "epoch": 3.087121212121212,
+      "grad_norm": 4.5625,
+      "learning_rate": 6.413504758607203e-05,
+      "loss": 4.5386,
+      "step": 358600
+    },
+    {
+      "epoch": 3.087551652892562,
+      "grad_norm": 2.8125,
+      "learning_rate": 6.410975923638246e-05,
+      "loss": 4.0379,
+      "step": 358650
+    },
+    {
+      "epoch": 3.087982093663912,
+      "grad_norm": 4.75,
+      "learning_rate": 6.408447352095935e-05,
+      "loss": 4.5101,
+      "step": 358700
+    },
+    {
+      "epoch": 3.0884125344352618,
+      "grad_norm": 3.5625,
+      "learning_rate": 6.405919044165855e-05,
+      "loss": 4.022,
+      "step": 358750
+    },
+    {
+      "epoch": 3.0888429752066116,
+      "grad_norm": 5.4375,
+      "learning_rate": 6.403391000033587e-05,
+      "loss": 4.4119,
+      "step": 358800
+    },
+    {
+      "epoch": 3.0892734159779613,
+      "grad_norm": 3.34375,
+      "learning_rate": 6.400863219884684e-05,
+      "loss": 4.3505,
+      "step": 358850
+    },
+    {
+      "epoch": 3.089703856749311,
+      "grad_norm": 3.0625,
+      "learning_rate": 6.398335703904673e-05,
+      "loss": 4.2578,
+      "step": 358900
+    },
+    {
+      "epoch": 3.0901342975206614,
+      "grad_norm": 4.59375,
+      "learning_rate": 6.395808452279076e-05,
+      "loss": 4.4511,
+      "step": 358950
+    },
+    {
+      "epoch": 3.090564738292011,
+      "grad_norm": 5.21875,
+      "learning_rate": 6.393281465193379e-05,
+      "loss": 4.031,
+      "step": 359000
+    },
+    {
+      "epoch": 3.090995179063361,
+      "grad_norm": 2.984375,
+      "learning_rate": 6.390754742833066e-05,
+      "loss": 4.3142,
+      "step": 359050
+    },
+    {
+      "epoch": 3.0914256198347108,
+      "grad_norm": 4.21875,
+      "learning_rate": 6.388228285383588e-05,
+      "loss": 4.2652,
+      "step": 359100
+    },
+    {
+      "epoch": 3.0918560606060606,
+      "grad_norm": 4.09375,
+      "learning_rate": 6.385702093030383e-05,
+      "loss": 4.68,
+      "step": 359150
+    },
+    {
+      "epoch": 3.0922865013774103,
+      "grad_norm": 2.96875,
+      "learning_rate": 6.383176165958872e-05,
+      "loss": 4.2393,
+      "step": 359200
+    },
+    {
+      "epoch": 3.09271694214876,
+      "grad_norm": 4.65625,
+      "learning_rate": 6.380650504354444e-05,
+      "loss": 4.192,
+      "step": 359250
+    },
+    {
+      "epoch": 3.0931473829201104,
+      "grad_norm": 2.28125,
+      "learning_rate": 6.378125108402482e-05,
+      "loss": 4.1481,
+      "step": 359300
+    },
+    {
+      "epoch": 3.09357782369146,
+      "grad_norm": 1.3359375,
+      "learning_rate": 6.375599978288345e-05,
+      "loss": 4.484,
+      "step": 359350
+    },
+    {
+      "epoch": 3.09400826446281,
+      "grad_norm": 3.1875,
+      "learning_rate": 6.373075114197371e-05,
+      "loss": 4.7217,
+      "step": 359400
+    },
+    {
+      "epoch": 3.0944387052341598,
+      "grad_norm": 2.390625,
+      "learning_rate": 6.37055051631488e-05,
+      "loss": 4.199,
+      "step": 359450
+    },
+    {
+      "epoch": 3.0948691460055096,
+      "grad_norm": 3.671875,
+      "learning_rate": 6.368026184826174e-05,
+      "loss": 4.1352,
+      "step": 359500
+    },
+    {
+      "epoch": 3.0952995867768593,
+      "grad_norm": 2.921875,
+      "learning_rate": 6.365502119916532e-05,
+      "loss": 4.5603,
+      "step": 359550
+    },
+    {
+      "epoch": 3.0957300275482096,
+      "grad_norm": 2.296875,
+      "learning_rate": 6.362978321771212e-05,
+      "loss": 4.1977,
+      "step": 359600
+    },
+    {
+      "epoch": 3.0961604683195594,
+      "grad_norm": 5.34375,
+      "learning_rate": 6.36045479057546e-05,
+      "loss": 4.4953,
+      "step": 359650
+    },
+    {
+      "epoch": 3.096590909090909,
+      "grad_norm": 3.0,
+      "learning_rate": 6.357931526514493e-05,
+      "loss": 4.2018,
+      "step": 359700
+    },
+    {
+      "epoch": 3.097021349862259,
+      "grad_norm": 0.828125,
+      "learning_rate": 6.35540852977352e-05,
+      "loss": 4.3821,
+      "step": 359750
+    },
+    {
+      "epoch": 3.0974517906336088,
+      "grad_norm": 4.4375,
+      "learning_rate": 6.352885800537716e-05,
+      "loss": 4.1022,
+      "step": 359800
+    },
+    {
+      "epoch": 3.0978822314049586,
+      "grad_norm": 2.375,
+      "learning_rate": 6.350363338992249e-05,
+      "loss": 4.3221,
+      "step": 359850
+    },
+    {
+      "epoch": 3.0983126721763083,
+      "grad_norm": 3.4375,
+      "learning_rate": 6.347841145322259e-05,
+      "loss": 4.2567,
+      "step": 359900
+    },
+    {
+      "epoch": 3.0987431129476586,
+      "grad_norm": 3.765625,
+      "learning_rate": 6.345319219712871e-05,
+      "loss": 4.1406,
+      "step": 359950
+    },
+    {
+      "epoch": 3.0991735537190084,
+      "grad_norm": 4.5,
+      "learning_rate": 6.342797562349192e-05,
+      "loss": 4.3052,
+      "step": 360000
+    },
+    {
+      "epoch": 3.0991735537190084,
+      "eval_loss": 4.993627071380615,
+      "eval_runtime": 23.8195,
+      "eval_samples_per_second": 26.869,
+      "eval_steps_per_second": 13.434,
+      "eval_tts_loss": 7.470391318419026,
+      "step": 360000
+    },
+    {
+      "epoch": 3.099603994490358,
+      "grad_norm": 5.15625,
+      "learning_rate": 6.340276173416302e-05,
+      "loss": 4.6036,
+      "step": 360050
+    },
+    {
+      "epoch": 3.100034435261708,
+      "grad_norm": 2.40625,
+      "learning_rate": 6.337755053099263e-05,
+      "loss": 4.3408,
+      "step": 360100
+    },
+    {
+      "epoch": 3.1004648760330578,
+      "grad_norm": 4.5625,
+      "learning_rate": 6.335234201583126e-05,
+      "loss": 4.4338,
+      "step": 360150
+    },
+    {
+      "epoch": 3.1008953168044076,
+      "grad_norm": 3.296875,
+      "learning_rate": 6.332713619052915e-05,
+      "loss": 4.3446,
+      "step": 360200
+    },
+    {
+      "epoch": 3.101325757575758,
+      "grad_norm": 3.953125,
+      "learning_rate": 6.330193305693637e-05,
+      "loss": 4.3714,
+      "step": 360250
+    },
+    {
+      "epoch": 3.1017561983471076,
+      "grad_norm": 2.421875,
+      "learning_rate": 6.327673261690267e-05,
+      "loss": 4.2715,
+      "step": 360300
+    },
+    {
+      "epoch": 3.1021866391184574,
+      "grad_norm": 5.90625,
+      "learning_rate": 6.325153487227782e-05,
+      "loss": 4.1991,
+      "step": 360350
+    },
+    {
+      "epoch": 3.102617079889807,
+      "grad_norm": 2.609375,
+      "learning_rate": 6.322633982491126e-05,
+      "loss": 4.2997,
+      "step": 360400
+    },
+    {
+      "epoch": 3.103047520661157,
+      "grad_norm": 2.484375,
+      "learning_rate": 6.32011474766522e-05,
+      "loss": 4.2015,
+      "step": 360450
+    },
+    {
+      "epoch": 3.1034779614325068,
+      "grad_norm": 1.96875,
+      "learning_rate": 6.31759578293498e-05,
+      "loss": 4.2898,
+      "step": 360500
+    },
+    {
+      "epoch": 3.1039084022038566,
+      "grad_norm": 3.421875,
+      "learning_rate": 6.315077088485282e-05,
+      "loss": 4.2906,
+      "step": 360550
+    },
+    {
+      "epoch": 3.104338842975207,
+      "grad_norm": 2.171875,
+      "learning_rate": 6.312558664500998e-05,
+      "loss": 4.2858,
+      "step": 360600
+    },
+    {
+      "epoch": 3.1047692837465566,
+      "grad_norm": 3.1875,
+      "learning_rate": 6.310040511166973e-05,
+      "loss": 4.3682,
+      "step": 360650
+    },
+    {
+      "epoch": 3.1051997245179064,
+      "grad_norm": 3.140625,
+      "learning_rate": 6.307522628668038e-05,
+      "loss": 4.65,
+      "step": 360700
+    },
+    {
+      "epoch": 3.105630165289256,
+      "grad_norm": 3.25,
+      "learning_rate": 6.305005017188999e-05,
+      "loss": 4.1018,
+      "step": 360750
+    },
+    {
+      "epoch": 3.106060606060606,
+      "grad_norm": 3.328125,
+      "learning_rate": 6.302487676914636e-05,
+      "loss": 4.1349,
+      "step": 360800
+    },
+    {
+      "epoch": 3.1064910468319558,
+      "grad_norm": 1.6953125,
+      "learning_rate": 6.299970608029726e-05,
+      "loss": 4.2589,
+      "step": 360850
+    },
+    {
+      "epoch": 3.106921487603306,
+      "grad_norm": 3.640625,
+      "learning_rate": 6.297453810719011e-05,
+      "loss": 3.9939,
+      "step": 360900
+    },
+    {
+      "epoch": 3.107351928374656,
+      "grad_norm": 2.8125,
+      "learning_rate": 6.29493728516722e-05,
+      "loss": 4.3162,
+      "step": 360950
+    },
+    {
+      "epoch": 3.1077823691460056,
+      "grad_norm": 2.859375,
+      "learning_rate": 6.292421031559064e-05,
+      "loss": 4.274,
+      "step": 361000
+    },
+    {
+      "epoch": 3.1082128099173554,
+      "grad_norm": 4.28125,
+      "learning_rate": 6.289905050079224e-05,
+      "loss": 4.4353,
+      "step": 361050
+    },
+    {
+      "epoch": 3.108643250688705,
+      "grad_norm": 1.9375,
+      "learning_rate": 6.287389340912372e-05,
+      "loss": 3.9579,
+      "step": 361100
+    },
+    {
+      "epoch": 3.109073691460055,
+      "grad_norm": 3.859375,
+      "learning_rate": 6.284873904243152e-05,
+      "loss": 4.4099,
+      "step": 361150
+    },
+    {
+      "epoch": 3.1095041322314048,
+      "grad_norm": 2.109375,
+      "learning_rate": 6.282358740256198e-05,
+      "loss": 4.3133,
+      "step": 361200
+    },
+    {
+      "epoch": 3.109934573002755,
+      "grad_norm": 4.65625,
+      "learning_rate": 6.27984384913611e-05,
+      "loss": 4.071,
+      "step": 361250
+    },
+    {
+      "epoch": 3.110365013774105,
+      "grad_norm": 2.828125,
+      "learning_rate": 6.277329231067484e-05,
+      "loss": 4.1805,
+      "step": 361300
+    },
+    {
+      "epoch": 3.1107954545454546,
+      "grad_norm": 3.96875,
+      "learning_rate": 6.274814886234879e-05,
+      "loss": 4.2217,
+      "step": 361350
+    },
+    {
+      "epoch": 3.1112258953168044,
+      "grad_norm": 4.84375,
+      "learning_rate": 6.272300814822847e-05,
+      "loss": 4.1471,
+      "step": 361400
+    },
+    {
+      "epoch": 3.111656336088154,
+      "grad_norm": 2.4375,
+      "learning_rate": 6.269787017015914e-05,
+      "loss": 4.2561,
+      "step": 361450
+    },
+    {
+      "epoch": 3.112086776859504,
+      "grad_norm": 4.75,
+      "learning_rate": 6.267273492998589e-05,
+      "loss": 4.0435,
+      "step": 361500
+    },
+    {
+      "epoch": 3.112517217630854,
+      "grad_norm": 1.3515625,
+      "learning_rate": 6.264760242955361e-05,
+      "loss": 4.0504,
+      "step": 361550
+    },
+    {
+      "epoch": 3.112947658402204,
+      "grad_norm": 3.140625,
+      "learning_rate": 6.262247267070693e-05,
+      "loss": 4.0465,
+      "step": 361600
+    },
+    {
+      "epoch": 3.113378099173554,
+      "grad_norm": 2.640625,
+      "learning_rate": 6.259734565529033e-05,
+      "loss": 4.3578,
+      "step": 361650
+    },
+    {
+      "epoch": 3.1138085399449036,
+      "grad_norm": 3.296875,
+      "learning_rate": 6.257222138514807e-05,
+      "loss": 4.3567,
+      "step": 361700
+    },
+    {
+      "epoch": 3.1142389807162534,
+      "grad_norm": 3.40625,
+      "learning_rate": 6.254709986212426e-05,
+      "loss": 4.8075,
+      "step": 361750
+    },
+    {
+      "epoch": 3.114669421487603,
+      "grad_norm": 2.40625,
+      "learning_rate": 6.252198108806275e-05,
+      "loss": 4.2787,
+      "step": 361800
+    },
+    {
+      "epoch": 3.115099862258953,
+      "grad_norm": 4.125,
+      "learning_rate": 6.249686506480716e-05,
+      "loss": 4.3385,
+      "step": 361850
+    },
+    {
+      "epoch": 3.115530303030303,
+      "grad_norm": 3.28125,
+      "learning_rate": 6.2471751794201e-05,
+      "loss": 4.6175,
+      "step": 361900
+    },
+    {
+      "epoch": 3.115960743801653,
+      "grad_norm": 1.484375,
+      "learning_rate": 6.244664127808751e-05,
+      "loss": 4.1945,
+      "step": 361950
+    },
+    {
+      "epoch": 3.116391184573003,
+      "grad_norm": 3.640625,
+      "learning_rate": 6.242153351830972e-05,
+      "loss": 4.0884,
+      "step": 362000
+    },
+    {
+      "epoch": 3.1168216253443526,
+      "grad_norm": 3.734375,
+      "learning_rate": 6.239642851671058e-05,
+      "loss": 4.5337,
+      "step": 362050
+    },
+    {
+      "epoch": 3.1172520661157024,
+      "grad_norm": 4.5,
+      "learning_rate": 6.237132627513265e-05,
+      "loss": 4.1964,
+      "step": 362100
+    },
+    {
+      "epoch": 3.117682506887052,
+      "grad_norm": 3.453125,
+      "learning_rate": 6.23462267954184e-05,
+      "loss": 4.3932,
+      "step": 362150
+    },
+    {
+      "epoch": 3.1181129476584024,
+      "grad_norm": 2.265625,
+      "learning_rate": 6.232113007941008e-05,
+      "loss": 4.3091,
+      "step": 362200
+    },
+    {
+      "epoch": 3.118543388429752,
+      "grad_norm": 1.6953125,
+      "learning_rate": 6.229603612894978e-05,
+      "loss": 4.2519,
+      "step": 362250
+    },
+    {
+      "epoch": 3.118973829201102,
+      "grad_norm": 3.546875,
+      "learning_rate": 6.22709449458793e-05,
+      "loss": 4.1282,
+      "step": 362300
+    },
+    {
+      "epoch": 3.119404269972452,
+      "grad_norm": 1.296875,
+      "learning_rate": 6.224585653204027e-05,
+      "loss": 4.364,
+      "step": 362350
+    },
+    {
+      "epoch": 3.1198347107438016,
+      "grad_norm": 3.53125,
+      "learning_rate": 6.222077088927414e-05,
+      "loss": 4.387,
+      "step": 362400
+    },
+    {
+      "epoch": 3.1202651515151514,
+      "grad_norm": 2.71875,
+      "learning_rate": 6.219568801942215e-05,
+      "loss": 4.174,
+      "step": 362450
+    },
+    {
+      "epoch": 3.120695592286501,
+      "grad_norm": 1.65625,
+      "learning_rate": 6.21706079243253e-05,
+      "loss": 4.1107,
+      "step": 362500
+    },
+    {
+      "epoch": 3.1211260330578514,
+      "grad_norm": 4.1875,
+      "learning_rate": 6.214553060582449e-05,
+      "loss": 4.327,
+      "step": 362550
+    },
+    {
+      "epoch": 3.121556473829201,
+      "grad_norm": 4.375,
+      "learning_rate": 6.212045606576023e-05,
+      "loss": 4.2824,
+      "step": 362600
+    },
+    {
+      "epoch": 3.121986914600551,
+      "grad_norm": 2.5625,
+      "learning_rate": 6.209538430597303e-05,
+      "loss": 4.3991,
+      "step": 362650
+    },
+    {
+      "epoch": 3.122417355371901,
+      "grad_norm": 1.7578125,
+      "learning_rate": 6.207031532830303e-05,
+      "loss": 4.3649,
+      "step": 362700
+    },
+    {
+      "epoch": 3.1228477961432506,
+      "grad_norm": 2.78125,
+      "learning_rate": 6.204524913459031e-05,
+      "loss": 3.7126,
+      "step": 362750
+    },
+    {
+      "epoch": 3.1232782369146004,
+      "grad_norm": 2.859375,
+      "learning_rate": 6.202018572667467e-05,
+      "loss": 3.9808,
+      "step": 362800
+    },
+    {
+      "epoch": 3.1237086776859506,
+      "grad_norm": 3.96875,
+      "learning_rate": 6.199512510639563e-05,
+      "loss": 4.2064,
+      "step": 362850
+    },
+    {
+      "epoch": 3.1241391184573004,
+      "grad_norm": 3.453125,
+      "learning_rate": 6.197006727559264e-05,
+      "loss": 3.9251,
+      "step": 362900
+    },
+    {
+      "epoch": 3.12456955922865,
+      "grad_norm": 3.09375,
+      "learning_rate": 6.19450122361049e-05,
+      "loss": 4.34,
+      "step": 362950
+    },
+    {
+      "epoch": 3.125,
+      "grad_norm": 4.6875,
+      "learning_rate": 6.191995998977138e-05,
+      "loss": 4.1998,
+      "step": 363000
+    },
+    {
+      "epoch": 3.125,
+      "eval_loss": 4.995603561401367,
+      "eval_runtime": 24.2612,
+      "eval_samples_per_second": 26.38,
+      "eval_steps_per_second": 13.19,
+      "eval_tts_loss": 7.500190487637266,
+      "step": 363000
+    },
+    {
+      "epoch": 3.12543044077135,
+      "grad_norm": 6.71875,
+      "learning_rate": 6.189491053843086e-05,
+      "loss": 4.439,
+      "step": 363050
+    },
+    {
+      "epoch": 3.1258608815426996,
+      "grad_norm": 1.625,
+      "learning_rate": 6.186986388392196e-05,
+      "loss": 4.4058,
+      "step": 363100
+    },
+    {
+      "epoch": 3.1262913223140494,
+      "grad_norm": 2.4375,
+      "learning_rate": 6.184482002808298e-05,
+      "loss": 4.3491,
+      "step": 363150
+    },
+    {
+      "epoch": 3.1267217630853996,
+      "grad_norm": 2.0625,
+      "learning_rate": 6.18197789727521e-05,
+      "loss": 4.4404,
+      "step": 363200
+    },
+    {
+      "epoch": 3.1271522038567494,
+      "grad_norm": 2.375,
+      "learning_rate": 6.179474071976734e-05,
+      "loss": 4.2152,
+      "step": 363250
+    },
+    {
+      "epoch": 3.127582644628099,
+      "grad_norm": 5.25,
+      "learning_rate": 6.176970527096637e-05,
+      "loss": 4.2062,
+      "step": 363300
+    },
+    {
+      "epoch": 3.128013085399449,
+      "grad_norm": 3.578125,
+      "learning_rate": 6.174467262818683e-05,
+      "loss": 4.0699,
+      "step": 363350
+    },
+    {
+      "epoch": 3.128443526170799,
+      "grad_norm": 3.140625,
+      "learning_rate": 6.171964279326596e-05,
+      "loss": 4.5137,
+      "step": 363400
+    },
+    {
+      "epoch": 3.1288739669421486,
+      "grad_norm": 4.28125,
+      "learning_rate": 6.169461576804098e-05,
+      "loss": 4.0273,
+      "step": 363450
+    },
+    {
+      "epoch": 3.129304407713499,
+      "grad_norm": 3.640625,
+      "learning_rate": 6.166959155434875e-05,
+      "loss": 4.1074,
+      "step": 363500
+    },
+    {
+      "epoch": 3.1297348484848486,
+      "grad_norm": 3.125,
+      "learning_rate": 6.164457015402604e-05,
+      "loss": 4.1651,
+      "step": 363550
+    },
+    {
+      "epoch": 3.1301652892561984,
+      "grad_norm": 3.359375,
+      "learning_rate": 6.16195515689094e-05,
+      "loss": 4.1972,
+      "step": 363600
+    },
+    {
+      "epoch": 3.130595730027548,
+      "grad_norm": 6.1875,
+      "learning_rate": 6.159453580083508e-05,
+      "loss": 4.3394,
+      "step": 363650
+    },
+    {
+      "epoch": 3.131026170798898,
+      "grad_norm": 4.0625,
+      "learning_rate": 6.156952285163915e-05,
+      "loss": 4.2125,
+      "step": 363700
+    },
+    {
+      "epoch": 3.131456611570248,
+      "grad_norm": 3.109375,
+      "learning_rate": 6.154451272315758e-05,
+      "loss": 4.0876,
+      "step": 363750
+    },
+    {
+      "epoch": 3.1318870523415976,
+      "grad_norm": 3.59375,
+      "learning_rate": 6.151950541722606e-05,
+      "loss": 4.4389,
+      "step": 363800
+    },
+    {
+      "epoch": 3.132317493112948,
+      "grad_norm": 3.640625,
+      "learning_rate": 6.149450093568008e-05,
+      "loss": 4.5061,
+      "step": 363850
+    },
+    {
+      "epoch": 3.1327479338842976,
+      "grad_norm": 2.34375,
+      "learning_rate": 6.146949928035482e-05,
+      "loss": 4.5249,
+      "step": 363900
+    },
+    {
+      "epoch": 3.1331783746556474,
+      "grad_norm": 1.75,
+      "learning_rate": 6.144450045308543e-05,
+      "loss": 4.3586,
+      "step": 363950
+    },
+    {
+      "epoch": 3.133608815426997,
+      "grad_norm": 3.96875,
+      "learning_rate": 6.141950445570677e-05,
+      "loss": 4.3803,
+      "step": 364000
+    },
+    {
+      "epoch": 3.134039256198347,
+      "grad_norm": 3.40625,
+      "learning_rate": 6.139451129005344e-05,
+      "loss": 4.4974,
+      "step": 364050
+    },
+    {
+      "epoch": 3.134469696969697,
+      "grad_norm": 4.28125,
+      "learning_rate": 6.136952095795997e-05,
+      "loss": 4.6624,
+      "step": 364100
+    },
+    {
+      "epoch": 3.134900137741047,
+      "grad_norm": 1.21875,
+      "learning_rate": 6.134453346126052e-05,
+      "loss": 4.1538,
+      "step": 364150
+    },
+    {
+      "epoch": 3.135330578512397,
+      "grad_norm": 1.9453125,
+      "learning_rate": 6.131954880178914e-05,
+      "loss": 4.0171,
+      "step": 364200
+    },
+    {
+      "epoch": 3.1357610192837466,
+      "grad_norm": 1.65625,
+      "learning_rate": 6.129456698137964e-05,
+      "loss": 4.448,
+      "step": 364250
+    },
+    {
+      "epoch": 3.1361914600550964,
+      "grad_norm": 3.578125,
+      "learning_rate": 6.126958800186568e-05,
+      "loss": 4.4268,
+      "step": 364300
+    },
+    {
+      "epoch": 3.136621900826446,
+      "grad_norm": 3.40625,
+      "learning_rate": 6.124461186508064e-05,
+      "loss": 4.1488,
+      "step": 364350
+    },
+    {
+      "epoch": 3.137052341597796,
+      "grad_norm": 1.9921875,
+      "learning_rate": 6.121963857285766e-05,
+      "loss": 4.3665,
+      "step": 364400
+    },
+    {
+      "epoch": 3.137482782369146,
+      "grad_norm": 2.734375,
+      "learning_rate": 6.119466812702977e-05,
+      "loss": 4.2319,
+      "step": 364450
+    },
+    {
+      "epoch": 3.137913223140496,
+      "grad_norm": 4.8125,
+      "learning_rate": 6.116970052942979e-05,
+      "loss": 4.4639,
+      "step": 364500
+    },
+    {
+      "epoch": 3.138343663911846,
+      "grad_norm": 4.1875,
+      "learning_rate": 6.11447357818902e-05,
+      "loss": 4.3677,
+      "step": 364550
+    },
+    {
+      "epoch": 3.1387741046831956,
+      "grad_norm": 2.765625,
+      "learning_rate": 6.111977388624345e-05,
+      "loss": 4.3788,
+      "step": 364600
+    },
+    {
+      "epoch": 3.1392045454545454,
+      "grad_norm": 3.640625,
+      "learning_rate": 6.109481484432162e-05,
+      "loss": 4.5474,
+      "step": 364650
+    },
+    {
+      "epoch": 3.139634986225895,
+      "grad_norm": 4.53125,
+      "learning_rate": 6.106985865795668e-05,
+      "loss": 4.372,
+      "step": 364700
+    },
+    {
+      "epoch": 3.140065426997245,
+      "grad_norm": 2.546875,
+      "learning_rate": 6.104490532898033e-05,
+      "loss": 4.2874,
+      "step": 364750
+    },
+    {
+      "epoch": 3.1404958677685952,
+      "grad_norm": 1.9921875,
+      "learning_rate": 6.1019954859224146e-05,
+      "loss": 3.8088,
+      "step": 364800
+    },
+    {
+      "epoch": 3.140926308539945,
+      "grad_norm": 2.90625,
+      "learning_rate": 6.0995007250519386e-05,
+      "loss": 4.1065,
+      "step": 364850
+    },
+    {
+      "epoch": 3.141356749311295,
+      "grad_norm": 3.015625,
+      "learning_rate": 6.097006250469721e-05,
+      "loss": 4.6339,
+      "step": 364900
+    },
+    {
+      "epoch": 3.1417871900826446,
+      "grad_norm": 3.390625,
+      "learning_rate": 6.094512062358842e-05,
+      "loss": 4.1012,
+      "step": 364950
+    },
+    {
+      "epoch": 3.1422176308539944,
+      "grad_norm": 2.875,
+      "learning_rate": 6.0920181609023785e-05,
+      "loss": 4.5716,
+      "step": 365000
+    },
+    {
+      "epoch": 3.142648071625344,
+      "grad_norm": 3.765625,
+      "learning_rate": 6.089524546283372e-05,
+      "loss": 4.461,
+      "step": 365050
+    },
+    {
+      "epoch": 3.143078512396694,
+      "grad_norm": 3.40625,
+      "learning_rate": 6.087031218684849e-05,
+      "loss": 4.3748,
+      "step": 365100
+    },
+    {
+      "epoch": 3.1435089531680442,
+      "grad_norm": 2.84375,
+      "learning_rate": 6.0845381782898194e-05,
+      "loss": 4.2182,
+      "step": 365150
+    },
+    {
+      "epoch": 3.143939393939394,
+      "grad_norm": 2.1875,
+      "learning_rate": 6.082045425281262e-05,
+      "loss": 4.5994,
+      "step": 365200
+    },
+    {
+      "epoch": 3.144369834710744,
+      "grad_norm": 3.921875,
+      "learning_rate": 6.0795529598421374e-05,
+      "loss": 4.3806,
+      "step": 365250
+    },
+    {
+      "epoch": 3.1448002754820936,
+      "grad_norm": 3.328125,
+      "learning_rate": 6.0770607821553915e-05,
+      "loss": 4.0178,
+      "step": 365300
+    },
+    {
+      "epoch": 3.1452307162534434,
+      "grad_norm": 4.15625,
+      "learning_rate": 6.0745688924039454e-05,
+      "loss": 4.3046,
+      "step": 365350
+    },
+    {
+      "epoch": 3.145661157024793,
+      "grad_norm": 3.734375,
+      "learning_rate": 6.072077290770699e-05,
+      "loss": 4.3052,
+      "step": 365400
+    },
+    {
+      "epoch": 3.1460915977961434,
+      "grad_norm": 2.96875,
+      "learning_rate": 6.069585977438525e-05,
+      "loss": 4.1699,
+      "step": 365450
+    },
+    {
+      "epoch": 3.1465220385674932,
+      "grad_norm": 2.796875,
+      "learning_rate": 6.0670949525902844e-05,
+      "loss": 4.5027,
+      "step": 365500
+    },
+    {
+      "epoch": 3.146952479338843,
+      "grad_norm": 2.515625,
+      "learning_rate": 6.0646042164088114e-05,
+      "loss": 4.3779,
+      "step": 365550
+    },
+    {
+      "epoch": 3.147382920110193,
+      "grad_norm": 3.8125,
+      "learning_rate": 6.062113769076923e-05,
+      "loss": 3.7515,
+      "step": 365600
+    },
+    {
+      "epoch": 3.1478133608815426,
+      "grad_norm": 3.328125,
+      "learning_rate": 6.059623610777412e-05,
+      "loss": 4.2055,
+      "step": 365650
+    },
+    {
+      "epoch": 3.1482438016528924,
+      "grad_norm": 3.15625,
+      "learning_rate": 6.057133741693051e-05,
+      "loss": 4.1914,
+      "step": 365700
+    },
+    {
+      "epoch": 3.148674242424242,
+      "grad_norm": 2.796875,
+      "learning_rate": 6.054644162006586e-05,
+      "loss": 4.3808,
+      "step": 365750
+    },
+    {
+      "epoch": 3.1491046831955924,
+      "grad_norm": 2.71875,
+      "learning_rate": 6.052154871900753e-05,
+      "loss": 4.1459,
+      "step": 365800
+    },
+    {
+      "epoch": 3.1495351239669422,
+      "grad_norm": 3.28125,
+      "learning_rate": 6.049665871558259e-05,
+      "loss": 3.9858,
+      "step": 365850
+    },
+    {
+      "epoch": 3.149965564738292,
+      "grad_norm": 5.0625,
+      "learning_rate": 6.047177161161792e-05,
+      "loss": 4.5558,
+      "step": 365900
+    },
+    {
+      "epoch": 3.150396005509642,
+      "grad_norm": 2.71875,
+      "learning_rate": 6.044688740894013e-05,
+      "loss": 4.4139,
+      "step": 365950
+    },
+    {
+      "epoch": 3.1508264462809916,
+      "grad_norm": 4.25,
+      "learning_rate": 6.042200610937572e-05,
+      "loss": 4.3586,
+      "step": 366000
+    },
+    {
+      "epoch": 3.1508264462809916,
+      "eval_loss": 4.995471000671387,
+      "eval_runtime": 24.3469,
+      "eval_samples_per_second": 26.287,
+      "eval_steps_per_second": 13.143,
+      "eval_tts_loss": 7.496336504952824,
+      "step": 366000
+    },
+    {
+      "epoch": 3.1512568870523414,
+      "grad_norm": 3.203125,
+      "learning_rate": 6.03971277147509e-05,
+      "loss": 4.6061,
+      "step": 366050
+    },
+    {
+      "epoch": 3.1516873278236917,
+      "grad_norm": 1.9375,
+      "learning_rate": 6.03722522268917e-05,
+      "loss": 4.4332,
+      "step": 366100
+    },
+    {
+      "epoch": 3.1521177685950414,
+      "grad_norm": 2.65625,
+      "learning_rate": 6.0347379647623956e-05,
+      "loss": 4.5218,
+      "step": 366150
+    },
+    {
+      "epoch": 3.1525482093663912,
+      "grad_norm": 3.5625,
+      "learning_rate": 6.03225099787732e-05,
+      "loss": 4.3174,
+      "step": 366200
+    },
+    {
+      "epoch": 3.152978650137741,
+      "grad_norm": 2.90625,
+      "learning_rate": 6.029764322216486e-05,
+      "loss": 4.2855,
+      "step": 366250
+    },
+    {
+      "epoch": 3.153409090909091,
+      "grad_norm": 6.15625,
+      "learning_rate": 6.027277937962406e-05,
+      "loss": 4.3564,
+      "step": 366300
+    },
+    {
+      "epoch": 3.1538395316804406,
+      "grad_norm": 3.484375,
+      "learning_rate": 6.024791845297579e-05,
+      "loss": 4.6584,
+      "step": 366350
+    },
+    {
+      "epoch": 3.1542699724517904,
+      "grad_norm": 1.328125,
+      "learning_rate": 6.022306044404482e-05,
+      "loss": 4.0403,
+      "step": 366400
+    },
+    {
+      "epoch": 3.1547004132231407,
+      "grad_norm": 5.3125,
+      "learning_rate": 6.019820535465557e-05,
+      "loss": 4.1525,
+      "step": 366450
+    },
+    {
+      "epoch": 3.1551308539944904,
+      "grad_norm": 4.78125,
+      "learning_rate": 6.017335318663242e-05,
+      "loss": 4.5943,
+      "step": 366500
+    },
+    {
+      "epoch": 3.1555612947658402,
+      "grad_norm": 4.25,
+      "learning_rate": 6.0148503941799485e-05,
+      "loss": 4.0463,
+      "step": 366550
+    },
+    {
+      "epoch": 3.15599173553719,
+      "grad_norm": 4.375,
+      "learning_rate": 6.012365762198059e-05,
+      "loss": 4.2938,
+      "step": 366600
+    },
+    {
+      "epoch": 3.15642217630854,
+      "grad_norm": 5.65625,
+      "learning_rate": 6.009881422899942e-05,
+      "loss": 4.6583,
+      "step": 366650
+    },
+    {
+      "epoch": 3.1568526170798896,
+      "grad_norm": 3.109375,
+      "learning_rate": 6.007397376467949e-05,
+      "loss": 4.3703,
+      "step": 366700
+    },
+    {
+      "epoch": 3.15728305785124,
+      "grad_norm": 7.96875,
+      "learning_rate": 6.004913623084396e-05,
+      "loss": 4.2425,
+      "step": 366750
+    },
+    {
+      "epoch": 3.1577134986225897,
+      "grad_norm": 3.671875,
+      "learning_rate": 6.002430162931585e-05,
+      "loss": 4.2259,
+      "step": 366800
+    },
+    {
+      "epoch": 3.1581439393939394,
+      "grad_norm": 2.515625,
+      "learning_rate": 5.9999469961917987e-05,
+      "loss": 4.1448,
+      "step": 366850
+    },
+    {
+      "epoch": 3.1585743801652892,
+      "grad_norm": 4.90625,
+      "learning_rate": 5.9974641230472995e-05,
+      "loss": 4.3983,
+      "step": 366900
+    },
+    {
+      "epoch": 3.159004820936639,
+      "grad_norm": 4.78125,
+      "learning_rate": 5.994981543680324e-05,
+      "loss": 4.5295,
+      "step": 366950
+    },
+    {
+      "epoch": 3.159435261707989,
+      "grad_norm": 3.0,
+      "learning_rate": 5.992499258273081e-05,
+      "loss": 4.1152,
+      "step": 367000
+    },
+    {
+      "epoch": 3.1598657024793386,
+      "grad_norm": 3.109375,
+      "learning_rate": 5.9900172670077725e-05,
+      "loss": 4.1262,
+      "step": 367050
+    },
+    {
+      "epoch": 3.160296143250689,
+      "grad_norm": 1.125,
+      "learning_rate": 5.9875355700665684e-05,
+      "loss": 4.0769,
+      "step": 367100
+    },
+    {
+      "epoch": 3.1607265840220387,
+      "grad_norm": 0.953125,
+      "learning_rate": 5.9850541676316185e-05,
+      "loss": 4.2557,
+      "step": 367150
+    },
+    {
+      "epoch": 3.1611570247933884,
+      "grad_norm": 1.8828125,
+      "learning_rate": 5.9825730598850594e-05,
+      "loss": 4.3774,
+      "step": 367200
+    },
+    {
+      "epoch": 3.1615874655647382,
+      "grad_norm": 2.8125,
+      "learning_rate": 5.9800922470089915e-05,
+      "loss": 4.6077,
+      "step": 367250
+    },
+    {
+      "epoch": 3.162017906336088,
+      "grad_norm": 3.953125,
+      "learning_rate": 5.977611729185502e-05,
+      "loss": 4.4091,
+      "step": 367300
+    },
+    {
+      "epoch": 3.162448347107438,
+      "grad_norm": 2.90625,
+      "learning_rate": 5.975131506596656e-05,
+      "loss": 4.4654,
+      "step": 367350
+    },
+    {
+      "epoch": 3.162878787878788,
+      "grad_norm": 4.25,
+      "learning_rate": 5.972651579424502e-05,
+      "loss": 4.5842,
+      "step": 367400
+    },
+    {
+      "epoch": 3.163309228650138,
+      "grad_norm": 1.3984375,
+      "learning_rate": 5.970171947851058e-05,
+      "loss": 4.3589,
+      "step": 367450
+    },
+    {
+      "epoch": 3.1637396694214877,
+      "grad_norm": 2.96875,
+      "learning_rate": 5.967692612058318e-05,
+      "loss": 4.1508,
+      "step": 367500
+    },
+    {
+      "epoch": 3.1641701101928374,
+      "grad_norm": 3.734375,
+      "learning_rate": 5.965213572228264e-05,
+      "loss": 4.2954,
+      "step": 367550
+    },
+    {
+      "epoch": 3.1646005509641872,
+      "grad_norm": 4.84375,
+      "learning_rate": 5.962734828542855e-05,
+      "loss": 4.2682,
+      "step": 367600
+    },
+    {
+      "epoch": 3.165030991735537,
+      "grad_norm": 2.3125,
+      "learning_rate": 5.9602563811840215e-05,
+      "loss": 4.1953,
+      "step": 367650
+    },
+    {
+      "epoch": 3.165461432506887,
+      "grad_norm": 4.1875,
+      "learning_rate": 5.957778230333684e-05,
+      "loss": 4.307,
+      "step": 367700
+    },
+    {
+      "epoch": 3.165891873278237,
+      "grad_norm": 2.703125,
+      "learning_rate": 5.955300376173722e-05,
+      "loss": 4.3906,
+      "step": 367750
+    },
+    {
+      "epoch": 3.166322314049587,
+      "grad_norm": 4.125,
+      "learning_rate": 5.9528228188860104e-05,
+      "loss": 4.0493,
+      "step": 367800
+    },
+    {
+      "epoch": 3.1667527548209367,
+      "grad_norm": 3.21875,
+      "learning_rate": 5.950345558652395e-05,
+      "loss": 4.3504,
+      "step": 367850
+    },
+    {
+      "epoch": 3.1671831955922864,
+      "grad_norm": 3.296875,
+      "learning_rate": 5.947868595654704e-05,
+      "loss": 4.2462,
+      "step": 367900
+    },
+    {
+      "epoch": 3.1676136363636362,
+      "grad_norm": 3.984375,
+      "learning_rate": 5.945391930074743e-05,
+      "loss": 4.0279,
+      "step": 367950
+    },
+    {
+      "epoch": 3.168044077134986,
+      "grad_norm": 1.828125,
+      "learning_rate": 5.9429155620942866e-05,
+      "loss": 4.1845,
+      "step": 368000
+    },
+    {
+      "epoch": 3.1684745179063363,
+      "grad_norm": 1.9453125,
+      "learning_rate": 5.940439491895098e-05,
+      "loss": 4.1649,
+      "step": 368050
+    },
+    {
+      "epoch": 3.168904958677686,
+      "grad_norm": 2.90625,
+      "learning_rate": 5.937963719658919e-05,
+      "loss": 4.2094,
+      "step": 368100
+    },
+    {
+      "epoch": 3.169335399449036,
+      "grad_norm": 2.46875,
+      "learning_rate": 5.93548824556746e-05,
+      "loss": 4.1752,
+      "step": 368150
+    },
+    {
+      "epoch": 3.1697658402203857,
+      "grad_norm": 4.125,
+      "learning_rate": 5.9330130698024244e-05,
+      "loss": 4.126,
+      "step": 368200
+    },
+    {
+      "epoch": 3.1701962809917354,
+      "grad_norm": 2.765625,
+      "learning_rate": 5.930538192545474e-05,
+      "loss": 4.194,
+      "step": 368250
+    },
+    {
+      "epoch": 3.1706267217630852,
+      "grad_norm": 5.34375,
+      "learning_rate": 5.928063613978268e-05,
+      "loss": 4.2881,
+      "step": 368300
+    },
+    {
+      "epoch": 3.171057162534435,
+      "grad_norm": 4.03125,
+      "learning_rate": 5.9255893342824284e-05,
+      "loss": 4.4968,
+      "step": 368350
+    },
+    {
+      "epoch": 3.1714876033057853,
+      "grad_norm": 3.640625,
+      "learning_rate": 5.923115353639568e-05,
+      "loss": 4.2541,
+      "step": 368400
+    },
+    {
+      "epoch": 3.171918044077135,
+      "grad_norm": 3.109375,
+      "learning_rate": 5.9206416722312685e-05,
+      "loss": 4.1264,
+      "step": 368450
+    },
+    {
+      "epoch": 3.172348484848485,
+      "grad_norm": 2.046875,
+      "learning_rate": 5.9181682902390935e-05,
+      "loss": 4.5032,
+      "step": 368500
+    },
+    {
+      "epoch": 3.1727789256198347,
+      "grad_norm": 3.078125,
+      "learning_rate": 5.915695207844581e-05,
+      "loss": 4.1226,
+      "step": 368550
+    },
+    {
+      "epoch": 3.1732093663911844,
+      "grad_norm": 3.328125,
+      "learning_rate": 5.913222425229255e-05,
+      "loss": 4.424,
+      "step": 368600
+    },
+    {
+      "epoch": 3.1736398071625342,
+      "grad_norm": 2.5,
+      "learning_rate": 5.9107499425746095e-05,
+      "loss": 4.3525,
+      "step": 368650
+    },
+    {
+      "epoch": 3.1740702479338845,
+      "grad_norm": 3.171875,
+      "learning_rate": 5.908277760062118e-05,
+      "loss": 4.4041,
+      "step": 368700
+    },
+    {
+      "epoch": 3.1745006887052343,
+      "grad_norm": 4.4375,
+      "learning_rate": 5.9058058778732406e-05,
+      "loss": 4.3873,
+      "step": 368750
+    },
+    {
+      "epoch": 3.174931129476584,
+      "grad_norm": 3.84375,
+      "learning_rate": 5.9033342961894e-05,
+      "loss": 3.9546,
+      "step": 368800
+    },
+    {
+      "epoch": 3.175361570247934,
+      "grad_norm": 1.1484375,
+      "learning_rate": 5.900863015192007e-05,
+      "loss": 4.5598,
+      "step": 368850
+    },
+    {
+      "epoch": 3.1757920110192837,
+      "grad_norm": 4.53125,
+      "learning_rate": 5.8983920350624475e-05,
+      "loss": 4.0618,
+      "step": 368900
+    },
+    {
+      "epoch": 3.1762224517906334,
+      "grad_norm": 3.75,
+      "learning_rate": 5.8959213559820926e-05,
+      "loss": 4.2138,
+      "step": 368950
+    },
+    {
+      "epoch": 3.1766528925619832,
+      "grad_norm": 6.0,
+      "learning_rate": 5.89345097813228e-05,
+      "loss": 4.3419,
+      "step": 369000
+    },
+    {
+      "epoch": 3.1766528925619832,
+      "eval_loss": 4.995718955993652,
+      "eval_runtime": 24.0677,
+      "eval_samples_per_second": 26.592,
+      "eval_steps_per_second": 13.296,
+      "eval_tts_loss": 7.504618805636363,
+      "step": 369000
+    },
+    {
+      "epoch": 3.1770833333333335,
+      "grad_norm": 2.734375,
+      "learning_rate": 5.890980901694325e-05,
+      "loss": 4.319,
+      "step": 369050
+    },
+    {
+      "epoch": 3.1775137741046833,
+      "grad_norm": 2.265625,
+      "learning_rate": 5.888511126849535e-05,
+      "loss": 4.25,
+      "step": 369100
+    },
+    {
+      "epoch": 3.177944214876033,
+      "grad_norm": 2.734375,
+      "learning_rate": 5.8860416537791796e-05,
+      "loss": 4.1594,
+      "step": 369150
+    },
+    {
+      "epoch": 3.178374655647383,
+      "grad_norm": 2.8125,
+      "learning_rate": 5.883572482664516e-05,
+      "loss": 4.4726,
+      "step": 369200
+    },
+    {
+      "epoch": 3.1788050964187327,
+      "grad_norm": 3.5,
+      "learning_rate": 5.881103613686778e-05,
+      "loss": 4.3734,
+      "step": 369250
+    },
+    {
+      "epoch": 3.1792355371900825,
+      "grad_norm": 3.5625,
+      "learning_rate": 5.878635047027172e-05,
+      "loss": 4.4663,
+      "step": 369300
+    },
+    {
+      "epoch": 3.1796659779614327,
+      "grad_norm": 4.4375,
+      "learning_rate": 5.876166782866883e-05,
+      "loss": 4.1323,
+      "step": 369350
+    },
+    {
+      "epoch": 3.1800964187327825,
+      "grad_norm": 4.375,
+      "learning_rate": 5.8736988213870816e-05,
+      "loss": 4.2094,
+      "step": 369400
+    },
+    {
+      "epoch": 3.1805268595041323,
+      "grad_norm": 3.828125,
+      "learning_rate": 5.871231162768909e-05,
+      "loss": 4.7159,
+      "step": 369450
+    },
+    {
+      "epoch": 3.180957300275482,
+      "grad_norm": 4.84375,
+      "learning_rate": 5.8687638071934884e-05,
+      "loss": 4.0699,
+      "step": 369500
+    },
+    {
+      "epoch": 3.181387741046832,
+      "grad_norm": 3.046875,
+      "learning_rate": 5.866296754841911e-05,
+      "loss": 3.8833,
+      "step": 369550
+    },
+    {
+      "epoch": 3.1818181818181817,
+      "grad_norm": 3.4375,
+      "learning_rate": 5.863830005895259e-05,
+      "loss": 4.1504,
+      "step": 369600
+    },
+    {
+      "epoch": 3.1822486225895315,
+      "grad_norm": 4.4375,
+      "learning_rate": 5.8613635605345865e-05,
+      "loss": 4.7323,
+      "step": 369650
+    },
+    {
+      "epoch": 3.1826790633608817,
+      "grad_norm": 4.125,
+      "learning_rate": 5.858897418940923e-05,
+      "loss": 3.8436,
+      "step": 369700
+    },
+    {
+      "epoch": 3.1831095041322315,
+      "grad_norm": 1.2421875,
+      "learning_rate": 5.856431581295282e-05,
+      "loss": 3.9756,
+      "step": 369750
+    },
+    {
+      "epoch": 3.1835399449035813,
+      "grad_norm": 4.875,
+      "learning_rate": 5.853966047778645e-05,
+      "loss": 4.1581,
+      "step": 369800
+    },
+    {
+      "epoch": 3.183970385674931,
+      "grad_norm": 3.921875,
+      "learning_rate": 5.85150081857198e-05,
+      "loss": 4.2672,
+      "step": 369850
+    },
+    {
+      "epoch": 3.184400826446281,
+      "grad_norm": 3.578125,
+      "learning_rate": 5.8490358938562294e-05,
+      "loss": 4.1397,
+      "step": 369900
+    },
+    {
+      "epoch": 3.1848312672176307,
+      "grad_norm": 2.890625,
+      "learning_rate": 5.8465712738123136e-05,
+      "loss": 4.2818,
+      "step": 369950
+    },
+    {
+      "epoch": 3.185261707988981,
+      "grad_norm": 3.90625,
+      "learning_rate": 5.844106958621132e-05,
+      "loss": 4.1011,
+      "step": 370000
+    },
+    {
+      "epoch": 3.1856921487603307,
+      "grad_norm": 4.25,
+      "learning_rate": 5.841642948463555e-05,
+      "loss": 4.4151,
+      "step": 370050
+    },
+    {
+      "epoch": 3.1861225895316805,
+      "grad_norm": 7.1875,
+      "learning_rate": 5.839179243520439e-05,
+      "loss": 4.2047,
+      "step": 370100
+    },
+    {
+      "epoch": 3.1865530303030303,
+      "grad_norm": 2.90625,
+      "learning_rate": 5.8367158439726134e-05,
+      "loss": 4.4279,
+      "step": 370150
+    },
+    {
+      "epoch": 3.18698347107438,
+      "grad_norm": 4.21875,
+      "learning_rate": 5.83425275000089e-05,
+      "loss": 4.1264,
+      "step": 370200
+    },
+    {
+      "epoch": 3.18741391184573,
+      "grad_norm": 5.25,
+      "learning_rate": 5.8317899617860516e-05,
+      "loss": 4.0315,
+      "step": 370250
+    },
+    {
+      "epoch": 3.1878443526170797,
+      "grad_norm": 4.03125,
+      "learning_rate": 5.829327479508861e-05,
+      "loss": 4.2168,
+      "step": 370300
+    },
+    {
+      "epoch": 3.18827479338843,
+      "grad_norm": 1.5859375,
+      "learning_rate": 5.8268653033500564e-05,
+      "loss": 4.127,
+      "step": 370350
+    },
+    {
+      "epoch": 3.1887052341597797,
+      "grad_norm": 3.265625,
+      "learning_rate": 5.824403433490364e-05,
+      "loss": 4.2178,
+      "step": 370400
+    },
+    {
+      "epoch": 3.1891356749311295,
+      "grad_norm": 3.046875,
+      "learning_rate": 5.8219418701104745e-05,
+      "loss": 4.5114,
+      "step": 370450
+    },
+    {
+      "epoch": 3.1895661157024793,
+      "grad_norm": 3.796875,
+      "learning_rate": 5.819480613391059e-05,
+      "loss": 4.2566,
+      "step": 370500
+    },
+    {
+      "epoch": 3.189996556473829,
+      "grad_norm": 3.71875,
+      "learning_rate": 5.817019663512782e-05,
+      "loss": 4.6002,
+      "step": 370550
+    },
+    {
+      "epoch": 3.190426997245179,
+      "grad_norm": 1.8984375,
+      "learning_rate": 5.814559020656255e-05,
+      "loss": 4.8422,
+      "step": 370600
+    },
+    {
+      "epoch": 3.190857438016529,
+      "grad_norm": 5.46875,
+      "learning_rate": 5.8120986850020876e-05,
+      "loss": 4.3317,
+      "step": 370650
+    },
+    {
+      "epoch": 3.191287878787879,
+      "grad_norm": 5.25,
+      "learning_rate": 5.80963865673087e-05,
+      "loss": 4.0244,
+      "step": 370700
+    },
+    {
+      "epoch": 3.1917183195592287,
+      "grad_norm": 4.9375,
+      "learning_rate": 5.80717893602316e-05,
+      "loss": 4.6798,
+      "step": 370750
+    },
+    {
+      "epoch": 3.1921487603305785,
+      "grad_norm": 4.34375,
+      "learning_rate": 5.8047195230594944e-05,
+      "loss": 4.1681,
+      "step": 370800
+    },
+    {
+      "epoch": 3.1925792011019283,
+      "grad_norm": 2.984375,
+      "learning_rate": 5.802260418020389e-05,
+      "loss": 4.3521,
+      "step": 370850
+    },
+    {
+      "epoch": 3.193009641873278,
+      "grad_norm": 3.78125,
+      "learning_rate": 5.7998016210863336e-05,
+      "loss": 4.3071,
+      "step": 370900
+    },
+    {
+      "epoch": 3.193440082644628,
+      "grad_norm": 1.953125,
+      "learning_rate": 5.797343132437807e-05,
+      "loss": 4.3148,
+      "step": 370950
+    },
+    {
+      "epoch": 3.193870523415978,
+      "grad_norm": 6.15625,
+      "learning_rate": 5.7948849522552505e-05,
+      "loss": 4.2374,
+      "step": 371000
+    },
+    {
+      "epoch": 3.194300964187328,
+      "grad_norm": 1.9609375,
+      "learning_rate": 5.7924270807190925e-05,
+      "loss": 4.4721,
+      "step": 371050
+    },
+    {
+      "epoch": 3.1947314049586777,
+      "grad_norm": 1.4375,
+      "learning_rate": 5.7899695180097335e-05,
+      "loss": 4.2304,
+      "step": 371100
+    },
+    {
+      "epoch": 3.1951618457300275,
+      "grad_norm": 8.6875,
+      "learning_rate": 5.787512264307554e-05,
+      "loss": 4.1293,
+      "step": 371150
+    },
+    {
+      "epoch": 3.1955922865013773,
+      "grad_norm": 3.78125,
+      "learning_rate": 5.785055319792907e-05,
+      "loss": 4.3817,
+      "step": 371200
+    },
+    {
+      "epoch": 3.196022727272727,
+      "grad_norm": 2.890625,
+      "learning_rate": 5.782598684646134e-05,
+      "loss": 4.5016,
+      "step": 371250
+    },
+    {
+      "epoch": 3.1964531680440773,
+      "grad_norm": 4.21875,
+      "learning_rate": 5.780142359047549e-05,
+      "loss": 4.2858,
+      "step": 371300
+    },
+    {
+      "epoch": 3.196883608815427,
+      "grad_norm": 3.375,
+      "learning_rate": 5.7776863431774263e-05,
+      "loss": 4.2233,
+      "step": 371350
+    },
+    {
+      "epoch": 3.197314049586777,
+      "grad_norm": 4.78125,
+      "learning_rate": 5.775230637216045e-05,
+      "loss": 4.2313,
+      "step": 371400
+    },
+    {
+      "epoch": 3.1977444903581267,
+      "grad_norm": 3.828125,
+      "learning_rate": 5.7727752413436465e-05,
+      "loss": 3.9859,
+      "step": 371450
+    },
+    {
+      "epoch": 3.1981749311294765,
+      "grad_norm": 4.375,
+      "learning_rate": 5.7703201557404454e-05,
+      "loss": 4.2741,
+      "step": 371500
+    },
+    {
+      "epoch": 3.1986053719008263,
+      "grad_norm": 2.8125,
+      "learning_rate": 5.767865380586652e-05,
+      "loss": 4.0336,
+      "step": 371550
+    },
+    {
+      "epoch": 3.1990358126721765,
+      "grad_norm": 4.34375,
+      "learning_rate": 5.765410916062428e-05,
+      "loss": 4.1959,
+      "step": 371600
+    },
+    {
+      "epoch": 3.1994662534435263,
+      "grad_norm": 3.03125,
+      "learning_rate": 5.762956762347934e-05,
+      "loss": 4.3028,
+      "step": 371650
+    },
+    {
+      "epoch": 3.199896694214876,
+      "grad_norm": 2.90625,
+      "learning_rate": 5.7605029196232984e-05,
+      "loss": 4.0621,
+      "step": 371700
+    },
+    {
+      "epoch": 3.200327134986226,
+      "grad_norm": 3.890625,
+      "learning_rate": 5.758049388068626e-05,
+      "loss": 4.618,
+      "step": 371750
+    },
+    {
+      "epoch": 3.2007575757575757,
+      "grad_norm": 4.84375,
+      "learning_rate": 5.7555961678640015e-05,
+      "loss": 4.2074,
+      "step": 371800
+    },
+    {
+      "epoch": 3.2011880165289255,
+      "grad_norm": 1.6328125,
+      "learning_rate": 5.753143259189487e-05,
+      "loss": 4.3642,
+      "step": 371850
+    },
+    {
+      "epoch": 3.2016184573002753,
+      "grad_norm": 2.53125,
+      "learning_rate": 5.7506906622251165e-05,
+      "loss": 4.0858,
+      "step": 371900
+    },
+    {
+      "epoch": 3.2020488980716255,
+      "grad_norm": 5.03125,
+      "learning_rate": 5.7482383771509115e-05,
+      "loss": 4.495,
+      "step": 371950
+    },
+    {
+      "epoch": 3.2024793388429753,
+      "grad_norm": 3.65625,
+      "learning_rate": 5.7457864041468625e-05,
+      "loss": 4.0665,
+      "step": 372000
+    },
+    {
+      "epoch": 3.2024793388429753,
+      "eval_loss": 4.995144844055176,
+      "eval_runtime": 24.1419,
+      "eval_samples_per_second": 26.51,
+      "eval_steps_per_second": 13.255,
+      "eval_tts_loss": 7.490159339418232,
+      "step": 372000
+    },
+    {
+      "epoch": 3.202909779614325,
+      "grad_norm": 4.28125,
+      "learning_rate": 5.743334743392939e-05,
+      "loss": 4.3072,
+      "step": 372050
+    },
+    {
+      "epoch": 3.203340220385675,
+      "grad_norm": 3.375,
+      "learning_rate": 5.7408833950690856e-05,
+      "loss": 4.2735,
+      "step": 372100
+    },
+    {
+      "epoch": 3.2037706611570247,
+      "grad_norm": 1.4296875,
+      "learning_rate": 5.738432359355228e-05,
+      "loss": 4.3538,
+      "step": 372150
+    },
+    {
+      "epoch": 3.2042011019283745,
+      "grad_norm": 2.046875,
+      "learning_rate": 5.7359816364312614e-05,
+      "loss": 4.5132,
+      "step": 372200
+    },
+    {
+      "epoch": 3.2046315426997247,
+      "grad_norm": 2.859375,
+      "learning_rate": 5.7335312264770735e-05,
+      "loss": 4.2675,
+      "step": 372250
+    },
+    {
+      "epoch": 3.2050619834710745,
+      "grad_norm": 1.546875,
+      "learning_rate": 5.731081129672513e-05,
+      "loss": 4.345,
+      "step": 372300
+    },
+    {
+      "epoch": 3.2054924242424243,
+      "grad_norm": 3.953125,
+      "learning_rate": 5.728631346197413e-05,
+      "loss": 4.4702,
+      "step": 372350
+    },
+    {
+      "epoch": 3.205922865013774,
+      "grad_norm": 5.625,
+      "learning_rate": 5.726181876231582e-05,
+      "loss": 4.4556,
+      "step": 372400
+    },
+    {
+      "epoch": 3.206353305785124,
+      "grad_norm": 3.5625,
+      "learning_rate": 5.723732719954806e-05,
+      "loss": 4.4963,
+      "step": 372450
+    },
+    {
+      "epoch": 3.2067837465564737,
+      "grad_norm": 4.96875,
+      "learning_rate": 5.7212838775468434e-05,
+      "loss": 4.2145,
+      "step": 372500
+    },
+    {
+      "epoch": 3.2072141873278235,
+      "grad_norm": 3.109375,
+      "learning_rate": 5.718835349187444e-05,
+      "loss": 4.0102,
+      "step": 372550
+    },
+    {
+      "epoch": 3.2076446280991737,
+      "grad_norm": 4.0625,
+      "learning_rate": 5.716387135056318e-05,
+      "loss": 4.087,
+      "step": 372600
+    },
+    {
+      "epoch": 3.2080750688705235,
+      "grad_norm": 4.46875,
+      "learning_rate": 5.713939235333159e-05,
+      "loss": 4.5784,
+      "step": 372650
+    },
+    {
+      "epoch": 3.2085055096418733,
+      "grad_norm": 3.15625,
+      "learning_rate": 5.711491650197639e-05,
+      "loss": 4.6105,
+      "step": 372700
+    },
+    {
+      "epoch": 3.208935950413223,
+      "grad_norm": 3.390625,
+      "learning_rate": 5.7090443798294e-05,
+      "loss": 4.1451,
+      "step": 372750
+    },
+    {
+      "epoch": 3.209366391184573,
+      "grad_norm": 4.1875,
+      "learning_rate": 5.7065974244080764e-05,
+      "loss": 4.231,
+      "step": 372800
+    },
+    {
+      "epoch": 3.2097968319559227,
+      "grad_norm": 4.59375,
+      "learning_rate": 5.704150784113269e-05,
+      "loss": 4.4618,
+      "step": 372850
+    },
+    {
+      "epoch": 3.210227272727273,
+      "grad_norm": 2.90625,
+      "learning_rate": 5.7017044591245416e-05,
+      "loss": 4.5273,
+      "step": 372900
+    },
+    {
+      "epoch": 3.2106577134986227,
+      "grad_norm": 3.625,
+      "learning_rate": 5.699258449621464e-05,
+      "loss": 4.3227,
+      "step": 372950
+    },
+    {
+      "epoch": 3.2110881542699725,
+      "grad_norm": 2.828125,
+      "learning_rate": 5.6968127557835626e-05,
+      "loss": 3.9205,
+      "step": 373000
+    },
+    {
+      "epoch": 3.2115185950413223,
+      "grad_norm": 4.09375,
+      "learning_rate": 5.694367377790342e-05,
+      "loss": 4.5589,
+      "step": 373050
+    },
+    {
+      "epoch": 3.211949035812672,
+      "grad_norm": 1.765625,
+      "learning_rate": 5.691922315821301e-05,
+      "loss": 4.1797,
+      "step": 373100
+    },
+    {
+      "epoch": 3.212379476584022,
+      "grad_norm": 2.765625,
+      "learning_rate": 5.6894775700558875e-05,
+      "loss": 4.585,
+      "step": 373150
+    },
+    {
+      "epoch": 3.212809917355372,
+      "grad_norm": 4.34375,
+      "learning_rate": 5.687033140673543e-05,
+      "loss": 4.1926,
+      "step": 373200
+    },
+    {
+      "epoch": 3.213240358126722,
+      "grad_norm": 3.90625,
+      "learning_rate": 5.684589027853691e-05,
+      "loss": 4.228,
+      "step": 373250
+    },
+    {
+      "epoch": 3.2136707988980717,
+      "grad_norm": 1.7109375,
+      "learning_rate": 5.682145231775719e-05,
+      "loss": 4.4378,
+      "step": 373300
+    },
+    {
+      "epoch": 3.2141012396694215,
+      "grad_norm": 4.96875,
+      "learning_rate": 5.679701752618996e-05,
+      "loss": 4.3353,
+      "step": 373350
+    },
+    {
+      "epoch": 3.2145316804407713,
+      "grad_norm": 2.296875,
+      "learning_rate": 5.6772585905628706e-05,
+      "loss": 4.4708,
+      "step": 373400
+    },
+    {
+      "epoch": 3.214962121212121,
+      "grad_norm": 1.953125,
+      "learning_rate": 5.674815745786659e-05,
+      "loss": 4.0621,
+      "step": 373450
+    },
+    {
+      "epoch": 3.215392561983471,
+      "grad_norm": 5.15625,
+      "learning_rate": 5.672373218469669e-05,
+      "loss": 4.0959,
+      "step": 373500
+    },
+    {
+      "epoch": 3.215823002754821,
+      "grad_norm": 1.9609375,
+      "learning_rate": 5.669931008791175e-05,
+      "loss": 4.3164,
+      "step": 373550
+    },
+    {
+      "epoch": 3.216253443526171,
+      "grad_norm": 2.125,
+      "learning_rate": 5.667489116930428e-05,
+      "loss": 4.2294,
+      "step": 373600
+    },
+    {
+      "epoch": 3.2166838842975207,
+      "grad_norm": 3.484375,
+      "learning_rate": 5.6650475430666584e-05,
+      "loss": 4.1345,
+      "step": 373650
+    },
+    {
+      "epoch": 3.2171143250688705,
+      "grad_norm": 1.515625,
+      "learning_rate": 5.662606287379072e-05,
+      "loss": 3.9975,
+      "step": 373700
+    },
+    {
+      "epoch": 3.2175447658402203,
+      "grad_norm": 2.0,
+      "learning_rate": 5.660165350046849e-05,
+      "loss": 4.4641,
+      "step": 373750
+    },
+    {
+      "epoch": 3.21797520661157,
+      "grad_norm": 3.015625,
+      "learning_rate": 5.657724731249154e-05,
+      "loss": 4.359,
+      "step": 373800
+    },
+    {
+      "epoch": 3.2184056473829203,
+      "grad_norm": 2.921875,
+      "learning_rate": 5.655284431165122e-05,
+      "loss": 4.5078,
+      "step": 373850
+    },
+    {
+      "epoch": 3.21883608815427,
+      "grad_norm": 2.734375,
+      "learning_rate": 5.6528444499738644e-05,
+      "loss": 4.358,
+      "step": 373900
+    },
+    {
+      "epoch": 3.21926652892562,
+      "grad_norm": 2.28125,
+      "learning_rate": 5.6504047878544716e-05,
+      "loss": 4.0136,
+      "step": 373950
+    },
+    {
+      "epoch": 3.2196969696969697,
+      "grad_norm": 1.625,
+      "learning_rate": 5.64796544498601e-05,
+      "loss": 4.3448,
+      "step": 374000
+    },
+    {
+      "epoch": 3.2201274104683195,
+      "grad_norm": 4.65625,
+      "learning_rate": 5.6455264215475156e-05,
+      "loss": 4.0919,
+      "step": 374050
+    },
+    {
+      "epoch": 3.2205578512396693,
+      "grad_norm": 1.78125,
+      "learning_rate": 5.6430877177180176e-05,
+      "loss": 4.3081,
+      "step": 374100
+    },
+    {
+      "epoch": 3.220988292011019,
+      "grad_norm": 5.875,
+      "learning_rate": 5.6406493336765064e-05,
+      "loss": 3.9731,
+      "step": 374150
+    },
+    {
+      "epoch": 3.2214187327823693,
+      "grad_norm": 8.75,
+      "learning_rate": 5.638211269601955e-05,
+      "loss": 4.5018,
+      "step": 374200
+    },
+    {
+      "epoch": 3.221849173553719,
+      "grad_norm": 3.265625,
+      "learning_rate": 5.635773525673311e-05,
+      "loss": 4.6859,
+      "step": 374250
+    },
+    {
+      "epoch": 3.222279614325069,
+      "grad_norm": 2.28125,
+      "learning_rate": 5.633336102069497e-05,
+      "loss": 4.0108,
+      "step": 374300
+    },
+    {
+      "epoch": 3.2227100550964187,
+      "grad_norm": 3.453125,
+      "learning_rate": 5.630898998969423e-05,
+      "loss": 4.2525,
+      "step": 374350
+    },
+    {
+      "epoch": 3.2231404958677685,
+      "grad_norm": 4.0625,
+      "learning_rate": 5.6284622165519664e-05,
+      "loss": 4.1158,
+      "step": 374400
+    },
+    {
+      "epoch": 3.2235709366391183,
+      "grad_norm": 3.328125,
+      "learning_rate": 5.626025754995967e-05,
+      "loss": 4.3399,
+      "step": 374450
+    },
+    {
+      "epoch": 3.2240013774104685,
+      "grad_norm": 1.6875,
+      "learning_rate": 5.6235896144802726e-05,
+      "loss": 4.0313,
+      "step": 374500
+    },
+    {
+      "epoch": 3.2244318181818183,
+      "grad_norm": 2.21875,
+      "learning_rate": 5.621153795183683e-05,
+      "loss": 4.2005,
+      "step": 374550
+    },
+    {
+      "epoch": 3.224862258953168,
+      "grad_norm": 3.65625,
+      "learning_rate": 5.6187182972849805e-05,
+      "loss": 4.5472,
+      "step": 374600
+    },
+    {
+      "epoch": 3.225292699724518,
+      "grad_norm": 3.34375,
+      "learning_rate": 5.616283120962936e-05,
+      "loss": 4.6579,
+      "step": 374650
+    },
+    {
+      "epoch": 3.2257231404958677,
+      "grad_norm": 4.1875,
+      "learning_rate": 5.613848266396274e-05,
+      "loss": 3.9426,
+      "step": 374700
+    },
+    {
+      "epoch": 3.2261535812672175,
+      "grad_norm": 3.28125,
+      "learning_rate": 5.61141373376371e-05,
+      "loss": 4.0532,
+      "step": 374750
+    },
+    {
+      "epoch": 3.2265840220385673,
+      "grad_norm": 3.65625,
+      "learning_rate": 5.608979523243938e-05,
+      "loss": 4.3047,
+      "step": 374800
+    },
+    {
+      "epoch": 3.2270144628099175,
+      "grad_norm": 3.5,
+      "learning_rate": 5.606545635015622e-05,
+      "loss": 4.0366,
+      "step": 374850
+    },
+    {
+      "epoch": 3.2274449035812673,
+      "grad_norm": 4.375,
+      "learning_rate": 5.604112069257403e-05,
+      "loss": 4.5421,
+      "step": 374900
+    },
+    {
+      "epoch": 3.227875344352617,
+      "grad_norm": 2.203125,
+      "learning_rate": 5.6016788261479006e-05,
+      "loss": 3.9863,
+      "step": 374950
+    },
+    {
+      "epoch": 3.228305785123967,
+      "grad_norm": 3.71875,
+      "learning_rate": 5.5992459058657056e-05,
+      "loss": 4.5136,
+      "step": 375000
+    },
+    {
+      "epoch": 3.228305785123967,
+      "eval_loss": 4.99493932723999,
+      "eval_runtime": 24.1536,
+      "eval_samples_per_second": 26.497,
+      "eval_steps_per_second": 13.249,
+      "eval_tts_loss": 7.483240953229233,
+      "step": 375000
+    },
+    {
+      "epoch": 3.2287362258953167,
+      "grad_norm": 1.5625,
+      "learning_rate": 5.596813308589396e-05,
+      "loss": 4.0813,
+      "step": 375050
+    },
+    {
+      "epoch": 3.2291666666666665,
+      "grad_norm": 3.296875,
+      "learning_rate": 5.594381034497516e-05,
+      "loss": 4.5951,
+      "step": 375100
+    },
+    {
+      "epoch": 3.2295971074380168,
+      "grad_norm": 6.375,
+      "learning_rate": 5.5919490837685896e-05,
+      "loss": 4.3345,
+      "step": 375150
+    },
+    {
+      "epoch": 3.2300275482093666,
+      "grad_norm": 7.09375,
+      "learning_rate": 5.589517456581116e-05,
+      "loss": 4.4634,
+      "step": 375200
+    },
+    {
+      "epoch": 3.2304579889807163,
+      "grad_norm": 3.3125,
+      "learning_rate": 5.5870861531135724e-05,
+      "loss": 4.3415,
+      "step": 375250
+    },
+    {
+      "epoch": 3.230888429752066,
+      "grad_norm": 3.125,
+      "learning_rate": 5.584655173544404e-05,
+      "loss": 4.7467,
+      "step": 375300
+    },
+    {
+      "epoch": 3.231318870523416,
+      "grad_norm": 4.21875,
+      "learning_rate": 5.582224518052053e-05,
+      "loss": 3.8459,
+      "step": 375350
+    },
+    {
+      "epoch": 3.2317493112947657,
+      "grad_norm": 3.328125,
+      "learning_rate": 5.5797941868149196e-05,
+      "loss": 4.0402,
+      "step": 375400
+    },
+    {
+      "epoch": 3.2321797520661155,
+      "grad_norm": 3.5,
+      "learning_rate": 5.577364180011375e-05,
+      "loss": 4.3711,
+      "step": 375450
+    },
+    {
+      "epoch": 3.2326101928374658,
+      "grad_norm": 3.53125,
+      "learning_rate": 5.574934497819788e-05,
+      "loss": 4.3598,
+      "step": 375500
+    },
+    {
+      "epoch": 3.2330406336088156,
+      "grad_norm": 4.03125,
+      "learning_rate": 5.572505140418488e-05,
+      "loss": 4.4656,
+      "step": 375550
+    },
+    {
+      "epoch": 3.2334710743801653,
+      "grad_norm": 1.640625,
+      "learning_rate": 5.57007610798578e-05,
+      "loss": 4.3525,
+      "step": 375600
+    },
+    {
+      "epoch": 3.233901515151515,
+      "grad_norm": 3.71875,
+      "learning_rate": 5.567647400699965e-05,
+      "loss": 4.3795,
+      "step": 375650
+    },
+    {
+      "epoch": 3.234331955922865,
+      "grad_norm": 2.703125,
+      "learning_rate": 5.565219018739283e-05,
+      "loss": 4.4562,
+      "step": 375700
+    },
+    {
+      "epoch": 3.2347623966942147,
+      "grad_norm": 6.5625,
+      "learning_rate": 5.56279096228199e-05,
+      "loss": 4.496,
+      "step": 375750
+    },
+    {
+      "epoch": 3.235192837465565,
+      "grad_norm": 3.71875,
+      "learning_rate": 5.5603632315062915e-05,
+      "loss": 4.3562,
+      "step": 375800
+    },
+    {
+      "epoch": 3.2356232782369148,
+      "grad_norm": 4.46875,
+      "learning_rate": 5.557935826590377e-05,
+      "loss": 4.4144,
+      "step": 375850
+    },
+    {
+      "epoch": 3.2360537190082646,
+      "grad_norm": 2.609375,
+      "learning_rate": 5.555508747712419e-05,
+      "loss": 4.448,
+      "step": 375900
+    },
+    {
+      "epoch": 3.2364841597796143,
+      "grad_norm": 2.609375,
+      "learning_rate": 5.5530819950505597e-05,
+      "loss": 4.4849,
+      "step": 375950
+    },
+    {
+      "epoch": 3.236914600550964,
+      "grad_norm": 1.921875,
+      "learning_rate": 5.550655568782908e-05,
+      "loss": 4.4583,
+      "step": 376000
+    },
+    {
+      "epoch": 3.237345041322314,
+      "grad_norm": 3.390625,
+      "learning_rate": 5.548229469087567e-05,
+      "loss": 4.219,
+      "step": 376050
+    },
+    {
+      "epoch": 3.2377754820936637,
+      "grad_norm": 7.65625,
+      "learning_rate": 5.545803696142605e-05,
+      "loss": 3.8585,
+      "step": 376100
+    },
+    {
+      "epoch": 3.238205922865014,
+      "grad_norm": 1.375,
+      "learning_rate": 5.543378250126064e-05,
+      "loss": 4.1569,
+      "step": 376150
+    },
+    {
+      "epoch": 3.2386363636363638,
+      "grad_norm": 1.5859375,
+      "learning_rate": 5.54095313121598e-05,
+      "loss": 4.2678,
+      "step": 376200
+    },
+    {
+      "epoch": 3.2390668044077136,
+      "grad_norm": 3.453125,
+      "learning_rate": 5.538528339590336e-05,
+      "loss": 4.7378,
+      "step": 376250
+    },
+    {
+      "epoch": 3.2394972451790633,
+      "grad_norm": 2.0,
+      "learning_rate": 5.536103875427111e-05,
+      "loss": 4.5216,
+      "step": 376300
+    },
+    {
+      "epoch": 3.239927685950413,
+      "grad_norm": 5.0625,
+      "learning_rate": 5.5336797389042606e-05,
+      "loss": 4.1535,
+      "step": 376350
+    },
+    {
+      "epoch": 3.240358126721763,
+      "grad_norm": 4.09375,
+      "learning_rate": 5.531255930199708e-05,
+      "loss": 4.4992,
+      "step": 376400
+    },
+    {
+      "epoch": 3.240788567493113,
+      "grad_norm": 2.203125,
+      "learning_rate": 5.528832449491355e-05,
+      "loss": 4.4439,
+      "step": 376450
+    },
+    {
+      "epoch": 3.241219008264463,
+      "grad_norm": 4.78125,
+      "learning_rate": 5.52640929695708e-05,
+      "loss": 4.3586,
+      "step": 376500
+    },
+    {
+      "epoch": 3.2416494490358128,
+      "grad_norm": 4.5625,
+      "learning_rate": 5.523986472774734e-05,
+      "loss": 4.3194,
+      "step": 376550
+    },
+    {
+      "epoch": 3.2420798898071626,
+      "grad_norm": 4.625,
+      "learning_rate": 5.521563977122154e-05,
+      "loss": 3.9062,
+      "step": 376600
+    },
+    {
+      "epoch": 3.2425103305785123,
+      "grad_norm": 3.78125,
+      "learning_rate": 5.519141810177143e-05,
+      "loss": 3.869,
+      "step": 376650
+    },
+    {
+      "epoch": 3.242940771349862,
+      "grad_norm": 3.546875,
+      "learning_rate": 5.516719972117481e-05,
+      "loss": 3.9001,
+      "step": 376700
+    },
+    {
+      "epoch": 3.243371212121212,
+      "grad_norm": 2.5,
+      "learning_rate": 5.5142984631209284e-05,
+      "loss": 4.2057,
+      "step": 376750
+    },
+    {
+      "epoch": 3.243801652892562,
+      "grad_norm": 3.953125,
+      "learning_rate": 5.511877283365218e-05,
+      "loss": 4.3461,
+      "step": 376800
+    },
+    {
+      "epoch": 3.244232093663912,
+      "grad_norm": 2.03125,
+      "learning_rate": 5.509456433028053e-05,
+      "loss": 4.219,
+      "step": 376850
+    },
+    {
+      "epoch": 3.2446625344352618,
+      "grad_norm": 1.5078125,
+      "learning_rate": 5.5070359122871285e-05,
+      "loss": 4.1926,
+      "step": 376900
+    },
+    {
+      "epoch": 3.2450929752066116,
+      "grad_norm": 2.296875,
+      "learning_rate": 5.5046157213201056e-05,
+      "loss": 4.325,
+      "step": 376950
+    },
+    {
+      "epoch": 3.2455234159779613,
+      "grad_norm": 4.59375,
+      "learning_rate": 5.5021958603046084e-05,
+      "loss": 3.9349,
+      "step": 377000
+    },
+    {
+      "epoch": 3.245953856749311,
+      "grad_norm": 2.25,
+      "learning_rate": 5.499776329418262e-05,
+      "loss": 4.812,
+      "step": 377050
+    },
+    {
+      "epoch": 3.2463842975206614,
+      "grad_norm": 8.4375,
+      "learning_rate": 5.497357128838651e-05,
+      "loss": 4.5487,
+      "step": 377100
+    },
+    {
+      "epoch": 3.246814738292011,
+      "grad_norm": 3.171875,
+      "learning_rate": 5.494938258743334e-05,
+      "loss": 4.0864,
+      "step": 377150
+    },
+    {
+      "epoch": 3.247245179063361,
+      "grad_norm": 2.6875,
+      "learning_rate": 5.492519719309864e-05,
+      "loss": 4.1475,
+      "step": 377200
+    },
+    {
+      "epoch": 3.2476756198347108,
+      "grad_norm": 3.203125,
+      "learning_rate": 5.4901015107157416e-05,
+      "loss": 4.3365,
+      "step": 377250
+    },
+    {
+      "epoch": 3.2481060606060606,
+      "grad_norm": 3.421875,
+      "learning_rate": 5.4876836331384674e-05,
+      "loss": 4.5483,
+      "step": 377300
+    },
+    {
+      "epoch": 3.2485365013774103,
+      "grad_norm": 3.0,
+      "learning_rate": 5.485266086755507e-05,
+      "loss": 4.3026,
+      "step": 377350
+    },
+    {
+      "epoch": 3.24896694214876,
+      "grad_norm": 2.1875,
+      "learning_rate": 5.4828488717443024e-05,
+      "loss": 4.1662,
+      "step": 377400
+    },
+    {
+      "epoch": 3.2493973829201104,
+      "grad_norm": 4.90625,
+      "learning_rate": 5.480431988282271e-05,
+      "loss": 4.2475,
+      "step": 377450
+    },
+    {
+      "epoch": 3.24982782369146,
+      "grad_norm": 8.9375,
+      "learning_rate": 5.478015436546808e-05,
+      "loss": 4.4411,
+      "step": 377500
+    },
+    {
+      "epoch": 3.25025826446281,
+      "grad_norm": 3.9375,
+      "learning_rate": 5.475599216715279e-05,
+      "loss": 4.3599,
+      "step": 377550
+    },
+    {
+      "epoch": 3.2506887052341598,
+      "grad_norm": 4.8125,
+      "learning_rate": 5.473183328965037e-05,
+      "loss": 4.2462,
+      "step": 377600
+    },
+    {
+      "epoch": 3.2511191460055096,
+      "grad_norm": 3.09375,
+      "learning_rate": 5.470767773473398e-05,
+      "loss": 4.1422,
+      "step": 377650
+    },
+    {
+      "epoch": 3.2515495867768593,
+      "grad_norm": 4.46875,
+      "learning_rate": 5.468352550417657e-05,
+      "loss": 4.2798,
+      "step": 377700
+    },
+    {
+      "epoch": 3.2519800275482096,
+      "grad_norm": 3.390625,
+      "learning_rate": 5.4659376599750957e-05,
+      "loss": 4.6952,
+      "step": 377750
+    },
+    {
+      "epoch": 3.2524104683195594,
+      "grad_norm": 2.390625,
+      "learning_rate": 5.4635231023229506e-05,
+      "loss": 4.5524,
+      "step": 377800
+    },
+    {
+      "epoch": 3.252840909090909,
+      "grad_norm": 5.90625,
+      "learning_rate": 5.4611088776384453e-05,
+      "loss": 4.0466,
+      "step": 377850
+    },
+    {
+      "epoch": 3.253271349862259,
+      "grad_norm": 3.40625,
+      "learning_rate": 5.458694986098787e-05,
+      "loss": 4.219,
+      "step": 377900
+    },
+    {
+      "epoch": 3.2537017906336088,
+      "grad_norm": 2.984375,
+      "learning_rate": 5.4562814278811456e-05,
+      "loss": 4.5835,
+      "step": 377950
+    },
+    {
+      "epoch": 3.2541322314049586,
+      "grad_norm": 3.4375,
+      "learning_rate": 5.453868203162672e-05,
+      "loss": 4.2677,
+      "step": 378000
+    },
+    {
+      "epoch": 3.2541322314049586,
+      "eval_loss": 4.995173454284668,
+      "eval_runtime": 23.6957,
+      "eval_samples_per_second": 27.009,
+      "eval_steps_per_second": 13.505,
+      "eval_tts_loss": 7.484762244246582,
+      "step": 378000
+    },
+    {
+      "epoch": 3.2545626721763083,
+      "grad_norm": 2.0,
+      "learning_rate": 5.4514553121204905e-05,
+      "loss": 4.379,
+      "step": 378050
+    },
+    {
+      "epoch": 3.2549931129476586,
+      "grad_norm": 3.09375,
+      "learning_rate": 5.449042754931699e-05,
+      "loss": 4.2898,
+      "step": 378100
+    },
+    {
+      "epoch": 3.2554235537190084,
+      "grad_norm": 5.78125,
+      "learning_rate": 5.4466305317733826e-05,
+      "loss": 4.0237,
+      "step": 378150
+    },
+    {
+      "epoch": 3.255853994490358,
+      "grad_norm": 3.0,
+      "learning_rate": 5.444218642822586e-05,
+      "loss": 4.3617,
+      "step": 378200
+    },
+    {
+      "epoch": 3.256284435261708,
+      "grad_norm": 3.75,
+      "learning_rate": 5.441807088256341e-05,
+      "loss": 4.5322,
+      "step": 378250
+    },
+    {
+      "epoch": 3.2567148760330578,
+      "grad_norm": 4.125,
+      "learning_rate": 5.439395868251647e-05,
+      "loss": 4.128,
+      "step": 378300
+    },
+    {
+      "epoch": 3.2571453168044076,
+      "grad_norm": 3.015625,
+      "learning_rate": 5.4369849829854844e-05,
+      "loss": 4.1571,
+      "step": 378350
+    },
+    {
+      "epoch": 3.257575757575758,
+      "grad_norm": 3.421875,
+      "learning_rate": 5.434574432634802e-05,
+      "loss": 4.2891,
+      "step": 378400
+    },
+    {
+      "epoch": 3.2580061983471076,
+      "grad_norm": 3.28125,
+      "learning_rate": 5.432164217376539e-05,
+      "loss": 4.0523,
+      "step": 378450
+    },
+    {
+      "epoch": 3.2584366391184574,
+      "grad_norm": 2.609375,
+      "learning_rate": 5.429754337387598e-05,
+      "loss": 3.9968,
+      "step": 378500
+    },
+    {
+      "epoch": 3.258867079889807,
+      "grad_norm": 2.15625,
+      "learning_rate": 5.4273447928448477e-05,
+      "loss": 4.333,
+      "step": 378550
+    },
+    {
+      "epoch": 3.259297520661157,
+      "grad_norm": 2.265625,
+      "learning_rate": 5.4249355839251547e-05,
+      "loss": 4.5407,
+      "step": 378600
+    },
+    {
+      "epoch": 3.2597279614325068,
+      "grad_norm": 1.4609375,
+      "learning_rate": 5.422526710805346e-05,
+      "loss": 4.3556,
+      "step": 378650
+    },
+    {
+      "epoch": 3.2601584022038566,
+      "grad_norm": 2.71875,
+      "learning_rate": 5.420118173662224e-05,
+      "loss": 4.3925,
+      "step": 378700
+    },
+    {
+      "epoch": 3.260588842975207,
+      "grad_norm": 2.9375,
+      "learning_rate": 5.417709972672582e-05,
+      "loss": 4.2892,
+      "step": 378750
+    },
+    {
+      "epoch": 3.2610192837465566,
+      "grad_norm": 4.125,
+      "learning_rate": 5.4153021080131626e-05,
+      "loss": 4.8368,
+      "step": 378800
+    },
+    {
+      "epoch": 3.2614497245179064,
+      "grad_norm": 3.578125,
+      "learning_rate": 5.412894579860707e-05,
+      "loss": 4.168,
+      "step": 378850
+    },
+    {
+      "epoch": 3.261880165289256,
+      "grad_norm": 2.84375,
+      "learning_rate": 5.41048738839192e-05,
+      "loss": 3.8279,
+      "step": 378900
+    },
+    {
+      "epoch": 3.262310606060606,
+      "grad_norm": 3.421875,
+      "learning_rate": 5.408080533783484e-05,
+      "loss": 4.7128,
+      "step": 378950
+    },
+    {
+      "epoch": 3.2627410468319558,
+      "grad_norm": 2.5,
+      "learning_rate": 5.4056740162120564e-05,
+      "loss": 4.3488,
+      "step": 379000
+    },
+    {
+      "epoch": 3.263171487603306,
+      "grad_norm": 1.3359375,
+      "learning_rate": 5.403267835854273e-05,
+      "loss": 4.5935,
+      "step": 379050
+    },
+    {
+      "epoch": 3.263601928374656,
+      "grad_norm": 2.625,
+      "learning_rate": 5.400861992886734e-05,
+      "loss": 4.423,
+      "step": 379100
+    },
+    {
+      "epoch": 3.2640323691460056,
+      "grad_norm": 4.0625,
+      "learning_rate": 5.398456487486036e-05,
+      "loss": 4.3281,
+      "step": 379150
+    },
+    {
+      "epoch": 3.2644628099173554,
+      "grad_norm": 2.421875,
+      "learning_rate": 5.39605131982873e-05,
+      "loss": 4.1154,
+      "step": 379200
+    },
+    {
+      "epoch": 3.264893250688705,
+      "grad_norm": 3.84375,
+      "learning_rate": 5.393646490091351e-05,
+      "loss": 3.7891,
+      "step": 379250
+    },
+    {
+      "epoch": 3.265323691460055,
+      "grad_norm": 2.59375,
+      "learning_rate": 5.391241998450409e-05,
+      "loss": 4.3135,
+      "step": 379300
+    },
+    {
+      "epoch": 3.2657541322314048,
+      "grad_norm": 2.5,
+      "learning_rate": 5.3888378450823885e-05,
+      "loss": 3.9096,
+      "step": 379350
+    },
+    {
+      "epoch": 3.266184573002755,
+      "grad_norm": 2.75,
+      "learning_rate": 5.386434030163745e-05,
+      "loss": 4.1958,
+      "step": 379400
+    },
+    {
+      "epoch": 3.266615013774105,
+      "grad_norm": 2.15625,
+      "learning_rate": 5.3840305538709214e-05,
+      "loss": 4.2789,
+      "step": 379450
+    },
+    {
+      "epoch": 3.2670454545454546,
+      "grad_norm": 4.15625,
+      "learning_rate": 5.381627416380324e-05,
+      "loss": 3.9915,
+      "step": 379500
+    },
+    {
+      "epoch": 3.2674758953168044,
+      "grad_norm": 4.375,
+      "learning_rate": 5.379224617868337e-05,
+      "loss": 4.5594,
+      "step": 379550
+    },
+    {
+      "epoch": 3.267906336088154,
+      "grad_norm": 4.9375,
+      "learning_rate": 5.376822158511321e-05,
+      "loss": 4.4055,
+      "step": 379600
+    },
+    {
+      "epoch": 3.268336776859504,
+      "grad_norm": 2.328125,
+      "learning_rate": 5.374420038485613e-05,
+      "loss": 4.3554,
+      "step": 379650
+    },
+    {
+      "epoch": 3.268767217630854,
+      "grad_norm": 3.984375,
+      "learning_rate": 5.372018257967515e-05,
+      "loss": 4.3826,
+      "step": 379700
+    },
+    {
+      "epoch": 3.269197658402204,
+      "grad_norm": 3.3125,
+      "learning_rate": 5.369616817133326e-05,
+      "loss": 4.3408,
+      "step": 379750
+    },
+    {
+      "epoch": 3.269628099173554,
+      "grad_norm": 3.859375,
+      "learning_rate": 5.3672157161592995e-05,
+      "loss": 4.2632,
+      "step": 379800
+    },
+    {
+      "epoch": 3.2700585399449036,
+      "grad_norm": 3.734375,
+      "learning_rate": 5.3648149552216706e-05,
+      "loss": 4.363,
+      "step": 379850
+    },
+    {
+      "epoch": 3.2704889807162534,
+      "grad_norm": 5.875,
+      "learning_rate": 5.362414534496654e-05,
+      "loss": 4.2094,
+      "step": 379900
+    },
+    {
+      "epoch": 3.270919421487603,
+      "grad_norm": 2.671875,
+      "learning_rate": 5.3600144541604267e-05,
+      "loss": 4.5149,
+      "step": 379950
+    },
+    {
+      "epoch": 3.271349862258953,
+      "grad_norm": 3.34375,
+      "learning_rate": 5.3576147143891584e-05,
+      "loss": 4.1589,
+      "step": 380000
+    },
+    {
+      "epoch": 3.271780303030303,
+      "grad_norm": 5.0625,
+      "learning_rate": 5.3552153153589865e-05,
+      "loss": 4.3196,
+      "step": 380050
+    },
+    {
+      "epoch": 3.272210743801653,
+      "grad_norm": 5.34375,
+      "learning_rate": 5.352816257246009e-05,
+      "loss": 4.4324,
+      "step": 380100
+    },
+    {
+      "epoch": 3.272641184573003,
+      "grad_norm": 4.03125,
+      "learning_rate": 5.3504175402263245e-05,
+      "loss": 4.1397,
+      "step": 380150
+    },
+    {
+      "epoch": 3.2730716253443526,
+      "grad_norm": 3.125,
+      "learning_rate": 5.348019164475988e-05,
+      "loss": 4.3999,
+      "step": 380200
+    },
+    {
+      "epoch": 3.2735020661157024,
+      "grad_norm": 5.78125,
+      "learning_rate": 5.345621130171031e-05,
+      "loss": 4.2404,
+      "step": 380250
+    },
+    {
+      "epoch": 3.273932506887052,
+      "grad_norm": 2.40625,
+      "learning_rate": 5.343223437487478e-05,
+      "loss": 4.5147,
+      "step": 380300
+    },
+    {
+      "epoch": 3.2743629476584024,
+      "grad_norm": 3.265625,
+      "learning_rate": 5.340826086601296e-05,
+      "loss": 4.3614,
+      "step": 380350
+    },
+    {
+      "epoch": 3.274793388429752,
+      "grad_norm": 2.28125,
+      "learning_rate": 5.33842907768846e-05,
+      "loss": 4.0675,
+      "step": 380400
+    },
+    {
+      "epoch": 3.275223829201102,
+      "grad_norm": 1.3671875,
+      "learning_rate": 5.3360324109249005e-05,
+      "loss": 4.2549,
+      "step": 380450
+    },
+    {
+      "epoch": 3.275654269972452,
+      "grad_norm": 3.390625,
+      "learning_rate": 5.3336360864865265e-05,
+      "loss": 4.3411,
+      "step": 380500
+    },
+    {
+      "epoch": 3.2760847107438016,
+      "grad_norm": 3.53125,
+      "learning_rate": 5.3312401045492246e-05,
+      "loss": 4.1957,
+      "step": 380550
+    },
+    {
+      "epoch": 3.2765151515151514,
+      "grad_norm": 4.9375,
+      "learning_rate": 5.328844465288853e-05,
+      "loss": 4.2552,
+      "step": 380600
+    },
+    {
+      "epoch": 3.276945592286501,
+      "grad_norm": 3.609375,
+      "learning_rate": 5.3264491688812424e-05,
+      "loss": 4.2976,
+      "step": 380650
+    },
+    {
+      "epoch": 3.2773760330578514,
+      "grad_norm": 2.609375,
+      "learning_rate": 5.3240542155022135e-05,
+      "loss": 4.3418,
+      "step": 380700
+    },
+    {
+      "epoch": 3.277806473829201,
+      "grad_norm": 3.046875,
+      "learning_rate": 5.3216596053275435e-05,
+      "loss": 4.5812,
+      "step": 380750
+    },
+    {
+      "epoch": 3.278236914600551,
+      "grad_norm": 3.0625,
+      "learning_rate": 5.3192653385329926e-05,
+      "loss": 4.5653,
+      "step": 380800
+    },
+    {
+      "epoch": 3.278667355371901,
+      "grad_norm": 2.3125,
+      "learning_rate": 5.3168714152942955e-05,
+      "loss": 4.0254,
+      "step": 380850
+    },
+    {
+      "epoch": 3.2790977961432506,
+      "grad_norm": 5.0,
+      "learning_rate": 5.31447783578716e-05,
+      "loss": 4.572,
+      "step": 380900
+    },
+    {
+      "epoch": 3.2795282369146004,
+      "grad_norm": 2.25,
+      "learning_rate": 5.312084600187267e-05,
+      "loss": 4.2414,
+      "step": 380950
+    },
+    {
+      "epoch": 3.2799586776859506,
+      "grad_norm": 1.7421875,
+      "learning_rate": 5.309691708670281e-05,
+      "loss": 4.288,
+      "step": 381000
+    },
+    {
+      "epoch": 3.2799586776859506,
+      "eval_loss": 4.992687225341797,
+      "eval_runtime": 24.042,
+      "eval_samples_per_second": 26.62,
+      "eval_steps_per_second": 13.31,
+      "eval_tts_loss": 7.472846155654953,
+      "step": 381000
+    },
+    {
+      "epoch": 3.2803891184573004,
+      "grad_norm": 3.46875,
+      "learning_rate": 5.307299161411834e-05,
+      "loss": 4.1551,
+      "step": 381050
+    },
+    {
+      "epoch": 3.28081955922865,
+      "grad_norm": 3.09375,
+      "learning_rate": 5.30490695858753e-05,
+      "loss": 4.4248,
+      "step": 381100
+    },
+    {
+      "epoch": 3.28125,
+      "grad_norm": 3.8125,
+      "learning_rate": 5.3025151003729535e-05,
+      "loss": 4.1041,
+      "step": 381150
+    },
+    {
+      "epoch": 3.28168044077135,
+      "grad_norm": 5.5625,
+      "learning_rate": 5.3001235869436636e-05,
+      "loss": 4.5563,
+      "step": 381200
+    },
+    {
+      "epoch": 3.2821108815426996,
+      "grad_norm": 3.09375,
+      "learning_rate": 5.2977324184751856e-05,
+      "loss": 4.2727,
+      "step": 381250
+    },
+    {
+      "epoch": 3.2825413223140494,
+      "grad_norm": 5.15625,
+      "learning_rate": 5.2953415951430376e-05,
+      "loss": 4.2165,
+      "step": 381300
+    },
+    {
+      "epoch": 3.2829717630853996,
+      "grad_norm": 3.09375,
+      "learning_rate": 5.292951117122686e-05,
+      "loss": 4.2481,
+      "step": 381350
+    },
+    {
+      "epoch": 3.2834022038567494,
+      "grad_norm": 4.96875,
+      "learning_rate": 5.290560984589601e-05,
+      "loss": 4.1418,
+      "step": 381400
+    },
+    {
+      "epoch": 3.283832644628099,
+      "grad_norm": 2.078125,
+      "learning_rate": 5.2881711977192064e-05,
+      "loss": 4.1323,
+      "step": 381450
+    },
+    {
+      "epoch": 3.284263085399449,
+      "grad_norm": 5.3125,
+      "learning_rate": 5.285781756686903e-05,
+      "loss": 4.4193,
+      "step": 381500
+    },
+    {
+      "epoch": 3.284693526170799,
+      "grad_norm": 5.59375,
+      "learning_rate": 5.2833926616680806e-05,
+      "loss": 3.9395,
+      "step": 381550
+    },
+    {
+      "epoch": 3.2851239669421486,
+      "grad_norm": 3.171875,
+      "learning_rate": 5.281003912838094e-05,
+      "loss": 4.2469,
+      "step": 381600
+    },
+    {
+      "epoch": 3.285554407713499,
+      "grad_norm": 3.484375,
+      "learning_rate": 5.278615510372257e-05,
+      "loss": 3.9632,
+      "step": 381650
+    },
+    {
+      "epoch": 3.2859848484848486,
+      "grad_norm": 2.515625,
+      "learning_rate": 5.276227454445888e-05,
+      "loss": 4.0695,
+      "step": 381700
+    },
+    {
+      "epoch": 3.2864152892561984,
+      "grad_norm": 4.03125,
+      "learning_rate": 5.2738397452342606e-05,
+      "loss": 4.5583,
+      "step": 381750
+    },
+    {
+      "epoch": 3.286845730027548,
+      "grad_norm": 3.5,
+      "learning_rate": 5.2714523829126225e-05,
+      "loss": 4.2797,
+      "step": 381800
+    },
+    {
+      "epoch": 3.287276170798898,
+      "grad_norm": 2.203125,
+      "learning_rate": 5.2690653676562144e-05,
+      "loss": 4.1788,
+      "step": 381850
+    },
+    {
+      "epoch": 3.287706611570248,
+      "grad_norm": 5.34375,
+      "learning_rate": 5.2666786996402265e-05,
+      "loss": 3.9448,
+      "step": 381900
+    },
+    {
+      "epoch": 3.2881370523415976,
+      "grad_norm": 3.0625,
+      "learning_rate": 5.2642923790398325e-05,
+      "loss": 4.487,
+      "step": 381950
+    },
+    {
+      "epoch": 3.288567493112948,
+      "grad_norm": 2.171875,
+      "learning_rate": 5.261906406030192e-05,
+      "loss": 4.3355,
+      "step": 382000
+    },
+    {
+      "epoch": 3.2889979338842976,
+      "grad_norm": 4.59375,
+      "learning_rate": 5.259520780786429e-05,
+      "loss": 4.1598,
+      "step": 382050
+    },
+    {
+      "epoch": 3.2894283746556474,
+      "grad_norm": 2.90625,
+      "learning_rate": 5.257135503483641e-05,
+      "loss": 4.6427,
+      "step": 382100
+    },
+    {
+      "epoch": 3.289858815426997,
+      "grad_norm": 3.75,
+      "learning_rate": 5.254750574296902e-05,
+      "loss": 3.9327,
+      "step": 382150
+    },
+    {
+      "epoch": 3.290289256198347,
+      "grad_norm": 1.84375,
+      "learning_rate": 5.252365993401256e-05,
+      "loss": 4.1238,
+      "step": 382200
+    },
+    {
+      "epoch": 3.290719696969697,
+      "grad_norm": 3.921875,
+      "learning_rate": 5.249981760971737e-05,
+      "loss": 4.294,
+      "step": 382250
+    },
+    {
+      "epoch": 3.291150137741047,
+      "grad_norm": 2.578125,
+      "learning_rate": 5.2475978771833366e-05,
+      "loss": 4.2283,
+      "step": 382300
+    },
+    {
+      "epoch": 3.291580578512397,
+      "grad_norm": 4.9375,
+      "learning_rate": 5.2452143422110267e-05,
+      "loss": 4.3361,
+      "step": 382350
+    },
+    {
+      "epoch": 3.2920110192837466,
+      "grad_norm": 3.6875,
+      "learning_rate": 5.242831156229753e-05,
+      "loss": 4.4711,
+      "step": 382400
+    },
+    {
+      "epoch": 3.2924414600550964,
+      "grad_norm": 0.95703125,
+      "learning_rate": 5.240448319414438e-05,
+      "loss": 4.3467,
+      "step": 382450
+    },
+    {
+      "epoch": 3.292871900826446,
+      "grad_norm": 3.03125,
+      "learning_rate": 5.238065831939972e-05,
+      "loss": 4.3403,
+      "step": 382500
+    },
+    {
+      "epoch": 3.293302341597796,
+      "grad_norm": 5.15625,
+      "learning_rate": 5.235683693981233e-05,
+      "loss": 4.5563,
+      "step": 382550
+    },
+    {
+      "epoch": 3.293732782369146,
+      "grad_norm": 5.28125,
+      "learning_rate": 5.23330190571306e-05,
+      "loss": 4.2845,
+      "step": 382600
+    },
+    {
+      "epoch": 3.294163223140496,
+      "grad_norm": 2.46875,
+      "learning_rate": 5.2309204673102695e-05,
+      "loss": 4.3014,
+      "step": 382650
+    },
+    {
+      "epoch": 3.294593663911846,
+      "grad_norm": 3.625,
+      "learning_rate": 5.2285393789476564e-05,
+      "loss": 4.1626,
+      "step": 382700
+    },
+    {
+      "epoch": 3.2950241046831956,
+      "grad_norm": 3.953125,
+      "learning_rate": 5.2261586407999877e-05,
+      "loss": 4.5075,
+      "step": 382750
+    },
+    {
+      "epoch": 3.2954545454545454,
+      "grad_norm": 2.765625,
+      "learning_rate": 5.223778253042e-05,
+      "loss": 4.2139,
+      "step": 382800
+    },
+    {
+      "epoch": 3.295884986225895,
+      "grad_norm": 3.765625,
+      "learning_rate": 5.221398215848419e-05,
+      "loss": 4.4173,
+      "step": 382850
+    },
+    {
+      "epoch": 3.296315426997245,
+      "grad_norm": 3.328125,
+      "learning_rate": 5.2190185293939196e-05,
+      "loss": 4.2536,
+      "step": 382900
+    },
+    {
+      "epoch": 3.2967458677685952,
+      "grad_norm": 2.953125,
+      "learning_rate": 5.216639193853179e-05,
+      "loss": 4.2438,
+      "step": 382950
+    },
+    {
+      "epoch": 3.297176308539945,
+      "grad_norm": 4.0,
+      "learning_rate": 5.214260209400829e-05,
+      "loss": 3.8653,
+      "step": 383000
+    },
+    {
+      "epoch": 3.297606749311295,
+      "grad_norm": 3.640625,
+      "learning_rate": 5.211881576211479e-05,
+      "loss": 4.391,
+      "step": 383050
+    },
+    {
+      "epoch": 3.2980371900826446,
+      "grad_norm": 2.53125,
+      "learning_rate": 5.209503294459727e-05,
+      "loss": 4.132,
+      "step": 383100
+    },
+    {
+      "epoch": 3.2984676308539944,
+      "grad_norm": 1.9375,
+      "learning_rate": 5.2071253643201227e-05,
+      "loss": 4.1338,
+      "step": 383150
+    },
+    {
+      "epoch": 3.298898071625344,
+      "grad_norm": 4.90625,
+      "learning_rate": 5.204747785967199e-05,
+      "loss": 4.2765,
+      "step": 383200
+    },
+    {
+      "epoch": 3.299328512396694,
+      "grad_norm": 1.6015625,
+      "learning_rate": 5.202370559575476e-05,
+      "loss": 4.1567,
+      "step": 383250
+    },
+    {
+      "epoch": 3.2997589531680442,
+      "grad_norm": 3.328125,
+      "learning_rate": 5.199993685319431e-05,
+      "loss": 4.3439,
+      "step": 383300
+    },
+    {
+      "epoch": 3.300189393939394,
+      "grad_norm": 1.953125,
+      "learning_rate": 5.1976171633735184e-05,
+      "loss": 3.9614,
+      "step": 383350
+    },
+    {
+      "epoch": 3.300619834710744,
+      "grad_norm": 3.28125,
+      "learning_rate": 5.195240993912181e-05,
+      "loss": 4.1035,
+      "step": 383400
+    },
+    {
+      "epoch": 3.3010502754820936,
+      "grad_norm": 3.21875,
+      "learning_rate": 5.192865177109813e-05,
+      "loss": 4.4349,
+      "step": 383450
+    },
+    {
+      "epoch": 3.3014807162534434,
+      "grad_norm": 3.046875,
+      "learning_rate": 5.1904897131407956e-05,
+      "loss": 4.6754,
+      "step": 383500
+    },
+    {
+      "epoch": 3.301911157024793,
+      "grad_norm": 3.296875,
+      "learning_rate": 5.188114602179488e-05,
+      "loss": 4.3999,
+      "step": 383550
+    },
+    {
+      "epoch": 3.3023415977961434,
+      "grad_norm": 3.53125,
+      "learning_rate": 5.1857398444002145e-05,
+      "loss": 4.356,
+      "step": 383600
+    },
+    {
+      "epoch": 3.3027720385674932,
+      "grad_norm": 2.859375,
+      "learning_rate": 5.183365439977281e-05,
+      "loss": 4.6765,
+      "step": 383650
+    },
+    {
+      "epoch": 3.303202479338843,
+      "grad_norm": 3.84375,
+      "learning_rate": 5.180991389084959e-05,
+      "loss": 3.9962,
+      "step": 383700
+    },
+    {
+      "epoch": 3.303632920110193,
+      "grad_norm": 5.125,
+      "learning_rate": 5.178617691897497e-05,
+      "loss": 4.6447,
+      "step": 383750
+    },
+    {
+      "epoch": 3.3040633608815426,
+      "grad_norm": 4.5,
+      "learning_rate": 5.176244348589128e-05,
+      "loss": 4.0382,
+      "step": 383800
+    },
+    {
+      "epoch": 3.3044938016528924,
+      "grad_norm": 4.125,
+      "learning_rate": 5.1738713593340436e-05,
+      "loss": 4.0448,
+      "step": 383850
+    },
+    {
+      "epoch": 3.304924242424242,
+      "grad_norm": 1.90625,
+      "learning_rate": 5.171498724306418e-05,
+      "loss": 4.2292,
+      "step": 383900
+    },
+    {
+      "epoch": 3.3053546831955924,
+      "grad_norm": 3.828125,
+      "learning_rate": 5.169126443680399e-05,
+      "loss": 4.3154,
+      "step": 383950
+    },
+    {
+      "epoch": 3.3057851239669422,
+      "grad_norm": 2.375,
+      "learning_rate": 5.1667545176301034e-05,
+      "loss": 4.4976,
+      "step": 384000
+    },
+    {
+      "epoch": 3.3057851239669422,
+      "eval_loss": 4.993453025817871,
+      "eval_runtime": 23.9704,
+      "eval_samples_per_second": 26.7,
+      "eval_steps_per_second": 13.35,
+      "eval_tts_loss": 7.479142194899681,
+      "step": 384000
+    },
+    {
+      "epoch": 3.306215564738292,
+      "grad_norm": 5.03125,
+      "learning_rate": 5.164382946329623e-05,
+      "loss": 4.1415,
+      "step": 384050
+    },
+    {
+      "epoch": 3.306646005509642,
+      "grad_norm": 1.546875,
+      "learning_rate": 5.1620117299530356e-05,
+      "loss": 4.234,
+      "step": 384100
+    },
+    {
+      "epoch": 3.3070764462809916,
+      "grad_norm": 3.890625,
+      "learning_rate": 5.15964086867438e-05,
+      "loss": 4.1881,
+      "step": 384150
+    },
+    {
+      "epoch": 3.3075068870523414,
+      "grad_norm": 3.421875,
+      "learning_rate": 5.1572703626676645e-05,
+      "loss": 3.9042,
+      "step": 384200
+    },
+    {
+      "epoch": 3.3079373278236917,
+      "grad_norm": 5.65625,
+      "learning_rate": 5.154900212106888e-05,
+      "loss": 4.3213,
+      "step": 384250
+    },
+    {
+      "epoch": 3.3083677685950414,
+      "grad_norm": 3.640625,
+      "learning_rate": 5.1525304171660114e-05,
+      "loss": 4.4492,
+      "step": 384300
+    },
+    {
+      "epoch": 3.3087982093663912,
+      "grad_norm": 4.625,
+      "learning_rate": 5.1501609780189694e-05,
+      "loss": 4.3373,
+      "step": 384350
+    },
+    {
+      "epoch": 3.309228650137741,
+      "grad_norm": 4.28125,
+      "learning_rate": 5.1477918948396845e-05,
+      "loss": 4.5075,
+      "step": 384400
+    },
+    {
+      "epoch": 3.309659090909091,
+      "grad_norm": 5.375,
+      "learning_rate": 5.145423167802026e-05,
+      "loss": 4.3757,
+      "step": 384450
+    },
+    {
+      "epoch": 3.3100895316804406,
+      "grad_norm": 4.0,
+      "learning_rate": 5.143054797079867e-05,
+      "loss": 4.4652,
+      "step": 384500
+    },
+    {
+      "epoch": 3.3105199724517904,
+      "grad_norm": 2.15625,
+      "learning_rate": 5.140686782847035e-05,
+      "loss": 4.329,
+      "step": 384550
+    },
+    {
+      "epoch": 3.3109504132231407,
+      "grad_norm": 2.640625,
+      "learning_rate": 5.138319125277334e-05,
+      "loss": 4.4536,
+      "step": 384600
+    },
+    {
+      "epoch": 3.3113808539944904,
+      "grad_norm": 3.359375,
+      "learning_rate": 5.135951824544558e-05,
+      "loss": 4.4816,
+      "step": 384650
+    },
+    {
+      "epoch": 3.3118112947658402,
+      "grad_norm": 2.828125,
+      "learning_rate": 5.133584880822449e-05,
+      "loss": 4.3902,
+      "step": 384700
+    },
+    {
+      "epoch": 3.31224173553719,
+      "grad_norm": 3.765625,
+      "learning_rate": 5.131218294284734e-05,
+      "loss": 4.1089,
+      "step": 384750
+    },
+    {
+      "epoch": 3.31267217630854,
+      "grad_norm": 2.265625,
+      "learning_rate": 5.128852065105125e-05,
+      "loss": 4.2618,
+      "step": 384800
+    },
+    {
+      "epoch": 3.3131026170798896,
+      "grad_norm": 2.765625,
+      "learning_rate": 5.126486193457295e-05,
+      "loss": 4.2408,
+      "step": 384850
+    },
+    {
+      "epoch": 3.31353305785124,
+      "grad_norm": 3.84375,
+      "learning_rate": 5.124120679514892e-05,
+      "loss": 4.1616,
+      "step": 384900
+    },
+    {
+      "epoch": 3.3139634986225897,
+      "grad_norm": 1.96875,
+      "learning_rate": 5.12175552345154e-05,
+      "loss": 4.368,
+      "step": 384950
+    },
+    {
+      "epoch": 3.3143939393939394,
+      "grad_norm": 2.953125,
+      "learning_rate": 5.119390725440838e-05,
+      "loss": 4.2134,
+      "step": 385000
+    },
+    {
+      "epoch": 3.3148243801652892,
+      "grad_norm": 4.46875,
+      "learning_rate": 5.1170262856563514e-05,
+      "loss": 4.2689,
+      "step": 385050
+    },
+    {
+      "epoch": 3.315254820936639,
+      "grad_norm": 5.4375,
+      "learning_rate": 5.114662204271633e-05,
+      "loss": 4.1006,
+      "step": 385100
+    },
+    {
+      "epoch": 3.315685261707989,
+      "grad_norm": 3.25,
+      "learning_rate": 5.112298481460198e-05,
+      "loss": 4.443,
+      "step": 385150
+    },
+    {
+      "epoch": 3.3161157024793386,
+      "grad_norm": 2.609375,
+      "learning_rate": 5.109935117395538e-05,
+      "loss": 4.486,
+      "step": 385200
+    },
+    {
+      "epoch": 3.316546143250689,
+      "grad_norm": 4.40625,
+      "learning_rate": 5.10757211225112e-05,
+      "loss": 4.4206,
+      "step": 385250
+    },
+    {
+      "epoch": 3.3169765840220387,
+      "grad_norm": 2.578125,
+      "learning_rate": 5.105209466200378e-05,
+      "loss": 4.2105,
+      "step": 385300
+    },
+    {
+      "epoch": 3.3174070247933884,
+      "grad_norm": 3.390625,
+      "learning_rate": 5.102847179416733e-05,
+      "loss": 4.2119,
+      "step": 385350
+    },
+    {
+      "epoch": 3.3178374655647382,
+      "grad_norm": 2.828125,
+      "learning_rate": 5.10048525207357e-05,
+      "loss": 4.0055,
+      "step": 385400
+    },
+    {
+      "epoch": 3.318267906336088,
+      "grad_norm": 1.625,
+      "learning_rate": 5.098123684344247e-05,
+      "loss": 4.3641,
+      "step": 385450
+    },
+    {
+      "epoch": 3.318698347107438,
+      "grad_norm": 4.0,
+      "learning_rate": 5.0957624764020995e-05,
+      "loss": 4.2787,
+      "step": 385500
+    },
+    {
+      "epoch": 3.319128787878788,
+      "grad_norm": 2.796875,
+      "learning_rate": 5.0934016284204336e-05,
+      "loss": 4.2553,
+      "step": 385550
+    },
+    {
+      "epoch": 3.319559228650138,
+      "grad_norm": 2.65625,
+      "learning_rate": 5.0910411405725275e-05,
+      "loss": 4.1602,
+      "step": 385600
+    },
+    {
+      "epoch": 3.3199896694214877,
+      "grad_norm": 2.234375,
+      "learning_rate": 5.088681013031644e-05,
+      "loss": 4.0053,
+      "step": 385650
+    },
+    {
+      "epoch": 3.3204201101928374,
+      "grad_norm": 5.125,
+      "learning_rate": 5.086321245971011e-05,
+      "loss": 3.9055,
+      "step": 385700
+    },
+    {
+      "epoch": 3.3208505509641872,
+      "grad_norm": 3.84375,
+      "learning_rate": 5.08396183956382e-05,
+      "loss": 4.9393,
+      "step": 385750
+    },
+    {
+      "epoch": 3.321280991735537,
+      "grad_norm": 2.78125,
+      "learning_rate": 5.081602793983256e-05,
+      "loss": 4.3131,
+      "step": 385800
+    },
+    {
+      "epoch": 3.321711432506887,
+      "grad_norm": 2.125,
+      "learning_rate": 5.079244109402463e-05,
+      "loss": 4.8291,
+      "step": 385850
+    },
+    {
+      "epoch": 3.322141873278237,
+      "grad_norm": 3.109375,
+      "learning_rate": 5.076885785994563e-05,
+      "loss": 3.9214,
+      "step": 385900
+    },
+    {
+      "epoch": 3.322572314049587,
+      "grad_norm": 3.265625,
+      "learning_rate": 5.074527823932662e-05,
+      "loss": 4.1668,
+      "step": 385950
+    },
+    {
+      "epoch": 3.3230027548209367,
+      "grad_norm": 3.046875,
+      "learning_rate": 5.072170223389812e-05,
+      "loss": 4.3386,
+      "step": 386000
+    },
+    {
+      "epoch": 3.3234331955922864,
+      "grad_norm": 4.1875,
+      "learning_rate": 5.069812984539071e-05,
+      "loss": 4.2399,
+      "step": 386050
+    },
+    {
+      "epoch": 3.3238636363636362,
+      "grad_norm": 1.8515625,
+      "learning_rate": 5.067456107553448e-05,
+      "loss": 4.6372,
+      "step": 386100
+    },
+    {
+      "epoch": 3.324294077134986,
+      "grad_norm": 2.171875,
+      "learning_rate": 5.065099592605933e-05,
+      "loss": 3.8546,
+      "step": 386150
+    },
+    {
+      "epoch": 3.3247245179063363,
+      "grad_norm": 3.078125,
+      "learning_rate": 5.062743439869492e-05,
+      "loss": 4.4209,
+      "step": 386200
+    },
+    {
+      "epoch": 3.325154958677686,
+      "grad_norm": 4.28125,
+      "learning_rate": 5.060387649517059e-05,
+      "loss": 4.0753,
+      "step": 386250
+    },
+    {
+      "epoch": 3.325585399449036,
+      "grad_norm": 2.8125,
+      "learning_rate": 5.05803222172154e-05,
+      "loss": 4.0681,
+      "step": 386300
+    },
+    {
+      "epoch": 3.3260158402203857,
+      "grad_norm": 4.875,
+      "learning_rate": 5.055677156655829e-05,
+      "loss": 4.1523,
+      "step": 386350
+    },
+    {
+      "epoch": 3.3264462809917354,
+      "grad_norm": 2.859375,
+      "learning_rate": 5.053322454492775e-05,
+      "loss": 4.3313,
+      "step": 386400
+    },
+    {
+      "epoch": 3.3268767217630852,
+      "grad_norm": 3.125,
+      "learning_rate": 5.05096811540521e-05,
+      "loss": 4.4541,
+      "step": 386450
+    },
+    {
+      "epoch": 3.327307162534435,
+      "grad_norm": 4.5625,
+      "learning_rate": 5.048614139565938e-05,
+      "loss": 4.4754,
+      "step": 386500
+    },
+    {
+      "epoch": 3.3277376033057853,
+      "grad_norm": 4.375,
+      "learning_rate": 5.046260527147735e-05,
+      "loss": 4.4157,
+      "step": 386550
+    },
+    {
+      "epoch": 3.328168044077135,
+      "grad_norm": 3.625,
+      "learning_rate": 5.043907278323346e-05,
+      "loss": 4.3153,
+      "step": 386600
+    },
+    {
+      "epoch": 3.328598484848485,
+      "grad_norm": 3.046875,
+      "learning_rate": 5.041554393265504e-05,
+      "loss": 4.2392,
+      "step": 386650
+    },
+    {
+      "epoch": 3.3290289256198347,
+      "grad_norm": 2.828125,
+      "learning_rate": 5.039201872146902e-05,
+      "loss": 4.7573,
+      "step": 386700
+    },
+    {
+      "epoch": 3.3294593663911844,
+      "grad_norm": 2.3125,
+      "learning_rate": 5.036849715140208e-05,
+      "loss": 4.5726,
+      "step": 386750
+    },
+    {
+      "epoch": 3.3298898071625342,
+      "grad_norm": 6.53125,
+      "learning_rate": 5.034497922418067e-05,
+      "loss": 4.2354,
+      "step": 386800
+    },
+    {
+      "epoch": 3.3303202479338845,
+      "grad_norm": 4.125,
+      "learning_rate": 5.032146494153095e-05,
+      "loss": 4.4768,
+      "step": 386850
+    },
+    {
+      "epoch": 3.3307506887052343,
+      "grad_norm": 5.09375,
+      "learning_rate": 5.0297954305178774e-05,
+      "loss": 3.8892,
+      "step": 386900
+    },
+    {
+      "epoch": 3.331181129476584,
+      "grad_norm": 0.9921875,
+      "learning_rate": 5.027444731684985e-05,
+      "loss": 4.2422,
+      "step": 386950
+    },
+    {
+      "epoch": 3.331611570247934,
+      "grad_norm": 2.28125,
+      "learning_rate": 5.0250943978269516e-05,
+      "loss": 4.283,
+      "step": 387000
+    },
+    {
+      "epoch": 3.331611570247934,
+      "eval_loss": 4.994412422180176,
+      "eval_runtime": 23.7253,
+      "eval_samples_per_second": 26.975,
+      "eval_steps_per_second": 13.488,
+      "eval_tts_loss": 7.491261116750935,
+      "step": 387000
+    },
+    {
+      "epoch": 3.3320420110192837,
+      "grad_norm": 2.0625,
+      "learning_rate": 5.022744429116285e-05,
+      "loss": 4.4999,
+      "step": 387050
+    },
+    {
+      "epoch": 3.3324724517906334,
+      "grad_norm": 3.921875,
+      "learning_rate": 5.0203948257254694e-05,
+      "loss": 4.2192,
+      "step": 387100
+    },
+    {
+      "epoch": 3.3329028925619832,
+      "grad_norm": 2.390625,
+      "learning_rate": 5.018045587826954e-05,
+      "loss": 4.296,
+      "step": 387150
+    },
+    {
+      "epoch": 3.3333333333333335,
+      "grad_norm": 3.625,
+      "learning_rate": 5.015696715593179e-05,
+      "loss": 4.2384,
+      "step": 387200
+    },
+    {
+      "epoch": 3.3337637741046833,
+      "grad_norm": 5.0625,
+      "learning_rate": 5.0133482091965445e-05,
+      "loss": 4.1351,
+      "step": 387250
+    },
+    {
+      "epoch": 3.334194214876033,
+      "grad_norm": 4.21875,
+      "learning_rate": 5.0110000688094146e-05,
+      "loss": 4.5279,
+      "step": 387300
+    },
+    {
+      "epoch": 3.334624655647383,
+      "grad_norm": 3.9375,
+      "learning_rate": 5.008652294604149e-05,
+      "loss": 3.993,
+      "step": 387350
+    },
+    {
+      "epoch": 3.3350550964187327,
+      "grad_norm": 3.046875,
+      "learning_rate": 5.006304886753066e-05,
+      "loss": 4.1831,
+      "step": 387400
+    },
+    {
+      "epoch": 3.335485537190083,
+      "grad_norm": 4.96875,
+      "learning_rate": 5.003957845428457e-05,
+      "loss": 4.36,
+      "step": 387450
+    },
+    {
+      "epoch": 3.3359159779614327,
+      "grad_norm": 2.96875,
+      "learning_rate": 5.001611170802599e-05,
+      "loss": 4.2505,
+      "step": 387500
+    },
+    {
+      "epoch": 3.3363464187327825,
+      "grad_norm": 2.453125,
+      "learning_rate": 4.99926486304772e-05,
+      "loss": 4.2206,
+      "step": 387550
+    },
+    {
+      "epoch": 3.3367768595041323,
+      "grad_norm": 2.734375,
+      "learning_rate": 4.9969189223360446e-05,
+      "loss": 4.5257,
+      "step": 387600
+    },
+    {
+      "epoch": 3.337207300275482,
+      "grad_norm": 2.734375,
+      "learning_rate": 4.994573348839756e-05,
+      "loss": 4.6288,
+      "step": 387650
+    },
+    {
+      "epoch": 3.337637741046832,
+      "grad_norm": 3.59375,
+      "learning_rate": 4.992228142731015e-05,
+      "loss": 4.3768,
+      "step": 387700
+    },
+    {
+      "epoch": 3.3380681818181817,
+      "grad_norm": 2.28125,
+      "learning_rate": 4.9898833041819515e-05,
+      "loss": 4.4389,
+      "step": 387750
+    },
+    {
+      "epoch": 3.3384986225895315,
+      "grad_norm": 1.5703125,
+      "learning_rate": 4.987538833364675e-05,
+      "loss": 4.0628,
+      "step": 387800
+    },
+    {
+      "epoch": 3.3389290633608817,
+      "grad_norm": 3.796875,
+      "learning_rate": 4.9851947304512605e-05,
+      "loss": 4.2834,
+      "step": 387850
+    },
+    {
+      "epoch": 3.3393595041322315,
+      "grad_norm": 3.59375,
+      "learning_rate": 4.982850995613767e-05,
+      "loss": 3.8162,
+      "step": 387900
+    },
+    {
+      "epoch": 3.3397899449035813,
+      "grad_norm": 3.0,
+      "learning_rate": 4.9805076290242144e-05,
+      "loss": 3.9103,
+      "step": 387950
+    },
+    {
+      "epoch": 3.340220385674931,
+      "grad_norm": 3.34375,
+      "learning_rate": 4.978164630854604e-05,
+      "loss": 4.3271,
+      "step": 388000
+    },
+    {
+      "epoch": 3.340650826446281,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.975822001276903e-05,
+      "loss": 4.4111,
+      "step": 388050
+    },
+    {
+      "epoch": 3.341081267217631,
+      "grad_norm": 5.0625,
+      "learning_rate": 4.973479740463058e-05,
+      "loss": 4.4683,
+      "step": 388100
+    },
+    {
+      "epoch": 3.341511707988981,
+      "grad_norm": 3.828125,
+      "learning_rate": 4.971137848584981e-05,
+      "loss": 4.1036,
+      "step": 388150
+    },
+    {
+      "epoch": 3.3419421487603307,
+      "grad_norm": 2.578125,
+      "learning_rate": 4.9687963258145706e-05,
+      "loss": 4.5362,
+      "step": 388200
+    },
+    {
+      "epoch": 3.3423725895316805,
+      "grad_norm": 3.234375,
+      "learning_rate": 4.966455172323685e-05,
+      "loss": 4.0423,
+      "step": 388250
+    },
+    {
+      "epoch": 3.3428030303030303,
+      "grad_norm": 5.15625,
+      "learning_rate": 4.96411438828416e-05,
+      "loss": 4.1517,
+      "step": 388300
+    },
+    {
+      "epoch": 3.34323347107438,
+      "grad_norm": 5.125,
+      "learning_rate": 4.961773973867804e-05,
+      "loss": 4.1902,
+      "step": 388350
+    },
+    {
+      "epoch": 3.34366391184573,
+      "grad_norm": 4.125,
+      "learning_rate": 4.959433929246398e-05,
+      "loss": 4.4626,
+      "step": 388400
+    },
+    {
+      "epoch": 3.3440943526170797,
+      "grad_norm": 3.53125,
+      "learning_rate": 4.957094254591692e-05,
+      "loss": 4.2152,
+      "step": 388450
+    },
+    {
+      "epoch": 3.34452479338843,
+      "grad_norm": 3.296875,
+      "learning_rate": 4.954754950075425e-05,
+      "loss": 4.1573,
+      "step": 388500
+    },
+    {
+      "epoch": 3.3449552341597797,
+      "grad_norm": 1.8984375,
+      "learning_rate": 4.9524160158692835e-05,
+      "loss": 4.2415,
+      "step": 388550
+    },
+    {
+      "epoch": 3.3453856749311295,
+      "grad_norm": 2.421875,
+      "learning_rate": 4.950077452144949e-05,
+      "loss": 4.6862,
+      "step": 388600
+    },
+    {
+      "epoch": 3.3458161157024793,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.9477392590740645e-05,
+      "loss": 4.3831,
+      "step": 388650
+    },
+    {
+      "epoch": 3.346246556473829,
+      "grad_norm": 5.90625,
+      "learning_rate": 4.945401436828244e-05,
+      "loss": 4.5584,
+      "step": 388700
+    },
+    {
+      "epoch": 3.3466769972451793,
+      "grad_norm": 2.34375,
+      "learning_rate": 4.9430639855790876e-05,
+      "loss": 4.2195,
+      "step": 388750
+    },
+    {
+      "epoch": 3.347107438016529,
+      "grad_norm": 1.890625,
+      "learning_rate": 4.940726905498158e-05,
+      "loss": 4.4348,
+      "step": 388800
+    },
+    {
+      "epoch": 3.347537878787879,
+      "grad_norm": 4.03125,
+      "learning_rate": 4.9383901967569804e-05,
+      "loss": 4.0903,
+      "step": 388850
+    },
+    {
+      "epoch": 3.3479683195592287,
+      "grad_norm": 3.703125,
+      "learning_rate": 4.936053859527075e-05,
+      "loss": 4.1093,
+      "step": 388900
+    },
+    {
+      "epoch": 3.3483987603305785,
+      "grad_norm": 3.625,
+      "learning_rate": 4.933717893979921e-05,
+      "loss": 4.081,
+      "step": 388950
+    },
+    {
+      "epoch": 3.3488292011019283,
+      "grad_norm": 0.90234375,
+      "learning_rate": 4.9313823002869705e-05,
+      "loss": 4.1994,
+      "step": 389000
+    },
+    {
+      "epoch": 3.349259641873278,
+      "grad_norm": 2.75,
+      "learning_rate": 4.9290470786196605e-05,
+      "loss": 3.9767,
+      "step": 389050
+    },
+    {
+      "epoch": 3.349690082644628,
+      "grad_norm": 6.09375,
+      "learning_rate": 4.9267122291493814e-05,
+      "loss": 4.4585,
+      "step": 389100
+    },
+    {
+      "epoch": 3.350120523415978,
+      "grad_norm": 2.34375,
+      "learning_rate": 4.9243777520475056e-05,
+      "loss": 4.5203,
+      "step": 389150
+    },
+    {
+      "epoch": 3.350550964187328,
+      "grad_norm": 2.578125,
+      "learning_rate": 4.9220436474853845e-05,
+      "loss": 4.4309,
+      "step": 389200
+    },
+    {
+      "epoch": 3.3509814049586777,
+      "grad_norm": 2.265625,
+      "learning_rate": 4.919709915634335e-05,
+      "loss": 4.4047,
+      "step": 389250
+    },
+    {
+      "epoch": 3.3514118457300275,
+      "grad_norm": 4.40625,
+      "learning_rate": 4.917376556665648e-05,
+      "loss": 4.4213,
+      "step": 389300
+    },
+    {
+      "epoch": 3.3518422865013773,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.915043570750586e-05,
+      "loss": 4.8675,
+      "step": 389350
+    },
+    {
+      "epoch": 3.3522727272727275,
+      "grad_norm": 3.75,
+      "learning_rate": 4.91271095806038e-05,
+      "loss": 4.5339,
+      "step": 389400
+    },
+    {
+      "epoch": 3.3527031680440773,
+      "grad_norm": 4.15625,
+      "learning_rate": 4.91037871876625e-05,
+      "loss": 3.929,
+      "step": 389450
+    },
+    {
+      "epoch": 3.353133608815427,
+      "grad_norm": 4.1875,
+      "learning_rate": 4.9080468530393706e-05,
+      "loss": 4.4526,
+      "step": 389500
+    },
+    {
+      "epoch": 3.353564049586777,
+      "grad_norm": 2.515625,
+      "learning_rate": 4.9057153610508965e-05,
+      "loss": 4.3747,
+      "step": 389550
+    },
+    {
+      "epoch": 3.3539944903581267,
+      "grad_norm": 2.25,
+      "learning_rate": 4.903384242971953e-05,
+      "loss": 4.083,
+      "step": 389600
+    },
+    {
+      "epoch": 3.3544249311294765,
+      "grad_norm": 2.546875,
+      "learning_rate": 4.9010534989736424e-05,
+      "loss": 4.0807,
+      "step": 389650
+    },
+    {
+      "epoch": 3.3548553719008263,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.89872312922703e-05,
+      "loss": 4.4303,
+      "step": 389700
+    },
+    {
+      "epoch": 3.355285812672176,
+      "grad_norm": 2.390625,
+      "learning_rate": 4.896393133903168e-05,
+      "loss": 4.4677,
+      "step": 389750
+    },
+    {
+      "epoch": 3.3557162534435263,
+      "grad_norm": 4.40625,
+      "learning_rate": 4.8940635131730685e-05,
+      "loss": 4.2841,
+      "step": 389800
+    },
+    {
+      "epoch": 3.356146694214876,
+      "grad_norm": 3.921875,
+      "learning_rate": 4.891734267207723e-05,
+      "loss": 4.0092,
+      "step": 389850
+    },
+    {
+      "epoch": 3.356577134986226,
+      "grad_norm": 5.40625,
+      "learning_rate": 4.8894053961780895e-05,
+      "loss": 4.345,
+      "step": 389900
+    },
+    {
+      "epoch": 3.3570075757575757,
+      "grad_norm": 2.0,
+      "learning_rate": 4.8870769002551055e-05,
+      "loss": 4.0638,
+      "step": 389950
+    },
+    {
+      "epoch": 3.3574380165289255,
+      "grad_norm": 2.40625,
+      "learning_rate": 4.884748779609672e-05,
+      "loss": 4.4728,
+      "step": 390000
+    },
+    {
+      "epoch": 3.3574380165289255,
+      "eval_loss": 4.994321346282959,
+      "eval_runtime": 24.6569,
+      "eval_samples_per_second": 25.956,
+      "eval_steps_per_second": 12.978,
+      "eval_tts_loss": 7.48053660015481,
+      "step": 390000
+    },
+    {
+      "epoch": 3.3578684573002757,
+      "grad_norm": 2.46875,
+      "learning_rate": 4.88242103441268e-05,
+      "loss": 4.0664,
+      "step": 390050
+    },
+    {
+      "epoch": 3.3582988980716255,
+      "grad_norm": 3.9375,
+      "learning_rate": 4.880093664834966e-05,
+      "loss": 3.9159,
+      "step": 390100
+    },
+    {
+      "epoch": 3.3587293388429753,
+      "grad_norm": 3.625,
+      "learning_rate": 4.8777666710473644e-05,
+      "loss": 4.2394,
+      "step": 390150
+    },
+    {
+      "epoch": 3.359159779614325,
+      "grad_norm": 2.296875,
+      "learning_rate": 4.875440053220669e-05,
+      "loss": 3.9599,
+      "step": 390200
+    },
+    {
+      "epoch": 3.359590220385675,
+      "grad_norm": 3.4375,
+      "learning_rate": 4.8731138115256446e-05,
+      "loss": 4.5411,
+      "step": 390250
+    },
+    {
+      "epoch": 3.3600206611570247,
+      "grad_norm": 6.78125,
+      "learning_rate": 4.870787946133043e-05,
+      "loss": 4.3078,
+      "step": 390300
+    },
+    {
+      "epoch": 3.3604511019283745,
+      "grad_norm": 3.59375,
+      "learning_rate": 4.868462457213566e-05,
+      "loss": 4.6105,
+      "step": 390350
+    },
+    {
+      "epoch": 3.3608815426997243,
+      "grad_norm": 3.625,
+      "learning_rate": 4.8661373449379e-05,
+      "loss": 4.2969,
+      "step": 390400
+    },
+    {
+      "epoch": 3.3613119834710745,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.863812609476711e-05,
+      "loss": 4.2849,
+      "step": 390450
+    },
+    {
+      "epoch": 3.3617424242424243,
+      "grad_norm": 5.3125,
+      "learning_rate": 4.861488251000627e-05,
+      "loss": 4.3593,
+      "step": 390500
+    },
+    {
+      "epoch": 3.362172865013774,
+      "grad_norm": 3.375,
+      "learning_rate": 4.8591642696802444e-05,
+      "loss": 4.1489,
+      "step": 390550
+    },
+    {
+      "epoch": 3.362603305785124,
+      "grad_norm": 1.3984375,
+      "learning_rate": 4.856840665686151e-05,
+      "loss": 4.3814,
+      "step": 390600
+    },
+    {
+      "epoch": 3.3630337465564737,
+      "grad_norm": 3.390625,
+      "learning_rate": 4.854517439188884e-05,
+      "loss": 4.3655,
+      "step": 390650
+    },
+    {
+      "epoch": 3.363464187327824,
+      "grad_norm": 1.9609375,
+      "learning_rate": 4.852194590358962e-05,
+      "loss": 4.0472,
+      "step": 390700
+    },
+    {
+      "epoch": 3.3638946280991737,
+      "grad_norm": 3.03125,
+      "learning_rate": 4.849872119366884e-05,
+      "loss": 4.3848,
+      "step": 390750
+    },
+    {
+      "epoch": 3.3643250688705235,
+      "grad_norm": 3.609375,
+      "learning_rate": 4.847550026383113e-05,
+      "loss": 4.5576,
+      "step": 390800
+    },
+    {
+      "epoch": 3.3647555096418733,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.8452283115780836e-05,
+      "loss": 4.5376,
+      "step": 390850
+    },
+    {
+      "epoch": 3.365185950413223,
+      "grad_norm": 3.546875,
+      "learning_rate": 4.842906975122206e-05,
+      "loss": 4.2247,
+      "step": 390900
+    },
+    {
+      "epoch": 3.365616391184573,
+      "grad_norm": 4.0625,
+      "learning_rate": 4.840586017185856e-05,
+      "loss": 4.1984,
+      "step": 390950
+    },
+    {
+      "epoch": 3.3660468319559227,
+      "grad_norm": 4.09375,
+      "learning_rate": 4.8382654379393966e-05,
+      "loss": 4.3393,
+      "step": 391000
+    },
+    {
+      "epoch": 3.3664772727272725,
+      "grad_norm": 3.75,
+      "learning_rate": 4.835945237553148e-05,
+      "loss": 4.4571,
+      "step": 391050
+    },
+    {
+      "epoch": 3.3669077134986227,
+      "grad_norm": 2.0625,
+      "learning_rate": 4.833625416197407e-05,
+      "loss": 4.0708,
+      "step": 391100
+    },
+    {
+      "epoch": 3.3673381542699725,
+      "grad_norm": 3.171875,
+      "learning_rate": 4.831305974042446e-05,
+      "loss": 4.4227,
+      "step": 391150
+    },
+    {
+      "epoch": 3.3677685950413223,
+      "grad_norm": 4.40625,
+      "learning_rate": 4.828986911258505e-05,
+      "loss": 4.5377,
+      "step": 391200
+    },
+    {
+      "epoch": 3.368199035812672,
+      "grad_norm": 3.6875,
+      "learning_rate": 4.826668228015796e-05,
+      "loss": 4.6127,
+      "step": 391250
+    },
+    {
+      "epoch": 3.368629476584022,
+      "grad_norm": 1.859375,
+      "learning_rate": 4.8243499244845116e-05,
+      "loss": 4.1092,
+      "step": 391300
+    },
+    {
+      "epoch": 3.369059917355372,
+      "grad_norm": 4.84375,
+      "learning_rate": 4.822032000834811e-05,
+      "loss": 4.3181,
+      "step": 391350
+    },
+    {
+      "epoch": 3.369490358126722,
+      "grad_norm": 3.375,
+      "learning_rate": 4.8197144572368135e-05,
+      "loss": 4.045,
+      "step": 391400
+    },
+    {
+      "epoch": 3.3699207988980717,
+      "grad_norm": 3.28125,
+      "learning_rate": 4.817397293860633e-05,
+      "loss": 4.3422,
+      "step": 391450
+    },
+    {
+      "epoch": 3.3703512396694215,
+      "grad_norm": 2.734375,
+      "learning_rate": 4.8150805108763395e-05,
+      "loss": 4.2938,
+      "step": 391500
+    },
+    {
+      "epoch": 3.3707816804407713,
+      "grad_norm": 2.875,
+      "learning_rate": 4.812764108453978e-05,
+      "loss": 4.3602,
+      "step": 391550
+    },
+    {
+      "epoch": 3.371212121212121,
+      "grad_norm": 2.578125,
+      "learning_rate": 4.810448086763577e-05,
+      "loss": 3.9826,
+      "step": 391600
+    },
+    {
+      "epoch": 3.371642561983471,
+      "grad_norm": 2.15625,
+      "learning_rate": 4.808132445975114e-05,
+      "loss": 4.3741,
+      "step": 391650
+    },
+    {
+      "epoch": 3.3720730027548207,
+      "grad_norm": 3.875,
+      "learning_rate": 4.805817186258562e-05,
+      "loss": 4.2754,
+      "step": 391700
+    },
+    {
+      "epoch": 3.372503443526171,
+      "grad_norm": 3.828125,
+      "learning_rate": 4.803502307783854e-05,
+      "loss": 4.1645,
+      "step": 391750
+    },
+    {
+      "epoch": 3.3729338842975207,
+      "grad_norm": 1.15625,
+      "learning_rate": 4.801187810720892e-05,
+      "loss": 4.5861,
+      "step": 391800
+    },
+    {
+      "epoch": 3.3733643250688705,
+      "grad_norm": 3.46875,
+      "learning_rate": 4.798873695239566e-05,
+      "loss": 4.2072,
+      "step": 391850
+    },
+    {
+      "epoch": 3.3737947658402203,
+      "grad_norm": 1.8046875,
+      "learning_rate": 4.796559961509717e-05,
+      "loss": 4.4066,
+      "step": 391900
+    },
+    {
+      "epoch": 3.37422520661157,
+      "grad_norm": 4.09375,
+      "learning_rate": 4.794246609701167e-05,
+      "loss": 4.4111,
+      "step": 391950
+    },
+    {
+      "epoch": 3.3746556473829203,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.79193363998372e-05,
+      "loss": 4.467,
+      "step": 392000
+    },
+    {
+      "epoch": 3.37508608815427,
+      "grad_norm": 2.359375,
+      "learning_rate": 4.789621052527138e-05,
+      "loss": 3.8915,
+      "step": 392050
+    },
+    {
+      "epoch": 3.37551652892562,
+      "grad_norm": 3.125,
+      "learning_rate": 4.7873088475011606e-05,
+      "loss": 4.3434,
+      "step": 392100
+    },
+    {
+      "epoch": 3.3759469696969697,
+      "grad_norm": 2.75,
+      "learning_rate": 4.7849970250754985e-05,
+      "loss": 4.0741,
+      "step": 392150
+    },
+    {
+      "epoch": 3.3763774104683195,
+      "grad_norm": 4.1875,
+      "learning_rate": 4.7826855854198336e-05,
+      "loss": 4.1819,
+      "step": 392200
+    },
+    {
+      "epoch": 3.3768078512396693,
+      "grad_norm": 3.0625,
+      "learning_rate": 4.780374528703818e-05,
+      "loss": 4.1821,
+      "step": 392250
+    },
+    {
+      "epoch": 3.377238292011019,
+      "grad_norm": 1.953125,
+      "learning_rate": 4.7780638550970855e-05,
+      "loss": 4.1299,
+      "step": 392300
+    },
+    {
+      "epoch": 3.377668732782369,
+      "grad_norm": 3.734375,
+      "learning_rate": 4.775753564769231e-05,
+      "loss": 4.2985,
+      "step": 392350
+    },
+    {
+      "epoch": 3.378099173553719,
+      "grad_norm": 2.796875,
+      "learning_rate": 4.773443657889824e-05,
+      "loss": 3.9785,
+      "step": 392400
+    },
+    {
+      "epoch": 3.378529614325069,
+      "grad_norm": 2.640625,
+      "learning_rate": 4.7711341346284074e-05,
+      "loss": 4.6907,
+      "step": 392450
+    },
+    {
+      "epoch": 3.3789600550964187,
+      "grad_norm": 3.0625,
+      "learning_rate": 4.7688249951544916e-05,
+      "loss": 4.2184,
+      "step": 392500
+    },
+    {
+      "epoch": 3.3793904958677685,
+      "grad_norm": 1.0859375,
+      "learning_rate": 4.7665162396375697e-05,
+      "loss": 4.2271,
+      "step": 392550
+    },
+    {
+      "epoch": 3.3798209366391183,
+      "grad_norm": 3.3125,
+      "learning_rate": 4.764207868247096e-05,
+      "loss": 4.3007,
+      "step": 392600
+    },
+    {
+      "epoch": 3.3802513774104685,
+      "grad_norm": 2.609375,
+      "learning_rate": 4.7618998811524995e-05,
+      "loss": 4.5501,
+      "step": 392650
+    },
+    {
+      "epoch": 3.3806818181818183,
+      "grad_norm": 4.15625,
+      "learning_rate": 4.7595922785231814e-05,
+      "loss": 4.5851,
+      "step": 392700
+    },
+    {
+      "epoch": 3.381112258953168,
+      "grad_norm": 3.34375,
+      "learning_rate": 4.7572850605285146e-05,
+      "loss": 4.3675,
+      "step": 392750
+    },
+    {
+      "epoch": 3.381542699724518,
+      "grad_norm": 4.90625,
+      "learning_rate": 4.754978227337842e-05,
+      "loss": 4.4713,
+      "step": 392800
+    },
+    {
+      "epoch": 3.3819731404958677,
+      "grad_norm": 2.484375,
+      "learning_rate": 4.752671779120486e-05,
+      "loss": 4.3006,
+      "step": 392850
+    },
+    {
+      "epoch": 3.3824035812672175,
+      "grad_norm": 4.875,
+      "learning_rate": 4.750365716045734e-05,
+      "loss": 4.1813,
+      "step": 392900
+    },
+    {
+      "epoch": 3.3828340220385673,
+      "grad_norm": 1.2421875,
+      "learning_rate": 4.7480600382828366e-05,
+      "loss": 4.2718,
+      "step": 392950
+    },
+    {
+      "epoch": 3.383264462809917,
+      "grad_norm": 5.15625,
+      "learning_rate": 4.7457547460010366e-05,
+      "loss": 4.1846,
+      "step": 393000
+    },
+    {
+      "epoch": 3.383264462809917,
+      "eval_loss": 4.993402004241943,
+      "eval_runtime": 23.4826,
+      "eval_samples_per_second": 27.254,
+      "eval_steps_per_second": 13.627,
+      "eval_tts_loss": 7.468892651127944,
+      "step": 393000
+    },
+    {
+      "epoch": 3.3836949035812673,
+      "grad_norm": 2.359375,
+      "learning_rate": 4.743449839369534e-05,
+      "loss": 4.4091,
+      "step": 393050
+    },
+    {
+      "epoch": 3.384125344352617,
+      "grad_norm": 5.0625,
+      "learning_rate": 4.741145318557498e-05,
+      "loss": 3.9521,
+      "step": 393100
+    },
+    {
+      "epoch": 3.384555785123967,
+      "grad_norm": 3.625,
+      "learning_rate": 4.73884118373409e-05,
+      "loss": 4.4356,
+      "step": 393150
+    },
+    {
+      "epoch": 3.3849862258953167,
+      "grad_norm": 3.328125,
+      "learning_rate": 4.736537435068411e-05,
+      "loss": 4.3135,
+      "step": 393200
+    },
+    {
+      "epoch": 3.3854166666666665,
+      "grad_norm": 3.296875,
+      "learning_rate": 4.734234072729564e-05,
+      "loss": 4.3765,
+      "step": 393250
+    },
+    {
+      "epoch": 3.3858471074380168,
+      "grad_norm": 4.1875,
+      "learning_rate": 4.731931096886606e-05,
+      "loss": 3.9165,
+      "step": 393300
+    },
+    {
+      "epoch": 3.3862775482093666,
+      "grad_norm": 3.796875,
+      "learning_rate": 4.729628507708571e-05,
+      "loss": 4.1185,
+      "step": 393350
+    },
+    {
+      "epoch": 3.3867079889807163,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.727326305364464e-05,
+      "loss": 4.5291,
+      "step": 393400
+    },
+    {
+      "epoch": 3.387138429752066,
+      "grad_norm": 5.3125,
+      "learning_rate": 4.7250244900232624e-05,
+      "loss": 4.4192,
+      "step": 393450
+    },
+    {
+      "epoch": 3.387568870523416,
+      "grad_norm": 1.296875,
+      "learning_rate": 4.72272306185391e-05,
+      "loss": 4.4869,
+      "step": 393500
+    },
+    {
+      "epoch": 3.3879993112947657,
+      "grad_norm": 2.234375,
+      "learning_rate": 4.7204220210253336e-05,
+      "loss": 4.494,
+      "step": 393550
+    },
+    {
+      "epoch": 3.3884297520661155,
+      "grad_norm": 5.25,
+      "learning_rate": 4.7181213677064216e-05,
+      "loss": 4.2612,
+      "step": 393600
+    },
+    {
+      "epoch": 3.3888601928374658,
+      "grad_norm": 3.46875,
+      "learning_rate": 4.715821102066037e-05,
+      "loss": 4.1171,
+      "step": 393650
+    },
+    {
+      "epoch": 3.3892906336088156,
+      "grad_norm": 1.8046875,
+      "learning_rate": 4.7135212242730154e-05,
+      "loss": 4.2632,
+      "step": 393700
+    },
+    {
+      "epoch": 3.3897210743801653,
+      "grad_norm": 3.75,
+      "learning_rate": 4.711221734496161e-05,
+      "loss": 3.9147,
+      "step": 393750
+    },
+    {
+      "epoch": 3.390151515151515,
+      "grad_norm": 0.92578125,
+      "learning_rate": 4.708922632904248e-05,
+      "loss": 4.3943,
+      "step": 393800
+    },
+    {
+      "epoch": 3.390581955922865,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.7066239196660335e-05,
+      "loss": 4.584,
+      "step": 393850
+    },
+    {
+      "epoch": 3.3910123966942147,
+      "grad_norm": 3.75,
+      "learning_rate": 4.704325594950234e-05,
+      "loss": 4.3678,
+      "step": 393900
+    },
+    {
+      "epoch": 3.391442837465565,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.702027658925543e-05,
+      "loss": 4.0652,
+      "step": 393950
+    },
+    {
+      "epoch": 3.3918732782369148,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.699730111760623e-05,
+      "loss": 4.1975,
+      "step": 394000
+    },
+    {
+      "epoch": 3.3923037190082646,
+      "grad_norm": 2.078125,
+      "learning_rate": 4.6974329536241036e-05,
+      "loss": 4.2021,
+      "step": 394050
+    },
+    {
+      "epoch": 3.3927341597796143,
+      "grad_norm": 1.9140625,
+      "learning_rate": 4.695136184684602e-05,
+      "loss": 4.38,
+      "step": 394100
+    },
+    {
+      "epoch": 3.393164600550964,
+      "grad_norm": 2.546875,
+      "learning_rate": 4.6928398051106936e-05,
+      "loss": 4.2613,
+      "step": 394150
+    },
+    {
+      "epoch": 3.393595041322314,
+      "grad_norm": 4.9375,
+      "learning_rate": 4.690543815070918e-05,
+      "loss": 4.5153,
+      "step": 394200
+    },
+    {
+      "epoch": 3.3940254820936637,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.6882482147338055e-05,
+      "loss": 4.5173,
+      "step": 394250
+    },
+    {
+      "epoch": 3.394455922865014,
+      "grad_norm": 4.65625,
+      "learning_rate": 4.685953004267847e-05,
+      "loss": 4.566,
+      "step": 394300
+    },
+    {
+      "epoch": 3.3948863636363638,
+      "grad_norm": 2.96875,
+      "learning_rate": 4.6836581838414996e-05,
+      "loss": 4.1144,
+      "step": 394350
+    },
+    {
+      "epoch": 3.3953168044077136,
+      "grad_norm": 4.40625,
+      "learning_rate": 4.6813637536232066e-05,
+      "loss": 4.4777,
+      "step": 394400
+    },
+    {
+      "epoch": 3.3957472451790633,
+      "grad_norm": 2.328125,
+      "learning_rate": 4.6790697137813756e-05,
+      "loss": 4.7136,
+      "step": 394450
+    },
+    {
+      "epoch": 3.396177685950413,
+      "grad_norm": 2.359375,
+      "learning_rate": 4.6767760644843716e-05,
+      "loss": 4.3581,
+      "step": 394500
+    },
+    {
+      "epoch": 3.396608126721763,
+      "grad_norm": 3.125,
+      "learning_rate": 4.674482805900555e-05,
+      "loss": 4.1576,
+      "step": 394550
+    },
+    {
+      "epoch": 3.397038567493113,
+      "grad_norm": 2.125,
+      "learning_rate": 4.672189938198242e-05,
+      "loss": 4.2351,
+      "step": 394600
+    },
+    {
+      "epoch": 3.397469008264463,
+      "grad_norm": 4.375,
+      "learning_rate": 4.6698974615457206e-05,
+      "loss": 4.4396,
+      "step": 394650
+    },
+    {
+      "epoch": 3.3978994490358128,
+      "grad_norm": 2.46875,
+      "learning_rate": 4.6676053761112646e-05,
+      "loss": 4.0337,
+      "step": 394700
+    },
+    {
+      "epoch": 3.3983298898071626,
+      "grad_norm": 2.640625,
+      "learning_rate": 4.665313682063095e-05,
+      "loss": 4.3153,
+      "step": 394750
+    },
+    {
+      "epoch": 3.3987603305785123,
+      "grad_norm": 2.265625,
+      "learning_rate": 4.663022379569426e-05,
+      "loss": 4.577,
+      "step": 394800
+    },
+    {
+      "epoch": 3.399190771349862,
+      "grad_norm": 4.25,
+      "learning_rate": 4.660731468798432e-05,
+      "loss": 4.3847,
+      "step": 394850
+    },
+    {
+      "epoch": 3.399621212121212,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.65844094991826e-05,
+      "loss": 4.4085,
+      "step": 394900
+    },
+    {
+      "epoch": 3.400051652892562,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.65615082309703e-05,
+      "loss": 4.1411,
+      "step": 394950
+    },
+    {
+      "epoch": 3.400482093663912,
+      "grad_norm": 2.8125,
+      "learning_rate": 4.6538610885028314e-05,
+      "loss": 4.2553,
+      "step": 395000
+    },
+    {
+      "epoch": 3.4009125344352618,
+      "grad_norm": 2.640625,
+      "learning_rate": 4.651571746303722e-05,
+      "loss": 4.3857,
+      "step": 395050
+    },
+    {
+      "epoch": 3.4013429752066116,
+      "grad_norm": 6.25,
+      "learning_rate": 4.6492827966677446e-05,
+      "loss": 4.2306,
+      "step": 395100
+    },
+    {
+      "epoch": 3.4017734159779613,
+      "grad_norm": 1.96875,
+      "learning_rate": 4.646994239762896e-05,
+      "loss": 4.0199,
+      "step": 395150
+    },
+    {
+      "epoch": 3.402203856749311,
+      "grad_norm": 2.4375,
+      "learning_rate": 4.644706075757154e-05,
+      "loss": 4.6054,
+      "step": 395200
+    },
+    {
+      "epoch": 3.4026342975206614,
+      "grad_norm": 3.921875,
+      "learning_rate": 4.642418304818463e-05,
+      "loss": 3.8235,
+      "step": 395250
+    },
+    {
+      "epoch": 3.403064738292011,
+      "grad_norm": 4.1875,
+      "learning_rate": 4.640130927114742e-05,
+      "loss": 3.994,
+      "step": 395300
+    },
+    {
+      "epoch": 3.403495179063361,
+      "grad_norm": 3.125,
+      "learning_rate": 4.6378439428138756e-05,
+      "loss": 4.4975,
+      "step": 395350
+    },
+    {
+      "epoch": 3.4039256198347108,
+      "grad_norm": 1.828125,
+      "learning_rate": 4.635557352083729e-05,
+      "loss": 4.2129,
+      "step": 395400
+    },
+    {
+      "epoch": 3.4043560606060606,
+      "grad_norm": 4.125,
+      "learning_rate": 4.6332711550921316e-05,
+      "loss": 4.5797,
+      "step": 395450
+    },
+    {
+      "epoch": 3.4047865013774103,
+      "grad_norm": 3.265625,
+      "learning_rate": 4.6309853520068844e-05,
+      "loss": 3.8621,
+      "step": 395500
+    },
+    {
+      "epoch": 3.40521694214876,
+      "grad_norm": 2.015625,
+      "learning_rate": 4.6286999429957613e-05,
+      "loss": 3.9599,
+      "step": 395550
+    },
+    {
+      "epoch": 3.4056473829201104,
+      "grad_norm": 4.65625,
+      "learning_rate": 4.626414928226506e-05,
+      "loss": 4.1441,
+      "step": 395600
+    },
+    {
+      "epoch": 3.40607782369146,
+      "grad_norm": 3.078125,
+      "learning_rate": 4.624130307866829e-05,
+      "loss": 4.5329,
+      "step": 395650
+    },
+    {
+      "epoch": 3.40650826446281,
+      "grad_norm": 2.5625,
+      "learning_rate": 4.6218460820844276e-05,
+      "loss": 4.0584,
+      "step": 395700
+    },
+    {
+      "epoch": 3.4069387052341598,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.619562251046946e-05,
+      "loss": 3.9191,
+      "step": 395750
+    },
+    {
+      "epoch": 3.4073691460055096,
+      "grad_norm": 3.078125,
+      "learning_rate": 4.617278814922022e-05,
+      "loss": 4.521,
+      "step": 395800
+    },
+    {
+      "epoch": 3.4077995867768593,
+      "grad_norm": 4.15625,
+      "learning_rate": 4.614995773877252e-05,
+      "loss": 4.2975,
+      "step": 395850
+    },
+    {
+      "epoch": 3.4082300275482096,
+      "grad_norm": 3.734375,
+      "learning_rate": 4.6127131280802026e-05,
+      "loss": 4.1153,
+      "step": 395900
+    },
+    {
+      "epoch": 3.4086604683195594,
+      "grad_norm": 2.03125,
+      "learning_rate": 4.6104308776984265e-05,
+      "loss": 4.3501,
+      "step": 395950
+    },
+    {
+      "epoch": 3.409090909090909,
+      "grad_norm": 5.09375,
+      "learning_rate": 4.608149022899423e-05,
+      "loss": 4.4347,
+      "step": 396000
+    },
+    {
+      "epoch": 3.409090909090909,
+      "eval_loss": 4.994675636291504,
+      "eval_runtime": 24.6395,
+      "eval_samples_per_second": 25.975,
+      "eval_steps_per_second": 12.987,
+      "eval_tts_loss": 7.505496205374016,
+      "step": 396000
+    },
+    {
+      "epoch": 3.409521349862259,
+      "grad_norm": 2.140625,
+      "learning_rate": 4.6058675638506776e-05,
+      "loss": 4.3295,
+      "step": 396050
+    },
+    {
+      "epoch": 3.4099517906336088,
+      "grad_norm": 3.875,
+      "learning_rate": 4.603586500719651e-05,
+      "loss": 4.2509,
+      "step": 396100
+    },
+    {
+      "epoch": 3.4103822314049586,
+      "grad_norm": 4.09375,
+      "learning_rate": 4.6013058336737637e-05,
+      "loss": 4.5618,
+      "step": 396150
+    },
+    {
+      "epoch": 3.4108126721763083,
+      "grad_norm": 7.6875,
+      "learning_rate": 4.599025562880409e-05,
+      "loss": 4.2612,
+      "step": 396200
+    },
+    {
+      "epoch": 3.4112431129476586,
+      "grad_norm": 2.171875,
+      "learning_rate": 4.596745688506966e-05,
+      "loss": 4.2725,
+      "step": 396250
+    },
+    {
+      "epoch": 3.4116735537190084,
+      "grad_norm": 2.609375,
+      "learning_rate": 4.594466210720756e-05,
+      "loss": 4.3079,
+      "step": 396300
+    },
+    {
+      "epoch": 3.412103994490358,
+      "grad_norm": 3.921875,
+      "learning_rate": 4.5921871296890996e-05,
+      "loss": 4.2141,
+      "step": 396350
+    },
+    {
+      "epoch": 3.412534435261708,
+      "grad_norm": 4.0,
+      "learning_rate": 4.589908445579272e-05,
+      "loss": 4.3151,
+      "step": 396400
+    },
+    {
+      "epoch": 3.4129648760330578,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.5876301585585246e-05,
+      "loss": 4.5092,
+      "step": 396450
+    },
+    {
+      "epoch": 3.4133953168044076,
+      "grad_norm": 4.15625,
+      "learning_rate": 4.5853522687940784e-05,
+      "loss": 4.2517,
+      "step": 396500
+    },
+    {
+      "epoch": 3.413825757575758,
+      "grad_norm": 3.390625,
+      "learning_rate": 4.583074776453127e-05,
+      "loss": 4.3639,
+      "step": 396550
+    },
+    {
+      "epoch": 3.4142561983471076,
+      "grad_norm": 5.34375,
+      "learning_rate": 4.5807976817028276e-05,
+      "loss": 4.5705,
+      "step": 396600
+    },
+    {
+      "epoch": 3.4146866391184574,
+      "grad_norm": 6.8125,
+      "learning_rate": 4.5785209847103225e-05,
+      "loss": 4.1948,
+      "step": 396650
+    },
+    {
+      "epoch": 3.415117079889807,
+      "grad_norm": 3.640625,
+      "learning_rate": 4.576244685642712e-05,
+      "loss": 4.4557,
+      "step": 396700
+    },
+    {
+      "epoch": 3.415547520661157,
+      "grad_norm": 3.734375,
+      "learning_rate": 4.573968784667073e-05,
+      "loss": 3.8626,
+      "step": 396750
+    },
+    {
+      "epoch": 3.4159779614325068,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.57169328195045e-05,
+      "loss": 4.1515,
+      "step": 396800
+    },
+    {
+      "epoch": 3.4164084022038566,
+      "grad_norm": 1.890625,
+      "learning_rate": 4.569418177659861e-05,
+      "loss": 3.9,
+      "step": 396850
+    },
+    {
+      "epoch": 3.416838842975207,
+      "grad_norm": 3.078125,
+      "learning_rate": 4.5671434719622906e-05,
+      "loss": 4.3722,
+      "step": 396900
+    },
+    {
+      "epoch": 3.4172692837465566,
+      "grad_norm": 3.046875,
+      "learning_rate": 4.564869165024703e-05,
+      "loss": 4.2138,
+      "step": 396950
+    },
+    {
+      "epoch": 3.4176997245179064,
+      "grad_norm": 4.59375,
+      "learning_rate": 4.562595257014025e-05,
+      "loss": 4.2076,
+      "step": 397000
+    },
+    {
+      "epoch": 3.418130165289256,
+      "grad_norm": 2.625,
+      "learning_rate": 4.560321748097156e-05,
+      "loss": 4.2064,
+      "step": 397050
+    },
+    {
+      "epoch": 3.418560606060606,
+      "grad_norm": 3.921875,
+      "learning_rate": 4.558048638440967e-05,
+      "loss": 4.5278,
+      "step": 397100
+    },
+    {
+      "epoch": 3.4189910468319558,
+      "grad_norm": 2.75,
+      "learning_rate": 4.5557759282123006e-05,
+      "loss": 4.1936,
+      "step": 397150
+    },
+    {
+      "epoch": 3.419421487603306,
+      "grad_norm": 4.75,
+      "learning_rate": 4.553503617577962e-05,
+      "loss": 4.4387,
+      "step": 397200
+    },
+    {
+      "epoch": 3.419851928374656,
+      "grad_norm": 0.9453125,
+      "learning_rate": 4.5512317067047484e-05,
+      "loss": 3.9222,
+      "step": 397250
+    },
+    {
+      "epoch": 3.4202823691460056,
+      "grad_norm": 3.984375,
+      "learning_rate": 4.548960195759395e-05,
+      "loss": 4.4349,
+      "step": 397300
+    },
+    {
+      "epoch": 3.4207128099173554,
+      "grad_norm": 2.625,
+      "learning_rate": 4.5466890849086395e-05,
+      "loss": 4.4317,
+      "step": 397350
+    },
+    {
+      "epoch": 3.421143250688705,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.5444183743191724e-05,
+      "loss": 4.2755,
+      "step": 397400
+    },
+    {
+      "epoch": 3.421573691460055,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.542148064157655e-05,
+      "loss": 4.467,
+      "step": 397450
+    },
+    {
+      "epoch": 3.4220041322314048,
+      "grad_norm": 4.90625,
+      "learning_rate": 4.5398781545907345e-05,
+      "loss": 4.0659,
+      "step": 397500
+    },
+    {
+      "epoch": 3.422434573002755,
+      "grad_norm": 3.125,
+      "learning_rate": 4.5376086457850055e-05,
+      "loss": 4.3646,
+      "step": 397550
+    },
+    {
+      "epoch": 3.422865013774105,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.535339537907045e-05,
+      "loss": 4.3562,
+      "step": 397600
+    },
+    {
+      "epoch": 3.4232954545454546,
+      "grad_norm": 2.296875,
+      "learning_rate": 4.5330708311234096e-05,
+      "loss": 4.47,
+      "step": 397650
+    },
+    {
+      "epoch": 3.4237258953168044,
+      "grad_norm": 4.78125,
+      "learning_rate": 4.530802525600613e-05,
+      "loss": 4.0313,
+      "step": 397700
+    },
+    {
+      "epoch": 3.424156336088154,
+      "grad_norm": 3.4375,
+      "learning_rate": 4.528534621505142e-05,
+      "loss": 4.5859,
+      "step": 397750
+    },
+    {
+      "epoch": 3.424586776859504,
+      "grad_norm": 4.75,
+      "learning_rate": 4.526267119003459e-05,
+      "loss": 4.327,
+      "step": 397800
+    },
+    {
+      "epoch": 3.425017217630854,
+      "grad_norm": 3.078125,
+      "learning_rate": 4.5240000182619924e-05,
+      "loss": 4.2197,
+      "step": 397850
+    },
+    {
+      "epoch": 3.425447658402204,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.521733319447138e-05,
+      "loss": 4.1267,
+      "step": 397900
+    },
+    {
+      "epoch": 3.425878099173554,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.5194670227252754e-05,
+      "loss": 4.3995,
+      "step": 397950
+    },
+    {
+      "epoch": 3.4263085399449036,
+      "grad_norm": 3.859375,
+      "learning_rate": 4.5172011282627414e-05,
+      "loss": 4.0103,
+      "step": 398000
+    },
+    {
+      "epoch": 3.4267389807162534,
+      "grad_norm": 0.85546875,
+      "learning_rate": 4.5149356362258486e-05,
+      "loss": 4.0549,
+      "step": 398050
+    },
+    {
+      "epoch": 3.427169421487603,
+      "grad_norm": 2.765625,
+      "learning_rate": 4.5126705467808785e-05,
+      "loss": 4.5051,
+      "step": 398100
+    },
+    {
+      "epoch": 3.427599862258953,
+      "grad_norm": 3.078125,
+      "learning_rate": 4.5104058600940805e-05,
+      "loss": 4.4772,
+      "step": 398150
+    },
+    {
+      "epoch": 3.428030303030303,
+      "grad_norm": 3.234375,
+      "learning_rate": 4.5081415763316845e-05,
+      "loss": 4.5994,
+      "step": 398200
+    },
+    {
+      "epoch": 3.428460743801653,
+      "grad_norm": 5.1875,
+      "learning_rate": 4.5058776956598804e-05,
+      "loss": 4.4596,
+      "step": 398250
+    },
+    {
+      "epoch": 3.428891184573003,
+      "grad_norm": 3.5625,
+      "learning_rate": 4.503614218244834e-05,
+      "loss": 4.3909,
+      "step": 398300
+    },
+    {
+      "epoch": 3.4293216253443526,
+      "grad_norm": 1.4296875,
+      "learning_rate": 4.501351144252677e-05,
+      "loss": 4.6276,
+      "step": 398350
+    },
+    {
+      "epoch": 3.4297520661157024,
+      "grad_norm": 2.125,
+      "learning_rate": 4.4990884738495154e-05,
+      "loss": 4.5762,
+      "step": 398400
+    },
+    {
+      "epoch": 3.430182506887052,
+      "grad_norm": 4.8125,
+      "learning_rate": 4.49682620720142e-05,
+      "loss": 4.1249,
+      "step": 398450
+    },
+    {
+      "epoch": 3.4306129476584024,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.494564344474446e-05,
+      "loss": 4.5648,
+      "step": 398500
+    },
+    {
+      "epoch": 3.431043388429752,
+      "grad_norm": 4.46875,
+      "learning_rate": 4.492302885834606e-05,
+      "loss": 4.3008,
+      "step": 398550
+    },
+    {
+      "epoch": 3.431473829201102,
+      "grad_norm": 3.078125,
+      "learning_rate": 4.490041831447877e-05,
+      "loss": 4.5481,
+      "step": 398600
+    },
+    {
+      "epoch": 3.431904269972452,
+      "grad_norm": 3.28125,
+      "learning_rate": 4.4877811814802276e-05,
+      "loss": 4.3814,
+      "step": 398650
+    },
+    {
+      "epoch": 3.4323347107438016,
+      "grad_norm": 3.734375,
+      "learning_rate": 4.485520936097578e-05,
+      "loss": 4.3508,
+      "step": 398700
+    },
+    {
+      "epoch": 3.4327651515151514,
+      "grad_norm": 1.3046875,
+      "learning_rate": 4.483261095465824e-05,
+      "loss": 4.3446,
+      "step": 398750
+    },
+    {
+      "epoch": 3.433195592286501,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.481001659750843e-05,
+      "loss": 4.0748,
+      "step": 398800
+    },
+    {
+      "epoch": 3.4336260330578514,
+      "grad_norm": 2.40625,
+      "learning_rate": 4.4787426291184586e-05,
+      "loss": 3.8616,
+      "step": 398850
+    },
+    {
+      "epoch": 3.434056473829201,
+      "grad_norm": 3.875,
+      "learning_rate": 4.47648400373449e-05,
+      "loss": 4.5028,
+      "step": 398900
+    },
+    {
+      "epoch": 3.434486914600551,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.4742257837647107e-05,
+      "loss": 4.0689,
+      "step": 398950
+    },
+    {
+      "epoch": 3.434917355371901,
+      "grad_norm": 2.390625,
+      "learning_rate": 4.471967969374866e-05,
+      "loss": 4.047,
+      "step": 399000
+    },
+    {
+      "epoch": 3.434917355371901,
+      "eval_loss": 4.9932861328125,
+      "eval_runtime": 26.286,
+      "eval_samples_per_second": 24.348,
+      "eval_steps_per_second": 12.174,
+      "eval_tts_loss": 7.48067210551469,
+      "step": 399000
+    },
+    {
+      "epoch": 3.4353477961432506,
+      "grad_norm": 2.15625,
+      "learning_rate": 4.4697105607306865e-05,
+      "loss": 4.5423,
+      "step": 399050
+    },
+    {
+      "epoch": 3.4357782369146004,
+      "grad_norm": 2.921875,
+      "learning_rate": 4.467453557997849e-05,
+      "loss": 4.1795,
+      "step": 399100
+    },
+    {
+      "epoch": 3.4362086776859506,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.465196961342014e-05,
+      "loss": 4.4131,
+      "step": 399150
+    },
+    {
+      "epoch": 3.4366391184573004,
+      "grad_norm": 3.234375,
+      "learning_rate": 4.4629407709288174e-05,
+      "loss": 4.1543,
+      "step": 399200
+    },
+    {
+      "epoch": 3.43706955922865,
+      "grad_norm": 3.28125,
+      "learning_rate": 4.460684986923854e-05,
+      "loss": 4.359,
+      "step": 399250
+    },
+    {
+      "epoch": 3.4375,
+      "grad_norm": 2.46875,
+      "learning_rate": 4.458429609492695e-05,
+      "loss": 4.0929,
+      "step": 399300
+    },
+    {
+      "epoch": 3.43793044077135,
+      "grad_norm": 2.796875,
+      "learning_rate": 4.456174638800879e-05,
+      "loss": 4.3975,
+      "step": 399350
+    },
+    {
+      "epoch": 3.4383608815426996,
+      "grad_norm": 3.0625,
+      "learning_rate": 4.453920075013916e-05,
+      "loss": 4.4661,
+      "step": 399400
+    },
+    {
+      "epoch": 3.4387913223140494,
+      "grad_norm": 4.25,
+      "learning_rate": 4.451665918297284e-05,
+      "loss": 4.5318,
+      "step": 399450
+    },
+    {
+      "epoch": 3.4392217630853996,
+      "grad_norm": 10.0,
+      "learning_rate": 4.4494121688164405e-05,
+      "loss": 4.4995,
+      "step": 399500
+    },
+    {
+      "epoch": 3.4396522038567494,
+      "grad_norm": 4.375,
+      "learning_rate": 4.4471588267368e-05,
+      "loss": 4.5414,
+      "step": 399550
+    },
+    {
+      "epoch": 3.440082644628099,
+      "grad_norm": 2.109375,
+      "learning_rate": 4.444905892223756e-05,
+      "loss": 3.9789,
+      "step": 399600
+    },
+    {
+      "epoch": 3.440513085399449,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.442653365442666e-05,
+      "loss": 4.5508,
+      "step": 399650
+    },
+    {
+      "epoch": 3.440943526170799,
+      "grad_norm": 2.3125,
+      "learning_rate": 4.44040124655886e-05,
+      "loss": 4.3807,
+      "step": 399700
+    },
+    {
+      "epoch": 3.4413739669421486,
+      "grad_norm": 3.25,
+      "learning_rate": 4.438149535737642e-05,
+      "loss": 4.6795,
+      "step": 399750
+    },
+    {
+      "epoch": 3.441804407713499,
+      "grad_norm": 4.5,
+      "learning_rate": 4.435898233144283e-05,
+      "loss": 4.1523,
+      "step": 399800
+    },
+    {
+      "epoch": 3.4422348484848486,
+      "grad_norm": 2.765625,
+      "learning_rate": 4.433647338944023e-05,
+      "loss": 4.4798,
+      "step": 399850
+    },
+    {
+      "epoch": 3.4426652892561984,
+      "grad_norm": 2.890625,
+      "learning_rate": 4.43139685330207e-05,
+      "loss": 4.3965,
+      "step": 399900
+    },
+    {
+      "epoch": 3.443095730027548,
+      "grad_norm": 4.28125,
+      "learning_rate": 4.429146776383608e-05,
+      "loss": 4.6089,
+      "step": 399950
+    },
+    {
+      "epoch": 3.443526170798898,
+      "grad_norm": 5.78125,
+      "learning_rate": 4.4268971083537825e-05,
+      "loss": 4.2955,
+      "step": 400000
+    },
+    {
+      "epoch": 3.443956611570248,
+      "grad_norm": 3.828125,
+      "learning_rate": 4.4246478493777223e-05,
+      "loss": 3.9327,
+      "step": 400050
+    },
+    {
+      "epoch": 3.4443870523415976,
+      "grad_norm": 3.921875,
+      "learning_rate": 4.422398999620517e-05,
+      "loss": 4.4019,
+      "step": 400100
+    },
+    {
+      "epoch": 3.444817493112948,
+      "grad_norm": 3.9375,
+      "learning_rate": 4.420150559247219e-05,
+      "loss": 4.2561,
+      "step": 400150
+    },
+    {
+      "epoch": 3.4452479338842976,
+      "grad_norm": 2.3125,
+      "learning_rate": 4.417902528422867e-05,
+      "loss": 4.1378,
+      "step": 400200
+    },
+    {
+      "epoch": 3.4456783746556474,
+      "grad_norm": 2.296875,
+      "learning_rate": 4.4156549073124595e-05,
+      "loss": 4.1599,
+      "step": 400250
+    },
+    {
+      "epoch": 3.446108815426997,
+      "grad_norm": 6.5625,
+      "learning_rate": 4.4134076960809634e-05,
+      "loss": 4.5083,
+      "step": 400300
+    },
+    {
+      "epoch": 3.446539256198347,
+      "grad_norm": 5.15625,
+      "learning_rate": 4.41116089489333e-05,
+      "loss": 3.8569,
+      "step": 400350
+    },
+    {
+      "epoch": 3.446969696969697,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.4089145039144544e-05,
+      "loss": 4.0468,
+      "step": 400400
+    },
+    {
+      "epoch": 3.447400137741047,
+      "grad_norm": 3.53125,
+      "learning_rate": 4.406668523309229e-05,
+      "loss": 3.9891,
+      "step": 400450
+    },
+    {
+      "epoch": 3.447830578512397,
+      "grad_norm": 6.0625,
+      "learning_rate": 4.404422953242499e-05,
+      "loss": 4.4894,
+      "step": 400500
+    },
+    {
+      "epoch": 3.4482610192837466,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.402177793879085e-05,
+      "loss": 4.185,
+      "step": 400550
+    },
+    {
+      "epoch": 3.4486914600550964,
+      "grad_norm": 3.515625,
+      "learning_rate": 4.399933045383777e-05,
+      "loss": 4.3618,
+      "step": 400600
+    },
+    {
+      "epoch": 3.449121900826446,
+      "grad_norm": 3.28125,
+      "learning_rate": 4.397688707921335e-05,
+      "loss": 4.3885,
+      "step": 400650
+    },
+    {
+      "epoch": 3.449552341597796,
+      "grad_norm": 1.890625,
+      "learning_rate": 4.3954447816564836e-05,
+      "loss": 4.6036,
+      "step": 400700
+    },
+    {
+      "epoch": 3.449982782369146,
+      "grad_norm": 5.0,
+      "learning_rate": 4.393201266753931e-05,
+      "loss": 4.3906,
+      "step": 400750
+    },
+    {
+      "epoch": 3.450413223140496,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.390958163378342e-05,
+      "loss": 4.8371,
+      "step": 400800
+    },
+    {
+      "epoch": 3.450843663911846,
+      "grad_norm": 4.15625,
+      "learning_rate": 4.388715471694357e-05,
+      "loss": 4.4369,
+      "step": 400850
+    },
+    {
+      "epoch": 3.4512741046831956,
+      "grad_norm": 3.140625,
+      "learning_rate": 4.386473191866581e-05,
+      "loss": 4.5882,
+      "step": 400900
+    },
+    {
+      "epoch": 3.4517045454545454,
+      "grad_norm": 1.7734375,
+      "learning_rate": 4.384231324059597e-05,
+      "loss": 4.1069,
+      "step": 400950
+    },
+    {
+      "epoch": 3.452134986225895,
+      "grad_norm": 3.890625,
+      "learning_rate": 4.381989868437946e-05,
+      "loss": 4.0956,
+      "step": 401000
+    },
+    {
+      "epoch": 3.452565426997245,
+      "grad_norm": 4.53125,
+      "learning_rate": 4.3797488251661555e-05,
+      "loss": 4.4168,
+      "step": 401050
+    },
+    {
+      "epoch": 3.4529958677685952,
+      "grad_norm": 2.421875,
+      "learning_rate": 4.3775081944087095e-05,
+      "loss": 4.2854,
+      "step": 401100
+    },
+    {
+      "epoch": 3.453426308539945,
+      "grad_norm": 1.8828125,
+      "learning_rate": 4.375267976330064e-05,
+      "loss": 3.9802,
+      "step": 401150
+    },
+    {
+      "epoch": 3.453856749311295,
+      "grad_norm": 2.40625,
+      "learning_rate": 4.373028171094648e-05,
+      "loss": 4.121,
+      "step": 401200
+    },
+    {
+      "epoch": 3.4542871900826446,
+      "grad_norm": 3.734375,
+      "learning_rate": 4.370788778866852e-05,
+      "loss": 4.2481,
+      "step": 401250
+    },
+    {
+      "epoch": 3.4547176308539944,
+      "grad_norm": 4.875,
+      "learning_rate": 4.3685497998110535e-05,
+      "loss": 4.3699,
+      "step": 401300
+    },
+    {
+      "epoch": 3.455148071625344,
+      "grad_norm": 3.90625,
+      "learning_rate": 4.366311234091586e-05,
+      "loss": 4.1837,
+      "step": 401350
+    },
+    {
+      "epoch": 3.455578512396694,
+      "grad_norm": 2.703125,
+      "learning_rate": 4.364073081872746e-05,
+      "loss": 4.6542,
+      "step": 401400
+    },
+    {
+      "epoch": 3.4560089531680442,
+      "grad_norm": 3.890625,
+      "learning_rate": 4.3618353433188194e-05,
+      "loss": 4.238,
+      "step": 401450
+    },
+    {
+      "epoch": 3.456439393939394,
+      "grad_norm": 3.984375,
+      "learning_rate": 4.3595980185940475e-05,
+      "loss": 4.3141,
+      "step": 401500
+    },
+    {
+      "epoch": 3.456869834710744,
+      "grad_norm": 5.4375,
+      "learning_rate": 4.357361107862642e-05,
+      "loss": 4.3467,
+      "step": 401550
+    },
+    {
+      "epoch": 3.4573002754820936,
+      "grad_norm": 2.578125,
+      "learning_rate": 4.3551246112887947e-05,
+      "loss": 4.3645,
+      "step": 401600
+    },
+    {
+      "epoch": 3.4577307162534434,
+      "grad_norm": 2.109375,
+      "learning_rate": 4.352888529036658e-05,
+      "loss": 4.2373,
+      "step": 401650
+    },
+    {
+      "epoch": 3.458161157024793,
+      "grad_norm": 4.40625,
+      "learning_rate": 4.350652861270347e-05,
+      "loss": 4.2433,
+      "step": 401700
+    },
+    {
+      "epoch": 3.4585915977961434,
+      "grad_norm": 4.03125,
+      "learning_rate": 4.3484176081539654e-05,
+      "loss": 4.3552,
+      "step": 401750
+    },
+    {
+      "epoch": 3.4590220385674932,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.34618276985157e-05,
+      "loss": 4.3612,
+      "step": 401800
+    },
+    {
+      "epoch": 3.459452479338843,
+      "grad_norm": 1.4609375,
+      "learning_rate": 4.343948346527192e-05,
+      "loss": 4.4421,
+      "step": 401850
+    },
+    {
+      "epoch": 3.459882920110193,
+      "grad_norm": 3.265625,
+      "learning_rate": 4.341714338344843e-05,
+      "loss": 4.4969,
+      "step": 401900
+    },
+    {
+      "epoch": 3.4603133608815426,
+      "grad_norm": 5.21875,
+      "learning_rate": 4.3394807454684805e-05,
+      "loss": 3.903,
+      "step": 401950
+    },
+    {
+      "epoch": 3.4607438016528924,
+      "grad_norm": 5.1875,
+      "learning_rate": 4.337247568062056e-05,
+      "loss": 4.1216,
+      "step": 402000
+    },
+    {
+      "epoch": 3.4607438016528924,
+      "eval_loss": 4.993624210357666,
+      "eval_runtime": 23.5215,
+      "eval_samples_per_second": 27.209,
+      "eval_steps_per_second": 13.605,
+      "eval_tts_loss": 7.48834117052633,
+      "step": 402000
+    },
+    {
+      "epoch": 3.461174242424242,
+      "grad_norm": 2.265625,
+      "learning_rate": 4.3350148062894766e-05,
+      "loss": 4.5171,
+      "step": 402050
+    },
+    {
+      "epoch": 3.4616046831955924,
+      "grad_norm": 3.078125,
+      "learning_rate": 4.3327824603146217e-05,
+      "loss": 4.1878,
+      "step": 402100
+    },
+    {
+      "epoch": 3.4620351239669422,
+      "grad_norm": 3.65625,
+      "learning_rate": 4.330550530301341e-05,
+      "loss": 4.2243,
+      "step": 402150
+    },
+    {
+      "epoch": 3.462465564738292,
+      "grad_norm": 3.34375,
+      "learning_rate": 4.3283190164134526e-05,
+      "loss": 4.3377,
+      "step": 402200
+    },
+    {
+      "epoch": 3.462896005509642,
+      "grad_norm": 2.828125,
+      "learning_rate": 4.326087918814742e-05,
+      "loss": 4.2108,
+      "step": 402250
+    },
+    {
+      "epoch": 3.4633264462809916,
+      "grad_norm": 5.0625,
+      "learning_rate": 4.323857237668975e-05,
+      "loss": 4.1,
+      "step": 402300
+    },
+    {
+      "epoch": 3.4637568870523414,
+      "grad_norm": 4.5625,
+      "learning_rate": 4.321626973139873e-05,
+      "loss": 4.2767,
+      "step": 402350
+    },
+    {
+      "epoch": 3.4641873278236917,
+      "grad_norm": 1.28125,
+      "learning_rate": 4.319397125391135e-05,
+      "loss": 4.0428,
+      "step": 402400
+    },
+    {
+      "epoch": 3.4646177685950414,
+      "grad_norm": 5.34375,
+      "learning_rate": 4.317167694586425e-05,
+      "loss": 4.4062,
+      "step": 402450
+    },
+    {
+      "epoch": 3.4650482093663912,
+      "grad_norm": 3.1875,
+      "learning_rate": 4.314938680889379e-05,
+      "loss": 4.4809,
+      "step": 402500
+    },
+    {
+      "epoch": 3.465478650137741,
+      "grad_norm": 4.09375,
+      "learning_rate": 4.312710084463598e-05,
+      "loss": 4.3265,
+      "step": 402550
+    },
+    {
+      "epoch": 3.465909090909091,
+      "grad_norm": 5.46875,
+      "learning_rate": 4.3104819054726656e-05,
+      "loss": 4.1198,
+      "step": 402600
+    },
+    {
+      "epoch": 3.4663395316804406,
+      "grad_norm": 2.1875,
+      "learning_rate": 4.3082541440801196e-05,
+      "loss": 4.3553,
+      "step": 402650
+    },
+    {
+      "epoch": 3.4667699724517904,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.3060268004494734e-05,
+      "loss": 4.4858,
+      "step": 402700
+    },
+    {
+      "epoch": 3.4672004132231407,
+      "grad_norm": 2.484375,
+      "learning_rate": 4.303799874744208e-05,
+      "loss": 4.2628,
+      "step": 402750
+    },
+    {
+      "epoch": 3.4676308539944904,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.301573367127777e-05,
+      "loss": 4.0898,
+      "step": 402800
+    },
+    {
+      "epoch": 3.4680612947658402,
+      "grad_norm": 2.203125,
+      "learning_rate": 4.2993472777635956e-05,
+      "loss": 4.3174,
+      "step": 402850
+    },
+    {
+      "epoch": 3.46849173553719,
+      "grad_norm": 4.8125,
+      "learning_rate": 4.2971216068150675e-05,
+      "loss": 4.4431,
+      "step": 402900
+    },
+    {
+      "epoch": 3.46892217630854,
+      "grad_norm": 2.765625,
+      "learning_rate": 4.2948963544455354e-05,
+      "loss": 3.8362,
+      "step": 402950
+    },
+    {
+      "epoch": 3.4693526170798896,
+      "grad_norm": 3.375,
+      "learning_rate": 4.2926715208183396e-05,
+      "loss": 4.0822,
+      "step": 403000
+    },
+    {
+      "epoch": 3.46978305785124,
+      "grad_norm": 6.09375,
+      "learning_rate": 4.2904471060967754e-05,
+      "loss": 3.732,
+      "step": 403050
+    },
+    {
+      "epoch": 3.4702134986225897,
+      "grad_norm": 3.375,
+      "learning_rate": 4.288223110444105e-05,
+      "loss": 4.4691,
+      "step": 403100
+    },
+    {
+      "epoch": 3.4706439393939394,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.285999534023576e-05,
+      "loss": 4.2009,
+      "step": 403150
+    },
+    {
+      "epoch": 3.4710743801652892,
+      "grad_norm": 2.546875,
+      "learning_rate": 4.283776376998384e-05,
+      "loss": 4.266,
+      "step": 403200
+    },
+    {
+      "epoch": 3.471504820936639,
+      "grad_norm": 5.5,
+      "learning_rate": 4.2815536395317035e-05,
+      "loss": 4.082,
+      "step": 403250
+    },
+    {
+      "epoch": 3.471935261707989,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.279331321786686e-05,
+      "loss": 4.244,
+      "step": 403300
+    },
+    {
+      "epoch": 3.4723657024793386,
+      "grad_norm": 1.65625,
+      "learning_rate": 4.277109423926441e-05,
+      "loss": 4.423,
+      "step": 403350
+    },
+    {
+      "epoch": 3.472796143250689,
+      "grad_norm": 4.28125,
+      "learning_rate": 4.274887946114048e-05,
+      "loss": 4.6118,
+      "step": 403400
+    },
+    {
+      "epoch": 3.4732265840220387,
+      "grad_norm": 3.3125,
+      "learning_rate": 4.2726668885125684e-05,
+      "loss": 4.447,
+      "step": 403450
+    },
+    {
+      "epoch": 3.4736570247933884,
+      "grad_norm": 2.765625,
+      "learning_rate": 4.27044625128501e-05,
+      "loss": 4.0383,
+      "step": 403500
+    },
+    {
+      "epoch": 3.4740874655647382,
+      "grad_norm": 1.7265625,
+      "learning_rate": 4.268226034594371e-05,
+      "loss": 4.3027,
+      "step": 403550
+    },
+    {
+      "epoch": 3.474517906336088,
+      "grad_norm": 2.5625,
+      "learning_rate": 4.266006238603612e-05,
+      "loss": 4.303,
+      "step": 403600
+    },
+    {
+      "epoch": 3.474948347107438,
+      "grad_norm": 2.546875,
+      "learning_rate": 4.263786863475655e-05,
+      "loss": 4.2431,
+      "step": 403650
+    },
+    {
+      "epoch": 3.475378787878788,
+      "grad_norm": 3.234375,
+      "learning_rate": 4.2615679093734016e-05,
+      "loss": 3.9285,
+      "step": 403700
+    },
+    {
+      "epoch": 3.475809228650138,
+      "grad_norm": 2.5,
+      "learning_rate": 4.259349376459718e-05,
+      "loss": 4.5531,
+      "step": 403750
+    },
+    {
+      "epoch": 3.4762396694214877,
+      "grad_norm": 2.4375,
+      "learning_rate": 4.2571312648974347e-05,
+      "loss": 4.343,
+      "step": 403800
+    },
+    {
+      "epoch": 3.4766701101928374,
+      "grad_norm": 3.90625,
+      "learning_rate": 4.254913574849364e-05,
+      "loss": 4.1313,
+      "step": 403850
+    },
+    {
+      "epoch": 3.4771005509641872,
+      "grad_norm": 4.875,
+      "learning_rate": 4.2526963064782754e-05,
+      "loss": 4.2954,
+      "step": 403900
+    },
+    {
+      "epoch": 3.477530991735537,
+      "grad_norm": 6.15625,
+      "learning_rate": 4.2504794599469134e-05,
+      "loss": 4.2832,
+      "step": 403950
+    },
+    {
+      "epoch": 3.477961432506887,
+      "grad_norm": 2.171875,
+      "learning_rate": 4.2482630354179875e-05,
+      "loss": 3.9697,
+      "step": 404000
+    },
+    {
+      "epoch": 3.478391873278237,
+      "grad_norm": 2.78125,
+      "learning_rate": 4.24604703305418e-05,
+      "loss": 4.263,
+      "step": 404050
+    },
+    {
+      "epoch": 3.478822314049587,
+      "grad_norm": 4.1875,
+      "learning_rate": 4.243831453018137e-05,
+      "loss": 4.1289,
+      "step": 404100
+    },
+    {
+      "epoch": 3.4792527548209367,
+      "grad_norm": 3.734375,
+      "learning_rate": 4.2416162954724836e-05,
+      "loss": 4.5328,
+      "step": 404150
+    },
+    {
+      "epoch": 3.4796831955922864,
+      "grad_norm": 1.78125,
+      "learning_rate": 4.2394015605798054e-05,
+      "loss": 4.4685,
+      "step": 404200
+    },
+    {
+      "epoch": 3.4801136363636362,
+      "grad_norm": 3.5625,
+      "learning_rate": 4.237187248502659e-05,
+      "loss": 4.0868,
+      "step": 404250
+    },
+    {
+      "epoch": 3.480544077134986,
+      "grad_norm": 3.0625,
+      "learning_rate": 4.234973359403568e-05,
+      "loss": 4.0506,
+      "step": 404300
+    },
+    {
+      "epoch": 3.4809745179063363,
+      "grad_norm": 3.515625,
+      "learning_rate": 4.23275989344503e-05,
+      "loss": 4.3471,
+      "step": 404350
+    },
+    {
+      "epoch": 3.481404958677686,
+      "grad_norm": 4.09375,
+      "learning_rate": 4.230546850789502e-05,
+      "loss": 4.051,
+      "step": 404400
+    },
+    {
+      "epoch": 3.481835399449036,
+      "grad_norm": 3.90625,
+      "learning_rate": 4.22833423159943e-05,
+      "loss": 4.4472,
+      "step": 404450
+    },
+    {
+      "epoch": 3.4822658402203857,
+      "grad_norm": 3.03125,
+      "learning_rate": 4.2261220360372e-05,
+      "loss": 4.3289,
+      "step": 404500
+    },
+    {
+      "epoch": 3.4826962809917354,
+      "grad_norm": 2.5625,
+      "learning_rate": 4.223910264265194e-05,
+      "loss": 3.6602,
+      "step": 404550
+    },
+    {
+      "epoch": 3.4831267217630852,
+      "grad_norm": 3.328125,
+      "learning_rate": 4.2216989164457466e-05,
+      "loss": 4.0248,
+      "step": 404600
+    },
+    {
+      "epoch": 3.483557162534435,
+      "grad_norm": 0.68359375,
+      "learning_rate": 4.2194879927411625e-05,
+      "loss": 4.2495,
+      "step": 404650
+    },
+    {
+      "epoch": 3.4839876033057853,
+      "grad_norm": 4.15625,
+      "learning_rate": 4.21727749331373e-05,
+      "loss": 4.3664,
+      "step": 404700
+    },
+    {
+      "epoch": 3.484418044077135,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.2150674183256844e-05,
+      "loss": 4.5422,
+      "step": 404750
+    },
+    {
+      "epoch": 3.484848484848485,
+      "grad_norm": 2.734375,
+      "learning_rate": 4.21285776793924e-05,
+      "loss": 4.0481,
+      "step": 404800
+    },
+    {
+      "epoch": 3.4852789256198347,
+      "grad_norm": 4.71875,
+      "learning_rate": 4.2106485423165876e-05,
+      "loss": 4.6,
+      "step": 404850
+    },
+    {
+      "epoch": 3.4857093663911844,
+      "grad_norm": 4.3125,
+      "learning_rate": 4.2084397416198764e-05,
+      "loss": 4.309,
+      "step": 404900
+    },
+    {
+      "epoch": 3.4861398071625342,
+      "grad_norm": 3.234375,
+      "learning_rate": 4.2062313660112265e-05,
+      "loss": 4.3773,
+      "step": 404950
+    },
+    {
+      "epoch": 3.4865702479338845,
+      "grad_norm": 2.046875,
+      "learning_rate": 4.2040234156527295e-05,
+      "loss": 4.2435,
+      "step": 405000
+    },
+    {
+      "epoch": 3.4865702479338845,
+      "eval_loss": 4.992835998535156,
+      "eval_runtime": 23.5141,
+      "eval_samples_per_second": 27.218,
+      "eval_steps_per_second": 13.609,
+      "eval_tts_loss": 7.475708443343599,
+      "step": 405000
+    },
+    {
+      "epoch": 3.4870006887052343,
+      "grad_norm": 1.4296875,
+      "learning_rate": 4.201815890706442e-05,
+      "loss": 4.4341,
+      "step": 405050
+    },
+    {
+      "epoch": 3.487431129476584,
+      "grad_norm": 3.4375,
+      "learning_rate": 4.1996087913343905e-05,
+      "loss": 4.4096,
+      "step": 405100
+    },
+    {
+      "epoch": 3.487861570247934,
+      "grad_norm": 3.0625,
+      "learning_rate": 4.197402117698578e-05,
+      "loss": 4.2599,
+      "step": 405150
+    },
+    {
+      "epoch": 3.4882920110192837,
+      "grad_norm": 2.015625,
+      "learning_rate": 4.195195869960964e-05,
+      "loss": 4.16,
+      "step": 405200
+    },
+    {
+      "epoch": 3.4887224517906334,
+      "grad_norm": 3.140625,
+      "learning_rate": 4.192990048283485e-05,
+      "loss": 4.1305,
+      "step": 405250
+    },
+    {
+      "epoch": 3.4891528925619832,
+      "grad_norm": 3.15625,
+      "learning_rate": 4.190784652828043e-05,
+      "loss": 4.7154,
+      "step": 405300
+    },
+    {
+      "epoch": 3.4895833333333335,
+      "grad_norm": 2.328125,
+      "learning_rate": 4.188579683756503e-05,
+      "loss": 4.6028,
+      "step": 405350
+    },
+    {
+      "epoch": 3.4900137741046833,
+      "grad_norm": 2.75,
+      "learning_rate": 4.186375141230715e-05,
+      "loss": 4.2713,
+      "step": 405400
+    },
+    {
+      "epoch": 3.490444214876033,
+      "grad_norm": 4.1875,
+      "learning_rate": 4.1841710254124836e-05,
+      "loss": 4.6832,
+      "step": 405450
+    },
+    {
+      "epoch": 3.490874655647383,
+      "grad_norm": 4.21875,
+      "learning_rate": 4.181967336463586e-05,
+      "loss": 4.3429,
+      "step": 405500
+    },
+    {
+      "epoch": 3.4913050964187327,
+      "grad_norm": 3.0625,
+      "learning_rate": 4.1797640745457675e-05,
+      "loss": 4.5516,
+      "step": 405550
+    },
+    {
+      "epoch": 3.491735537190083,
+      "grad_norm": 3.140625,
+      "learning_rate": 4.177561239820744e-05,
+      "loss": 4.2398,
+      "step": 405600
+    },
+    {
+      "epoch": 3.4921659779614327,
+      "grad_norm": 1.4375,
+      "learning_rate": 4.175358832450195e-05,
+      "loss": 4.4105,
+      "step": 405650
+    },
+    {
+      "epoch": 3.4925964187327825,
+      "grad_norm": 4.3125,
+      "learning_rate": 4.1731568525957774e-05,
+      "loss": 4.5158,
+      "step": 405700
+    },
+    {
+      "epoch": 3.4930268595041323,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.1709553004191115e-05,
+      "loss": 4.3405,
+      "step": 405750
+    },
+    {
+      "epoch": 3.493457300275482,
+      "grad_norm": 4.15625,
+      "learning_rate": 4.168754176081785e-05,
+      "loss": 4.1078,
+      "step": 405800
+    },
+    {
+      "epoch": 3.493887741046832,
+      "grad_norm": 3.28125,
+      "learning_rate": 4.166553479745356e-05,
+      "loss": 4.3578,
+      "step": 405850
+    },
+    {
+      "epoch": 3.4943181818181817,
+      "grad_norm": 3.328125,
+      "learning_rate": 4.16435321157135e-05,
+      "loss": 4.5388,
+      "step": 405900
+    },
+    {
+      "epoch": 3.4947486225895315,
+      "grad_norm": 3.53125,
+      "learning_rate": 4.162153371721259e-05,
+      "loss": 4.3248,
+      "step": 405950
+    },
+    {
+      "epoch": 3.4951790633608817,
+      "grad_norm": 5.46875,
+      "learning_rate": 4.159953960356557e-05,
+      "loss": 4.2265,
+      "step": 406000
+    },
+    {
+      "epoch": 3.4956095041322315,
+      "grad_norm": 3.0,
+      "learning_rate": 4.1577549776386625e-05,
+      "loss": 4.2892,
+      "step": 406050
+    },
+    {
+      "epoch": 3.4960399449035813,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.155556423728986e-05,
+      "loss": 4.5105,
+      "step": 406100
+    },
+    {
+      "epoch": 3.496470385674931,
+      "grad_norm": 4.75,
+      "learning_rate": 4.153358298788893e-05,
+      "loss": 4.5752,
+      "step": 406150
+    },
+    {
+      "epoch": 3.496900826446281,
+      "grad_norm": 4.21875,
+      "learning_rate": 4.151160602979719e-05,
+      "loss": 4.4671,
+      "step": 406200
+    },
+    {
+      "epoch": 3.497331267217631,
+      "grad_norm": 3.0625,
+      "learning_rate": 4.148963336462779e-05,
+      "loss": 4.4357,
+      "step": 406250
+    },
+    {
+      "epoch": 3.497761707988981,
+      "grad_norm": 4.75,
+      "learning_rate": 4.146766499399337e-05,
+      "loss": 4.3913,
+      "step": 406300
+    },
+    {
+      "epoch": 3.4981921487603307,
+      "grad_norm": 5.03125,
+      "learning_rate": 4.144570091950638e-05,
+      "loss": 4.2077,
+      "step": 406350
+    },
+    {
+      "epoch": 3.4986225895316805,
+      "grad_norm": 2.234375,
+      "learning_rate": 4.142374114277899e-05,
+      "loss": 4.0173,
+      "step": 406400
+    },
+    {
+      "epoch": 3.4990530303030303,
+      "grad_norm": 2.34375,
+      "learning_rate": 4.140178566542298e-05,
+      "loss": 4.4509,
+      "step": 406450
+    },
+    {
+      "epoch": 3.49948347107438,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.1379834489049806e-05,
+      "loss": 4.2229,
+      "step": 406500
+    },
+    {
+      "epoch": 3.49991391184573,
+      "grad_norm": 4.40625,
+      "learning_rate": 4.1357887615270676e-05,
+      "loss": 4.0933,
+      "step": 406550
+    },
+    {
+      "epoch": 3.5003443526170797,
+      "grad_norm": 2.171875,
+      "learning_rate": 4.133594504569641e-05,
+      "loss": 4.3448,
+      "step": 406600
+    },
+    {
+      "epoch": 3.50077479338843,
+      "grad_norm": 2.09375,
+      "learning_rate": 4.131400678193753e-05,
+      "loss": 4.1662,
+      "step": 406650
+    },
+    {
+      "epoch": 3.5012052341597797,
+      "grad_norm": 2.75,
+      "learning_rate": 4.129207282560431e-05,
+      "loss": 3.9231,
+      "step": 406700
+    },
+    {
+      "epoch": 3.5016356749311295,
+      "grad_norm": 3.0,
+      "learning_rate": 4.127014317830664e-05,
+      "loss": 4.3492,
+      "step": 406750
+    },
+    {
+      "epoch": 3.5020661157024793,
+      "grad_norm": 4.4375,
+      "learning_rate": 4.12482178416541e-05,
+      "loss": 4.5736,
+      "step": 406800
+    },
+    {
+      "epoch": 3.502496556473829,
+      "grad_norm": 4.34375,
+      "learning_rate": 4.122629681725597e-05,
+      "loss": 4.471,
+      "step": 406850
+    },
+    {
+      "epoch": 3.5029269972451793,
+      "grad_norm": 2.53125,
+      "learning_rate": 4.120438010672115e-05,
+      "loss": 4.046,
+      "step": 406900
+    },
+    {
+      "epoch": 3.503357438016529,
+      "grad_norm": 1.2734375,
+      "learning_rate": 4.118246771165838e-05,
+      "loss": 4.2873,
+      "step": 406950
+    },
+    {
+      "epoch": 3.503787878787879,
+      "grad_norm": 5.3125,
+      "learning_rate": 4.1160559633675954e-05,
+      "loss": 4.3485,
+      "step": 407000
+    },
+    {
+      "epoch": 3.5042183195592287,
+      "grad_norm": 2.84375,
+      "learning_rate": 4.113865587438178e-05,
+      "loss": 4.4273,
+      "step": 407050
+    },
+    {
+      "epoch": 3.5046487603305785,
+      "grad_norm": 2.734375,
+      "learning_rate": 4.111675643538366e-05,
+      "loss": 4.5408,
+      "step": 407100
+    },
+    {
+      "epoch": 3.5050792011019283,
+      "grad_norm": 3.625,
+      "learning_rate": 4.109486131828893e-05,
+      "loss": 4.4453,
+      "step": 407150
+    },
+    {
+      "epoch": 3.505509641873278,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.107297052470461e-05,
+      "loss": 4.3029,
+      "step": 407200
+    },
+    {
+      "epoch": 3.505940082644628,
+      "grad_norm": 4.8125,
+      "learning_rate": 4.10510840562375e-05,
+      "loss": 4.2998,
+      "step": 407250
+    },
+    {
+      "epoch": 3.506370523415978,
+      "grad_norm": 4.46875,
+      "learning_rate": 4.102920191449402e-05,
+      "loss": 4.3553,
+      "step": 407300
+    },
+    {
+      "epoch": 3.506800964187328,
+      "grad_norm": 4.71875,
+      "learning_rate": 4.100732410108018e-05,
+      "loss": 4.1658,
+      "step": 407350
+    },
+    {
+      "epoch": 3.5072314049586777,
+      "grad_norm": 1.3984375,
+      "learning_rate": 4.098545061760185e-05,
+      "loss": 4.2452,
+      "step": 407400
+    },
+    {
+      "epoch": 3.5076618457300275,
+      "grad_norm": 1.5625,
+      "learning_rate": 4.096358146566447e-05,
+      "loss": 4.3528,
+      "step": 407450
+    },
+    {
+      "epoch": 3.5080922865013773,
+      "grad_norm": 1.640625,
+      "learning_rate": 4.0941716646873165e-05,
+      "loss": 3.9162,
+      "step": 407500
+    },
+    {
+      "epoch": 3.5085227272727275,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.091985616283286e-05,
+      "loss": 4.3284,
+      "step": 407550
+    },
+    {
+      "epoch": 3.5089531680440773,
+      "grad_norm": 4.4375,
+      "learning_rate": 4.089800001514792e-05,
+      "loss": 4.3556,
+      "step": 407600
+    },
+    {
+      "epoch": 3.509383608815427,
+      "grad_norm": 1.265625,
+      "learning_rate": 4.087614820542266e-05,
+      "loss": 4.0045,
+      "step": 407650
+    },
+    {
+      "epoch": 3.509814049586777,
+      "grad_norm": 4.625,
+      "learning_rate": 4.085430073526091e-05,
+      "loss": 3.9775,
+      "step": 407700
+    },
+    {
+      "epoch": 3.5102444903581267,
+      "grad_norm": 4.78125,
+      "learning_rate": 4.083245760626619e-05,
+      "loss": 4.5104,
+      "step": 407750
+    },
+    {
+      "epoch": 3.5106749311294765,
+      "grad_norm": 6.3125,
+      "learning_rate": 4.081061882004186e-05,
+      "loss": 4.5451,
+      "step": 407800
+    },
+    {
+      "epoch": 3.5111053719008263,
+      "grad_norm": 3.015625,
+      "learning_rate": 4.0788784378190725e-05,
+      "loss": 3.7941,
+      "step": 407850
+    },
+    {
+      "epoch": 3.511535812672176,
+      "grad_norm": 3.859375,
+      "learning_rate": 4.0766954282315375e-05,
+      "loss": 4.0777,
+      "step": 407900
+    },
+    {
+      "epoch": 3.5119662534435263,
+      "grad_norm": 1.625,
+      "learning_rate": 4.074512853401818e-05,
+      "loss": 4.3275,
+      "step": 407950
+    },
+    {
+      "epoch": 3.512396694214876,
+      "grad_norm": 2.390625,
+      "learning_rate": 4.072330713490106e-05,
+      "loss": 4.2221,
+      "step": 408000
+    },
+    {
+      "epoch": 3.512396694214876,
+      "eval_loss": 4.992944240570068,
+      "eval_runtime": 24.6473,
+      "eval_samples_per_second": 25.966,
+      "eval_steps_per_second": 12.983,
+      "eval_tts_loss": 7.486227808068295,
+      "step": 408000
+    },
+    {
+      "epoch": 3.512827134986226,
+      "grad_norm": 1.890625,
+      "learning_rate": 4.0701490086565665e-05,
+      "loss": 4.4555,
+      "step": 408050
+    },
+    {
+      "epoch": 3.5132575757575757,
+      "grad_norm": 5.09375,
+      "learning_rate": 4.06796773906133e-05,
+      "loss": 4.5868,
+      "step": 408100
+    },
+    {
+      "epoch": 3.5136880165289255,
+      "grad_norm": 3.984375,
+      "learning_rate": 4.065786904864499e-05,
+      "loss": 4.5604,
+      "step": 408150
+    },
+    {
+      "epoch": 3.5141184573002757,
+      "grad_norm": 2.671875,
+      "learning_rate": 4.063606506226137e-05,
+      "loss": 4.4013,
+      "step": 408200
+    },
+    {
+      "epoch": 3.5145488980716255,
+      "grad_norm": 2.5,
+      "learning_rate": 4.06142654330629e-05,
+      "loss": 4.4633,
+      "step": 408250
+    },
+    {
+      "epoch": 3.5149793388429753,
+      "grad_norm": 3.296875,
+      "learning_rate": 4.0592470162649577e-05,
+      "loss": 4.115,
+      "step": 408300
+    },
+    {
+      "epoch": 3.515409779614325,
+      "grad_norm": 3.234375,
+      "learning_rate": 4.0570679252621116e-05,
+      "loss": 4.6039,
+      "step": 408350
+    },
+    {
+      "epoch": 3.515840220385675,
+      "grad_norm": 3.171875,
+      "learning_rate": 4.054889270457693e-05,
+      "loss": 4.0557,
+      "step": 408400
+    },
+    {
+      "epoch": 3.5162706611570247,
+      "grad_norm": 1.6640625,
+      "learning_rate": 4.0527110520116076e-05,
+      "loss": 4.4601,
+      "step": 408450
+    },
+    {
+      "epoch": 3.5167011019283745,
+      "grad_norm": 2.90625,
+      "learning_rate": 4.0505332700837386e-05,
+      "loss": 4.0398,
+      "step": 408500
+    },
+    {
+      "epoch": 3.5171315426997243,
+      "grad_norm": 4.25,
+      "learning_rate": 4.048355924833931e-05,
+      "loss": 4.4516,
+      "step": 408550
+    },
+    {
+      "epoch": 3.5175619834710745,
+      "grad_norm": 3.40625,
+      "learning_rate": 4.0461790164219846e-05,
+      "loss": 4.5029,
+      "step": 408600
+    },
+    {
+      "epoch": 3.5179924242424243,
+      "grad_norm": 1.9375,
+      "learning_rate": 4.0440025450076934e-05,
+      "loss": 4.0985,
+      "step": 408650
+    },
+    {
+      "epoch": 3.518422865013774,
+      "grad_norm": 4.03125,
+      "learning_rate": 4.0418265107508e-05,
+      "loss": 3.9489,
+      "step": 408700
+    },
+    {
+      "epoch": 3.518853305785124,
+      "grad_norm": 4.3125,
+      "learning_rate": 4.039650913811018e-05,
+      "loss": 4.624,
+      "step": 408750
+    },
+    {
+      "epoch": 3.5192837465564737,
+      "grad_norm": 4.5,
+      "learning_rate": 4.037475754348042e-05,
+      "loss": 4.6938,
+      "step": 408800
+    },
+    {
+      "epoch": 3.519714187327824,
+      "grad_norm": 3.59375,
+      "learning_rate": 4.035301032521515e-05,
+      "loss": 4.0894,
+      "step": 408850
+    },
+    {
+      "epoch": 3.5201446280991737,
+      "grad_norm": 3.59375,
+      "learning_rate": 4.033126748491053e-05,
+      "loss": 4.375,
+      "step": 408900
+    },
+    {
+      "epoch": 3.5205750688705235,
+      "grad_norm": 6.53125,
+      "learning_rate": 4.030952902416254e-05,
+      "loss": 4.6975,
+      "step": 408950
+    },
+    {
+      "epoch": 3.5210055096418733,
+      "grad_norm": 3.640625,
+      "learning_rate": 4.028779494456669e-05,
+      "loss": 4.6206,
+      "step": 409000
+    },
+    {
+      "epoch": 3.521435950413223,
+      "grad_norm": 2.40625,
+      "learning_rate": 4.026606524771819e-05,
+      "loss": 4.1971,
+      "step": 409050
+    },
+    {
+      "epoch": 3.521866391184573,
+      "grad_norm": 1.734375,
+      "learning_rate": 4.024433993521204e-05,
+      "loss": 4.042,
+      "step": 409100
+    },
+    {
+      "epoch": 3.5222968319559227,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.0222619008642706e-05,
+      "loss": 4.1495,
+      "step": 409150
+    },
+    {
+      "epoch": 3.5227272727272725,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.020090246960456e-05,
+      "loss": 4.449,
+      "step": 409200
+    },
+    {
+      "epoch": 3.5231577134986227,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.017919031969151e-05,
+      "loss": 4.4549,
+      "step": 409250
+    },
+    {
+      "epoch": 3.5235881542699725,
+      "grad_norm": 1.7109375,
+      "learning_rate": 4.015748256049717e-05,
+      "loss": 4.3652,
+      "step": 409300
+    },
+    {
+      "epoch": 3.5240185950413223,
+      "grad_norm": 2.234375,
+      "learning_rate": 4.013577919361488e-05,
+      "loss": 4.4345,
+      "step": 409350
+    },
+    {
+      "epoch": 3.524449035812672,
+      "grad_norm": 3.921875,
+      "learning_rate": 4.0114080220637576e-05,
+      "loss": 4.5185,
+      "step": 409400
+    },
+    {
+      "epoch": 3.524879476584022,
+      "grad_norm": 1.6796875,
+      "learning_rate": 4.0092385643157905e-05,
+      "loss": 4.3577,
+      "step": 409450
+    },
+    {
+      "epoch": 3.525309917355372,
+      "grad_norm": 3.40625,
+      "learning_rate": 4.0070695462768274e-05,
+      "loss": 4.2808,
+      "step": 409500
+    },
+    {
+      "epoch": 3.525740358126722,
+      "grad_norm": 6.34375,
+      "learning_rate": 4.004900968106064e-05,
+      "loss": 4.2734,
+      "step": 409550
+    },
+    {
+      "epoch": 3.5261707988980717,
+      "grad_norm": 2.328125,
+      "learning_rate": 4.002732829962673e-05,
+      "loss": 4.27,
+      "step": 409600
+    },
+    {
+      "epoch": 3.5266012396694215,
+      "grad_norm": 2.609375,
+      "learning_rate": 4.000565132005788e-05,
+      "loss": 4.4383,
+      "step": 409650
+    },
+    {
+      "epoch": 3.5270316804407713,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.9983978743945136e-05,
+      "loss": 3.9027,
+      "step": 409700
+    },
+    {
+      "epoch": 3.527462121212121,
+      "grad_norm": 2.765625,
+      "learning_rate": 3.996231057287919e-05,
+      "loss": 4.0727,
+      "step": 409750
+    },
+    {
+      "epoch": 3.527892561983471,
+      "grad_norm": 2.046875,
+      "learning_rate": 3.994064680845051e-05,
+      "loss": 4.214,
+      "step": 409800
+    },
+    {
+      "epoch": 3.5283230027548207,
+      "grad_norm": 3.09375,
+      "learning_rate": 3.991898745224913e-05,
+      "loss": 4.6483,
+      "step": 409850
+    },
+    {
+      "epoch": 3.528753443526171,
+      "grad_norm": 2.96875,
+      "learning_rate": 3.9897332505864816e-05,
+      "loss": 4.1834,
+      "step": 409900
+    },
+    {
+      "epoch": 3.5291838842975207,
+      "grad_norm": 2.265625,
+      "learning_rate": 3.987568197088697e-05,
+      "loss": 4.1743,
+      "step": 409950
+    },
+    {
+      "epoch": 3.5296143250688705,
+      "grad_norm": 3.390625,
+      "learning_rate": 3.985403584890468e-05,
+      "loss": 4.3048,
+      "step": 410000
+    },
+    {
+      "epoch": 3.5300447658402203,
+      "grad_norm": 4.40625,
+      "learning_rate": 3.983239414150678e-05,
+      "loss": 4.2697,
+      "step": 410050
+    },
+    {
+      "epoch": 3.53047520661157,
+      "grad_norm": 2.265625,
+      "learning_rate": 3.9810756850281727e-05,
+      "loss": 4.031,
+      "step": 410100
+    },
+    {
+      "epoch": 3.5309056473829203,
+      "grad_norm": 2.296875,
+      "learning_rate": 3.9789123976817556e-05,
+      "loss": 4.0189,
+      "step": 410150
+    },
+    {
+      "epoch": 3.53133608815427,
+      "grad_norm": 2.375,
+      "learning_rate": 3.976749552270218e-05,
+      "loss": 4.4259,
+      "step": 410200
+    },
+    {
+      "epoch": 3.53176652892562,
+      "grad_norm": 2.296875,
+      "learning_rate": 3.974587148952302e-05,
+      "loss": 4.2886,
+      "step": 410250
+    },
+    {
+      "epoch": 3.5321969696969697,
+      "grad_norm": 4.375,
+      "learning_rate": 3.972425187886724e-05,
+      "loss": 4.5847,
+      "step": 410300
+    },
+    {
+      "epoch": 3.5326274104683195,
+      "grad_norm": 3.359375,
+      "learning_rate": 3.970263669232175e-05,
+      "loss": 4.3887,
+      "step": 410350
+    },
+    {
+      "epoch": 3.5330578512396693,
+      "grad_norm": 2.515625,
+      "learning_rate": 3.968102593147295e-05,
+      "loss": 4.2217,
+      "step": 410400
+    },
+    {
+      "epoch": 3.533488292011019,
+      "grad_norm": 3.21875,
+      "learning_rate": 3.965941959790703e-05,
+      "loss": 3.98,
+      "step": 410450
+    },
+    {
+      "epoch": 3.533918732782369,
+      "grad_norm": 3.609375,
+      "learning_rate": 3.9637817693209936e-05,
+      "loss": 4.2807,
+      "step": 410500
+    },
+    {
+      "epoch": 3.534349173553719,
+      "grad_norm": 1.484375,
+      "learning_rate": 3.9616220218967134e-05,
+      "loss": 4.2255,
+      "step": 410550
+    },
+    {
+      "epoch": 3.534779614325069,
+      "grad_norm": 2.71875,
+      "learning_rate": 3.959462717676385e-05,
+      "loss": 4.3583,
+      "step": 410600
+    },
+    {
+      "epoch": 3.5352100550964187,
+      "grad_norm": 5.21875,
+      "learning_rate": 3.957303856818497e-05,
+      "loss": 4.314,
+      "step": 410650
+    },
+    {
+      "epoch": 3.5356404958677685,
+      "grad_norm": 4.65625,
+      "learning_rate": 3.955145439481502e-05,
+      "loss": 4.5294,
+      "step": 410700
+    },
+    {
+      "epoch": 3.5360709366391183,
+      "grad_norm": 3.84375,
+      "learning_rate": 3.9529874658238287e-05,
+      "loss": 4.3699,
+      "step": 410750
+    },
+    {
+      "epoch": 3.5365013774104685,
+      "grad_norm": 3.5625,
+      "learning_rate": 3.9508299360038645e-05,
+      "loss": 4.4296,
+      "step": 410800
+    },
+    {
+      "epoch": 3.5369318181818183,
+      "grad_norm": 1.9453125,
+      "learning_rate": 3.9486728501799685e-05,
+      "loss": 4.2292,
+      "step": 410850
+    },
+    {
+      "epoch": 3.537362258953168,
+      "grad_norm": 3.328125,
+      "learning_rate": 3.946516208510466e-05,
+      "loss": 4.158,
+      "step": 410900
+    },
+    {
+      "epoch": 3.537792699724518,
+      "grad_norm": 2.71875,
+      "learning_rate": 3.944360011153648e-05,
+      "loss": 4.2304,
+      "step": 410950
+    },
+    {
+      "epoch": 3.5382231404958677,
+      "grad_norm": 5.3125,
+      "learning_rate": 3.9422042582677743e-05,
+      "loss": 4.4766,
+      "step": 411000
+    },
+    {
+      "epoch": 3.5382231404958677,
+      "eval_loss": 4.992468357086182,
+      "eval_runtime": 23.829,
+      "eval_samples_per_second": 26.858,
+      "eval_steps_per_second": 13.429,
+      "eval_tts_loss": 7.48197030355066,
+      "step": 411000
+    },
+    {
+      "epoch": 3.5386535812672175,
+      "grad_norm": 2.859375,
+      "learning_rate": 3.940048950011077e-05,
+      "loss": 4.786,
+      "step": 411050
+    },
+    {
+      "epoch": 3.5390840220385673,
+      "grad_norm": 2.671875,
+      "learning_rate": 3.937894086541748e-05,
+      "loss": 4.2958,
+      "step": 411100
+    },
+    {
+      "epoch": 3.539514462809917,
+      "grad_norm": 4.34375,
+      "learning_rate": 3.9357396680179504e-05,
+      "loss": 4.6541,
+      "step": 411150
+    },
+    {
+      "epoch": 3.5399449035812673,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.933585694597814e-05,
+      "loss": 4.1217,
+      "step": 411200
+    },
+    {
+      "epoch": 3.540375344352617,
+      "grad_norm": 3.40625,
+      "learning_rate": 3.931432166439435e-05,
+      "loss": 4.431,
+      "step": 411250
+    },
+    {
+      "epoch": 3.540805785123967,
+      "grad_norm": 3.265625,
+      "learning_rate": 3.929279083700875e-05,
+      "loss": 4.3732,
+      "step": 411300
+    },
+    {
+      "epoch": 3.5412362258953167,
+      "grad_norm": 3.015625,
+      "learning_rate": 3.9271264465401724e-05,
+      "loss": 4.3619,
+      "step": 411350
+    },
+    {
+      "epoch": 3.5416666666666665,
+      "grad_norm": 1.265625,
+      "learning_rate": 3.924974255115321e-05,
+      "loss": 4.2284,
+      "step": 411400
+    },
+    {
+      "epoch": 3.5420971074380168,
+      "grad_norm": 3.3125,
+      "learning_rate": 3.922822509584291e-05,
+      "loss": 4.2226,
+      "step": 411450
+    },
+    {
+      "epoch": 3.5425275482093666,
+      "grad_norm": 4.375,
+      "learning_rate": 3.9206712101050114e-05,
+      "loss": 4.5041,
+      "step": 411500
+    },
+    {
+      "epoch": 3.5429579889807163,
+      "grad_norm": 3.953125,
+      "learning_rate": 3.9185203568353856e-05,
+      "loss": 4.0905,
+      "step": 411550
+    },
+    {
+      "epoch": 3.543388429752066,
+      "grad_norm": 4.03125,
+      "learning_rate": 3.9163699499332776e-05,
+      "loss": 4.6507,
+      "step": 411600
+    },
+    {
+      "epoch": 3.543818870523416,
+      "grad_norm": 3.203125,
+      "learning_rate": 3.914219989556532e-05,
+      "loss": 4.4417,
+      "step": 411650
+    },
+    {
+      "epoch": 3.5442493112947657,
+      "grad_norm": 1.9765625,
+      "learning_rate": 3.91207047586294e-05,
+      "loss": 4.3321,
+      "step": 411700
+    },
+    {
+      "epoch": 3.5446797520661155,
+      "grad_norm": 4.71875,
+      "learning_rate": 3.909921409010279e-05,
+      "loss": 4.1627,
+      "step": 411750
+    },
+    {
+      "epoch": 3.5451101928374653,
+      "grad_norm": 2.8125,
+      "learning_rate": 3.907772789156283e-05,
+      "loss": 4.2878,
+      "step": 411800
+    },
+    {
+      "epoch": 3.5455406336088156,
+      "grad_norm": 2.625,
+      "learning_rate": 3.905624616458654e-05,
+      "loss": 3.9459,
+      "step": 411850
+    },
+    {
+      "epoch": 3.5459710743801653,
+      "grad_norm": 2.765625,
+      "learning_rate": 3.903476891075073e-05,
+      "loss": 4.1009,
+      "step": 411900
+    },
+    {
+      "epoch": 3.546401515151515,
+      "grad_norm": 1.40625,
+      "learning_rate": 3.901329613163166e-05,
+      "loss": 4.3531,
+      "step": 411950
+    },
+    {
+      "epoch": 3.546831955922865,
+      "grad_norm": 4.375,
+      "learning_rate": 3.899182782880542e-05,
+      "loss": 4.6201,
+      "step": 412000
+    },
+    {
+      "epoch": 3.5472623966942147,
+      "grad_norm": 4.5625,
+      "learning_rate": 3.897036400384778e-05,
+      "loss": 4.2447,
+      "step": 412050
+    },
+    {
+      "epoch": 3.547692837465565,
+      "grad_norm": 2.3125,
+      "learning_rate": 3.894890465833412e-05,
+      "loss": 4.6572,
+      "step": 412100
+    },
+    {
+      "epoch": 3.5481232782369148,
+      "grad_norm": 6.46875,
+      "learning_rate": 3.89274497938395e-05,
+      "loss": 4.2096,
+      "step": 412150
+    },
+    {
+      "epoch": 3.5485537190082646,
+      "grad_norm": 7.5,
+      "learning_rate": 3.8905999411938665e-05,
+      "loss": 4.158,
+      "step": 412200
+    },
+    {
+      "epoch": 3.5489841597796143,
+      "grad_norm": 3.53125,
+      "learning_rate": 3.888455351420602e-05,
+      "loss": 4.4452,
+      "step": 412250
+    },
+    {
+      "epoch": 3.549414600550964,
+      "grad_norm": 2.03125,
+      "learning_rate": 3.886311210221562e-05,
+      "loss": 4.3399,
+      "step": 412300
+    },
+    {
+      "epoch": 3.549845041322314,
+      "grad_norm": 1.578125,
+      "learning_rate": 3.88416751775413e-05,
+      "loss": 4.2448,
+      "step": 412350
+    },
+    {
+      "epoch": 3.5502754820936637,
+      "grad_norm": 3.421875,
+      "learning_rate": 3.882024274175643e-05,
+      "loss": 4.5586,
+      "step": 412400
+    },
+    {
+      "epoch": 3.5507059228650135,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.87988147964341e-05,
+      "loss": 4.7373,
+      "step": 412450
+    },
+    {
+      "epoch": 3.5511363636363638,
+      "grad_norm": 4.28125,
+      "learning_rate": 3.8777391343147095e-05,
+      "loss": 4.2711,
+      "step": 412500
+    },
+    {
+      "epoch": 3.5515668044077136,
+      "grad_norm": 5.84375,
+      "learning_rate": 3.875597238346781e-05,
+      "loss": 4.3551,
+      "step": 412550
+    },
+    {
+      "epoch": 3.5519972451790633,
+      "grad_norm": 4.1875,
+      "learning_rate": 3.8734557918968415e-05,
+      "loss": 4.3224,
+      "step": 412600
+    },
+    {
+      "epoch": 3.552427685950413,
+      "grad_norm": 2.71875,
+      "learning_rate": 3.871314795122065e-05,
+      "loss": 4.2193,
+      "step": 412650
+    },
+    {
+      "epoch": 3.552858126721763,
+      "grad_norm": 1.703125,
+      "learning_rate": 3.869174248179597e-05,
+      "loss": 4.3477,
+      "step": 412700
+    },
+    {
+      "epoch": 3.553288567493113,
+      "grad_norm": 2.671875,
+      "learning_rate": 3.8670341512265476e-05,
+      "loss": 4.2872,
+      "step": 412750
+    },
+    {
+      "epoch": 3.553719008264463,
+      "grad_norm": 4.71875,
+      "learning_rate": 3.864894504419996e-05,
+      "loss": 4.5247,
+      "step": 412800
+    },
+    {
+      "epoch": 3.5541494490358128,
+      "grad_norm": 5.28125,
+      "learning_rate": 3.862755307916985e-05,
+      "loss": 4.4421,
+      "step": 412850
+    },
+    {
+      "epoch": 3.5545798898071626,
+      "grad_norm": 1.53125,
+      "learning_rate": 3.8606165618745324e-05,
+      "loss": 4.1341,
+      "step": 412900
+    },
+    {
+      "epoch": 3.5550103305785123,
+      "grad_norm": 2.859375,
+      "learning_rate": 3.8584782664496156e-05,
+      "loss": 4.5346,
+      "step": 412950
+    },
+    {
+      "epoch": 3.555440771349862,
+      "grad_norm": 5.75,
+      "learning_rate": 3.8563404217991805e-05,
+      "loss": 4.103,
+      "step": 413000
+    },
+    {
+      "epoch": 3.555871212121212,
+      "grad_norm": 2.875,
+      "learning_rate": 3.854203028080139e-05,
+      "loss": 4.3643,
+      "step": 413050
+    },
+    {
+      "epoch": 3.5563016528925617,
+      "grad_norm": 2.8125,
+      "learning_rate": 3.852066085449373e-05,
+      "loss": 4.3044,
+      "step": 413100
+    },
+    {
+      "epoch": 3.556732093663912,
+      "grad_norm": 2.75,
+      "learning_rate": 3.849929594063725e-05,
+      "loss": 4.7237,
+      "step": 413150
+    },
+    {
+      "epoch": 3.5571625344352618,
+      "grad_norm": 2.34375,
+      "learning_rate": 3.8477935540800195e-05,
+      "loss": 4.3593,
+      "step": 413200
+    },
+    {
+      "epoch": 3.5575929752066116,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.845657965655023e-05,
+      "loss": 4.0331,
+      "step": 413250
+    },
+    {
+      "epoch": 3.5580234159779613,
+      "grad_norm": 2.125,
+      "learning_rate": 3.843522828945495e-05,
+      "loss": 4.5532,
+      "step": 413300
+    },
+    {
+      "epoch": 3.558453856749311,
+      "grad_norm": 3.09375,
+      "learning_rate": 3.8413881441081446e-05,
+      "loss": 4.3786,
+      "step": 413350
+    },
+    {
+      "epoch": 3.5588842975206614,
+      "grad_norm": 5.03125,
+      "learning_rate": 3.8392539112996494e-05,
+      "loss": 4.1899,
+      "step": 413400
+    },
+    {
+      "epoch": 3.559314738292011,
+      "grad_norm": 3.59375,
+      "learning_rate": 3.83712013067667e-05,
+      "loss": 4.7605,
+      "step": 413450
+    },
+    {
+      "epoch": 3.559745179063361,
+      "grad_norm": 5.78125,
+      "learning_rate": 3.83498680239581e-05,
+      "loss": 3.7052,
+      "step": 413500
+    },
+    {
+      "epoch": 3.5601756198347108,
+      "grad_norm": 5.125,
+      "learning_rate": 3.8328539266136496e-05,
+      "loss": 4.6599,
+      "step": 413550
+    },
+    {
+      "epoch": 3.5606060606060606,
+      "grad_norm": 2.53125,
+      "learning_rate": 3.8307215034867473e-05,
+      "loss": 4.3137,
+      "step": 413600
+    },
+    {
+      "epoch": 3.5610365013774103,
+      "grad_norm": 2.28125,
+      "learning_rate": 3.828589533171611e-05,
+      "loss": 4.1429,
+      "step": 413650
+    },
+    {
+      "epoch": 3.56146694214876,
+      "grad_norm": 1.8671875,
+      "learning_rate": 3.826458015824727e-05,
+      "loss": 4.2299,
+      "step": 413700
+    },
+    {
+      "epoch": 3.56189738292011,
+      "grad_norm": 3.40625,
+      "learning_rate": 3.82432695160254e-05,
+      "loss": 4.3519,
+      "step": 413750
+    },
+    {
+      "epoch": 3.56232782369146,
+      "grad_norm": 2.453125,
+      "learning_rate": 3.822196340661469e-05,
+      "loss": 4.2761,
+      "step": 413800
+    },
+    {
+      "epoch": 3.56275826446281,
+      "grad_norm": 3.109375,
+      "learning_rate": 3.820066183157891e-05,
+      "loss": 4.4955,
+      "step": 413850
+    },
+    {
+      "epoch": 3.5631887052341598,
+      "grad_norm": 4.03125,
+      "learning_rate": 3.817936479248162e-05,
+      "loss": 4.0771,
+      "step": 413900
+    },
+    {
+      "epoch": 3.5636191460055096,
+      "grad_norm": 3.421875,
+      "learning_rate": 3.815807229088594e-05,
+      "loss": 4.1555,
+      "step": 413950
+    },
+    {
+      "epoch": 3.5640495867768593,
+      "grad_norm": 3.65625,
+      "learning_rate": 3.8136784328354694e-05,
+      "loss": 4.2392,
+      "step": 414000
+    },
+    {
+      "epoch": 3.5640495867768593,
+      "eval_loss": 4.993220329284668,
+      "eval_runtime": 23.761,
+      "eval_samples_per_second": 26.935,
+      "eval_steps_per_second": 13.467,
+      "eval_tts_loss": 7.490751677720266,
+      "step": 414000
+    },
+    {
+      "epoch": 3.5644800275482096,
+      "grad_norm": 3.09375,
+      "learning_rate": 3.811550090645039e-05,
+      "loss": 4.0995,
+      "step": 414050
+    },
+    {
+      "epoch": 3.5649104683195594,
+      "grad_norm": 4.625,
+      "learning_rate": 3.8094222026735117e-05,
+      "loss": 4.399,
+      "step": 414100
+    },
+    {
+      "epoch": 3.565340909090909,
+      "grad_norm": 3.5625,
+      "learning_rate": 3.80729476907708e-05,
+      "loss": 4.14,
+      "step": 414150
+    },
+    {
+      "epoch": 3.565771349862259,
+      "grad_norm": 3.0625,
+      "learning_rate": 3.805167790011892e-05,
+      "loss": 4.3395,
+      "step": 414200
+    },
+    {
+      "epoch": 3.5662017906336088,
+      "grad_norm": 2.765625,
+      "learning_rate": 3.803041265634052e-05,
+      "loss": 4.046,
+      "step": 414250
+    },
+    {
+      "epoch": 3.5666322314049586,
+      "grad_norm": 1.3046875,
+      "learning_rate": 3.800915196099655e-05,
+      "loss": 4.0017,
+      "step": 414300
+    },
+    {
+      "epoch": 3.5670626721763083,
+      "grad_norm": 5.5,
+      "learning_rate": 3.798789581564743e-05,
+      "loss": 4.5304,
+      "step": 414350
+    },
+    {
+      "epoch": 3.567493112947658,
+      "grad_norm": 3.921875,
+      "learning_rate": 3.796664422185331e-05,
+      "loss": 4.2132,
+      "step": 414400
+    },
+    {
+      "epoch": 3.5679235537190084,
+      "grad_norm": 3.375,
+      "learning_rate": 3.794539718117407e-05,
+      "loss": 4.2827,
+      "step": 414450
+    },
+    {
+      "epoch": 3.568353994490358,
+      "grad_norm": 2.875,
+      "learning_rate": 3.792415469516919e-05,
+      "loss": 4.3844,
+      "step": 414500
+    },
+    {
+      "epoch": 3.568784435261708,
+      "grad_norm": 3.171875,
+      "learning_rate": 3.790291676539773e-05,
+      "loss": 4.9016,
+      "step": 414550
+    },
+    {
+      "epoch": 3.5692148760330578,
+      "grad_norm": 0.5859375,
+      "learning_rate": 3.788168339341862e-05,
+      "loss": 4.4461,
+      "step": 414600
+    },
+    {
+      "epoch": 3.5696453168044076,
+      "grad_norm": 1.828125,
+      "learning_rate": 3.786045458079028e-05,
+      "loss": 4.2271,
+      "step": 414650
+    },
+    {
+      "epoch": 3.570075757575758,
+      "grad_norm": 3.515625,
+      "learning_rate": 3.783923032907086e-05,
+      "loss": 4.3428,
+      "step": 414700
+    },
+    {
+      "epoch": 3.5705061983471076,
+      "grad_norm": 2.375,
+      "learning_rate": 3.781801063981826e-05,
+      "loss": 4.5283,
+      "step": 414750
+    },
+    {
+      "epoch": 3.5709366391184574,
+      "grad_norm": 3.25,
+      "learning_rate": 3.779679551458981e-05,
+      "loss": 4.3794,
+      "step": 414800
+    },
+    {
+      "epoch": 3.571367079889807,
+      "grad_norm": 4.8125,
+      "learning_rate": 3.7775584954942764e-05,
+      "loss": 4.3613,
+      "step": 414850
+    },
+    {
+      "epoch": 3.571797520661157,
+      "grad_norm": 5.1875,
+      "learning_rate": 3.775437896243392e-05,
+      "loss": 4.2454,
+      "step": 414900
+    },
+    {
+      "epoch": 3.5722279614325068,
+      "grad_norm": 3.203125,
+      "learning_rate": 3.773317753861969e-05,
+      "loss": 4.3172,
+      "step": 414950
+    },
+    {
+      "epoch": 3.5726584022038566,
+      "grad_norm": 2.859375,
+      "learning_rate": 3.7711980685056334e-05,
+      "loss": 4.1723,
+      "step": 415000
+    },
+    {
+      "epoch": 3.5730888429752063,
+      "grad_norm": 2.78125,
+      "learning_rate": 3.769078840329955e-05,
+      "loss": 4.3558,
+      "step": 415050
+    },
+    {
+      "epoch": 3.5735192837465566,
+      "grad_norm": 4.03125,
+      "learning_rate": 3.7669600694904796e-05,
+      "loss": 4.5202,
+      "step": 415100
+    },
+    {
+      "epoch": 3.5739497245179064,
+      "grad_norm": 8.5625,
+      "learning_rate": 3.764841756142727e-05,
+      "loss": 4.2215,
+      "step": 415150
+    },
+    {
+      "epoch": 3.574380165289256,
+      "grad_norm": 5.0625,
+      "learning_rate": 3.762723900442175e-05,
+      "loss": 3.8953,
+      "step": 415200
+    },
+    {
+      "epoch": 3.574810606060606,
+      "grad_norm": 4.5625,
+      "learning_rate": 3.7606065025442684e-05,
+      "loss": 4.4706,
+      "step": 415250
+    },
+    {
+      "epoch": 3.5752410468319558,
+      "grad_norm": 4.78125,
+      "learning_rate": 3.7584895626044206e-05,
+      "loss": 4.4278,
+      "step": 415300
+    },
+    {
+      "epoch": 3.575671487603306,
+      "grad_norm": 3.328125,
+      "learning_rate": 3.7563730807780095e-05,
+      "loss": 4.1582,
+      "step": 415350
+    },
+    {
+      "epoch": 3.576101928374656,
+      "grad_norm": 3.8125,
+      "learning_rate": 3.754257057220377e-05,
+      "loss": 4.0407,
+      "step": 415400
+    },
+    {
+      "epoch": 3.5765323691460056,
+      "grad_norm": 6.15625,
+      "learning_rate": 3.7521414920868424e-05,
+      "loss": 4.5944,
+      "step": 415450
+    },
+    {
+      "epoch": 3.5769628099173554,
+      "grad_norm": 2.515625,
+      "learning_rate": 3.75002638553268e-05,
+      "loss": 4.7636,
+      "step": 415500
+    },
+    {
+      "epoch": 3.577393250688705,
+      "grad_norm": 1.5546875,
+      "learning_rate": 3.7479117377131335e-05,
+      "loss": 4.4981,
+      "step": 415550
+    },
+    {
+      "epoch": 3.577823691460055,
+      "grad_norm": 2.15625,
+      "learning_rate": 3.745797548783414e-05,
+      "loss": 4.2684,
+      "step": 415600
+    },
+    {
+      "epoch": 3.5782541322314048,
+      "grad_norm": 3.8125,
+      "learning_rate": 3.7436838188986946e-05,
+      "loss": 4.6423,
+      "step": 415650
+    },
+    {
+      "epoch": 3.5786845730027546,
+      "grad_norm": 2.65625,
+      "learning_rate": 3.741570548214125e-05,
+      "loss": 4.3073,
+      "step": 415700
+    },
+    {
+      "epoch": 3.579115013774105,
+      "grad_norm": 2.0,
+      "learning_rate": 3.739457736884815e-05,
+      "loss": 4.863,
+      "step": 415750
+    },
+    {
+      "epoch": 3.5795454545454546,
+      "grad_norm": 2.765625,
+      "learning_rate": 3.737345385065831e-05,
+      "loss": 4.546,
+      "step": 415800
+    },
+    {
+      "epoch": 3.5799758953168044,
+      "grad_norm": 1.8359375,
+      "learning_rate": 3.735233492912225e-05,
+      "loss": 4.3051,
+      "step": 415850
+    },
+    {
+      "epoch": 3.580406336088154,
+      "grad_norm": 5.78125,
+      "learning_rate": 3.733122060579001e-05,
+      "loss": 4.361,
+      "step": 415900
+    },
+    {
+      "epoch": 3.580836776859504,
+      "grad_norm": 5.28125,
+      "learning_rate": 3.731011088221131e-05,
+      "loss": 4.1161,
+      "step": 415950
+    },
+    {
+      "epoch": 3.581267217630854,
+      "grad_norm": 4.5625,
+      "learning_rate": 3.7289005759935655e-05,
+      "loss": 4.4665,
+      "step": 416000
+    },
+    {
+      "epoch": 3.581697658402204,
+      "grad_norm": 3.03125,
+      "learning_rate": 3.726790524051203e-05,
+      "loss": 4.4466,
+      "step": 416050
+    },
+    {
+      "epoch": 3.582128099173554,
+      "grad_norm": 2.609375,
+      "learning_rate": 3.724680932548914e-05,
+      "loss": 4.4199,
+      "step": 416100
+    },
+    {
+      "epoch": 3.5825585399449036,
+      "grad_norm": 3.0625,
+      "learning_rate": 3.7225718016415466e-05,
+      "loss": 4.2236,
+      "step": 416150
+    },
+    {
+      "epoch": 3.5829889807162534,
+      "grad_norm": 3.03125,
+      "learning_rate": 3.7204631314839034e-05,
+      "loss": 4.3118,
+      "step": 416200
+    },
+    {
+      "epoch": 3.583419421487603,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.718354922230751e-05,
+      "loss": 4.0586,
+      "step": 416250
+    },
+    {
+      "epoch": 3.583849862258953,
+      "grad_norm": 2.4375,
+      "learning_rate": 3.71624717403684e-05,
+      "loss": 4.3164,
+      "step": 416300
+    },
+    {
+      "epoch": 3.5842803030303028,
+      "grad_norm": 4.25,
+      "learning_rate": 3.714139887056859e-05,
+      "loss": 4.0762,
+      "step": 416350
+    },
+    {
+      "epoch": 3.584710743801653,
+      "grad_norm": 3.40625,
+      "learning_rate": 3.7120330614454905e-05,
+      "loss": 4.4412,
+      "step": 416400
+    },
+    {
+      "epoch": 3.585141184573003,
+      "grad_norm": 4.8125,
+      "learning_rate": 3.7099266973573645e-05,
+      "loss": 4.0982,
+      "step": 416450
+    },
+    {
+      "epoch": 3.5855716253443526,
+      "grad_norm": 4.96875,
+      "learning_rate": 3.707820794947087e-05,
+      "loss": 4.4131,
+      "step": 416500
+    },
+    {
+      "epoch": 3.5860020661157024,
+      "grad_norm": 3.734375,
+      "learning_rate": 3.705715354369225e-05,
+      "loss": 4.4848,
+      "step": 416550
+    },
+    {
+      "epoch": 3.5864325068870526,
+      "grad_norm": 1.3671875,
+      "learning_rate": 3.703610375778312e-05,
+      "loss": 4.0339,
+      "step": 416600
+    },
+    {
+      "epoch": 3.5868629476584024,
+      "grad_norm": 2.859375,
+      "learning_rate": 3.701505859328849e-05,
+      "loss": 4.3279,
+      "step": 416650
+    },
+    {
+      "epoch": 3.587293388429752,
+      "grad_norm": 4.875,
+      "learning_rate": 3.6994018051753064e-05,
+      "loss": 4.1989,
+      "step": 416700
+    },
+    {
+      "epoch": 3.587723829201102,
+      "grad_norm": 2.875,
+      "learning_rate": 3.697298213472116e-05,
+      "loss": 4.5478,
+      "step": 416750
+    },
+    {
+      "epoch": 3.588154269972452,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.695195084373676e-05,
+      "loss": 4.4153,
+      "step": 416800
+    },
+    {
+      "epoch": 3.5885847107438016,
+      "grad_norm": 3.203125,
+      "learning_rate": 3.693092418034351e-05,
+      "loss": 4.0509,
+      "step": 416850
+    },
+    {
+      "epoch": 3.5890151515151514,
+      "grad_norm": 2.328125,
+      "learning_rate": 3.690990214608473e-05,
+      "loss": 4.5398,
+      "step": 416900
+    },
+    {
+      "epoch": 3.589445592286501,
+      "grad_norm": 3.234375,
+      "learning_rate": 3.688888474250336e-05,
+      "loss": 4.135,
+      "step": 416950
+    },
+    {
+      "epoch": 3.589876033057851,
+      "grad_norm": 2.453125,
+      "learning_rate": 3.68678719711421e-05,
+      "loss": 4.402,
+      "step": 417000
+    },
+    {
+      "epoch": 3.589876033057851,
+      "eval_loss": 4.99120569229126,
+      "eval_runtime": 24.1585,
+      "eval_samples_per_second": 26.492,
+      "eval_steps_per_second": 13.246,
+      "eval_tts_loss": 7.476320691126799,
+      "step": 417000
+    },
+    {
+      "epoch": 3.590306473829201,
+      "grad_norm": 2.65625,
+      "learning_rate": 3.68468638335432e-05,
+      "loss": 4.7107,
+      "step": 417050
+    },
+    {
+      "epoch": 3.590736914600551,
+      "grad_norm": 3.890625,
+      "learning_rate": 3.682586033124862e-05,
+      "loss": 4.182,
+      "step": 417100
+    },
+    {
+      "epoch": 3.591167355371901,
+      "grad_norm": 3.6875,
+      "learning_rate": 3.680486146579997e-05,
+      "loss": 4.2218,
+      "step": 417150
+    },
+    {
+      "epoch": 3.5915977961432506,
+      "grad_norm": 3.421875,
+      "learning_rate": 3.678386723873848e-05,
+      "loss": 4.2281,
+      "step": 417200
+    },
+    {
+      "epoch": 3.592028236914601,
+      "grad_norm": 6.78125,
+      "learning_rate": 3.6762877651605164e-05,
+      "loss": 4.2575,
+      "step": 417250
+    },
+    {
+      "epoch": 3.5924586776859506,
+      "grad_norm": 3.0,
+      "learning_rate": 3.674189270594061e-05,
+      "loss": 4.1099,
+      "step": 417300
+    },
+    {
+      "epoch": 3.5928891184573004,
+      "grad_norm": 5.1875,
+      "learning_rate": 3.6720912403284945e-05,
+      "loss": 4.5233,
+      "step": 417350
+    },
+    {
+      "epoch": 3.59331955922865,
+      "grad_norm": 3.078125,
+      "learning_rate": 3.66999367451782e-05,
+      "loss": 4.1764,
+      "step": 417400
+    },
+    {
+      "epoch": 3.59375,
+      "grad_norm": 3.375,
+      "learning_rate": 3.66789657331599e-05,
+      "loss": 3.8756,
+      "step": 417450
+    },
+    {
+      "epoch": 3.59418044077135,
+      "grad_norm": 3.390625,
+      "learning_rate": 3.665799936876925e-05,
+      "loss": 4.4614,
+      "step": 417500
+    },
+    {
+      "epoch": 3.5946108815426996,
+      "grad_norm": 3.28125,
+      "learning_rate": 3.663703765354523e-05,
+      "loss": 4.3267,
+      "step": 417550
+    },
+    {
+      "epoch": 3.5950413223140494,
+      "grad_norm": 3.328125,
+      "learning_rate": 3.661608058902628e-05,
+      "loss": 4.4989,
+      "step": 417600
+    },
+    {
+      "epoch": 3.595471763085399,
+      "grad_norm": 3.40625,
+      "learning_rate": 3.6595128176750604e-05,
+      "loss": 4.2691,
+      "step": 417650
+    },
+    {
+      "epoch": 3.5959022038567494,
+      "grad_norm": 3.828125,
+      "learning_rate": 3.6574180418256146e-05,
+      "loss": 4.4983,
+      "step": 417700
+    },
+    {
+      "epoch": 3.596332644628099,
+      "grad_norm": 2.359375,
+      "learning_rate": 3.655323731508037e-05,
+      "loss": 4.0985,
+      "step": 417750
+    },
+    {
+      "epoch": 3.596763085399449,
+      "grad_norm": 3.65625,
+      "learning_rate": 3.653229886876047e-05,
+      "loss": 4.1892,
+      "step": 417800
+    },
+    {
+      "epoch": 3.597193526170799,
+      "grad_norm": 2.890625,
+      "learning_rate": 3.651136508083327e-05,
+      "loss": 4.4907,
+      "step": 417850
+    },
+    {
+      "epoch": 3.597623966942149,
+      "grad_norm": 2.15625,
+      "learning_rate": 3.649043595283526e-05,
+      "loss": 4.2404,
+      "step": 417900
+    },
+    {
+      "epoch": 3.598054407713499,
+      "grad_norm": 3.09375,
+      "learning_rate": 3.646951148630262e-05,
+      "loss": 4.2375,
+      "step": 417950
+    },
+    {
+      "epoch": 3.5984848484848486,
+      "grad_norm": 2.28125,
+      "learning_rate": 3.644859168277116e-05,
+      "loss": 4.5455,
+      "step": 418000
+    },
+    {
+      "epoch": 3.5989152892561984,
+      "grad_norm": 2.515625,
+      "learning_rate": 3.642767654377633e-05,
+      "loss": 4.3991,
+      "step": 418050
+    },
+    {
+      "epoch": 3.599345730027548,
+      "grad_norm": 3.078125,
+      "learning_rate": 3.640676607085327e-05,
+      "loss": 4.578,
+      "step": 418100
+    },
+    {
+      "epoch": 3.599776170798898,
+      "grad_norm": 4.125,
+      "learning_rate": 3.638586026553674e-05,
+      "loss": 4.2867,
+      "step": 418150
+    },
+    {
+      "epoch": 3.600206611570248,
+      "grad_norm": 4.375,
+      "learning_rate": 3.636495912936118e-05,
+      "loss": 4.2437,
+      "step": 418200
+    },
+    {
+      "epoch": 3.6006370523415976,
+      "grad_norm": 4.65625,
+      "learning_rate": 3.634406266386074e-05,
+      "loss": 4.6503,
+      "step": 418250
+    },
+    {
+      "epoch": 3.6010674931129474,
+      "grad_norm": 4.28125,
+      "learning_rate": 3.6323170870569136e-05,
+      "loss": 4.0717,
+      "step": 418300
+    },
+    {
+      "epoch": 3.6014979338842976,
+      "grad_norm": 3.28125,
+      "learning_rate": 3.630228375101977e-05,
+      "loss": 4.0399,
+      "step": 418350
+    },
+    {
+      "epoch": 3.6019283746556474,
+      "grad_norm": 3.640625,
+      "learning_rate": 3.6281401306745756e-05,
+      "loss": 4.5156,
+      "step": 418400
+    },
+    {
+      "epoch": 3.602358815426997,
+      "grad_norm": 4.03125,
+      "learning_rate": 3.626052353927977e-05,
+      "loss": 4.09,
+      "step": 418450
+    },
+    {
+      "epoch": 3.602789256198347,
+      "grad_norm": 3.515625,
+      "learning_rate": 3.62396504501542e-05,
+      "loss": 4.2808,
+      "step": 418500
+    },
+    {
+      "epoch": 3.6032196969696972,
+      "grad_norm": 4.21875,
+      "learning_rate": 3.6218782040901124e-05,
+      "loss": 4.3584,
+      "step": 418550
+    },
+    {
+      "epoch": 3.603650137741047,
+      "grad_norm": 3.546875,
+      "learning_rate": 3.6197918313052214e-05,
+      "loss": 4.3041,
+      "step": 418600
+    },
+    {
+      "epoch": 3.604080578512397,
+      "grad_norm": 3.03125,
+      "learning_rate": 3.6177059268138834e-05,
+      "loss": 4.4148,
+      "step": 418650
+    },
+    {
+      "epoch": 3.6045110192837466,
+      "grad_norm": 2.375,
+      "learning_rate": 3.6156204907691994e-05,
+      "loss": 4.1302,
+      "step": 418700
+    },
+    {
+      "epoch": 3.6049414600550964,
+      "grad_norm": 2.359375,
+      "learning_rate": 3.613535523324234e-05,
+      "loss": 4.3817,
+      "step": 418750
+    },
+    {
+      "epoch": 3.605371900826446,
+      "grad_norm": 2.75,
+      "learning_rate": 3.6114510246320165e-05,
+      "loss": 4.2568,
+      "step": 418800
+    },
+    {
+      "epoch": 3.605802341597796,
+      "grad_norm": 2.84375,
+      "learning_rate": 3.609366994845555e-05,
+      "loss": 4.509,
+      "step": 418850
+    },
+    {
+      "epoch": 3.606232782369146,
+      "grad_norm": 4.09375,
+      "learning_rate": 3.607283434117802e-05,
+      "loss": 4.3664,
+      "step": 418900
+    },
+    {
+      "epoch": 3.606663223140496,
+      "grad_norm": 2.5625,
+      "learning_rate": 3.605200342601692e-05,
+      "loss": 4.2576,
+      "step": 418950
+    },
+    {
+      "epoch": 3.607093663911846,
+      "grad_norm": 4.8125,
+      "learning_rate": 3.6031177204501176e-05,
+      "loss": 4.2341,
+      "step": 419000
+    },
+    {
+      "epoch": 3.6075241046831956,
+      "grad_norm": 3.03125,
+      "learning_rate": 3.6010355678159365e-05,
+      "loss": 4.2752,
+      "step": 419050
+    },
+    {
+      "epoch": 3.6079545454545454,
+      "grad_norm": 1.21875,
+      "learning_rate": 3.5989538848519845e-05,
+      "loss": 4.3877,
+      "step": 419100
+    },
+    {
+      "epoch": 3.608384986225895,
+      "grad_norm": 2.96875,
+      "learning_rate": 3.596872671711041e-05,
+      "loss": 4.1981,
+      "step": 419150
+    },
+    {
+      "epoch": 3.6088154269972454,
+      "grad_norm": 1.953125,
+      "learning_rate": 3.594791928545864e-05,
+      "loss": 3.8554,
+      "step": 419200
+    },
+    {
+      "epoch": 3.6092458677685952,
+      "grad_norm": 3.1875,
+      "learning_rate": 3.592711655509181e-05,
+      "loss": 4.3153,
+      "step": 419250
+    },
+    {
+      "epoch": 3.609676308539945,
+      "grad_norm": 3.703125,
+      "learning_rate": 3.5906318527536776e-05,
+      "loss": 4.1656,
+      "step": 419300
+    },
+    {
+      "epoch": 3.610106749311295,
+      "grad_norm": 4.3125,
+      "learning_rate": 3.5885525204320057e-05,
+      "loss": 4.5029,
+      "step": 419350
+    },
+    {
+      "epoch": 3.6105371900826446,
+      "grad_norm": 3.1875,
+      "learning_rate": 3.586473658696784e-05,
+      "loss": 4.369,
+      "step": 419400
+    },
+    {
+      "epoch": 3.6109676308539944,
+      "grad_norm": 2.9375,
+      "learning_rate": 3.584395267700593e-05,
+      "loss": 3.7832,
+      "step": 419450
+    },
+    {
+      "epoch": 3.611398071625344,
+      "grad_norm": 2.71875,
+      "learning_rate": 3.58231734759599e-05,
+      "loss": 4.3527,
+      "step": 419500
+    },
+    {
+      "epoch": 3.611828512396694,
+      "grad_norm": 2.609375,
+      "learning_rate": 3.580239898535486e-05,
+      "loss": 4.0425,
+      "step": 419550
+    },
+    {
+      "epoch": 3.6122589531680442,
+      "grad_norm": 3.328125,
+      "learning_rate": 3.5781629206715595e-05,
+      "loss": 4.6007,
+      "step": 419600
+    },
+    {
+      "epoch": 3.612689393939394,
+      "grad_norm": 4.1875,
+      "learning_rate": 3.576086414156659e-05,
+      "loss": 4.528,
+      "step": 419650
+    },
+    {
+      "epoch": 3.613119834710744,
+      "grad_norm": 2.359375,
+      "learning_rate": 3.574010379143193e-05,
+      "loss": 4.3826,
+      "step": 419700
+    },
+    {
+      "epoch": 3.6135502754820936,
+      "grad_norm": 3.421875,
+      "learning_rate": 3.5719348157835374e-05,
+      "loss": 4.4953,
+      "step": 419750
+    },
+    {
+      "epoch": 3.6139807162534434,
+      "grad_norm": 3.984375,
+      "learning_rate": 3.569859724230038e-05,
+      "loss": 4.0398,
+      "step": 419800
+    },
+    {
+      "epoch": 3.6144111570247937,
+      "grad_norm": 2.15625,
+      "learning_rate": 3.567785104635004e-05,
+      "loss": 4.1638,
+      "step": 419850
+    },
+    {
+      "epoch": 3.6148415977961434,
+      "grad_norm": 3.8125,
+      "learning_rate": 3.565710957150697e-05,
+      "loss": 4.3688,
+      "step": 419900
+    },
+    {
+      "epoch": 3.6152720385674932,
+      "grad_norm": 4.59375,
+      "learning_rate": 3.563637281929366e-05,
+      "loss": 4.5083,
+      "step": 419950
+    },
+    {
+      "epoch": 3.615702479338843,
+      "grad_norm": 5.5625,
+      "learning_rate": 3.5615640791232105e-05,
+      "loss": 3.8894,
+      "step": 420000
+    },
+    {
+      "epoch": 3.615702479338843,
+      "eval_loss": 4.991680145263672,
+      "eval_runtime": 23.5326,
+      "eval_samples_per_second": 27.196,
+      "eval_steps_per_second": 13.598,
+      "eval_tts_loss": 7.4729306236027835,
+      "step": 420000
+    },
+    {
+      "epoch": 3.616132920110193,
+      "grad_norm": 4.9375,
+      "learning_rate": 3.559491348884395e-05,
+      "loss": 4.4534,
+      "step": 420050
+    },
+    {
+      "epoch": 3.6165633608815426,
+      "grad_norm": 2.609375,
+      "learning_rate": 3.557419091365062e-05,
+      "loss": 4.4996,
+      "step": 420100
+    },
+    {
+      "epoch": 3.6169938016528924,
+      "grad_norm": 6.1875,
+      "learning_rate": 3.555347306717305e-05,
+      "loss": 4.5165,
+      "step": 420150
+    },
+    {
+      "epoch": 3.617424242424242,
+      "grad_norm": 2.96875,
+      "learning_rate": 3.553275995093189e-05,
+      "loss": 4.0153,
+      "step": 420200
+    },
+    {
+      "epoch": 3.6178546831955924,
+      "grad_norm": 4.25,
+      "learning_rate": 3.551205156644746e-05,
+      "loss": 4.288,
+      "step": 420250
+    },
+    {
+      "epoch": 3.6182851239669422,
+      "grad_norm": 2.703125,
+      "learning_rate": 3.549134791523968e-05,
+      "loss": 4.447,
+      "step": 420300
+    },
+    {
+      "epoch": 3.618715564738292,
+      "grad_norm": 3.3125,
+      "learning_rate": 3.547064899882814e-05,
+      "loss": 4.284,
+      "step": 420350
+    },
+    {
+      "epoch": 3.619146005509642,
+      "grad_norm": 3.546875,
+      "learning_rate": 3.54499548187322e-05,
+      "loss": 4.1718,
+      "step": 420400
+    },
+    {
+      "epoch": 3.6195764462809916,
+      "grad_norm": 2.765625,
+      "learning_rate": 3.5429265376470625e-05,
+      "loss": 4.2339,
+      "step": 420450
+    },
+    {
+      "epoch": 3.620006887052342,
+      "grad_norm": 3.828125,
+      "learning_rate": 3.540858067356207e-05,
+      "loss": 4.7083,
+      "step": 420500
+    },
+    {
+      "epoch": 3.6204373278236917,
+      "grad_norm": 1.328125,
+      "learning_rate": 3.5387900711524725e-05,
+      "loss": 4.2373,
+      "step": 420550
+    },
+    {
+      "epoch": 3.6208677685950414,
+      "grad_norm": 4.375,
+      "learning_rate": 3.536722549187641e-05,
+      "loss": 4.2521,
+      "step": 420600
+    },
+    {
+      "epoch": 3.6212982093663912,
+      "grad_norm": 5.5,
+      "learning_rate": 3.5346555016134766e-05,
+      "loss": 4.6201,
+      "step": 420650
+    },
+    {
+      "epoch": 3.621728650137741,
+      "grad_norm": 3.484375,
+      "learning_rate": 3.5325889285816824e-05,
+      "loss": 4.067,
+      "step": 420700
+    },
+    {
+      "epoch": 3.622159090909091,
+      "grad_norm": 3.25,
+      "learning_rate": 3.530522830243944e-05,
+      "loss": 4.4582,
+      "step": 420750
+    },
+    {
+      "epoch": 3.6225895316804406,
+      "grad_norm": 2.890625,
+      "learning_rate": 3.528457206751914e-05,
+      "loss": 4.2861,
+      "step": 420800
+    },
+    {
+      "epoch": 3.6230199724517904,
+      "grad_norm": 2.8125,
+      "learning_rate": 3.526392058257201e-05,
+      "loss": 4.7673,
+      "step": 420850
+    },
+    {
+      "epoch": 3.6234504132231407,
+      "grad_norm": 3.328125,
+      "learning_rate": 3.5243273849113824e-05,
+      "loss": 4.2167,
+      "step": 420900
+    },
+    {
+      "epoch": 3.6238808539944904,
+      "grad_norm": 4.625,
+      "learning_rate": 3.5222631868660004e-05,
+      "loss": 3.787,
+      "step": 420950
+    },
+    {
+      "epoch": 3.6243112947658402,
+      "grad_norm": 1.53125,
+      "learning_rate": 3.5201994642725635e-05,
+      "loss": 4.4714,
+      "step": 421000
+    },
+    {
+      "epoch": 3.62474173553719,
+      "grad_norm": 2.578125,
+      "learning_rate": 3.5181362172825415e-05,
+      "loss": 4.3027,
+      "step": 421050
+    },
+    {
+      "epoch": 3.62517217630854,
+      "grad_norm": 6.84375,
+      "learning_rate": 3.516073446047378e-05,
+      "loss": 4.3151,
+      "step": 421100
+    },
+    {
+      "epoch": 3.62560261707989,
+      "grad_norm": 3.703125,
+      "learning_rate": 3.5140111507184724e-05,
+      "loss": 4.1441,
+      "step": 421150
+    },
+    {
+      "epoch": 3.62603305785124,
+      "grad_norm": 4.59375,
+      "learning_rate": 3.511949331447194e-05,
+      "loss": 4.2146,
+      "step": 421200
+    },
+    {
+      "epoch": 3.6264634986225897,
+      "grad_norm": 5.1875,
+      "learning_rate": 3.509887988384876e-05,
+      "loss": 4.6127,
+      "step": 421250
+    },
+    {
+      "epoch": 3.6268939393939394,
+      "grad_norm": 3.625,
+      "learning_rate": 3.5078271216828115e-05,
+      "loss": 4.4266,
+      "step": 421300
+    },
+    {
+      "epoch": 3.6273243801652892,
+      "grad_norm": 3.96875,
+      "learning_rate": 3.5057667314922714e-05,
+      "loss": 4.0089,
+      "step": 421350
+    },
+    {
+      "epoch": 3.627754820936639,
+      "grad_norm": 4.25,
+      "learning_rate": 3.503706817964485e-05,
+      "loss": 3.9434,
+      "step": 421400
+    },
+    {
+      "epoch": 3.628185261707989,
+      "grad_norm": 3.203125,
+      "learning_rate": 3.5016473812506324e-05,
+      "loss": 4.3359,
+      "step": 421450
+    },
+    {
+      "epoch": 3.6286157024793386,
+      "grad_norm": 2.8125,
+      "learning_rate": 3.499588421501885e-05,
+      "loss": 4.1194,
+      "step": 421500
+    },
+    {
+      "epoch": 3.629046143250689,
+      "grad_norm": 2.453125,
+      "learning_rate": 3.4975299388693615e-05,
+      "loss": 4.5715,
+      "step": 421550
+    },
+    {
+      "epoch": 3.6294765840220387,
+      "grad_norm": 2.78125,
+      "learning_rate": 3.495471933504146e-05,
+      "loss": 4.6143,
+      "step": 421600
+    },
+    {
+      "epoch": 3.6299070247933884,
+      "grad_norm": 2.234375,
+      "learning_rate": 3.4934144055573034e-05,
+      "loss": 4.4325,
+      "step": 421650
+    },
+    {
+      "epoch": 3.6303374655647382,
+      "grad_norm": 2.78125,
+      "learning_rate": 3.491357355179837e-05,
+      "loss": 4.3959,
+      "step": 421700
+    },
+    {
+      "epoch": 3.630767906336088,
+      "grad_norm": 1.4921875,
+      "learning_rate": 3.489300782522741e-05,
+      "loss": 4.4999,
+      "step": 421750
+    },
+    {
+      "epoch": 3.6311983471074383,
+      "grad_norm": 3.078125,
+      "learning_rate": 3.487244687736958e-05,
+      "loss": 4.3383,
+      "step": 421800
+    },
+    {
+      "epoch": 3.631628787878788,
+      "grad_norm": 3.0625,
+      "learning_rate": 3.485189070973404e-05,
+      "loss": 4.5649,
+      "step": 421850
+    },
+    {
+      "epoch": 3.632059228650138,
+      "grad_norm": 2.640625,
+      "learning_rate": 3.4831339323829505e-05,
+      "loss": 3.8946,
+      "step": 421900
+    },
+    {
+      "epoch": 3.6324896694214877,
+      "grad_norm": 2.734375,
+      "learning_rate": 3.481079272116452e-05,
+      "loss": 4.2906,
+      "step": 421950
+    },
+    {
+      "epoch": 3.6329201101928374,
+      "grad_norm": 3.46875,
+      "learning_rate": 3.479025090324702e-05,
+      "loss": 4.5026,
+      "step": 422000
+    },
+    {
+      "epoch": 3.6333505509641872,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.4769713871584843e-05,
+      "loss": 4.1248,
+      "step": 422050
+    },
+    {
+      "epoch": 3.633780991735537,
+      "grad_norm": 3.09375,
+      "learning_rate": 3.4749181627685326e-05,
+      "loss": 4.4575,
+      "step": 422100
+    },
+    {
+      "epoch": 3.634211432506887,
+      "grad_norm": 2.484375,
+      "learning_rate": 3.4728654173055444e-05,
+      "loss": 4.3899,
+      "step": 422150
+    },
+    {
+      "epoch": 3.634641873278237,
+      "grad_norm": 1.6015625,
+      "learning_rate": 3.470813150920198e-05,
+      "loss": 3.8903,
+      "step": 422200
+    },
+    {
+      "epoch": 3.635072314049587,
+      "grad_norm": 3.640625,
+      "learning_rate": 3.468761363763114e-05,
+      "loss": 4.1154,
+      "step": 422250
+    },
+    {
+      "epoch": 3.6355027548209367,
+      "grad_norm": 2.53125,
+      "learning_rate": 3.466710055984892e-05,
+      "loss": 4.3853,
+      "step": 422300
+    },
+    {
+      "epoch": 3.6359331955922864,
+      "grad_norm": 2.578125,
+      "learning_rate": 3.464659227736098e-05,
+      "loss": 4.445,
+      "step": 422350
+    },
+    {
+      "epoch": 3.6363636363636362,
+      "grad_norm": 1.7109375,
+      "learning_rate": 3.462608879167255e-05,
+      "loss": 4.3955,
+      "step": 422400
+    },
+    {
+      "epoch": 3.6367940771349865,
+      "grad_norm": 2.953125,
+      "learning_rate": 3.460559010428855e-05,
+      "loss": 4.1305,
+      "step": 422450
+    },
+    {
+      "epoch": 3.6372245179063363,
+      "grad_norm": 1.4609375,
+      "learning_rate": 3.458509621671354e-05,
+      "loss": 4.2699,
+      "step": 422500
+    },
+    {
+      "epoch": 3.637654958677686,
+      "grad_norm": 2.453125,
+      "learning_rate": 3.456460713045172e-05,
+      "loss": 4.5622,
+      "step": 422550
+    },
+    {
+      "epoch": 3.638085399449036,
+      "grad_norm": 4.0625,
+      "learning_rate": 3.454412284700691e-05,
+      "loss": 4.1964,
+      "step": 422600
+    },
+    {
+      "epoch": 3.6385158402203857,
+      "grad_norm": 4.625,
+      "learning_rate": 3.45236433678827e-05,
+      "loss": 4.1271,
+      "step": 422650
+    },
+    {
+      "epoch": 3.6389462809917354,
+      "grad_norm": 4.375,
+      "learning_rate": 3.450316869458218e-05,
+      "loss": 4.4434,
+      "step": 422700
+    },
+    {
+      "epoch": 3.6393767217630852,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.448269882860815e-05,
+      "loss": 4.5871,
+      "step": 422750
+    },
+    {
+      "epoch": 3.639807162534435,
+      "grad_norm": 4.125,
+      "learning_rate": 3.446223377146308e-05,
+      "loss": 4.3567,
+      "step": 422800
+    },
+    {
+      "epoch": 3.6402376033057853,
+      "grad_norm": 3.21875,
+      "learning_rate": 3.4441773524649e-05,
+      "loss": 4.5778,
+      "step": 422850
+    },
+    {
+      "epoch": 3.640668044077135,
+      "grad_norm": 4.28125,
+      "learning_rate": 3.442131808966772e-05,
+      "loss": 4.3927,
+      "step": 422900
+    },
+    {
+      "epoch": 3.641098484848485,
+      "grad_norm": 1.6328125,
+      "learning_rate": 3.4400867468020636e-05,
+      "loss": 4.1146,
+      "step": 422950
+    },
+    {
+      "epoch": 3.6415289256198347,
+      "grad_norm": 4.96875,
+      "learning_rate": 3.4380421661208675e-05,
+      "loss": 4.2752,
+      "step": 423000
+    },
+    {
+      "epoch": 3.6415289256198347,
+      "eval_loss": 4.99205207824707,
+      "eval_runtime": 24.1641,
+      "eval_samples_per_second": 26.486,
+      "eval_steps_per_second": 13.243,
+      "eval_tts_loss": 7.487585057957488,
+      "step": 423000
+    },
+    {
+      "epoch": 3.6419593663911844,
+      "grad_norm": 4.28125,
+      "learning_rate": 3.43599806707326e-05,
+      "loss": 4.4091,
+      "step": 423050
+    },
+    {
+      "epoch": 3.6423898071625347,
+      "grad_norm": 2.078125,
+      "learning_rate": 3.4339544498092736e-05,
+      "loss": 4.0776,
+      "step": 423100
+    },
+    {
+      "epoch": 3.6428202479338845,
+      "grad_norm": 4.28125,
+      "learning_rate": 3.431911314478898e-05,
+      "loss": 4.5538,
+      "step": 423150
+    },
+    {
+      "epoch": 3.6432506887052343,
+      "grad_norm": 2.578125,
+      "learning_rate": 3.4298686612321086e-05,
+      "loss": 4.2152,
+      "step": 423200
+    },
+    {
+      "epoch": 3.643681129476584,
+      "grad_norm": 3.03125,
+      "learning_rate": 3.42782649021882e-05,
+      "loss": 4.2584,
+      "step": 423250
+    },
+    {
+      "epoch": 3.644111570247934,
+      "grad_norm": 6.21875,
+      "learning_rate": 3.425784801588923e-05,
+      "loss": 4.2701,
+      "step": 423300
+    },
+    {
+      "epoch": 3.6445420110192837,
+      "grad_norm": 2.09375,
+      "learning_rate": 3.4237435954922794e-05,
+      "loss": 4.4707,
+      "step": 423350
+    },
+    {
+      "epoch": 3.6449724517906334,
+      "grad_norm": 2.1875,
+      "learning_rate": 3.421702872078707e-05,
+      "loss": 4.1121,
+      "step": 423400
+    },
+    {
+      "epoch": 3.6454028925619832,
+      "grad_norm": 3.890625,
+      "learning_rate": 3.419662631497991e-05,
+      "loss": 4.4922,
+      "step": 423450
+    },
+    {
+      "epoch": 3.6458333333333335,
+      "grad_norm": 4.6875,
+      "learning_rate": 3.4176228738998784e-05,
+      "loss": 4.2265,
+      "step": 423500
+    },
+    {
+      "epoch": 3.6462637741046833,
+      "grad_norm": 1.46875,
+      "learning_rate": 3.4155835994340825e-05,
+      "loss": 4.4298,
+      "step": 423550
+    },
+    {
+      "epoch": 3.646694214876033,
+      "grad_norm": 3.171875,
+      "learning_rate": 3.413544808250286e-05,
+      "loss": 4.4119,
+      "step": 423600
+    },
+    {
+      "epoch": 3.647124655647383,
+      "grad_norm": 3.765625,
+      "learning_rate": 3.4115065004981305e-05,
+      "loss": 4.3498,
+      "step": 423650
+    },
+    {
+      "epoch": 3.6475550964187327,
+      "grad_norm": 2.34375,
+      "learning_rate": 3.409468676327219e-05,
+      "loss": 3.8549,
+      "step": 423700
+    },
+    {
+      "epoch": 3.647985537190083,
+      "grad_norm": 4.46875,
+      "learning_rate": 3.407431335887132e-05,
+      "loss": 4.3146,
+      "step": 423750
+    },
+    {
+      "epoch": 3.6484159779614327,
+      "grad_norm": 5.34375,
+      "learning_rate": 3.4053944793273985e-05,
+      "loss": 4.7968,
+      "step": 423800
+    },
+    {
+      "epoch": 3.6488464187327825,
+      "grad_norm": 1.421875,
+      "learning_rate": 3.403358106797519e-05,
+      "loss": 4.5443,
+      "step": 423850
+    },
+    {
+      "epoch": 3.6492768595041323,
+      "grad_norm": 3.859375,
+      "learning_rate": 3.401322218446963e-05,
+      "loss": 4.2175,
+      "step": 423900
+    },
+    {
+      "epoch": 3.649707300275482,
+      "grad_norm": 3.75,
+      "learning_rate": 3.39928681442516e-05,
+      "loss": 4.1739,
+      "step": 423950
+    },
+    {
+      "epoch": 3.650137741046832,
+      "grad_norm": 2.890625,
+      "learning_rate": 3.3972518948815025e-05,
+      "loss": 4.5082,
+      "step": 424000
+    },
+    {
+      "epoch": 3.6505681818181817,
+      "grad_norm": 1.5859375,
+      "learning_rate": 3.39521745996535e-05,
+      "loss": 4.026,
+      "step": 424050
+    },
+    {
+      "epoch": 3.6509986225895315,
+      "grad_norm": 3.203125,
+      "learning_rate": 3.393183509826024e-05,
+      "loss": 4.3433,
+      "step": 424100
+    },
+    {
+      "epoch": 3.6514290633608817,
+      "grad_norm": 4.875,
+      "learning_rate": 3.391150044612812e-05,
+      "loss": 4.2548,
+      "step": 424150
+    },
+    {
+      "epoch": 3.6518595041322315,
+      "grad_norm": 4.0625,
+      "learning_rate": 3.3891170644749705e-05,
+      "loss": 4.6771,
+      "step": 424200
+    },
+    {
+      "epoch": 3.6522899449035813,
+      "grad_norm": 1.6328125,
+      "learning_rate": 3.387084569561712e-05,
+      "loss": 4.0842,
+      "step": 424250
+    },
+    {
+      "epoch": 3.652720385674931,
+      "grad_norm": 6.9375,
+      "learning_rate": 3.385052560022218e-05,
+      "loss": 4.2175,
+      "step": 424300
+    },
+    {
+      "epoch": 3.653150826446281,
+      "grad_norm": 2.890625,
+      "learning_rate": 3.383021036005634e-05,
+      "loss": 4.0116,
+      "step": 424350
+    },
+    {
+      "epoch": 3.653581267217631,
+      "grad_norm": 4.8125,
+      "learning_rate": 3.380989997661066e-05,
+      "loss": 4.3937,
+      "step": 424400
+    },
+    {
+      "epoch": 3.654011707988981,
+      "grad_norm": 5.4375,
+      "learning_rate": 3.378959445137594e-05,
+      "loss": 4.0663,
+      "step": 424450
+    },
+    {
+      "epoch": 3.6544421487603307,
+      "grad_norm": 3.46875,
+      "learning_rate": 3.376929378584257e-05,
+      "loss": 4.1896,
+      "step": 424500
+    },
+    {
+      "epoch": 3.6548725895316805,
+      "grad_norm": 2.453125,
+      "learning_rate": 3.374899798150046e-05,
+      "loss": 4.2375,
+      "step": 424550
+    },
+    {
+      "epoch": 3.6553030303030303,
+      "grad_norm": 5.0,
+      "learning_rate": 3.37287070398394e-05,
+      "loss": 4.3769,
+      "step": 424600
+    },
+    {
+      "epoch": 3.65573347107438,
+      "grad_norm": 3.265625,
+      "learning_rate": 3.3708420962348655e-05,
+      "loss": 4.132,
+      "step": 424650
+    },
+    {
+      "epoch": 3.65616391184573,
+      "grad_norm": 1.8828125,
+      "learning_rate": 3.368813975051715e-05,
+      "loss": 4.1329,
+      "step": 424700
+    },
+    {
+      "epoch": 3.6565943526170797,
+      "grad_norm": 3.609375,
+      "learning_rate": 3.366786340583358e-05,
+      "loss": 4.1082,
+      "step": 424750
+    },
+    {
+      "epoch": 3.65702479338843,
+      "grad_norm": 2.875,
+      "learning_rate": 3.364759192978607e-05,
+      "loss": 4.3278,
+      "step": 424800
+    },
+    {
+      "epoch": 3.6574552341597797,
+      "grad_norm": 3.953125,
+      "learning_rate": 3.3627325323862537e-05,
+      "loss": 4.2373,
+      "step": 424850
+    },
+    {
+      "epoch": 3.6578856749311295,
+      "grad_norm": 2.15625,
+      "learning_rate": 3.360706358955055e-05,
+      "loss": 4.2467,
+      "step": 424900
+    },
+    {
+      "epoch": 3.6583161157024793,
+      "grad_norm": 1.265625,
+      "learning_rate": 3.3586806728337236e-05,
+      "loss": 3.951,
+      "step": 424950
+    },
+    {
+      "epoch": 3.658746556473829,
+      "grad_norm": 2.28125,
+      "learning_rate": 3.3566554741709434e-05,
+      "loss": 4.3198,
+      "step": 425000
+    },
+    {
+      "epoch": 3.6591769972451793,
+      "grad_norm": 3.9375,
+      "learning_rate": 3.354630763115356e-05,
+      "loss": 4.327,
+      "step": 425050
+    },
+    {
+      "epoch": 3.659607438016529,
+      "grad_norm": 2.78125,
+      "learning_rate": 3.3526065398155706e-05,
+      "loss": 4.1036,
+      "step": 425100
+    },
+    {
+      "epoch": 3.660037878787879,
+      "grad_norm": 2.96875,
+      "learning_rate": 3.350582804420166e-05,
+      "loss": 4.4225,
+      "step": 425150
+    },
+    {
+      "epoch": 3.6604683195592287,
+      "grad_norm": 1.359375,
+      "learning_rate": 3.348559557077677e-05,
+      "loss": 4.3596,
+      "step": 425200
+    },
+    {
+      "epoch": 3.6608987603305785,
+      "grad_norm": 3.859375,
+      "learning_rate": 3.346536797936606e-05,
+      "loss": 4.3288,
+      "step": 425250
+    },
+    {
+      "epoch": 3.6613292011019283,
+      "grad_norm": 2.5,
+      "learning_rate": 3.3445145271454184e-05,
+      "loss": 4.4486,
+      "step": 425300
+    },
+    {
+      "epoch": 3.661759641873278,
+      "grad_norm": 1.4453125,
+      "learning_rate": 3.3424927448525446e-05,
+      "loss": 4.2422,
+      "step": 425350
+    },
+    {
+      "epoch": 3.662190082644628,
+      "grad_norm": 3.046875,
+      "learning_rate": 3.340471451206376e-05,
+      "loss": 4.6697,
+      "step": 425400
+    },
+    {
+      "epoch": 3.662620523415978,
+      "grad_norm": 3.828125,
+      "learning_rate": 3.3384506463552775e-05,
+      "loss": 4.5314,
+      "step": 425450
+    },
+    {
+      "epoch": 3.663050964187328,
+      "grad_norm": 2.296875,
+      "learning_rate": 3.33643033044757e-05,
+      "loss": 4.0822,
+      "step": 425500
+    },
+    {
+      "epoch": 3.6634814049586777,
+      "grad_norm": 2.25,
+      "learning_rate": 3.3344105036315384e-05,
+      "loss": 4.1879,
+      "step": 425550
+    },
+    {
+      "epoch": 3.6639118457300275,
+      "grad_norm": 3.25,
+      "learning_rate": 3.332391166055435e-05,
+      "loss": 4.4086,
+      "step": 425600
+    },
+    {
+      "epoch": 3.6643422865013773,
+      "grad_norm": 3.65625,
+      "learning_rate": 3.3303723178674736e-05,
+      "loss": 4.0846,
+      "step": 425650
+    },
+    {
+      "epoch": 3.6647727272727275,
+      "grad_norm": 3.453125,
+      "learning_rate": 3.328353959215831e-05,
+      "loss": 4.0723,
+      "step": 425700
+    },
+    {
+      "epoch": 3.6652031680440773,
+      "grad_norm": 2.296875,
+      "learning_rate": 3.326336090248656e-05,
+      "loss": 4.3801,
+      "step": 425750
+    },
+    {
+      "epoch": 3.665633608815427,
+      "grad_norm": 3.28125,
+      "learning_rate": 3.324318711114055e-05,
+      "loss": 4.4875,
+      "step": 425800
+    },
+    {
+      "epoch": 3.666064049586777,
+      "grad_norm": 1.765625,
+      "learning_rate": 3.322301821960097e-05,
+      "loss": 4.11,
+      "step": 425850
+    },
+    {
+      "epoch": 3.6664944903581267,
+      "grad_norm": 3.375,
+      "learning_rate": 3.320285422934817e-05,
+      "loss": 4.3325,
+      "step": 425900
+    },
+    {
+      "epoch": 3.6669249311294765,
+      "grad_norm": 3.4375,
+      "learning_rate": 3.318269514186215e-05,
+      "loss": 4.3303,
+      "step": 425950
+    },
+    {
+      "epoch": 3.6673553719008263,
+      "grad_norm": 2.296875,
+      "learning_rate": 3.3162540958622514e-05,
+      "loss": 4.3698,
+      "step": 426000
+    },
+    {
+      "epoch": 3.6673553719008263,
+      "eval_loss": 4.991609573364258,
+      "eval_runtime": 23.8703,
+      "eval_samples_per_second": 26.812,
+      "eval_steps_per_second": 13.406,
+      "eval_tts_loss": 7.474097057115932,
+      "step": 426000
+    },
+    {
+      "epoch": 3.667785812672176,
+      "grad_norm": 2.25,
+      "learning_rate": 3.314239168110863e-05,
+      "loss": 4.4304,
+      "step": 426050
+    },
+    {
+      "epoch": 3.6682162534435263,
+      "grad_norm": 2.171875,
+      "learning_rate": 3.3122247310799293e-05,
+      "loss": 3.7824,
+      "step": 426100
+    },
+    {
+      "epoch": 3.668646694214876,
+      "grad_norm": 2.765625,
+      "learning_rate": 3.310210784917315e-05,
+      "loss": 4.4677,
+      "step": 426150
+    },
+    {
+      "epoch": 3.669077134986226,
+      "grad_norm": 7.0,
+      "learning_rate": 3.3081973297708336e-05,
+      "loss": 4.2738,
+      "step": 426200
+    },
+    {
+      "epoch": 3.6695075757575757,
+      "grad_norm": 3.015625,
+      "learning_rate": 3.3061843657882684e-05,
+      "loss": 4.1526,
+      "step": 426250
+    },
+    {
+      "epoch": 3.6699380165289255,
+      "grad_norm": 2.125,
+      "learning_rate": 3.3041718931173746e-05,
+      "loss": 4.3083,
+      "step": 426300
+    },
+    {
+      "epoch": 3.6703684573002757,
+      "grad_norm": 3.140625,
+      "learning_rate": 3.302159911905856e-05,
+      "loss": 4.2445,
+      "step": 426350
+    },
+    {
+      "epoch": 3.6707988980716255,
+      "grad_norm": 5.8125,
+      "learning_rate": 3.3001484223013836e-05,
+      "loss": 4.4528,
+      "step": 426400
+    },
+    {
+      "epoch": 3.6712293388429753,
+      "grad_norm": 1.3828125,
+      "learning_rate": 3.298137424451606e-05,
+      "loss": 4.3865,
+      "step": 426450
+    },
+    {
+      "epoch": 3.671659779614325,
+      "grad_norm": 2.640625,
+      "learning_rate": 3.296126918504122e-05,
+      "loss": 4.2466,
+      "step": 426500
+    },
+    {
+      "epoch": 3.672090220385675,
+      "grad_norm": 5.09375,
+      "learning_rate": 3.2941169046064976e-05,
+      "loss": 4.1227,
+      "step": 426550
+    },
+    {
+      "epoch": 3.6725206611570247,
+      "grad_norm": 3.765625,
+      "learning_rate": 3.2921073829062654e-05,
+      "loss": 4.5522,
+      "step": 426600
+    },
+    {
+      "epoch": 3.6729511019283745,
+      "grad_norm": 5.03125,
+      "learning_rate": 3.290098353550915e-05,
+      "loss": 4.3921,
+      "step": 426650
+    },
+    {
+      "epoch": 3.6733815426997243,
+      "grad_norm": 2.8125,
+      "learning_rate": 3.2880898166879125e-05,
+      "loss": 4.0194,
+      "step": 426700
+    },
+    {
+      "epoch": 3.6738119834710745,
+      "grad_norm": 5.46875,
+      "learning_rate": 3.2860817724646765e-05,
+      "loss": 4.365,
+      "step": 426750
+    },
+    {
+      "epoch": 3.6742424242424243,
+      "grad_norm": 3.421875,
+      "learning_rate": 3.2840742210285925e-05,
+      "loss": 4.1843,
+      "step": 426800
+    },
+    {
+      "epoch": 3.674672865013774,
+      "grad_norm": 3.109375,
+      "learning_rate": 3.2820671625270116e-05,
+      "loss": 4.4287,
+      "step": 426850
+    },
+    {
+      "epoch": 3.675103305785124,
+      "grad_norm": 4.09375,
+      "learning_rate": 3.2800605971072465e-05,
+      "loss": 4.2402,
+      "step": 426900
+    },
+    {
+      "epoch": 3.6755337465564737,
+      "grad_norm": 1.2578125,
+      "learning_rate": 3.278054524916572e-05,
+      "loss": 4.1549,
+      "step": 426950
+    },
+    {
+      "epoch": 3.675964187327824,
+      "grad_norm": 3.71875,
+      "learning_rate": 3.2760489461022356e-05,
+      "loss": 4.1764,
+      "step": 427000
+    },
+    {
+      "epoch": 3.6763946280991737,
+      "grad_norm": 4.21875,
+      "learning_rate": 3.274043860811444e-05,
+      "loss": 4.3668,
+      "step": 427050
+    },
+    {
+      "epoch": 3.6768250688705235,
+      "grad_norm": 1.328125,
+      "learning_rate": 3.272039269191355e-05,
+      "loss": 4.0412,
+      "step": 427100
+    },
+    {
+      "epoch": 3.6772555096418733,
+      "grad_norm": 4.40625,
+      "learning_rate": 3.2700351713891107e-05,
+      "loss": 4.38,
+      "step": 427150
+    },
+    {
+      "epoch": 3.677685950413223,
+      "grad_norm": 2.984375,
+      "learning_rate": 3.2680315675518046e-05,
+      "loss": 3.9532,
+      "step": 427200
+    },
+    {
+      "epoch": 3.678116391184573,
+      "grad_norm": 2.953125,
+      "learning_rate": 3.2660284578264954e-05,
+      "loss": 4.3657,
+      "step": 427250
+    },
+    {
+      "epoch": 3.6785468319559227,
+      "grad_norm": 4.4375,
+      "learning_rate": 3.264025842360211e-05,
+      "loss": 4.3934,
+      "step": 427300
+    },
+    {
+      "epoch": 3.6789772727272725,
+      "grad_norm": 3.875,
+      "learning_rate": 3.2620237212999375e-05,
+      "loss": 4.365,
+      "step": 427350
+    },
+    {
+      "epoch": 3.6794077134986227,
+      "grad_norm": 3.0,
+      "learning_rate": 3.2600220947926255e-05,
+      "loss": 4.0804,
+      "step": 427400
+    },
+    {
+      "epoch": 3.6798381542699725,
+      "grad_norm": 2.671875,
+      "learning_rate": 3.25802096298519e-05,
+      "loss": 4.2792,
+      "step": 427450
+    },
+    {
+      "epoch": 3.6802685950413223,
+      "grad_norm": 2.09375,
+      "learning_rate": 3.25602032602451e-05,
+      "loss": 4.6233,
+      "step": 427500
+    },
+    {
+      "epoch": 3.680699035812672,
+      "grad_norm": 2.578125,
+      "learning_rate": 3.2540201840574246e-05,
+      "loss": 4.3971,
+      "step": 427550
+    },
+    {
+      "epoch": 3.681129476584022,
+      "grad_norm": 2.125,
+      "learning_rate": 3.252020537230751e-05,
+      "loss": 4.1721,
+      "step": 427600
+    },
+    {
+      "epoch": 3.681559917355372,
+      "grad_norm": 3.109375,
+      "learning_rate": 3.250021385691244e-05,
+      "loss": 4.462,
+      "step": 427650
+    },
+    {
+      "epoch": 3.681990358126722,
+      "grad_norm": 3.921875,
+      "learning_rate": 3.248022729585648e-05,
+      "loss": 4.2483,
+      "step": 427700
+    },
+    {
+      "epoch": 3.6824207988980717,
+      "grad_norm": 2.609375,
+      "learning_rate": 3.246024569060656e-05,
+      "loss": 4.359,
+      "step": 427750
+    },
+    {
+      "epoch": 3.6828512396694215,
+      "grad_norm": 2.78125,
+      "learning_rate": 3.2440269042629254e-05,
+      "loss": 4.6668,
+      "step": 427800
+    },
+    {
+      "epoch": 3.6832816804407713,
+      "grad_norm": 2.671875,
+      "learning_rate": 3.242029735339092e-05,
+      "loss": 4.3851,
+      "step": 427850
+    },
+    {
+      "epoch": 3.683712121212121,
+      "grad_norm": 2.71875,
+      "learning_rate": 3.2400330624357314e-05,
+      "loss": 4.3193,
+      "step": 427900
+    },
+    {
+      "epoch": 3.684142561983471,
+      "grad_norm": 1.5390625,
+      "learning_rate": 3.238036885699397e-05,
+      "loss": 4.1023,
+      "step": 427950
+    },
+    {
+      "epoch": 3.6845730027548207,
+      "grad_norm": 2.84375,
+      "learning_rate": 3.2360412052766096e-05,
+      "loss": 4.6179,
+      "step": 428000
+    },
+    {
+      "epoch": 3.685003443526171,
+      "grad_norm": 4.03125,
+      "learning_rate": 3.2340460213138443e-05,
+      "loss": 4.2287,
+      "step": 428050
+    },
+    {
+      "epoch": 3.6854338842975207,
+      "grad_norm": 3.3125,
+      "learning_rate": 3.232051333957544e-05,
+      "loss": 4.1586,
+      "step": 428100
+    },
+    {
+      "epoch": 3.6858643250688705,
+      "grad_norm": 3.5625,
+      "learning_rate": 3.230057143354114e-05,
+      "loss": 4.2369,
+      "step": 428150
+    },
+    {
+      "epoch": 3.6862947658402203,
+      "grad_norm": 3.90625,
+      "learning_rate": 3.228063449649924e-05,
+      "loss": 4.5667,
+      "step": 428200
+    },
+    {
+      "epoch": 3.68672520661157,
+      "grad_norm": 3.765625,
+      "learning_rate": 3.2260702529913034e-05,
+      "loss": 4.7902,
+      "step": 428250
+    },
+    {
+      "epoch": 3.6871556473829203,
+      "grad_norm": 4.53125,
+      "learning_rate": 3.224077553524555e-05,
+      "loss": 4.1604,
+      "step": 428300
+    },
+    {
+      "epoch": 3.68758608815427,
+      "grad_norm": 2.6875,
+      "learning_rate": 3.2220853513959345e-05,
+      "loss": 4.2735,
+      "step": 428350
+    },
+    {
+      "epoch": 3.68801652892562,
+      "grad_norm": 2.984375,
+      "learning_rate": 3.2200936467516675e-05,
+      "loss": 4.2855,
+      "step": 428400
+    },
+    {
+      "epoch": 3.6884469696969697,
+      "grad_norm": 3.84375,
+      "learning_rate": 3.218102439737939e-05,
+      "loss": 4.4194,
+      "step": 428450
+    },
+    {
+      "epoch": 3.6888774104683195,
+      "grad_norm": 2.171875,
+      "learning_rate": 3.216111730500896e-05,
+      "loss": 4.3832,
+      "step": 428500
+    },
+    {
+      "epoch": 3.6893078512396693,
+      "grad_norm": 3.140625,
+      "learning_rate": 3.21412151918666e-05,
+      "loss": 4.1877,
+      "step": 428550
+    },
+    {
+      "epoch": 3.689738292011019,
+      "grad_norm": 2.953125,
+      "learning_rate": 3.2121318059413076e-05,
+      "loss": 4.7643,
+      "step": 428600
+    },
+    {
+      "epoch": 3.690168732782369,
+      "grad_norm": 4.3125,
+      "learning_rate": 3.21014259091087e-05,
+      "loss": 4.0911,
+      "step": 428650
+    },
+    {
+      "epoch": 3.690599173553719,
+      "grad_norm": 2.140625,
+      "learning_rate": 3.2081538742413596e-05,
+      "loss": 4.1611,
+      "step": 428700
+    },
+    {
+      "epoch": 3.691029614325069,
+      "grad_norm": 3.9375,
+      "learning_rate": 3.2061656560787424e-05,
+      "loss": 4.243,
+      "step": 428750
+    },
+    {
+      "epoch": 3.6914600550964187,
+      "grad_norm": 1.5703125,
+      "learning_rate": 3.204177936568945e-05,
+      "loss": 4.2006,
+      "step": 428800
+    },
+    {
+      "epoch": 3.6918904958677685,
+      "grad_norm": 2.578125,
+      "learning_rate": 3.202190715857872e-05,
+      "loss": 4.3707,
+      "step": 428850
+    },
+    {
+      "epoch": 3.6923209366391183,
+      "grad_norm": 4.53125,
+      "learning_rate": 3.2002039940913684e-05,
+      "loss": 4.3832,
+      "step": 428900
+    },
+    {
+      "epoch": 3.6927513774104685,
+      "grad_norm": 3.53125,
+      "learning_rate": 3.198217771415265e-05,
+      "loss": 4.2319,
+      "step": 428950
+    },
+    {
+      "epoch": 3.6931818181818183,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.196232047975343e-05,
+      "loss": 4.0473,
+      "step": 429000
+    },
+    {
+      "epoch": 3.6931818181818183,
+      "eval_loss": 4.99127197265625,
+      "eval_runtime": 24.3883,
+      "eval_samples_per_second": 26.242,
+      "eval_steps_per_second": 13.121,
+      "eval_tts_loss": 7.473439829219363,
+      "step": 429000
+    },
+    {
+      "epoch": 3.693612258953168,
+      "grad_norm": 4.875,
+      "learning_rate": 3.1942468239173495e-05,
+      "loss": 4.2664,
+      "step": 429050
+    },
+    {
+      "epoch": 3.694042699724518,
+      "grad_norm": 3.9375,
+      "learning_rate": 3.192262099386992e-05,
+      "loss": 4.5117,
+      "step": 429100
+    },
+    {
+      "epoch": 3.6944731404958677,
+      "grad_norm": 3.5625,
+      "learning_rate": 3.1902778745299576e-05,
+      "loss": 4.3189,
+      "step": 429150
+    },
+    {
+      "epoch": 3.6949035812672175,
+      "grad_norm": 2.53125,
+      "learning_rate": 3.1882941494918686e-05,
+      "loss": 4.1603,
+      "step": 429200
+    },
+    {
+      "epoch": 3.6953340220385673,
+      "grad_norm": 3.59375,
+      "learning_rate": 3.186310924418335e-05,
+      "loss": 4.0825,
+      "step": 429250
+    },
+    {
+      "epoch": 3.695764462809917,
+      "grad_norm": 3.171875,
+      "learning_rate": 3.184328199454921e-05,
+      "loss": 4.4443,
+      "step": 429300
+    },
+    {
+      "epoch": 3.6961949035812673,
+      "grad_norm": 2.875,
+      "learning_rate": 3.1823459747471494e-05,
+      "loss": 4.2424,
+      "step": 429350
+    },
+    {
+      "epoch": 3.696625344352617,
+      "grad_norm": 4.4375,
+      "learning_rate": 3.180364250440521e-05,
+      "loss": 4.4861,
+      "step": 429400
+    },
+    {
+      "epoch": 3.697055785123967,
+      "grad_norm": 3.109375,
+      "learning_rate": 3.17838302668048e-05,
+      "loss": 4.1681,
+      "step": 429450
+    },
+    {
+      "epoch": 3.6974862258953167,
+      "grad_norm": 4.25,
+      "learning_rate": 3.176402303612446e-05,
+      "loss": 4.7728,
+      "step": 429500
+    },
+    {
+      "epoch": 3.6979166666666665,
+      "grad_norm": 4.375,
+      "learning_rate": 3.1744220813818036e-05,
+      "loss": 4.06,
+      "step": 429550
+    },
+    {
+      "epoch": 3.6983471074380168,
+      "grad_norm": 1.9609375,
+      "learning_rate": 3.172442360133896e-05,
+      "loss": 4.3757,
+      "step": 429600
+    },
+    {
+      "epoch": 3.6987775482093666,
+      "grad_norm": 3.46875,
+      "learning_rate": 3.170463140014028e-05,
+      "loss": 4.4362,
+      "step": 429650
+    },
+    {
+      "epoch": 3.6992079889807163,
+      "grad_norm": 5.0,
+      "learning_rate": 3.168484421167471e-05,
+      "loss": 4.3744,
+      "step": 429700
+    },
+    {
+      "epoch": 3.699638429752066,
+      "grad_norm": 3.609375,
+      "learning_rate": 3.166506203739461e-05,
+      "loss": 4.4303,
+      "step": 429750
+    },
+    {
+      "epoch": 3.700068870523416,
+      "grad_norm": 3.625,
+      "learning_rate": 3.164528487875188e-05,
+      "loss": 4.492,
+      "step": 429800
+    },
+    {
+      "epoch": 3.7004993112947657,
+      "grad_norm": 3.6875,
+      "learning_rate": 3.1625512737198216e-05,
+      "loss": 4.1798,
+      "step": 429850
+    },
+    {
+      "epoch": 3.7009297520661155,
+      "grad_norm": 5.75,
+      "learning_rate": 3.16057456141848e-05,
+      "loss": 4.3871,
+      "step": 429900
+    },
+    {
+      "epoch": 3.7013601928374653,
+      "grad_norm": 4.71875,
+      "learning_rate": 3.1585983511162495e-05,
+      "loss": 4.2744,
+      "step": 429950
+    },
+    {
+      "epoch": 3.7017906336088156,
+      "grad_norm": 2.75,
+      "learning_rate": 3.156622642958181e-05,
+      "loss": 4.2747,
+      "step": 430000
+    },
+    {
+      "epoch": 3.7022210743801653,
+      "grad_norm": 3.78125,
+      "learning_rate": 3.154647437089284e-05,
+      "loss": 4.485,
+      "step": 430050
+    },
+    {
+      "epoch": 3.702651515151515,
+      "grad_norm": 1.7421875,
+      "learning_rate": 3.1526727336545405e-05,
+      "loss": 4.765,
+      "step": 430100
+    },
+    {
+      "epoch": 3.703081955922865,
+      "grad_norm": 2.15625,
+      "learning_rate": 3.150698532798888e-05,
+      "loss": 4.0358,
+      "step": 430150
+    },
+    {
+      "epoch": 3.7035123966942147,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.1487248346672205e-05,
+      "loss": 3.9946,
+      "step": 430200
+    },
+    {
+      "epoch": 3.703942837465565,
+      "grad_norm": 1.9453125,
+      "learning_rate": 3.146751639404412e-05,
+      "loss": 4.348,
+      "step": 430250
+    },
+    {
+      "epoch": 3.7043732782369148,
+      "grad_norm": 3.421875,
+      "learning_rate": 3.144778947155288e-05,
+      "loss": 4.3902,
+      "step": 430300
+    },
+    {
+      "epoch": 3.7048037190082646,
+      "grad_norm": 3.1875,
+      "learning_rate": 3.142806758064637e-05,
+      "loss": 4.1994,
+      "step": 430350
+    },
+    {
+      "epoch": 3.7052341597796143,
+      "grad_norm": 2.359375,
+      "learning_rate": 3.140835072277223e-05,
+      "loss": 4.4339,
+      "step": 430400
+    },
+    {
+      "epoch": 3.705664600550964,
+      "grad_norm": 5.0,
+      "learning_rate": 3.1388638899377534e-05,
+      "loss": 4.2263,
+      "step": 430450
+    },
+    {
+      "epoch": 3.706095041322314,
+      "grad_norm": 2.3125,
+      "learning_rate": 3.136893211190909e-05,
+      "loss": 4.3503,
+      "step": 430500
+    },
+    {
+      "epoch": 3.7065254820936637,
+      "grad_norm": 2.859375,
+      "learning_rate": 3.13492303618134e-05,
+      "loss": 3.8462,
+      "step": 430550
+    },
+    {
+      "epoch": 3.7069559228650135,
+      "grad_norm": 3.09375,
+      "learning_rate": 3.132953365053649e-05,
+      "loss": 4.2801,
+      "step": 430600
+    },
+    {
+      "epoch": 3.7073863636363638,
+      "grad_norm": 3.3125,
+      "learning_rate": 3.130984197952407e-05,
+      "loss": 4.2556,
+      "step": 430650
+    },
+    {
+      "epoch": 3.7078168044077136,
+      "grad_norm": 2.5,
+      "learning_rate": 3.129015535022145e-05,
+      "loss": 4.0256,
+      "step": 430700
+    },
+    {
+      "epoch": 3.7082472451790633,
+      "grad_norm": 2.546875,
+      "learning_rate": 3.127047376407356e-05,
+      "loss": 4.0735,
+      "step": 430750
+    },
+    {
+      "epoch": 3.708677685950413,
+      "grad_norm": 3.78125,
+      "learning_rate": 3.125079722252506e-05,
+      "loss": 4.3669,
+      "step": 430800
+    },
+    {
+      "epoch": 3.709108126721763,
+      "grad_norm": 5.84375,
+      "learning_rate": 3.123112572702013e-05,
+      "loss": 4.2698,
+      "step": 430850
+    },
+    {
+      "epoch": 3.709538567493113,
+      "grad_norm": 2.53125,
+      "learning_rate": 3.121145927900256e-05,
+      "loss": 4.3271,
+      "step": 430900
+    },
+    {
+      "epoch": 3.709969008264463,
+      "grad_norm": 3.140625,
+      "learning_rate": 3.1191797879915966e-05,
+      "loss": 4.0091,
+      "step": 430950
+    },
+    {
+      "epoch": 3.7103994490358128,
+      "grad_norm": 2.65625,
+      "learning_rate": 3.1172141531203314e-05,
+      "loss": 4.0807,
+      "step": 431000
+    },
+    {
+      "epoch": 3.7108298898071626,
+      "grad_norm": 2.9375,
+      "learning_rate": 3.115249023430736e-05,
+      "loss": 4.6708,
+      "step": 431050
+    },
+    {
+      "epoch": 3.7112603305785123,
+      "grad_norm": 1.609375,
+      "learning_rate": 3.113284399067052e-05,
+      "loss": 4.4226,
+      "step": 431100
+    },
+    {
+      "epoch": 3.711690771349862,
+      "grad_norm": 2.390625,
+      "learning_rate": 3.111320280173476e-05,
+      "loss": 4.3794,
+      "step": 431150
+    },
+    {
+      "epoch": 3.712121212121212,
+      "grad_norm": 4.59375,
+      "learning_rate": 3.109356666894171e-05,
+      "loss": 4.4494,
+      "step": 431200
+    },
+    {
+      "epoch": 3.7125516528925617,
+      "grad_norm": 1.90625,
+      "learning_rate": 3.1073935593732606e-05,
+      "loss": 4.2612,
+      "step": 431250
+    },
+    {
+      "epoch": 3.712982093663912,
+      "grad_norm": 3.015625,
+      "learning_rate": 3.1054309577548334e-05,
+      "loss": 4.3012,
+      "step": 431300
+    },
+    {
+      "epoch": 3.7134125344352618,
+      "grad_norm": 2.625,
+      "learning_rate": 3.103468862182935e-05,
+      "loss": 4.1803,
+      "step": 431350
+    },
+    {
+      "epoch": 3.7138429752066116,
+      "grad_norm": 2.578125,
+      "learning_rate": 3.101507272801589e-05,
+      "loss": 4.4064,
+      "step": 431400
+    },
+    {
+      "epoch": 3.7142734159779613,
+      "grad_norm": 2.390625,
+      "learning_rate": 3.099546189754766e-05,
+      "loss": 4.3279,
+      "step": 431450
+    },
+    {
+      "epoch": 3.714703856749311,
+      "grad_norm": 1.828125,
+      "learning_rate": 3.0975856131864065e-05,
+      "loss": 3.821,
+      "step": 431500
+    },
+    {
+      "epoch": 3.7151342975206614,
+      "grad_norm": 2.765625,
+      "learning_rate": 3.0956255432404117e-05,
+      "loss": 4.0931,
+      "step": 431550
+    },
+    {
+      "epoch": 3.715564738292011,
+      "grad_norm": 2.03125,
+      "learning_rate": 3.093665980060644e-05,
+      "loss": 4.6108,
+      "step": 431600
+    },
+    {
+      "epoch": 3.715995179063361,
+      "grad_norm": 2.53125,
+      "learning_rate": 3.091706923790936e-05,
+      "loss": 4.1827,
+      "step": 431650
+    },
+    {
+      "epoch": 3.7164256198347108,
+      "grad_norm": 2.515625,
+      "learning_rate": 3.08974837457508e-05,
+      "loss": 4.2806,
+      "step": 431700
+    },
+    {
+      "epoch": 3.7168560606060606,
+      "grad_norm": 3.515625,
+      "learning_rate": 3.087790332556819e-05,
+      "loss": 4.4051,
+      "step": 431750
+    },
+    {
+      "epoch": 3.7172865013774103,
+      "grad_norm": 3.40625,
+      "learning_rate": 3.08583279787988e-05,
+      "loss": 4.7127,
+      "step": 431800
+    },
+    {
+      "epoch": 3.71771694214876,
+      "grad_norm": 3.640625,
+      "learning_rate": 3.083875770687936e-05,
+      "loss": 4.0419,
+      "step": 431850
+    },
+    {
+      "epoch": 3.71814738292011,
+      "grad_norm": 3.0625,
+      "learning_rate": 3.0819192511246274e-05,
+      "loss": 4.2122,
+      "step": 431900
+    },
+    {
+      "epoch": 3.71857782369146,
+      "grad_norm": 3.5,
+      "learning_rate": 3.0799632393335677e-05,
+      "loss": 4.3132,
+      "step": 431950
+    },
+    {
+      "epoch": 3.71900826446281,
+      "grad_norm": 4.21875,
+      "learning_rate": 3.0780077354583134e-05,
+      "loss": 4.4361,
+      "step": 432000
+    },
+    {
+      "epoch": 3.71900826446281,
+      "eval_loss": 4.991272449493408,
+      "eval_runtime": 23.5652,
+      "eval_samples_per_second": 27.159,
+      "eval_steps_per_second": 13.579,
+      "eval_tts_loss": 7.483763048702864,
+      "step": 432000
+    },
+    {
+      "epoch": 3.7194387052341598,
+      "grad_norm": 3.3125,
+      "learning_rate": 3.0760527396423954e-05,
+      "loss": 4.3255,
+      "step": 432050
+    },
+    {
+      "epoch": 3.7198691460055096,
+      "grad_norm": 6.53125,
+      "learning_rate": 3.0740982520293125e-05,
+      "loss": 3.975,
+      "step": 432100
+    },
+    {
+      "epoch": 3.7202995867768593,
+      "grad_norm": 3.53125,
+      "learning_rate": 3.072144272762516e-05,
+      "loss": 4.3928,
+      "step": 432150
+    },
+    {
+      "epoch": 3.7207300275482096,
+      "grad_norm": 3.171875,
+      "learning_rate": 3.0701908019854256e-05,
+      "loss": 4.3393,
+      "step": 432200
+    },
+    {
+      "epoch": 3.7211604683195594,
+      "grad_norm": 2.5,
+      "learning_rate": 3.0682378398414194e-05,
+      "loss": 4.1379,
+      "step": 432250
+    },
+    {
+      "epoch": 3.721590909090909,
+      "grad_norm": 1.9609375,
+      "learning_rate": 3.066285386473838e-05,
+      "loss": 4.005,
+      "step": 432300
+    },
+    {
+      "epoch": 3.722021349862259,
+      "grad_norm": 1.9609375,
+      "learning_rate": 3.064333442025995e-05,
+      "loss": 4.2459,
+      "step": 432350
+    },
+    {
+      "epoch": 3.7224517906336088,
+      "grad_norm": 2.921875,
+      "learning_rate": 3.062382006641155e-05,
+      "loss": 3.925,
+      "step": 432400
+    },
+    {
+      "epoch": 3.7228822314049586,
+      "grad_norm": 3.28125,
+      "learning_rate": 3.060431080462548e-05,
+      "loss": 4.2413,
+      "step": 432450
+    },
+    {
+      "epoch": 3.7233126721763083,
+      "grad_norm": 5.0,
+      "learning_rate": 3.0584806636333693e-05,
+      "loss": 4.5442,
+      "step": 432500
+    },
+    {
+      "epoch": 3.723743112947658,
+      "grad_norm": 2.9375,
+      "learning_rate": 3.0565307562967746e-05,
+      "loss": 4.1942,
+      "step": 432550
+    },
+    {
+      "epoch": 3.7241735537190084,
+      "grad_norm": 5.71875,
+      "learning_rate": 3.05458135859588e-05,
+      "loss": 4.1234,
+      "step": 432600
+    },
+    {
+      "epoch": 3.724603994490358,
+      "grad_norm": 3.5625,
+      "learning_rate": 3.052632470673774e-05,
+      "loss": 4.0179,
+      "step": 432650
+    },
+    {
+      "epoch": 3.725034435261708,
+      "grad_norm": 2.171875,
+      "learning_rate": 3.050684092673497e-05,
+      "loss": 4.4334,
+      "step": 432700
+    },
+    {
+      "epoch": 3.7254648760330578,
+      "grad_norm": 3.171875,
+      "learning_rate": 3.0487362247380558e-05,
+      "loss": 4.1771,
+      "step": 432750
+    },
+    {
+      "epoch": 3.7258953168044076,
+      "grad_norm": 4.6875,
+      "learning_rate": 3.0467888670104194e-05,
+      "loss": 4.2968,
+      "step": 432800
+    },
+    {
+      "epoch": 3.726325757575758,
+      "grad_norm": 3.0625,
+      "learning_rate": 3.0448420196335203e-05,
+      "loss": 4.3296,
+      "step": 432850
+    },
+    {
+      "epoch": 3.7267561983471076,
+      "grad_norm": 1.890625,
+      "learning_rate": 3.0428956827502488e-05,
+      "loss": 4.0443,
+      "step": 432900
+    },
+    {
+      "epoch": 3.7271866391184574,
+      "grad_norm": 3.34375,
+      "learning_rate": 3.0409498565034688e-05,
+      "loss": 4.1331,
+      "step": 432950
+    },
+    {
+      "epoch": 3.727617079889807,
+      "grad_norm": 4.34375,
+      "learning_rate": 3.039004541035997e-05,
+      "loss": 4.5738,
+      "step": 433000
+    },
+    {
+      "epoch": 3.728047520661157,
+      "grad_norm": 2.96875,
+      "learning_rate": 3.037059736490614e-05,
+      "loss": 4.1179,
+      "step": 433050
+    },
+    {
+      "epoch": 3.7284779614325068,
+      "grad_norm": 2.421875,
+      "learning_rate": 3.035115443010066e-05,
+      "loss": 4.2804,
+      "step": 433100
+    },
+    {
+      "epoch": 3.7289084022038566,
+      "grad_norm": 3.03125,
+      "learning_rate": 3.0331716607370552e-05,
+      "loss": 4.435,
+      "step": 433150
+    },
+    {
+      "epoch": 3.7293388429752063,
+      "grad_norm": 3.171875,
+      "learning_rate": 3.031228389814259e-05,
+      "loss": 4.3343,
+      "step": 433200
+    },
+    {
+      "epoch": 3.7297692837465566,
+      "grad_norm": 6.9375,
+      "learning_rate": 3.0292856303843087e-05,
+      "loss": 4.3997,
+      "step": 433250
+    },
+    {
+      "epoch": 3.7301997245179064,
+      "grad_norm": 1.9921875,
+      "learning_rate": 3.027343382589788e-05,
+      "loss": 4.5484,
+      "step": 433300
+    },
+    {
+      "epoch": 3.730630165289256,
+      "grad_norm": 5.0,
+      "learning_rate": 3.025401646573265e-05,
+      "loss": 4.3565,
+      "step": 433350
+    },
+    {
+      "epoch": 3.731060606060606,
+      "grad_norm": 3.734375,
+      "learning_rate": 3.0234604224772543e-05,
+      "loss": 4.313,
+      "step": 433400
+    },
+    {
+      "epoch": 3.7314910468319558,
+      "grad_norm": 3.390625,
+      "learning_rate": 3.021519710444235e-05,
+      "loss": 4.0807,
+      "step": 433450
+    },
+    {
+      "epoch": 3.731921487603306,
+      "grad_norm": 3.15625,
+      "learning_rate": 3.01957951061666e-05,
+      "loss": 4.1778,
+      "step": 433500
+    },
+    {
+      "epoch": 3.732351928374656,
+      "grad_norm": 2.515625,
+      "learning_rate": 3.0176398231369285e-05,
+      "loss": 4.3013,
+      "step": 433550
+    },
+    {
+      "epoch": 3.7327823691460056,
+      "grad_norm": 6.28125,
+      "learning_rate": 3.0157006481474072e-05,
+      "loss": 4.745,
+      "step": 433600
+    },
+    {
+      "epoch": 3.7332128099173554,
+      "grad_norm": 2.859375,
+      "learning_rate": 3.0137619857904344e-05,
+      "loss": 4.2788,
+      "step": 433650
+    },
+    {
+      "epoch": 3.733643250688705,
+      "grad_norm": 4.8125,
+      "learning_rate": 3.0118238362083007e-05,
+      "loss": 4.1373,
+      "step": 433700
+    },
+    {
+      "epoch": 3.734073691460055,
+      "grad_norm": 2.46875,
+      "learning_rate": 3.0098861995432616e-05,
+      "loss": 4.4103,
+      "step": 433750
+    },
+    {
+      "epoch": 3.7345041322314048,
+      "grad_norm": 3.265625,
+      "learning_rate": 3.0079490759375362e-05,
+      "loss": 4.1113,
+      "step": 433800
+    },
+    {
+      "epoch": 3.7349345730027546,
+      "grad_norm": 3.078125,
+      "learning_rate": 3.0060124655333023e-05,
+      "loss": 4.3222,
+      "step": 433850
+    },
+    {
+      "epoch": 3.735365013774105,
+      "grad_norm": 2.0,
+      "learning_rate": 3.004076368472707e-05,
+      "loss": 4.2851,
+      "step": 433900
+    },
+    {
+      "epoch": 3.7357954545454546,
+      "grad_norm": 3.46875,
+      "learning_rate": 3.0021407848978554e-05,
+      "loss": 4.3138,
+      "step": 433950
+    },
+    {
+      "epoch": 3.7362258953168044,
+      "grad_norm": 4.0,
+      "learning_rate": 3.0002057149508133e-05,
+      "loss": 4.1978,
+      "step": 434000
+    },
+    {
+      "epoch": 3.736656336088154,
+      "grad_norm": 2.96875,
+      "learning_rate": 2.998271158773611e-05,
+      "loss": 4.6576,
+      "step": 434050
+    },
+    {
+      "epoch": 3.737086776859504,
+      "grad_norm": 2.875,
+      "learning_rate": 2.996337116508242e-05,
+      "loss": 4.314,
+      "step": 434100
+    },
+    {
+      "epoch": 3.737517217630854,
+      "grad_norm": 3.953125,
+      "learning_rate": 2.994403588296656e-05,
+      "loss": 4.3415,
+      "step": 434150
+    },
+    {
+      "epoch": 3.737947658402204,
+      "grad_norm": 4.0625,
+      "learning_rate": 2.9924705742807778e-05,
+      "loss": 4.3394,
+      "step": 434200
+    },
+    {
+      "epoch": 3.738378099173554,
+      "grad_norm": 1.7578125,
+      "learning_rate": 2.9905380746024846e-05,
+      "loss": 4.1042,
+      "step": 434250
+    },
+    {
+      "epoch": 3.7388085399449036,
+      "grad_norm": 1.671875,
+      "learning_rate": 2.9886060894036095e-05,
+      "loss": 4.4118,
+      "step": 434300
+    },
+    {
+      "epoch": 3.7392389807162534,
+      "grad_norm": 4.75,
+      "learning_rate": 2.9866746188259655e-05,
+      "loss": 4.0821,
+      "step": 434350
+    },
+    {
+      "epoch": 3.739669421487603,
+      "grad_norm": 4.65625,
+      "learning_rate": 2.9847436630113136e-05,
+      "loss": 4.2308,
+      "step": 434400
+    },
+    {
+      "epoch": 3.740099862258953,
+      "grad_norm": 3.8125,
+      "learning_rate": 2.9828132221013817e-05,
+      "loss": 4.1144,
+      "step": 434450
+    },
+    {
+      "epoch": 3.7405303030303028,
+      "grad_norm": 3.890625,
+      "learning_rate": 2.9808832962378664e-05,
+      "loss": 4.2559,
+      "step": 434500
+    },
+    {
+      "epoch": 3.740960743801653,
+      "grad_norm": 5.53125,
+      "learning_rate": 2.9789538855624098e-05,
+      "loss": 4.3842,
+      "step": 434550
+    },
+    {
+      "epoch": 3.741391184573003,
+      "grad_norm": 1.9609375,
+      "learning_rate": 2.9770249902166338e-05,
+      "loss": 4.4714,
+      "step": 434600
+    },
+    {
+      "epoch": 3.7418216253443526,
+      "grad_norm": 2.3125,
+      "learning_rate": 2.975096610342114e-05,
+      "loss": 4.2406,
+      "step": 434650
+    },
+    {
+      "epoch": 3.7422520661157024,
+      "grad_norm": 2.609375,
+      "learning_rate": 2.9731687460803894e-05,
+      "loss": 4.2844,
+      "step": 434700
+    },
+    {
+      "epoch": 3.7426825068870526,
+      "grad_norm": 2.21875,
+      "learning_rate": 2.9712413975729557e-05,
+      "loss": 4.4126,
+      "step": 434750
+    },
+    {
+      "epoch": 3.7431129476584024,
+      "grad_norm": 1.984375,
+      "learning_rate": 2.9693145649612876e-05,
+      "loss": 4.1002,
+      "step": 434800
+    },
+    {
+      "epoch": 3.743543388429752,
+      "grad_norm": 3.125,
+      "learning_rate": 2.9673882483867966e-05,
+      "loss": 4.5453,
+      "step": 434850
+    },
+    {
+      "epoch": 3.743973829201102,
+      "grad_norm": 4.21875,
+      "learning_rate": 2.965462447990881e-05,
+      "loss": 4.7175,
+      "step": 434900
+    },
+    {
+      "epoch": 3.744404269972452,
+      "grad_norm": 3.421875,
+      "learning_rate": 2.9635371639148858e-05,
+      "loss": 4.2061,
+      "step": 434950
+    },
+    {
+      "epoch": 3.7448347107438016,
+      "grad_norm": 4.84375,
+      "learning_rate": 2.9616123963001206e-05,
+      "loss": 3.9491,
+      "step": 435000
+    },
+    {
+      "epoch": 3.7448347107438016,
+      "eval_loss": 4.991069316864014,
+      "eval_runtime": 24.8181,
+      "eval_samples_per_second": 25.788,
+      "eval_steps_per_second": 12.894,
+      "eval_tts_loss": 7.4790225853498375,
+      "step": 435000
+    },
+    {
+      "epoch": 3.7452651515151514,
+      "grad_norm": 2.0625,
+      "learning_rate": 2.9596881452878687e-05,
+      "loss": 3.9077,
+      "step": 435050
+    },
+    {
+      "epoch": 3.745695592286501,
+      "grad_norm": 4.03125,
+      "learning_rate": 2.957764411019356e-05,
+      "loss": 4.2705,
+      "step": 435100
+    },
+    {
+      "epoch": 3.746126033057851,
+      "grad_norm": 3.140625,
+      "learning_rate": 2.95584119363578e-05,
+      "loss": 4.2979,
+      "step": 435150
+    },
+    {
+      "epoch": 3.746556473829201,
+      "grad_norm": 3.015625,
+      "learning_rate": 2.9539184932783083e-05,
+      "loss": 4.5898,
+      "step": 435200
+    },
+    {
+      "epoch": 3.746986914600551,
+      "grad_norm": 2.1875,
+      "learning_rate": 2.9519963100880588e-05,
+      "loss": 4.0995,
+      "step": 435250
+    },
+    {
+      "epoch": 3.747417355371901,
+      "grad_norm": 3.40625,
+      "learning_rate": 2.9500746442061165e-05,
+      "loss": 3.8345,
+      "step": 435300
+    },
+    {
+      "epoch": 3.7478477961432506,
+      "grad_norm": 5.59375,
+      "learning_rate": 2.948153495773527e-05,
+      "loss": 3.8499,
+      "step": 435350
+    },
+    {
+      "epoch": 3.748278236914601,
+      "grad_norm": 2.859375,
+      "learning_rate": 2.9462328649312944e-05,
+      "loss": 4.2361,
+      "step": 435400
+    },
+    {
+      "epoch": 3.7487086776859506,
+      "grad_norm": 2.484375,
+      "learning_rate": 2.9443127518203962e-05,
+      "loss": 4.0381,
+      "step": 435450
+    },
+    {
+      "epoch": 3.7491391184573004,
+      "grad_norm": 2.0625,
+      "learning_rate": 2.942393156581762e-05,
+      "loss": 4.11,
+      "step": 435500
+    },
+    {
+      "epoch": 3.74956955922865,
+      "grad_norm": 5.1875,
+      "learning_rate": 2.9404740793562857e-05,
+      "loss": 4.2611,
+      "step": 435550
+    },
+    {
+      "epoch": 3.75,
+      "grad_norm": 3.734375,
+      "learning_rate": 2.9385555202848215e-05,
+      "loss": 4.25,
+      "step": 435600
+    },
+    {
+      "epoch": 3.75043044077135,
+      "grad_norm": 4.4375,
+      "learning_rate": 2.936637479508191e-05,
+      "loss": 3.9644,
+      "step": 435650
+    },
+    {
+      "epoch": 3.7508608815426996,
+      "grad_norm": 3.296875,
+      "learning_rate": 2.9347199571671667e-05,
+      "loss": 4.2064,
+      "step": 435700
+    },
+    {
+      "epoch": 3.7512913223140494,
+      "grad_norm": 3.015625,
+      "learning_rate": 2.9328029534025014e-05,
+      "loss": 4.1764,
+      "step": 435750
+    },
+    {
+      "epoch": 3.751721763085399,
+      "grad_norm": 2.359375,
+      "learning_rate": 2.9308864683548966e-05,
+      "loss": 4.4975,
+      "step": 435800
+    },
+    {
+      "epoch": 3.7521522038567494,
+      "grad_norm": 2.828125,
+      "learning_rate": 2.9289705021650083e-05,
+      "loss": 4.2121,
+      "step": 435850
+    },
+    {
+      "epoch": 3.752582644628099,
+      "grad_norm": 2.890625,
+      "learning_rate": 2.9270550549734754e-05,
+      "loss": 4.3651,
+      "step": 435900
+    },
+    {
+      "epoch": 3.753013085399449,
+      "grad_norm": 2.03125,
+      "learning_rate": 2.9251401269208835e-05,
+      "loss": 4.3243,
+      "step": 435950
+    },
+    {
+      "epoch": 3.753443526170799,
+      "grad_norm": 8.875,
+      "learning_rate": 2.9232257181477805e-05,
+      "loss": 4.2977,
+      "step": 436000
+    },
+    {
+      "epoch": 3.753873966942149,
+      "grad_norm": 2.546875,
+      "learning_rate": 2.9213118287946905e-05,
+      "loss": 4.0136,
+      "step": 436050
+    },
+    {
+      "epoch": 3.754304407713499,
+      "grad_norm": 3.40625,
+      "learning_rate": 2.919398459002076e-05,
+      "loss": 3.9184,
+      "step": 436100
+    },
+    {
+      "epoch": 3.7547348484848486,
+      "grad_norm": 4.09375,
+      "learning_rate": 2.917485608910383e-05,
+      "loss": 3.9648,
+      "step": 436150
+    },
+    {
+      "epoch": 3.7551652892561984,
+      "grad_norm": 5.625,
+      "learning_rate": 2.9155732786600088e-05,
+      "loss": 4.4635,
+      "step": 436200
+    },
+    {
+      "epoch": 3.755595730027548,
+      "grad_norm": 4.09375,
+      "learning_rate": 2.913661468391312e-05,
+      "loss": 4.3398,
+      "step": 436250
+    },
+    {
+      "epoch": 3.756026170798898,
+      "grad_norm": 2.6875,
+      "learning_rate": 2.9117501782446177e-05,
+      "loss": 4.1965,
+      "step": 436300
+    },
+    {
+      "epoch": 3.756456611570248,
+      "grad_norm": 2.515625,
+      "learning_rate": 2.9098394083602097e-05,
+      "loss": 4.6177,
+      "step": 436350
+    },
+    {
+      "epoch": 3.7568870523415976,
+      "grad_norm": 3.09375,
+      "learning_rate": 2.9079291588783308e-05,
+      "loss": 4.3223,
+      "step": 436400
+    },
+    {
+      "epoch": 3.7573174931129474,
+      "grad_norm": 7.28125,
+      "learning_rate": 2.906019429939195e-05,
+      "loss": 4.4316,
+      "step": 436450
+    },
+    {
+      "epoch": 3.7577479338842976,
+      "grad_norm": 3.0625,
+      "learning_rate": 2.904110221682971e-05,
+      "loss": 3.8981,
+      "step": 436500
+    },
+    {
+      "epoch": 3.7581783746556474,
+      "grad_norm": 3.328125,
+      "learning_rate": 2.9022015342497866e-05,
+      "loss": 4.5275,
+      "step": 436550
+    },
+    {
+      "epoch": 3.758608815426997,
+      "grad_norm": 4.0,
+      "learning_rate": 2.9002933677797438e-05,
+      "loss": 4.8012,
+      "step": 436600
+    },
+    {
+      "epoch": 3.759039256198347,
+      "grad_norm": 4.0,
+      "learning_rate": 2.8983857224128897e-05,
+      "loss": 4.2455,
+      "step": 436650
+    },
+    {
+      "epoch": 3.7594696969696972,
+      "grad_norm": 1.6484375,
+      "learning_rate": 2.8964785982892407e-05,
+      "loss": 3.99,
+      "step": 436700
+    },
+    {
+      "epoch": 3.759900137741047,
+      "grad_norm": 4.71875,
+      "learning_rate": 2.8945719955487817e-05,
+      "loss": 4.393,
+      "step": 436750
+    },
+    {
+      "epoch": 3.760330578512397,
+      "grad_norm": 3.890625,
+      "learning_rate": 2.8926659143314515e-05,
+      "loss": 4.4553,
+      "step": 436800
+    },
+    {
+      "epoch": 3.7607610192837466,
+      "grad_norm": 3.21875,
+      "learning_rate": 2.89076035477715e-05,
+      "loss": 4.7125,
+      "step": 436850
+    },
+    {
+      "epoch": 3.7611914600550964,
+      "grad_norm": 3.765625,
+      "learning_rate": 2.888855317025744e-05,
+      "loss": 4.4291,
+      "step": 436900
+    },
+    {
+      "epoch": 3.761621900826446,
+      "grad_norm": 1.90625,
+      "learning_rate": 2.8869508012170566e-05,
+      "loss": 4.5483,
+      "step": 436950
+    },
+    {
+      "epoch": 3.762052341597796,
+      "grad_norm": 3.1875,
+      "learning_rate": 2.8850468074908745e-05,
+      "loss": 4.337,
+      "step": 437000
+    },
+    {
+      "epoch": 3.762482782369146,
+      "grad_norm": 3.859375,
+      "learning_rate": 2.8831433359869496e-05,
+      "loss": 4.722,
+      "step": 437050
+    },
+    {
+      "epoch": 3.762913223140496,
+      "grad_norm": 1.2421875,
+      "learning_rate": 2.8812403868449932e-05,
+      "loss": 4.155,
+      "step": 437100
+    },
+    {
+      "epoch": 3.763343663911846,
+      "grad_norm": 4.71875,
+      "learning_rate": 2.8793379602046753e-05,
+      "loss": 4.2119,
+      "step": 437150
+    },
+    {
+      "epoch": 3.7637741046831956,
+      "grad_norm": 4.75,
+      "learning_rate": 2.8774360562056312e-05,
+      "loss": 4.1054,
+      "step": 437200
+    },
+    {
+      "epoch": 3.7642045454545454,
+      "grad_norm": 3.4375,
+      "learning_rate": 2.8755346749874522e-05,
+      "loss": 3.8826,
+      "step": 437250
+    },
+    {
+      "epoch": 3.764634986225895,
+      "grad_norm": 5.53125,
+      "learning_rate": 2.8736338166897026e-05,
+      "loss": 4.2956,
+      "step": 437300
+    },
+    {
+      "epoch": 3.7650654269972454,
+      "grad_norm": 3.25,
+      "learning_rate": 2.8717334814519015e-05,
+      "loss": 4.2503,
+      "step": 437350
+    },
+    {
+      "epoch": 3.7654958677685952,
+      "grad_norm": 4.75,
+      "learning_rate": 2.8698336694135197e-05,
+      "loss": 4.3729,
+      "step": 437400
+    },
+    {
+      "epoch": 3.765926308539945,
+      "grad_norm": 5.0,
+      "learning_rate": 2.867934380714008e-05,
+      "loss": 4.1687,
+      "step": 437450
+    },
+    {
+      "epoch": 3.766356749311295,
+      "grad_norm": 5.03125,
+      "learning_rate": 2.8660356154927693e-05,
+      "loss": 4.1661,
+      "step": 437500
+    },
+    {
+      "epoch": 3.7667871900826446,
+      "grad_norm": 3.1875,
+      "learning_rate": 2.864137373889163e-05,
+      "loss": 4.561,
+      "step": 437550
+    },
+    {
+      "epoch": 3.7672176308539944,
+      "grad_norm": 2.546875,
+      "learning_rate": 2.862239656042527e-05,
+      "loss": 4.07,
+      "step": 437600
+    },
+    {
+      "epoch": 3.767648071625344,
+      "grad_norm": 2.828125,
+      "learning_rate": 2.8603424620921392e-05,
+      "loss": 4.0358,
+      "step": 437650
+    },
+    {
+      "epoch": 3.768078512396694,
+      "grad_norm": 3.8125,
+      "learning_rate": 2.85844579217725e-05,
+      "loss": 4.1929,
+      "step": 437700
+    },
+    {
+      "epoch": 3.7685089531680442,
+      "grad_norm": 3.078125,
+      "learning_rate": 2.856549646437078e-05,
+      "loss": 4.3856,
+      "step": 437750
+    },
+    {
+      "epoch": 3.768939393939394,
+      "grad_norm": 2.53125,
+      "learning_rate": 2.8546540250107922e-05,
+      "loss": 4.3432,
+      "step": 437800
+    },
+    {
+      "epoch": 3.769369834710744,
+      "grad_norm": 4.03125,
+      "learning_rate": 2.852758928037528e-05,
+      "loss": 4.416,
+      "step": 437850
+    },
+    {
+      "epoch": 3.7698002754820936,
+      "grad_norm": 2.296875,
+      "learning_rate": 2.8508643556563795e-05,
+      "loss": 4.2531,
+      "step": 437900
+    },
+    {
+      "epoch": 3.7702307162534434,
+      "grad_norm": 2.359375,
+      "learning_rate": 2.8489703080064035e-05,
+      "loss": 4.179,
+      "step": 437950
+    },
+    {
+      "epoch": 3.7706611570247937,
+      "grad_norm": 4.125,
+      "learning_rate": 2.8470767852266244e-05,
+      "loss": 4.3769,
+      "step": 438000
+    },
+    {
+      "epoch": 3.7706611570247937,
+      "eval_loss": 4.991232395172119,
+      "eval_runtime": 23.3738,
+      "eval_samples_per_second": 27.381,
+      "eval_steps_per_second": 13.691,
+      "eval_tts_loss": 7.4769983337970585,
+      "step": 438000
+    },
+    {
+      "epoch": 3.7710915977961434,
+      "grad_norm": 2.265625,
+      "learning_rate": 2.8451837874560182e-05,
+      "loss": 4.5347,
+      "step": 438050
+    },
+    {
+      "epoch": 3.7715220385674932,
+      "grad_norm": 1.90625,
+      "learning_rate": 2.8432913148335295e-05,
+      "loss": 4.492,
+      "step": 438100
+    },
+    {
+      "epoch": 3.771952479338843,
+      "grad_norm": 2.8125,
+      "learning_rate": 2.841399367498059e-05,
+      "loss": 4.6457,
+      "step": 438150
+    },
+    {
+      "epoch": 3.772382920110193,
+      "grad_norm": 3.0625,
+      "learning_rate": 2.839507945588473e-05,
+      "loss": 4.1685,
+      "step": 438200
+    },
+    {
+      "epoch": 3.7728133608815426,
+      "grad_norm": 2.5625,
+      "learning_rate": 2.837617049243595e-05,
+      "loss": 4.1531,
+      "step": 438250
+    },
+    {
+      "epoch": 3.7732438016528924,
+      "grad_norm": 3.453125,
+      "learning_rate": 2.8357266786022174e-05,
+      "loss": 4.1016,
+      "step": 438300
+    },
+    {
+      "epoch": 3.773674242424242,
+      "grad_norm": 3.546875,
+      "learning_rate": 2.8338368338030885e-05,
+      "loss": 4.4416,
+      "step": 438350
+    },
+    {
+      "epoch": 3.7741046831955924,
+      "grad_norm": 4.0625,
+      "learning_rate": 2.831947514984916e-05,
+      "loss": 4.3861,
+      "step": 438400
+    },
+    {
+      "epoch": 3.7745351239669422,
+      "grad_norm": 2.53125,
+      "learning_rate": 2.830058722286375e-05,
+      "loss": 4.3383,
+      "step": 438450
+    },
+    {
+      "epoch": 3.774965564738292,
+      "grad_norm": 4.03125,
+      "learning_rate": 2.8281704558460963e-05,
+      "loss": 4.3597,
+      "step": 438500
+    },
+    {
+      "epoch": 3.775396005509642,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.826282715802674e-05,
+      "loss": 4.188,
+      "step": 438550
+    },
+    {
+      "epoch": 3.7758264462809916,
+      "grad_norm": 2.390625,
+      "learning_rate": 2.8243955022946668e-05,
+      "loss": 4.4031,
+      "step": 438600
+    },
+    {
+      "epoch": 3.776256887052342,
+      "grad_norm": 3.65625,
+      "learning_rate": 2.8225088154605928e-05,
+      "loss": 3.9243,
+      "step": 438650
+    },
+    {
+      "epoch": 3.7766873278236917,
+      "grad_norm": 3.796875,
+      "learning_rate": 2.820622655438927e-05,
+      "loss": 4.5401,
+      "step": 438700
+    },
+    {
+      "epoch": 3.7771177685950414,
+      "grad_norm": 2.71875,
+      "learning_rate": 2.8187370223681132e-05,
+      "loss": 4.2161,
+      "step": 438750
+    },
+    {
+      "epoch": 3.7775482093663912,
+      "grad_norm": 4.5625,
+      "learning_rate": 2.816851916386547e-05,
+      "loss": 4.5565,
+      "step": 438800
+    },
+    {
+      "epoch": 3.777978650137741,
+      "grad_norm": 2.984375,
+      "learning_rate": 2.814967337632599e-05,
+      "loss": 4.1442,
+      "step": 438850
+    },
+    {
+      "epoch": 3.778409090909091,
+      "grad_norm": 2.25,
+      "learning_rate": 2.813083286244592e-05,
+      "loss": 4.2059,
+      "step": 438900
+    },
+    {
+      "epoch": 3.7788395316804406,
+      "grad_norm": 3.453125,
+      "learning_rate": 2.811199762360802e-05,
+      "loss": 4.1219,
+      "step": 438950
+    },
+    {
+      "epoch": 3.7792699724517904,
+      "grad_norm": 3.671875,
+      "learning_rate": 2.8093167661194864e-05,
+      "loss": 4.454,
+      "step": 439000
+    },
+    {
+      "epoch": 3.7797004132231407,
+      "grad_norm": 4.46875,
+      "learning_rate": 2.8074342976588485e-05,
+      "loss": 4.2363,
+      "step": 439050
+    },
+    {
+      "epoch": 3.7801308539944904,
+      "grad_norm": 2.03125,
+      "learning_rate": 2.805552357117056e-05,
+      "loss": 4.4363,
+      "step": 439100
+    },
+    {
+      "epoch": 3.7805612947658402,
+      "grad_norm": 4.1875,
+      "learning_rate": 2.8036709446322463e-05,
+      "loss": 4.3007,
+      "step": 439150
+    },
+    {
+      "epoch": 3.78099173553719,
+      "grad_norm": 5.125,
+      "learning_rate": 2.801790060342503e-05,
+      "loss": 4.4281,
+      "step": 439200
+    },
+    {
+      "epoch": 3.78142217630854,
+      "grad_norm": 3.625,
+      "learning_rate": 2.799909704385879e-05,
+      "loss": 4.4363,
+      "step": 439250
+    },
+    {
+      "epoch": 3.78185261707989,
+      "grad_norm": 3.84375,
+      "learning_rate": 2.7980298769003942e-05,
+      "loss": 4.2758,
+      "step": 439300
+    },
+    {
+      "epoch": 3.78228305785124,
+      "grad_norm": 3.25,
+      "learning_rate": 2.7961505780240205e-05,
+      "loss": 4.338,
+      "step": 439350
+    },
+    {
+      "epoch": 3.7827134986225897,
+      "grad_norm": 3.421875,
+      "learning_rate": 2.7942718078946962e-05,
+      "loss": 4.2511,
+      "step": 439400
+    },
+    {
+      "epoch": 3.7831439393939394,
+      "grad_norm": 2.34375,
+      "learning_rate": 2.7923935666503166e-05,
+      "loss": 4.0931,
+      "step": 439450
+    },
+    {
+      "epoch": 3.7835743801652892,
+      "grad_norm": 3.296875,
+      "learning_rate": 2.790515854428739e-05,
+      "loss": 3.9331,
+      "step": 439500
+    },
+    {
+      "epoch": 3.784004820936639,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.7886386713677892e-05,
+      "loss": 4.2368,
+      "step": 439550
+    },
+    {
+      "epoch": 3.784435261707989,
+      "grad_norm": 5.3125,
+      "learning_rate": 2.7867620176052456e-05,
+      "loss": 4.5077,
+      "step": 439600
+    },
+    {
+      "epoch": 3.7848657024793386,
+      "grad_norm": 2.484375,
+      "learning_rate": 2.784885893278849e-05,
+      "loss": 4.1218,
+      "step": 439650
+    },
+    {
+      "epoch": 3.785296143250689,
+      "grad_norm": 3.3125,
+      "learning_rate": 2.783010298526305e-05,
+      "loss": 4.2164,
+      "step": 439700
+    },
+    {
+      "epoch": 3.7857265840220387,
+      "grad_norm": 2.84375,
+      "learning_rate": 2.781135233485277e-05,
+      "loss": 4.6578,
+      "step": 439750
+    },
+    {
+      "epoch": 3.7861570247933884,
+      "grad_norm": 3.640625,
+      "learning_rate": 2.779260698293389e-05,
+      "loss": 4.5096,
+      "step": 439800
+    },
+    {
+      "epoch": 3.7865874655647382,
+      "grad_norm": 4.6875,
+      "learning_rate": 2.777386693088232e-05,
+      "loss": 4.2233,
+      "step": 439850
+    },
+    {
+      "epoch": 3.787017906336088,
+      "grad_norm": 6.34375,
+      "learning_rate": 2.7755132180073552e-05,
+      "loss": 4.2595,
+      "step": 439900
+    },
+    {
+      "epoch": 3.7874483471074383,
+      "grad_norm": 2.84375,
+      "learning_rate": 2.7736402731882582e-05,
+      "loss": 3.8967,
+      "step": 439950
+    },
+    {
+      "epoch": 3.787878787878788,
+      "grad_norm": 5.3125,
+      "learning_rate": 2.7717678587684205e-05,
+      "loss": 4.4487,
+      "step": 440000
+    },
+    {
+      "epoch": 3.788309228650138,
+      "grad_norm": 2.53125,
+      "learning_rate": 2.7698959748852692e-05,
+      "loss": 4.2939,
+      "step": 440050
+    },
+    {
+      "epoch": 3.7887396694214877,
+      "grad_norm": 2.59375,
+      "learning_rate": 2.768024621676195e-05,
+      "loss": 4.1282,
+      "step": 440100
+    },
+    {
+      "epoch": 3.7891701101928374,
+      "grad_norm": 4.34375,
+      "learning_rate": 2.766153799278557e-05,
+      "loss": 4.4936,
+      "step": 440150
+    },
+    {
+      "epoch": 3.7896005509641872,
+      "grad_norm": 5.4375,
+      "learning_rate": 2.7642835078296647e-05,
+      "loss": 4.5217,
+      "step": 440200
+    },
+    {
+      "epoch": 3.790030991735537,
+      "grad_norm": 2.234375,
+      "learning_rate": 2.7624137474667966e-05,
+      "loss": 4.4994,
+      "step": 440250
+    },
+    {
+      "epoch": 3.790461432506887,
+      "grad_norm": 3.359375,
+      "learning_rate": 2.760544518327186e-05,
+      "loss": 4.2678,
+      "step": 440300
+    },
+    {
+      "epoch": 3.790891873278237,
+      "grad_norm": 3.0,
+      "learning_rate": 2.758675820548029e-05,
+      "loss": 4.1454,
+      "step": 440350
+    },
+    {
+      "epoch": 3.791322314049587,
+      "grad_norm": 4.65625,
+      "learning_rate": 2.75680765426649e-05,
+      "loss": 4.0165,
+      "step": 440400
+    },
+    {
+      "epoch": 3.7917527548209367,
+      "grad_norm": 2.140625,
+      "learning_rate": 2.7549400196196884e-05,
+      "loss": 4.1414,
+      "step": 440450
+    },
+    {
+      "epoch": 3.7921831955922864,
+      "grad_norm": 2.71875,
+      "learning_rate": 2.7530729167446945e-05,
+      "loss": 4.3019,
+      "step": 440500
+    },
+    {
+      "epoch": 3.7926136363636362,
+      "grad_norm": 4.28125,
+      "learning_rate": 2.7512063457785596e-05,
+      "loss": 4.6083,
+      "step": 440550
+    },
+    {
+      "epoch": 3.7930440771349865,
+      "grad_norm": 3.265625,
+      "learning_rate": 2.7493403068582823e-05,
+      "loss": 4.026,
+      "step": 440600
+    },
+    {
+      "epoch": 3.7934745179063363,
+      "grad_norm": 5.34375,
+      "learning_rate": 2.747474800120824e-05,
+      "loss": 3.9298,
+      "step": 440650
+    },
+    {
+      "epoch": 3.793904958677686,
+      "grad_norm": 1.7421875,
+      "learning_rate": 2.745609825703117e-05,
+      "loss": 4.2406,
+      "step": 440700
+    },
+    {
+      "epoch": 3.794335399449036,
+      "grad_norm": 3.953125,
+      "learning_rate": 2.7437453837420378e-05,
+      "loss": 4.3084,
+      "step": 440750
+    },
+    {
+      "epoch": 3.7947658402203857,
+      "grad_norm": 4.1875,
+      "learning_rate": 2.7418814743744314e-05,
+      "loss": 4.2397,
+      "step": 440800
+    },
+    {
+      "epoch": 3.7951962809917354,
+      "grad_norm": 2.40625,
+      "learning_rate": 2.7400180977371126e-05,
+      "loss": 4.3035,
+      "step": 440850
+    },
+    {
+      "epoch": 3.7956267217630852,
+      "grad_norm": 3.65625,
+      "learning_rate": 2.7381552539668442e-05,
+      "loss": 4.3716,
+      "step": 440900
+    },
+    {
+      "epoch": 3.796057162534435,
+      "grad_norm": 2.359375,
+      "learning_rate": 2.736292943200356e-05,
+      "loss": 4.4056,
+      "step": 440950
+    },
+    {
+      "epoch": 3.7964876033057853,
+      "grad_norm": 3.625,
+      "learning_rate": 2.7344311655743382e-05,
+      "loss": 4.2118,
+      "step": 441000
+    },
+    {
+      "epoch": 3.7964876033057853,
+      "eval_loss": 4.9910712242126465,
+      "eval_runtime": 24.0559,
+      "eval_samples_per_second": 26.605,
+      "eval_steps_per_second": 13.302,
+      "eval_tts_loss": 7.471262435327908,
+      "step": 441000
+    },
+    {
+      "epoch": 3.796918044077135,
+      "grad_norm": 3.375,
+      "learning_rate": 2.7325699212254363e-05,
+      "loss": 4.5749,
+      "step": 441050
+    },
+    {
+      "epoch": 3.797348484848485,
+      "grad_norm": 3.65625,
+      "learning_rate": 2.73070921029027e-05,
+      "loss": 4.0224,
+      "step": 441100
+    },
+    {
+      "epoch": 3.7977789256198347,
+      "grad_norm": 4.125,
+      "learning_rate": 2.7288490329054073e-05,
+      "loss": 4.1009,
+      "step": 441150
+    },
+    {
+      "epoch": 3.7982093663911844,
+      "grad_norm": 2.6875,
+      "learning_rate": 2.7269893892073804e-05,
+      "loss": 4.4229,
+      "step": 441200
+    },
+    {
+      "epoch": 3.7986398071625347,
+      "grad_norm": 2.484375,
+      "learning_rate": 2.725130279332685e-05,
+      "loss": 4.2913,
+      "step": 441250
+    },
+    {
+      "epoch": 3.7990702479338845,
+      "grad_norm": 4.25,
+      "learning_rate": 2.7232717034177746e-05,
+      "loss": 4.1804,
+      "step": 441300
+    },
+    {
+      "epoch": 3.7995006887052343,
+      "grad_norm": 1.8671875,
+      "learning_rate": 2.7214136615990626e-05,
+      "loss": 3.6511,
+      "step": 441350
+    },
+    {
+      "epoch": 3.799931129476584,
+      "grad_norm": 1.875,
+      "learning_rate": 2.71955615401293e-05,
+      "loss": 3.7034,
+      "step": 441400
+    },
+    {
+      "epoch": 3.800361570247934,
+      "grad_norm": 4.34375,
+      "learning_rate": 2.7176991807957142e-05,
+      "loss": 4.2727,
+      "step": 441450
+    },
+    {
+      "epoch": 3.8007920110192837,
+      "grad_norm": 5.25,
+      "learning_rate": 2.7158427420837042e-05,
+      "loss": 4.4436,
+      "step": 441500
+    },
+    {
+      "epoch": 3.8012224517906334,
+      "grad_norm": 2.34375,
+      "learning_rate": 2.7139868380131673e-05,
+      "loss": 4.2093,
+      "step": 441550
+    },
+    {
+      "epoch": 3.8016528925619832,
+      "grad_norm": 2.515625,
+      "learning_rate": 2.7121314687203203e-05,
+      "loss": 4.1859,
+      "step": 441600
+    },
+    {
+      "epoch": 3.8020833333333335,
+      "grad_norm": 2.53125,
+      "learning_rate": 2.710276634341341e-05,
+      "loss": 4.2751,
+      "step": 441650
+    },
+    {
+      "epoch": 3.8025137741046833,
+      "grad_norm": 3.078125,
+      "learning_rate": 2.7084223350123783e-05,
+      "loss": 4.4252,
+      "step": 441700
+    },
+    {
+      "epoch": 3.802944214876033,
+      "grad_norm": 4.53125,
+      "learning_rate": 2.706568570869521e-05,
+      "loss": 4.5279,
+      "step": 441750
+    },
+    {
+      "epoch": 3.803374655647383,
+      "grad_norm": 1.875,
+      "learning_rate": 2.7047153420488425e-05,
+      "loss": 4.2245,
+      "step": 441800
+    },
+    {
+      "epoch": 3.8038050964187327,
+      "grad_norm": 2.6875,
+      "learning_rate": 2.7028626486863617e-05,
+      "loss": 4.3283,
+      "step": 441850
+    },
+    {
+      "epoch": 3.804235537190083,
+      "grad_norm": 3.140625,
+      "learning_rate": 2.7010104909180624e-05,
+      "loss": 4.2854,
+      "step": 441900
+    },
+    {
+      "epoch": 3.8046659779614327,
+      "grad_norm": 5.0625,
+      "learning_rate": 2.6991588688798853e-05,
+      "loss": 4.2609,
+      "step": 441950
+    },
+    {
+      "epoch": 3.8050964187327825,
+      "grad_norm": 3.6875,
+      "learning_rate": 2.6973077827077442e-05,
+      "loss": 4.4615,
+      "step": 442000
+    },
+    {
+      "epoch": 3.8055268595041323,
+      "grad_norm": 5.09375,
+      "learning_rate": 2.6954572325374948e-05,
+      "loss": 4.2627,
+      "step": 442050
+    },
+    {
+      "epoch": 3.805957300275482,
+      "grad_norm": 4.96875,
+      "learning_rate": 2.693607218504972e-05,
+      "loss": 4.6391,
+      "step": 442100
+    },
+    {
+      "epoch": 3.806387741046832,
+      "grad_norm": 3.5,
+      "learning_rate": 2.6917577407459583e-05,
+      "loss": 4.4179,
+      "step": 442150
+    },
+    {
+      "epoch": 3.8068181818181817,
+      "grad_norm": 2.828125,
+      "learning_rate": 2.6899087993961992e-05,
+      "loss": 4.4048,
+      "step": 442200
+    },
+    {
+      "epoch": 3.8072486225895315,
+      "grad_norm": 2.109375,
+      "learning_rate": 2.688060394591413e-05,
+      "loss": 4.4438,
+      "step": 442250
+    },
+    {
+      "epoch": 3.8076790633608817,
+      "grad_norm": 2.75,
+      "learning_rate": 2.6862125264672578e-05,
+      "loss": 4.5975,
+      "step": 442300
+    },
+    {
+      "epoch": 3.8081095041322315,
+      "grad_norm": 1.5234375,
+      "learning_rate": 2.684365195159365e-05,
+      "loss": 4.197,
+      "step": 442350
+    },
+    {
+      "epoch": 3.8085399449035813,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.682518400803329e-05,
+      "loss": 4.0516,
+      "step": 442400
+    },
+    {
+      "epoch": 3.808970385674931,
+      "grad_norm": 1.7265625,
+      "learning_rate": 2.6806721435346992e-05,
+      "loss": 4.6244,
+      "step": 442450
+    },
+    {
+      "epoch": 3.809400826446281,
+      "grad_norm": 3.484375,
+      "learning_rate": 2.678826423488986e-05,
+      "loss": 4.2073,
+      "step": 442500
+    },
+    {
+      "epoch": 3.809831267217631,
+      "grad_norm": 2.984375,
+      "learning_rate": 2.6769812408016616e-05,
+      "loss": 4.5374,
+      "step": 442550
+    },
+    {
+      "epoch": 3.810261707988981,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.675136595608154e-05,
+      "loss": 4.2841,
+      "step": 442600
+    },
+    {
+      "epoch": 3.8106921487603307,
+      "grad_norm": 4.4375,
+      "learning_rate": 2.6732924880438648e-05,
+      "loss": 4.1611,
+      "step": 442650
+    },
+    {
+      "epoch": 3.8111225895316805,
+      "grad_norm": 3.390625,
+      "learning_rate": 2.6714489182441427e-05,
+      "loss": 4.5294,
+      "step": 442700
+    },
+    {
+      "epoch": 3.8115530303030303,
+      "grad_norm": 2.75,
+      "learning_rate": 2.669605886344303e-05,
+      "loss": 4.4323,
+      "step": 442750
+    },
+    {
+      "epoch": 3.81198347107438,
+      "grad_norm": 5.375,
+      "learning_rate": 2.6677633924796186e-05,
+      "loss": 4.4475,
+      "step": 442800
+    },
+    {
+      "epoch": 3.81241391184573,
+      "grad_norm": 2.984375,
+      "learning_rate": 2.6659214367853248e-05,
+      "loss": 4.5494,
+      "step": 442850
+    },
+    {
+      "epoch": 3.8128443526170797,
+      "grad_norm": 4.5,
+      "learning_rate": 2.6640800193966165e-05,
+      "loss": 4.3748,
+      "step": 442900
+    },
+    {
+      "epoch": 3.81327479338843,
+      "grad_norm": 4.4375,
+      "learning_rate": 2.6622391404486523e-05,
+      "loss": 4.3392,
+      "step": 442950
+    },
+    {
+      "epoch": 3.8137052341597797,
+      "grad_norm": 2.84375,
+      "learning_rate": 2.6603988000765523e-05,
+      "loss": 4.3399,
+      "step": 443000
+    },
+    {
+      "epoch": 3.8141356749311295,
+      "grad_norm": 2.015625,
+      "learning_rate": 2.6585589984153824e-05,
+      "loss": 4.4275,
+      "step": 443050
+    },
+    {
+      "epoch": 3.8145661157024793,
+      "grad_norm": 4.21875,
+      "learning_rate": 2.6567197356001883e-05,
+      "loss": 4.0843,
+      "step": 443100
+    },
+    {
+      "epoch": 3.814996556473829,
+      "grad_norm": 3.21875,
+      "learning_rate": 2.6548810117659673e-05,
+      "loss": 4.4453,
+      "step": 443150
+    },
+    {
+      "epoch": 3.8154269972451793,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.653042827047674e-05,
+      "loss": 4.1612,
+      "step": 443200
+    },
+    {
+      "epoch": 3.815857438016529,
+      "grad_norm": 3.40625,
+      "learning_rate": 2.6512051815802354e-05,
+      "loss": 4.4573,
+      "step": 443250
+    },
+    {
+      "epoch": 3.816287878787879,
+      "grad_norm": 4.03125,
+      "learning_rate": 2.6493680754985184e-05,
+      "loss": 4.8011,
+      "step": 443300
+    },
+    {
+      "epoch": 3.8167183195592287,
+      "grad_norm": 4.71875,
+      "learning_rate": 2.647531508937373e-05,
+      "loss": 4.6141,
+      "step": 443350
+    },
+    {
+      "epoch": 3.8171487603305785,
+      "grad_norm": 1.7578125,
+      "learning_rate": 2.6456954820315948e-05,
+      "loss": 4.0716,
+      "step": 443400
+    },
+    {
+      "epoch": 3.8175792011019283,
+      "grad_norm": 3.515625,
+      "learning_rate": 2.643859994915946e-05,
+      "loss": 4.3812,
+      "step": 443450
+    },
+    {
+      "epoch": 3.818009641873278,
+      "grad_norm": 1.078125,
+      "learning_rate": 2.6420250477251463e-05,
+      "loss": 4.2029,
+      "step": 443500
+    },
+    {
+      "epoch": 3.818440082644628,
+      "grad_norm": 4.0,
+      "learning_rate": 2.640190640593877e-05,
+      "loss": 4.2339,
+      "step": 443550
+    },
+    {
+      "epoch": 3.818870523415978,
+      "grad_norm": 1.5,
+      "learning_rate": 2.638356773656776e-05,
+      "loss": 4.5809,
+      "step": 443600
+    },
+    {
+      "epoch": 3.819300964187328,
+      "grad_norm": 3.546875,
+      "learning_rate": 2.6365234470484524e-05,
+      "loss": 4.0893,
+      "step": 443650
+    },
+    {
+      "epoch": 3.8197314049586777,
+      "grad_norm": 4.0625,
+      "learning_rate": 2.634690660903465e-05,
+      "loss": 4.9535,
+      "step": 443700
+    },
+    {
+      "epoch": 3.8201618457300275,
+      "grad_norm": 4.8125,
+      "learning_rate": 2.632858415356333e-05,
+      "loss": 4.1552,
+      "step": 443750
+    },
+    {
+      "epoch": 3.8205922865013773,
+      "grad_norm": 3.9375,
+      "learning_rate": 2.6310267105415477e-05,
+      "loss": 4.5734,
+      "step": 443800
+    },
+    {
+      "epoch": 3.8210227272727275,
+      "grad_norm": 4.21875,
+      "learning_rate": 2.629195546593545e-05,
+      "loss": 4.5253,
+      "step": 443850
+    },
+    {
+      "epoch": 3.8214531680440773,
+      "grad_norm": 2.65625,
+      "learning_rate": 2.6273649236467278e-05,
+      "loss": 4.3029,
+      "step": 443900
+    },
+    {
+      "epoch": 3.821883608815427,
+      "grad_norm": 4.9375,
+      "learning_rate": 2.6255348418354652e-05,
+      "loss": 4.3019,
+      "step": 443950
+    },
+    {
+      "epoch": 3.822314049586777,
+      "grad_norm": 1.828125,
+      "learning_rate": 2.623705301294078e-05,
+      "loss": 4.3868,
+      "step": 444000
+    },
+    {
+      "epoch": 3.822314049586777,
+      "eval_loss": 4.990033149719238,
+      "eval_runtime": 24.0036,
+      "eval_samples_per_second": 26.663,
+      "eval_steps_per_second": 13.331,
+      "eval_tts_loss": 7.464097043278323,
+      "step": 444000
+    },
+    {
+      "epoch": 3.8227444903581267,
+      "grad_norm": 2.296875,
+      "learning_rate": 2.6218763021568514e-05,
+      "loss": 3.7023,
+      "step": 444050
+    },
+    {
+      "epoch": 3.8231749311294765,
+      "grad_norm": 2.75,
+      "learning_rate": 2.62004784455803e-05,
+      "loss": 4.2649,
+      "step": 444100
+    },
+    {
+      "epoch": 3.8236053719008263,
+      "grad_norm": 4.84375,
+      "learning_rate": 2.618219928631819e-05,
+      "loss": 4.4702,
+      "step": 444150
+    },
+    {
+      "epoch": 3.824035812672176,
+      "grad_norm": 4.1875,
+      "learning_rate": 2.616392554512379e-05,
+      "loss": 4.4969,
+      "step": 444200
+    },
+    {
+      "epoch": 3.8244662534435263,
+      "grad_norm": 1.71875,
+      "learning_rate": 2.614565722333844e-05,
+      "loss": 4.12,
+      "step": 444250
+    },
+    {
+      "epoch": 3.824896694214876,
+      "grad_norm": 2.984375,
+      "learning_rate": 2.6127394322302935e-05,
+      "loss": 4.0162,
+      "step": 444300
+    },
+    {
+      "epoch": 3.825327134986226,
+      "grad_norm": 2.34375,
+      "learning_rate": 2.610913684335774e-05,
+      "loss": 3.9894,
+      "step": 444350
+    },
+    {
+      "epoch": 3.8257575757575757,
+      "grad_norm": 3.71875,
+      "learning_rate": 2.609088478784293e-05,
+      "loss": 4.4528,
+      "step": 444400
+    },
+    {
+      "epoch": 3.8261880165289255,
+      "grad_norm": 3.625,
+      "learning_rate": 2.6072638157098115e-05,
+      "loss": 4.411,
+      "step": 444450
+    },
+    {
+      "epoch": 3.8266184573002757,
+      "grad_norm": 3.8125,
+      "learning_rate": 2.6054396952462645e-05,
+      "loss": 4.6046,
+      "step": 444500
+    },
+    {
+      "epoch": 3.8270488980716255,
+      "grad_norm": 2.65625,
+      "learning_rate": 2.6036161175275365e-05,
+      "loss": 4.5418,
+      "step": 444550
+    },
+    {
+      "epoch": 3.8274793388429753,
+      "grad_norm": 4.28125,
+      "learning_rate": 2.601793082687466e-05,
+      "loss": 4.1446,
+      "step": 444600
+    },
+    {
+      "epoch": 3.827909779614325,
+      "grad_norm": 2.125,
+      "learning_rate": 2.599970590859868e-05,
+      "loss": 4.7732,
+      "step": 444650
+    },
+    {
+      "epoch": 3.828340220385675,
+      "grad_norm": 2.0,
+      "learning_rate": 2.5981486421785074e-05,
+      "loss": 4.4962,
+      "step": 444700
+    },
+    {
+      "epoch": 3.8287706611570247,
+      "grad_norm": 4.09375,
+      "learning_rate": 2.5963272367771073e-05,
+      "loss": 4.4409,
+      "step": 444750
+    },
+    {
+      "epoch": 3.8292011019283745,
+      "grad_norm": 3.59375,
+      "learning_rate": 2.594506374789365e-05,
+      "loss": 4.2017,
+      "step": 444800
+    },
+    {
+      "epoch": 3.8296315426997243,
+      "grad_norm": 3.234375,
+      "learning_rate": 2.5926860563489164e-05,
+      "loss": 3.9821,
+      "step": 444850
+    },
+    {
+      "epoch": 3.8300619834710745,
+      "grad_norm": 3.734375,
+      "learning_rate": 2.5908662815893757e-05,
+      "loss": 3.932,
+      "step": 444900
+    },
+    {
+      "epoch": 3.8304924242424243,
+      "grad_norm": 2.5,
+      "learning_rate": 2.58904705064431e-05,
+      "loss": 4.1595,
+      "step": 444950
+    },
+    {
+      "epoch": 3.830922865013774,
+      "grad_norm": 4.90625,
+      "learning_rate": 2.587228363647245e-05,
+      "loss": 4.1939,
+      "step": 445000
+    },
+    {
+      "epoch": 3.831353305785124,
+      "grad_norm": 4.875,
+      "learning_rate": 2.585410220731669e-05,
+      "loss": 3.9153,
+      "step": 445050
+    },
+    {
+      "epoch": 3.8317837465564737,
+      "grad_norm": 3.390625,
+      "learning_rate": 2.5835926220310304e-05,
+      "loss": 4.3097,
+      "step": 445100
+    },
+    {
+      "epoch": 3.832214187327824,
+      "grad_norm": 2.625,
+      "learning_rate": 2.5817755676787327e-05,
+      "loss": 4.4918,
+      "step": 445150
+    },
+    {
+      "epoch": 3.8326446280991737,
+      "grad_norm": 6.6875,
+      "learning_rate": 2.5799590578081513e-05,
+      "loss": 4.1698,
+      "step": 445200
+    },
+    {
+      "epoch": 3.8330750688705235,
+      "grad_norm": 3.75,
+      "learning_rate": 2.5781430925526107e-05,
+      "loss": 4.4249,
+      "step": 445250
+    },
+    {
+      "epoch": 3.8335055096418733,
+      "grad_norm": 3.0,
+      "learning_rate": 2.5763276720453975e-05,
+      "loss": 4.2956,
+      "step": 445300
+    },
+    {
+      "epoch": 3.833935950413223,
+      "grad_norm": 2.109375,
+      "learning_rate": 2.5745127964197613e-05,
+      "loss": 4.1823,
+      "step": 445350
+    },
+    {
+      "epoch": 3.834366391184573,
+      "grad_norm": 5.1875,
+      "learning_rate": 2.572698465808909e-05,
+      "loss": 4.2439,
+      "step": 445400
+    },
+    {
+      "epoch": 3.8347968319559227,
+      "grad_norm": 2.625,
+      "learning_rate": 2.5708846803460074e-05,
+      "loss": 4.2578,
+      "step": 445450
+    },
+    {
+      "epoch": 3.8352272727272725,
+      "grad_norm": 1.6875,
+      "learning_rate": 2.569071440164188e-05,
+      "loss": 4.0689,
+      "step": 445500
+    },
+    {
+      "epoch": 3.8356577134986227,
+      "grad_norm": 2.046875,
+      "learning_rate": 2.567258745396537e-05,
+      "loss": 4.2629,
+      "step": 445550
+    },
+    {
+      "epoch": 3.8360881542699725,
+      "grad_norm": 2.640625,
+      "learning_rate": 2.565446596176102e-05,
+      "loss": 4.3783,
+      "step": 445600
+    },
+    {
+      "epoch": 3.8365185950413223,
+      "grad_norm": 3.265625,
+      "learning_rate": 2.5636349926358916e-05,
+      "loss": 4.3314,
+      "step": 445650
+    },
+    {
+      "epoch": 3.836949035812672,
+      "grad_norm": 6.875,
+      "learning_rate": 2.5618239349088723e-05,
+      "loss": 4.6774,
+      "step": 445700
+    },
+    {
+      "epoch": 3.837379476584022,
+      "grad_norm": 2.21875,
+      "learning_rate": 2.5600134231279706e-05,
+      "loss": 4.1236,
+      "step": 445750
+    },
+    {
+      "epoch": 3.837809917355372,
+      "grad_norm": 4.53125,
+      "learning_rate": 2.5582034574260782e-05,
+      "loss": 4.3099,
+      "step": 445800
+    },
+    {
+      "epoch": 3.838240358126722,
+      "grad_norm": 3.40625,
+      "learning_rate": 2.5563940379360406e-05,
+      "loss": 4.1484,
+      "step": 445850
+    },
+    {
+      "epoch": 3.8386707988980717,
+      "grad_norm": 3.078125,
+      "learning_rate": 2.554585164790666e-05,
+      "loss": 4.4937,
+      "step": 445900
+    },
+    {
+      "epoch": 3.8391012396694215,
+      "grad_norm": 1.3359375,
+      "learning_rate": 2.55277683812272e-05,
+      "loss": 4.2706,
+      "step": 445950
+    },
+    {
+      "epoch": 3.8395316804407713,
+      "grad_norm": 3.21875,
+      "learning_rate": 2.550969058064928e-05,
+      "loss": 4.4372,
+      "step": 446000
+    },
+    {
+      "epoch": 3.839962121212121,
+      "grad_norm": 2.390625,
+      "learning_rate": 2.5491618247499837e-05,
+      "loss": 4.0736,
+      "step": 446050
+    },
+    {
+      "epoch": 3.840392561983471,
+      "grad_norm": 3.03125,
+      "learning_rate": 2.5473551383105344e-05,
+      "loss": 4.0506,
+      "step": 446100
+    },
+    {
+      "epoch": 3.8408230027548207,
+      "grad_norm": 2.640625,
+      "learning_rate": 2.5455489988791758e-05,
+      "loss": 4.1714,
+      "step": 446150
+    },
+    {
+      "epoch": 3.841253443526171,
+      "grad_norm": 4.21875,
+      "learning_rate": 2.5437434065884856e-05,
+      "loss": 4.0006,
+      "step": 446200
+    },
+    {
+      "epoch": 3.8416838842975207,
+      "grad_norm": 3.40625,
+      "learning_rate": 2.5419383615709868e-05,
+      "loss": 4.2765,
+      "step": 446250
+    },
+    {
+      "epoch": 3.8421143250688705,
+      "grad_norm": 2.703125,
+      "learning_rate": 2.540133863959162e-05,
+      "loss": 4.3294,
+      "step": 446300
+    },
+    {
+      "epoch": 3.8425447658402203,
+      "grad_norm": 4.84375,
+      "learning_rate": 2.538329913885469e-05,
+      "loss": 4.0776,
+      "step": 446350
+    },
+    {
+      "epoch": 3.84297520661157,
+      "grad_norm": 3.75,
+      "learning_rate": 2.536526511482301e-05,
+      "loss": 4.3104,
+      "step": 446400
+    },
+    {
+      "epoch": 3.8434056473829203,
+      "grad_norm": 2.453125,
+      "learning_rate": 2.5347236568820275e-05,
+      "loss": 4.222,
+      "step": 446450
+    },
+    {
+      "epoch": 3.84383608815427,
+      "grad_norm": 1.0703125,
+      "learning_rate": 2.5329213502169792e-05,
+      "loss": 4.2919,
+      "step": 446500
+    },
+    {
+      "epoch": 3.84426652892562,
+      "grad_norm": 1.4609375,
+      "learning_rate": 2.5311195916194362e-05,
+      "loss": 4.271,
+      "step": 446550
+    },
+    {
+      "epoch": 3.8446969696969697,
+      "grad_norm": 4.28125,
+      "learning_rate": 2.5293183812216458e-05,
+      "loss": 4.2928,
+      "step": 446600
+    },
+    {
+      "epoch": 3.8451274104683195,
+      "grad_norm": 3.09375,
+      "learning_rate": 2.5275177191558143e-05,
+      "loss": 3.933,
+      "step": 446650
+    },
+    {
+      "epoch": 3.8455578512396693,
+      "grad_norm": 4.84375,
+      "learning_rate": 2.5257176055540998e-05,
+      "loss": 3.9075,
+      "step": 446700
+    },
+    {
+      "epoch": 3.845988292011019,
+      "grad_norm": 4.25,
+      "learning_rate": 2.523918040548635e-05,
+      "loss": 4.2498,
+      "step": 446750
+    },
+    {
+      "epoch": 3.846418732782369,
+      "grad_norm": 3.21875,
+      "learning_rate": 2.5221190242715022e-05,
+      "loss": 4.2309,
+      "step": 446800
+    },
+    {
+      "epoch": 3.846849173553719,
+      "grad_norm": 3.515625,
+      "learning_rate": 2.5203205568547427e-05,
+      "loss": 4.3565,
+      "step": 446850
+    },
+    {
+      "epoch": 3.847279614325069,
+      "grad_norm": 2.1875,
+      "learning_rate": 2.5185226384303617e-05,
+      "loss": 4.1702,
+      "step": 446900
+    },
+    {
+      "epoch": 3.8477100550964187,
+      "grad_norm": 4.3125,
+      "learning_rate": 2.5167252691303223e-05,
+      "loss": 4.217,
+      "step": 446950
+    },
+    {
+      "epoch": 3.8481404958677685,
+      "grad_norm": 2.25,
+      "learning_rate": 2.5149284490865445e-05,
+      "loss": 4.3949,
+      "step": 447000
+    },
+    {
+      "epoch": 3.8481404958677685,
+      "eval_loss": 4.990220546722412,
+      "eval_runtime": 23.9495,
+      "eval_samples_per_second": 26.723,
+      "eval_steps_per_second": 13.361,
+      "eval_tts_loss": 7.46313589566455,
+      "step": 447000
+    },
+    {
+      "epoch": 3.8485709366391183,
+      "grad_norm": 4.09375,
+      "learning_rate": 2.513132178430917e-05,
+      "loss": 4.2625,
+      "step": 447050
+    },
+    {
+      "epoch": 3.8490013774104685,
+      "grad_norm": 3.453125,
+      "learning_rate": 2.511336457295279e-05,
+      "loss": 4.3678,
+      "step": 447100
+    },
+    {
+      "epoch": 3.8494318181818183,
+      "grad_norm": 3.375,
+      "learning_rate": 2.509541285811432e-05,
+      "loss": 4.1511,
+      "step": 447150
+    },
+    {
+      "epoch": 3.849862258953168,
+      "grad_norm": 4.59375,
+      "learning_rate": 2.5077466641111403e-05,
+      "loss": 4.3171,
+      "step": 447200
+    },
+    {
+      "epoch": 3.850292699724518,
+      "grad_norm": 3.59375,
+      "learning_rate": 2.5059525923261218e-05,
+      "loss": 4.447,
+      "step": 447250
+    },
+    {
+      "epoch": 3.8507231404958677,
+      "grad_norm": 3.5625,
+      "learning_rate": 2.504159070588057e-05,
+      "loss": 4.1131,
+      "step": 447300
+    },
+    {
+      "epoch": 3.8511535812672175,
+      "grad_norm": 4.625,
+      "learning_rate": 2.5023660990285947e-05,
+      "loss": 4.2023,
+      "step": 447350
+    },
+    {
+      "epoch": 3.8515840220385673,
+      "grad_norm": 5.1875,
+      "learning_rate": 2.500573677779323e-05,
+      "loss": 3.9092,
+      "step": 447400
+    },
+    {
+      "epoch": 3.852014462809917,
+      "grad_norm": 2.796875,
+      "learning_rate": 2.4987818069718104e-05,
+      "loss": 4.4807,
+      "step": 447450
+    },
+    {
+      "epoch": 3.8524449035812673,
+      "grad_norm": 3.28125,
+      "learning_rate": 2.496990486737575e-05,
+      "loss": 4.4532,
+      "step": 447500
+    },
+    {
+      "epoch": 3.852875344352617,
+      "grad_norm": 5.21875,
+      "learning_rate": 2.495199717208091e-05,
+      "loss": 4.6168,
+      "step": 447550
+    },
+    {
+      "epoch": 3.853305785123967,
+      "grad_norm": 4.15625,
+      "learning_rate": 2.4934094985148038e-05,
+      "loss": 4.3396,
+      "step": 447600
+    },
+    {
+      "epoch": 3.8537362258953167,
+      "grad_norm": 2.0,
+      "learning_rate": 2.4916198307891115e-05,
+      "loss": 4.4078,
+      "step": 447650
+    },
+    {
+      "epoch": 3.8541666666666665,
+      "grad_norm": 2.125,
+      "learning_rate": 2.4898307141623632e-05,
+      "loss": 4.1776,
+      "step": 447700
+    },
+    {
+      "epoch": 3.8545971074380168,
+      "grad_norm": 2.59375,
+      "learning_rate": 2.4880421487658844e-05,
+      "loss": 4.1043,
+      "step": 447750
+    },
+    {
+      "epoch": 3.8550275482093666,
+      "grad_norm": 4.59375,
+      "learning_rate": 2.48625413473095e-05,
+      "loss": 4.1875,
+      "step": 447800
+    },
+    {
+      "epoch": 3.8554579889807163,
+      "grad_norm": 3.90625,
+      "learning_rate": 2.4844666721887932e-05,
+      "loss": 4.36,
+      "step": 447850
+    },
+    {
+      "epoch": 3.855888429752066,
+      "grad_norm": 3.71875,
+      "learning_rate": 2.4826797612706178e-05,
+      "loss": 4.6485,
+      "step": 447900
+    },
+    {
+      "epoch": 3.856318870523416,
+      "grad_norm": 5.28125,
+      "learning_rate": 2.4808934021075713e-05,
+      "loss": 4.0192,
+      "step": 447950
+    },
+    {
+      "epoch": 3.8567493112947657,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.4791075948307685e-05,
+      "loss": 4.0292,
+      "step": 448000
+    },
+    {
+      "epoch": 3.8571797520661155,
+      "grad_norm": 2.078125,
+      "learning_rate": 2.4773223395712896e-05,
+      "loss": 4.2331,
+      "step": 448050
+    },
+    {
+      "epoch": 3.8576101928374653,
+      "grad_norm": 3.46875,
+      "learning_rate": 2.4755376364601647e-05,
+      "loss": 4.1477,
+      "step": 448100
+    },
+    {
+      "epoch": 3.8580406336088156,
+      "grad_norm": 3.625,
+      "learning_rate": 2.4737534856283893e-05,
+      "loss": 4.175,
+      "step": 448150
+    },
+    {
+      "epoch": 3.8584710743801653,
+      "grad_norm": 4.46875,
+      "learning_rate": 2.471969887206914e-05,
+      "loss": 4.117,
+      "step": 448200
+    },
+    {
+      "epoch": 3.858901515151515,
+      "grad_norm": 3.09375,
+      "learning_rate": 2.4701868413266495e-05,
+      "loss": 4.342,
+      "step": 448250
+    },
+    {
+      "epoch": 3.859331955922865,
+      "grad_norm": 3.609375,
+      "learning_rate": 2.4684043481184727e-05,
+      "loss": 4.3347,
+      "step": 448300
+    },
+    {
+      "epoch": 3.8597623966942147,
+      "grad_norm": 3.53125,
+      "learning_rate": 2.4666224077132118e-05,
+      "loss": 4.3601,
+      "step": 448350
+    },
+    {
+      "epoch": 3.860192837465565,
+      "grad_norm": 2.109375,
+      "learning_rate": 2.464841020241658e-05,
+      "loss": 4.233,
+      "step": 448400
+    },
+    {
+      "epoch": 3.8606232782369148,
+      "grad_norm": 3.34375,
+      "learning_rate": 2.4630601858345615e-05,
+      "loss": 4.203,
+      "step": 448450
+    },
+    {
+      "epoch": 3.8610537190082646,
+      "grad_norm": 2.875,
+      "learning_rate": 2.4612799046226308e-05,
+      "loss": 4.5164,
+      "step": 448500
+    },
+    {
+      "epoch": 3.8614841597796143,
+      "grad_norm": 3.546875,
+      "learning_rate": 2.459500176736532e-05,
+      "loss": 4.2748,
+      "step": 448550
+    },
+    {
+      "epoch": 3.861914600550964,
+      "grad_norm": 4.34375,
+      "learning_rate": 2.4577210023068996e-05,
+      "loss": 4.1038,
+      "step": 448600
+    },
+    {
+      "epoch": 3.862345041322314,
+      "grad_norm": 3.15625,
+      "learning_rate": 2.455942381464321e-05,
+      "loss": 4.2417,
+      "step": 448650
+    },
+    {
+      "epoch": 3.8627754820936637,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.4541643143393344e-05,
+      "loss": 4.1249,
+      "step": 448700
+    },
+    {
+      "epoch": 3.8632059228650135,
+      "grad_norm": 1.6796875,
+      "learning_rate": 2.452386801062454e-05,
+      "loss": 4.2709,
+      "step": 448750
+    },
+    {
+      "epoch": 3.8636363636363638,
+      "grad_norm": 3.3125,
+      "learning_rate": 2.4506098417641443e-05,
+      "loss": 4.4672,
+      "step": 448800
+    },
+    {
+      "epoch": 3.8640668044077136,
+      "grad_norm": 3.453125,
+      "learning_rate": 2.4488334365748254e-05,
+      "loss": 4.2192,
+      "step": 448850
+    },
+    {
+      "epoch": 3.8644972451790633,
+      "grad_norm": 2.0625,
+      "learning_rate": 2.4470575856248924e-05,
+      "loss": 3.8423,
+      "step": 448900
+    },
+    {
+      "epoch": 3.864927685950413,
+      "grad_norm": 3.109375,
+      "learning_rate": 2.4452822890446758e-05,
+      "loss": 4.5047,
+      "step": 448950
+    },
+    {
+      "epoch": 3.865358126721763,
+      "grad_norm": 2.296875,
+      "learning_rate": 2.4435075469644876e-05,
+      "loss": 4.6898,
+      "step": 449000
+    },
+    {
+      "epoch": 3.865788567493113,
+      "grad_norm": 2.421875,
+      "learning_rate": 2.4417333595145875e-05,
+      "loss": 4.2188,
+      "step": 449050
+    },
+    {
+      "epoch": 3.866219008264463,
+      "grad_norm": 4.34375,
+      "learning_rate": 2.4399597268251927e-05,
+      "loss": 4.6614,
+      "step": 449100
+    },
+    {
+      "epoch": 3.8666494490358128,
+      "grad_norm": 3.96875,
+      "learning_rate": 2.4381866490264948e-05,
+      "loss": 3.9614,
+      "step": 449150
+    },
+    {
+      "epoch": 3.8670798898071626,
+      "grad_norm": 3.859375,
+      "learning_rate": 2.4364141262486252e-05,
+      "loss": 4.0229,
+      "step": 449200
+    },
+    {
+      "epoch": 3.8675103305785123,
+      "grad_norm": 3.125,
+      "learning_rate": 2.4346421586216807e-05,
+      "loss": 4.5416,
+      "step": 449250
+    },
+    {
+      "epoch": 3.867940771349862,
+      "grad_norm": 5.34375,
+      "learning_rate": 2.4328707462757283e-05,
+      "loss": 4.5289,
+      "step": 449300
+    },
+    {
+      "epoch": 3.868371212121212,
+      "grad_norm": 3.375,
+      "learning_rate": 2.431099889340782e-05,
+      "loss": 4.4172,
+      "step": 449350
+    },
+    {
+      "epoch": 3.8688016528925617,
+      "grad_norm": 1.9921875,
+      "learning_rate": 2.429329587946816e-05,
+      "loss": 4.161,
+      "step": 449400
+    },
+    {
+      "epoch": 3.869232093663912,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.4275598422237745e-05,
+      "loss": 4.1826,
+      "step": 449450
+    },
+    {
+      "epoch": 3.8696625344352618,
+      "grad_norm": 3.015625,
+      "learning_rate": 2.4257906523015462e-05,
+      "loss": 4.1467,
+      "step": 449500
+    },
+    {
+      "epoch": 3.8700929752066116,
+      "grad_norm": 4.375,
+      "learning_rate": 2.4240220183099836e-05,
+      "loss": 4.3832,
+      "step": 449550
+    },
+    {
+      "epoch": 3.8705234159779613,
+      "grad_norm": 3.875,
+      "learning_rate": 2.422253940378909e-05,
+      "loss": 4.4008,
+      "step": 449600
+    },
+    {
+      "epoch": 3.870953856749311,
+      "grad_norm": 5.75,
+      "learning_rate": 2.4204864186380904e-05,
+      "loss": 4.3042,
+      "step": 449650
+    },
+    {
+      "epoch": 3.8713842975206614,
+      "grad_norm": 2.84375,
+      "learning_rate": 2.4187194532172618e-05,
+      "loss": 4.382,
+      "step": 449700
+    },
+    {
+      "epoch": 3.871814738292011,
+      "grad_norm": 5.96875,
+      "learning_rate": 2.416953044246113e-05,
+      "loss": 4.2328,
+      "step": 449750
+    },
+    {
+      "epoch": 3.872245179063361,
+      "grad_norm": 4.125,
+      "learning_rate": 2.4151871918542923e-05,
+      "loss": 4.2891,
+      "step": 449800
+    },
+    {
+      "epoch": 3.8726756198347108,
+      "grad_norm": 5.53125,
+      "learning_rate": 2.413421896171416e-05,
+      "loss": 4.2952,
+      "step": 449850
+    },
+    {
+      "epoch": 3.8731060606060606,
+      "grad_norm": 2.484375,
+      "learning_rate": 2.4116571573270495e-05,
+      "loss": 4.3608,
+      "step": 449900
+    },
+    {
+      "epoch": 3.8735365013774103,
+      "grad_norm": 2.828125,
+      "learning_rate": 2.4098929754507215e-05,
+      "loss": 4.4103,
+      "step": 449950
+    },
+    {
+      "epoch": 3.87396694214876,
+      "grad_norm": 1.3125,
+      "learning_rate": 2.4081293506719172e-05,
+      "loss": 4.0878,
+      "step": 450000
+    },
+    {
+      "epoch": 3.87396694214876,
+      "eval_loss": 4.990390777587891,
+      "eval_runtime": 24.439,
+      "eval_samples_per_second": 26.188,
+      "eval_steps_per_second": 13.094,
+      "eval_tts_loss": 7.472823273668061,
+      "step": 450000
+    },
+    {
+      "epoch": 3.87439738292011,
+      "grad_norm": 2.09375,
+      "learning_rate": 2.4063662831200852e-05,
+      "loss": 4.4684,
+      "step": 450050
+    },
+    {
+      "epoch": 3.87482782369146,
+      "grad_norm": 3.9375,
+      "learning_rate": 2.404603772924626e-05,
+      "loss": 4.207,
+      "step": 450100
+    },
+    {
+      "epoch": 3.87525826446281,
+      "grad_norm": 4.09375,
+      "learning_rate": 2.4028418202149127e-05,
+      "loss": 4.1974,
+      "step": 450150
+    },
+    {
+      "epoch": 3.8756887052341598,
+      "grad_norm": 2.09375,
+      "learning_rate": 2.4010804251202655e-05,
+      "loss": 4.3152,
+      "step": 450200
+    },
+    {
+      "epoch": 3.8761191460055096,
+      "grad_norm": 1.3046875,
+      "learning_rate": 2.3993195877699605e-05,
+      "loss": 4.1947,
+      "step": 450250
+    },
+    {
+      "epoch": 3.8765495867768593,
+      "grad_norm": 4.625,
+      "learning_rate": 2.397559308293247e-05,
+      "loss": 4.3928,
+      "step": 450300
+    },
+    {
+      "epoch": 3.8769800275482096,
+      "grad_norm": 3.46875,
+      "learning_rate": 2.395799586819323e-05,
+      "loss": 4.4172,
+      "step": 450350
+    },
+    {
+      "epoch": 3.8774104683195594,
+      "grad_norm": 3.34375,
+      "learning_rate": 2.394040423477346e-05,
+      "loss": 4.3648,
+      "step": 450400
+    },
+    {
+      "epoch": 3.877840909090909,
+      "grad_norm": 4.375,
+      "learning_rate": 2.3922818183964423e-05,
+      "loss": 4.2479,
+      "step": 450450
+    },
+    {
+      "epoch": 3.878271349862259,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.3905237717056796e-05,
+      "loss": 4.3042,
+      "step": 450500
+    },
+    {
+      "epoch": 3.8787017906336088,
+      "grad_norm": 3.21875,
+      "learning_rate": 2.388766283534103e-05,
+      "loss": 4.3755,
+      "step": 450550
+    },
+    {
+      "epoch": 3.8791322314049586,
+      "grad_norm": 3.265625,
+      "learning_rate": 2.3870093540107053e-05,
+      "loss": 4.7025,
+      "step": 450600
+    },
+    {
+      "epoch": 3.8795626721763083,
+      "grad_norm": 2.71875,
+      "learning_rate": 2.38525298326444e-05,
+      "loss": 3.9741,
+      "step": 450650
+    },
+    {
+      "epoch": 3.879993112947658,
+      "grad_norm": 3.0,
+      "learning_rate": 2.383497171424224e-05,
+      "loss": 3.7651,
+      "step": 450700
+    },
+    {
+      "epoch": 3.8804235537190084,
+      "grad_norm": 4.34375,
+      "learning_rate": 2.3817419186189262e-05,
+      "loss": 4.0537,
+      "step": 450750
+    },
+    {
+      "epoch": 3.880853994490358,
+      "grad_norm": 3.53125,
+      "learning_rate": 2.3799872249773793e-05,
+      "loss": 4.4243,
+      "step": 450800
+    },
+    {
+      "epoch": 3.881284435261708,
+      "grad_norm": 3.390625,
+      "learning_rate": 2.378233090628377e-05,
+      "loss": 3.9407,
+      "step": 450850
+    },
+    {
+      "epoch": 3.8817148760330578,
+      "grad_norm": 3.90625,
+      "learning_rate": 2.376479515700668e-05,
+      "loss": 4.3903,
+      "step": 450900
+    },
+    {
+      "epoch": 3.8821453168044076,
+      "grad_norm": 2.90625,
+      "learning_rate": 2.374726500322959e-05,
+      "loss": 4.3274,
+      "step": 450950
+    },
+    {
+      "epoch": 3.882575757575758,
+      "grad_norm": 2.390625,
+      "learning_rate": 2.37297404462392e-05,
+      "loss": 4.4075,
+      "step": 451000
+    },
+    {
+      "epoch": 3.8830061983471076,
+      "grad_norm": 5.28125,
+      "learning_rate": 2.371222148732175e-05,
+      "loss": 4.1965,
+      "step": 451050
+    },
+    {
+      "epoch": 3.8834366391184574,
+      "grad_norm": 3.71875,
+      "learning_rate": 2.369470812776309e-05,
+      "loss": 4.1834,
+      "step": 451100
+    },
+    {
+      "epoch": 3.883867079889807,
+      "grad_norm": 2.234375,
+      "learning_rate": 2.3677200368848697e-05,
+      "loss": 4.2672,
+      "step": 451150
+    },
+    {
+      "epoch": 3.884297520661157,
+      "grad_norm": 2.734375,
+      "learning_rate": 2.3659698211863592e-05,
+      "loss": 4.3593,
+      "step": 451200
+    },
+    {
+      "epoch": 3.8847279614325068,
+      "grad_norm": 2.875,
+      "learning_rate": 2.3642201658092377e-05,
+      "loss": 4.5176,
+      "step": 451250
+    },
+    {
+      "epoch": 3.8851584022038566,
+      "grad_norm": 3.21875,
+      "learning_rate": 2.3624710708819276e-05,
+      "loss": 4.2402,
+      "step": 451300
+    },
+    {
+      "epoch": 3.8855888429752063,
+      "grad_norm": 5.0,
+      "learning_rate": 2.360722536532809e-05,
+      "loss": 3.8982,
+      "step": 451350
+    },
+    {
+      "epoch": 3.8860192837465566,
+      "grad_norm": 2.828125,
+      "learning_rate": 2.358974562890216e-05,
+      "loss": 4.46,
+      "step": 451400
+    },
+    {
+      "epoch": 3.8864497245179064,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.357227150082454e-05,
+      "loss": 4.2613,
+      "step": 451450
+    },
+    {
+      "epoch": 3.886880165289256,
+      "grad_norm": 3.359375,
+      "learning_rate": 2.3554802982377742e-05,
+      "loss": 4.3461,
+      "step": 451500
+    },
+    {
+      "epoch": 3.887310606060606,
+      "grad_norm": 3.03125,
+      "learning_rate": 2.3537340074843938e-05,
+      "loss": 4.2654,
+      "step": 451550
+    },
+    {
+      "epoch": 3.8877410468319558,
+      "grad_norm": 5.28125,
+      "learning_rate": 2.3519882779504854e-05,
+      "loss": 4.5663,
+      "step": 451600
+    },
+    {
+      "epoch": 3.888171487603306,
+      "grad_norm": 3.734375,
+      "learning_rate": 2.3502431097641796e-05,
+      "loss": 4.2362,
+      "step": 451650
+    },
+    {
+      "epoch": 3.888601928374656,
+      "grad_norm": 2.46875,
+      "learning_rate": 2.3484985030535733e-05,
+      "loss": 4.3701,
+      "step": 451700
+    },
+    {
+      "epoch": 3.8890323691460056,
+      "grad_norm": 1.5859375,
+      "learning_rate": 2.346754457946717e-05,
+      "loss": 4.3001,
+      "step": 451750
+    },
+    {
+      "epoch": 3.8894628099173554,
+      "grad_norm": 1.6640625,
+      "learning_rate": 2.345010974571612e-05,
+      "loss": 4.0906,
+      "step": 451800
+    },
+    {
+      "epoch": 3.889893250688705,
+      "grad_norm": 5.90625,
+      "learning_rate": 2.3432680530562335e-05,
+      "loss": 4.6358,
+      "step": 451850
+    },
+    {
+      "epoch": 3.890323691460055,
+      "grad_norm": 3.296875,
+      "learning_rate": 2.341525693528507e-05,
+      "loss": 4.1502,
+      "step": 451900
+    },
+    {
+      "epoch": 3.8907541322314048,
+      "grad_norm": 3.796875,
+      "learning_rate": 2.339783896116313e-05,
+      "loss": 4.323,
+      "step": 451950
+    },
+    {
+      "epoch": 3.8911845730027546,
+      "grad_norm": 3.0625,
+      "learning_rate": 2.3380426609475058e-05,
+      "loss": 4.4724,
+      "step": 452000
+    },
+    {
+      "epoch": 3.891615013774105,
+      "grad_norm": 2.8125,
+      "learning_rate": 2.336301988149876e-05,
+      "loss": 4.3725,
+      "step": 452050
+    },
+    {
+      "epoch": 3.8920454545454546,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.3345618778511947e-05,
+      "loss": 4.1476,
+      "step": 452100
+    },
+    {
+      "epoch": 3.8924758953168044,
+      "grad_norm": 2.53125,
+      "learning_rate": 2.3328223301791796e-05,
+      "loss": 4.4637,
+      "step": 452150
+    },
+    {
+      "epoch": 3.892906336088154,
+      "grad_norm": 6.25,
+      "learning_rate": 2.331083345261508e-05,
+      "loss": 4.5984,
+      "step": 452200
+    },
+    {
+      "epoch": 3.893336776859504,
+      "grad_norm": 4.46875,
+      "learning_rate": 2.329344923225819e-05,
+      "loss": 4.0695,
+      "step": 452250
+    },
+    {
+      "epoch": 3.893767217630854,
+      "grad_norm": 2.34375,
+      "learning_rate": 2.3276070641997093e-05,
+      "loss": 4.1813,
+      "step": 452300
+    },
+    {
+      "epoch": 3.894197658402204,
+      "grad_norm": 5.15625,
+      "learning_rate": 2.32586976831073e-05,
+      "loss": 4.3672,
+      "step": 452350
+    },
+    {
+      "epoch": 3.894628099173554,
+      "grad_norm": 5.0,
+      "learning_rate": 2.3241330356864023e-05,
+      "loss": 4.2152,
+      "step": 452400
+    },
+    {
+      "epoch": 3.8950585399449036,
+      "grad_norm": 2.390625,
+      "learning_rate": 2.322396866454194e-05,
+      "loss": 4.3901,
+      "step": 452450
+    },
+    {
+      "epoch": 3.8954889807162534,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.320661260741537e-05,
+      "loss": 4.0748,
+      "step": 452500
+    },
+    {
+      "epoch": 3.895919421487603,
+      "grad_norm": 2.25,
+      "learning_rate": 2.318926218675821e-05,
+      "loss": 4.2602,
+      "step": 452550
+    },
+    {
+      "epoch": 3.896349862258953,
+      "grad_norm": 5.625,
+      "learning_rate": 2.3171917403843934e-05,
+      "loss": 4.8482,
+      "step": 452600
+    },
+    {
+      "epoch": 3.8967803030303028,
+      "grad_norm": 3.203125,
+      "learning_rate": 2.3154578259945602e-05,
+      "loss": 4.2367,
+      "step": 452650
+    },
+    {
+      "epoch": 3.897210743801653,
+      "grad_norm": 4.8125,
+      "learning_rate": 2.313724475633592e-05,
+      "loss": 4.4015,
+      "step": 452700
+    },
+    {
+      "epoch": 3.897641184573003,
+      "grad_norm": 1.4296875,
+      "learning_rate": 2.31199168942871e-05,
+      "loss": 4.5502,
+      "step": 452750
+    },
+    {
+      "epoch": 3.8980716253443526,
+      "grad_norm": 4.125,
+      "learning_rate": 2.3102594675070956e-05,
+      "loss": 4.15,
+      "step": 452800
+    },
+    {
+      "epoch": 3.8985020661157024,
+      "grad_norm": 1.6484375,
+      "learning_rate": 2.3085278099958918e-05,
+      "loss": 4.4132,
+      "step": 452850
+    },
+    {
+      "epoch": 3.8989325068870526,
+      "grad_norm": 1.8671875,
+      "learning_rate": 2.3067967170221984e-05,
+      "loss": 4.6712,
+      "step": 452900
+    },
+    {
+      "epoch": 3.8993629476584024,
+      "grad_norm": 2.0625,
+      "learning_rate": 2.305066188713071e-05,
+      "loss": 4.1893,
+      "step": 452950
+    },
+    {
+      "epoch": 3.899793388429752,
+      "grad_norm": 3.421875,
+      "learning_rate": 2.3033362251955315e-05,
+      "loss": 4.1439,
+      "step": 453000
+    },
+    {
+      "epoch": 3.899793388429752,
+      "eval_loss": 4.99050235748291,
+      "eval_runtime": 24.212,
+      "eval_samples_per_second": 26.433,
+      "eval_steps_per_second": 13.217,
+      "eval_tts_loss": 7.474577586485744,
+      "step": 453000
+    },
+    {
+      "epoch": 3.900223829201102,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.301606826596553e-05,
+      "loss": 4.1552,
+      "step": 453050
+    },
+    {
+      "epoch": 3.900654269972452,
+      "grad_norm": 3.015625,
+      "learning_rate": 2.2998779930430703e-05,
+      "loss": 4.2096,
+      "step": 453100
+    },
+    {
+      "epoch": 3.9010847107438016,
+      "grad_norm": 4.125,
+      "learning_rate": 2.2981497246619767e-05,
+      "loss": 4.3103,
+      "step": 453150
+    },
+    {
+      "epoch": 3.9015151515151514,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.296422021580118e-05,
+      "loss": 4.1876,
+      "step": 453200
+    },
+    {
+      "epoch": 3.901945592286501,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.294694883924312e-05,
+      "loss": 4.2966,
+      "step": 453250
+    },
+    {
+      "epoch": 3.902376033057851,
+      "grad_norm": 3.671875,
+      "learning_rate": 2.292968311821325e-05,
+      "loss": 4.3407,
+      "step": 453300
+    },
+    {
+      "epoch": 3.902806473829201,
+      "grad_norm": 1.328125,
+      "learning_rate": 2.291242305397877e-05,
+      "loss": 3.5915,
+      "step": 453350
+    },
+    {
+      "epoch": 3.903236914600551,
+      "grad_norm": 4.65625,
+      "learning_rate": 2.289516864780661e-05,
+      "loss": 4.1179,
+      "step": 453400
+    },
+    {
+      "epoch": 3.903667355371901,
+      "grad_norm": 1.296875,
+      "learning_rate": 2.2877919900963174e-05,
+      "loss": 4.1593,
+      "step": 453450
+    },
+    {
+      "epoch": 3.9040977961432506,
+      "grad_norm": 3.21875,
+      "learning_rate": 2.2860676814714453e-05,
+      "loss": 4.4113,
+      "step": 453500
+    },
+    {
+      "epoch": 3.904528236914601,
+      "grad_norm": 3.296875,
+      "learning_rate": 2.2843439390326137e-05,
+      "loss": 4.5458,
+      "step": 453550
+    },
+    {
+      "epoch": 3.9049586776859506,
+      "grad_norm": 2.21875,
+      "learning_rate": 2.2826207629063344e-05,
+      "loss": 4.3653,
+      "step": 453600
+    },
+    {
+      "epoch": 3.9053891184573004,
+      "grad_norm": 4.9375,
+      "learning_rate": 2.280898153219083e-05,
+      "loss": 4.6557,
+      "step": 453650
+    },
+    {
+      "epoch": 3.90581955922865,
+      "grad_norm": 3.453125,
+      "learning_rate": 2.279176110097303e-05,
+      "loss": 4.3007,
+      "step": 453700
+    },
+    {
+      "epoch": 3.90625,
+      "grad_norm": 4.46875,
+      "learning_rate": 2.277454633667384e-05,
+      "loss": 4.5776,
+      "step": 453750
+    },
+    {
+      "epoch": 3.90668044077135,
+      "grad_norm": 3.578125,
+      "learning_rate": 2.27573372405568e-05,
+      "loss": 4.2157,
+      "step": 453800
+    },
+    {
+      "epoch": 3.9071108815426996,
+      "grad_norm": 5.125,
+      "learning_rate": 2.2740133813885e-05,
+      "loss": 4.2166,
+      "step": 453850
+    },
+    {
+      "epoch": 3.9075413223140494,
+      "grad_norm": 4.84375,
+      "learning_rate": 2.272293605792114e-05,
+      "loss": 4.2229,
+      "step": 453900
+    },
+    {
+      "epoch": 3.907971763085399,
+      "grad_norm": 3.46875,
+      "learning_rate": 2.270574397392753e-05,
+      "loss": 4.0801,
+      "step": 453950
+    },
+    {
+      "epoch": 3.9084022038567494,
+      "grad_norm": 2.46875,
+      "learning_rate": 2.2688557563166003e-05,
+      "loss": 4.2192,
+      "step": 454000
+    },
+    {
+      "epoch": 3.908832644628099,
+      "grad_norm": 5.1875,
+      "learning_rate": 2.2671376826898017e-05,
+      "loss": 4.4716,
+      "step": 454050
+    },
+    {
+      "epoch": 3.909263085399449,
+      "grad_norm": 1.109375,
+      "learning_rate": 2.2654201766384586e-05,
+      "loss": 4.3365,
+      "step": 454100
+    },
+    {
+      "epoch": 3.909693526170799,
+      "grad_norm": 3.1875,
+      "learning_rate": 2.2637032382886337e-05,
+      "loss": 4.5118,
+      "step": 454150
+    },
+    {
+      "epoch": 3.910123966942149,
+      "grad_norm": 1.8203125,
+      "learning_rate": 2.2619868677663437e-05,
+      "loss": 4.3146,
+      "step": 454200
+    },
+    {
+      "epoch": 3.910554407713499,
+      "grad_norm": 5.21875,
+      "learning_rate": 2.260271065197571e-05,
+      "loss": 4.2847,
+      "step": 454250
+    },
+    {
+      "epoch": 3.9109848484848486,
+      "grad_norm": 3.46875,
+      "learning_rate": 2.25855583070825e-05,
+      "loss": 4.5388,
+      "step": 454300
+    },
+    {
+      "epoch": 3.9114152892561984,
+      "grad_norm": 3.5,
+      "learning_rate": 2.256841164424275e-05,
+      "loss": 4.0971,
+      "step": 454350
+    },
+    {
+      "epoch": 3.911845730027548,
+      "grad_norm": 4.34375,
+      "learning_rate": 2.2551270664714984e-05,
+      "loss": 3.8285,
+      "step": 454400
+    },
+    {
+      "epoch": 3.912276170798898,
+      "grad_norm": 3.9375,
+      "learning_rate": 2.2534135369757316e-05,
+      "loss": 4.4057,
+      "step": 454450
+    },
+    {
+      "epoch": 3.912706611570248,
+      "grad_norm": 1.953125,
+      "learning_rate": 2.251700576062742e-05,
+      "loss": 4.1583,
+      "step": 454500
+    },
+    {
+      "epoch": 3.9131370523415976,
+      "grad_norm": 2.734375,
+      "learning_rate": 2.2499881838582637e-05,
+      "loss": 4.3322,
+      "step": 454550
+    },
+    {
+      "epoch": 3.9135674931129474,
+      "grad_norm": 4.09375,
+      "learning_rate": 2.248276360487973e-05,
+      "loss": 4.31,
+      "step": 454600
+    },
+    {
+      "epoch": 3.9139979338842976,
+      "grad_norm": 1.6953125,
+      "learning_rate": 2.2465651060775217e-05,
+      "loss": 4.0418,
+      "step": 454650
+    },
+    {
+      "epoch": 3.9144283746556474,
+      "grad_norm": 1.34375,
+      "learning_rate": 2.2448544207525102e-05,
+      "loss": 4.7029,
+      "step": 454700
+    },
+    {
+      "epoch": 3.914858815426997,
+      "grad_norm": 2.546875,
+      "learning_rate": 2.2431443046384948e-05,
+      "loss": 4.0764,
+      "step": 454750
+    },
+    {
+      "epoch": 3.915289256198347,
+      "grad_norm": 3.625,
+      "learning_rate": 2.2414347578610017e-05,
+      "loss": 3.9479,
+      "step": 454800
+    },
+    {
+      "epoch": 3.9157196969696972,
+      "grad_norm": 1.984375,
+      "learning_rate": 2.2397257805455074e-05,
+      "loss": 4.3572,
+      "step": 454850
+    },
+    {
+      "epoch": 3.916150137741047,
+      "grad_norm": 4.78125,
+      "learning_rate": 2.238017372817438e-05,
+      "loss": 4.5265,
+      "step": 454900
+    },
+    {
+      "epoch": 3.916580578512397,
+      "grad_norm": 3.25,
+      "learning_rate": 2.2363095348021968e-05,
+      "loss": 4.1992,
+      "step": 454950
+    },
+    {
+      "epoch": 3.9170110192837466,
+      "grad_norm": 2.390625,
+      "learning_rate": 2.234602266625131e-05,
+      "loss": 4.07,
+      "step": 455000
+    },
+    {
+      "epoch": 3.9174414600550964,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.2328955684115483e-05,
+      "loss": 3.9802,
+      "step": 455050
+    },
+    {
+      "epoch": 3.917871900826446,
+      "grad_norm": 2.75,
+      "learning_rate": 2.231189440286725e-05,
+      "loss": 4.4272,
+      "step": 455100
+    },
+    {
+      "epoch": 3.918302341597796,
+      "grad_norm": 3.90625,
+      "learning_rate": 2.229483882375879e-05,
+      "loss": 4.3121,
+      "step": 455150
+    },
+    {
+      "epoch": 3.918732782369146,
+      "grad_norm": 4.03125,
+      "learning_rate": 2.2277788948041944e-05,
+      "loss": 4.4964,
+      "step": 455200
+    },
+    {
+      "epoch": 3.919163223140496,
+      "grad_norm": 2.84375,
+      "learning_rate": 2.22607447769682e-05,
+      "loss": 4.1009,
+      "step": 455250
+    },
+    {
+      "epoch": 3.919593663911846,
+      "grad_norm": 3.90625,
+      "learning_rate": 2.2243706311788527e-05,
+      "loss": 4.0371,
+      "step": 455300
+    },
+    {
+      "epoch": 3.9200241046831956,
+      "grad_norm": 3.546875,
+      "learning_rate": 2.2226673553753506e-05,
+      "loss": 4.6052,
+      "step": 455350
+    },
+    {
+      "epoch": 3.9204545454545454,
+      "grad_norm": 3.296875,
+      "learning_rate": 2.2209646504113323e-05,
+      "loss": 4.0969,
+      "step": 455400
+    },
+    {
+      "epoch": 3.920884986225895,
+      "grad_norm": 4.25,
+      "learning_rate": 2.2192625164117685e-05,
+      "loss": 4.1325,
+      "step": 455450
+    },
+    {
+      "epoch": 3.9213154269972454,
+      "grad_norm": 4.1875,
+      "learning_rate": 2.2175609535015984e-05,
+      "loss": 4.3922,
+      "step": 455500
+    },
+    {
+      "epoch": 3.9217458677685952,
+      "grad_norm": 3.21875,
+      "learning_rate": 2.2158599618057098e-05,
+      "loss": 4.2107,
+      "step": 455550
+    },
+    {
+      "epoch": 3.922176308539945,
+      "grad_norm": 3.125,
+      "learning_rate": 2.2141595414489513e-05,
+      "loss": 4.263,
+      "step": 455600
+    },
+    {
+      "epoch": 3.922606749311295,
+      "grad_norm": 1.3828125,
+      "learning_rate": 2.2124596925561315e-05,
+      "loss": 4.0325,
+      "step": 455650
+    },
+    {
+      "epoch": 3.9230371900826446,
+      "grad_norm": 3.25,
+      "learning_rate": 2.2107604152520147e-05,
+      "loss": 4.6308,
+      "step": 455700
+    },
+    {
+      "epoch": 3.9234676308539944,
+      "grad_norm": 3.046875,
+      "learning_rate": 2.209061709661322e-05,
+      "loss": 4.3302,
+      "step": 455750
+    },
+    {
+      "epoch": 3.923898071625344,
+      "grad_norm": 3.15625,
+      "learning_rate": 2.2073635759087397e-05,
+      "loss": 4.3726,
+      "step": 455800
+    },
+    {
+      "epoch": 3.924328512396694,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.2056660141189068e-05,
+      "loss": 4.2595,
+      "step": 455850
+    },
+    {
+      "epoch": 3.9247589531680442,
+      "grad_norm": 4.9375,
+      "learning_rate": 2.2039690244164136e-05,
+      "loss": 3.8418,
+      "step": 455900
+    },
+    {
+      "epoch": 3.925189393939394,
+      "grad_norm": 2.703125,
+      "learning_rate": 2.2022726069258238e-05,
+      "loss": 4.1195,
+      "step": 455950
+    },
+    {
+      "epoch": 3.925619834710744,
+      "grad_norm": 3.96875,
+      "learning_rate": 2.2005767617716465e-05,
+      "loss": 4.5111,
+      "step": 456000
+    },
+    {
+      "epoch": 3.925619834710744,
+      "eval_loss": 4.989672660827637,
+      "eval_runtime": 24.7502,
+      "eval_samples_per_second": 25.858,
+      "eval_steps_per_second": 12.929,
+      "eval_tts_loss": 7.461466540797048,
+      "step": 456000
+    },
+    {
+      "epoch": 3.9260502754820936,
+      "grad_norm": 5.0,
+      "learning_rate": 2.1988814890783517e-05,
+      "loss": 4.4894,
+      "step": 456050
+    },
+    {
+      "epoch": 3.9264807162534434,
+      "grad_norm": 2.5,
+      "learning_rate": 2.1971867889703755e-05,
+      "loss": 4.5378,
+      "step": 456100
+    },
+    {
+      "epoch": 3.9269111570247937,
+      "grad_norm": 1.8125,
+      "learning_rate": 2.195492661572096e-05,
+      "loss": 4.2555,
+      "step": 456150
+    },
+    {
+      "epoch": 3.9273415977961434,
+      "grad_norm": 4.875,
+      "learning_rate": 2.193799107007867e-05,
+      "loss": 4.3128,
+      "step": 456200
+    },
+    {
+      "epoch": 3.9277720385674932,
+      "grad_norm": 1.2421875,
+      "learning_rate": 2.192106125401987e-05,
+      "loss": 4.3103,
+      "step": 456250
+    },
+    {
+      "epoch": 3.928202479338843,
+      "grad_norm": 2.21875,
+      "learning_rate": 2.1904137168787155e-05,
+      "loss": 4.1488,
+      "step": 456300
+    },
+    {
+      "epoch": 3.928632920110193,
+      "grad_norm": 4.09375,
+      "learning_rate": 2.18872188156228e-05,
+      "loss": 4.0685,
+      "step": 456350
+    },
+    {
+      "epoch": 3.9290633608815426,
+      "grad_norm": 1.46875,
+      "learning_rate": 2.1870306195768497e-05,
+      "loss": 4.4033,
+      "step": 456400
+    },
+    {
+      "epoch": 3.9294938016528924,
+      "grad_norm": 3.21875,
+      "learning_rate": 2.1853399310465595e-05,
+      "loss": 4.2641,
+      "step": 456450
+    },
+    {
+      "epoch": 3.929924242424242,
+      "grad_norm": 2.21875,
+      "learning_rate": 2.1836498160955065e-05,
+      "loss": 4.382,
+      "step": 456500
+    },
+    {
+      "epoch": 3.9303546831955924,
+      "grad_norm": 1.6875,
+      "learning_rate": 2.1819602748477397e-05,
+      "loss": 4.3428,
+      "step": 456550
+    },
+    {
+      "epoch": 3.9307851239669422,
+      "grad_norm": 4.125,
+      "learning_rate": 2.1802713074272653e-05,
+      "loss": 4.3731,
+      "step": 456600
+    },
+    {
+      "epoch": 3.931215564738292,
+      "grad_norm": 5.15625,
+      "learning_rate": 2.1785829139580584e-05,
+      "loss": 4.4775,
+      "step": 456650
+    },
+    {
+      "epoch": 3.931646005509642,
+      "grad_norm": 2.078125,
+      "learning_rate": 2.1768950945640343e-05,
+      "loss": 4.4326,
+      "step": 456700
+    },
+    {
+      "epoch": 3.9320764462809916,
+      "grad_norm": 1.7578125,
+      "learning_rate": 2.1752078493690752e-05,
+      "loss": 4.5349,
+      "step": 456750
+    },
+    {
+      "epoch": 3.932506887052342,
+      "grad_norm": 2.796875,
+      "learning_rate": 2.173521178497028e-05,
+      "loss": 4.2966,
+      "step": 456800
+    },
+    {
+      "epoch": 3.9329373278236917,
+      "grad_norm": 3.09375,
+      "learning_rate": 2.171835082071686e-05,
+      "loss": 4.4935,
+      "step": 456850
+    },
+    {
+      "epoch": 3.9333677685950414,
+      "grad_norm": 3.21875,
+      "learning_rate": 2.1701495602168077e-05,
+      "loss": 4.2956,
+      "step": 456900
+    },
+    {
+      "epoch": 3.9337982093663912,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.1684646130561047e-05,
+      "loss": 4.4077,
+      "step": 456950
+    },
+    {
+      "epoch": 3.934228650137741,
+      "grad_norm": 3.03125,
+      "learning_rate": 2.1667802407132466e-05,
+      "loss": 3.9705,
+      "step": 457000
+    },
+    {
+      "epoch": 3.934659090909091,
+      "grad_norm": 2.078125,
+      "learning_rate": 2.165096443311867e-05,
+      "loss": 4.5346,
+      "step": 457050
+    },
+    {
+      "epoch": 3.9350895316804406,
+      "grad_norm": 3.234375,
+      "learning_rate": 2.163413220975552e-05,
+      "loss": 4.1243,
+      "step": 457100
+    },
+    {
+      "epoch": 3.9355199724517904,
+      "grad_norm": 3.21875,
+      "learning_rate": 2.1617305738278458e-05,
+      "loss": 4.3102,
+      "step": 457150
+    },
+    {
+      "epoch": 3.9359504132231407,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.16004850199225e-05,
+      "loss": 3.713,
+      "step": 457200
+    },
+    {
+      "epoch": 3.9363808539944904,
+      "grad_norm": 4.5625,
+      "learning_rate": 2.1583670055922277e-05,
+      "loss": 4.5762,
+      "step": 457250
+    },
+    {
+      "epoch": 3.9368112947658402,
+      "grad_norm": 3.0625,
+      "learning_rate": 2.1566860847511916e-05,
+      "loss": 4.5289,
+      "step": 457300
+    },
+    {
+      "epoch": 3.93724173553719,
+      "grad_norm": 1.703125,
+      "learning_rate": 2.155005739592525e-05,
+      "loss": 4.4189,
+      "step": 457350
+    },
+    {
+      "epoch": 3.93767217630854,
+      "grad_norm": 5.5,
+      "learning_rate": 2.1533259702395604e-05,
+      "loss": 4.616,
+      "step": 457400
+    },
+    {
+      "epoch": 3.93810261707989,
+      "grad_norm": 2.015625,
+      "learning_rate": 2.1516467768155812e-05,
+      "loss": 4.2837,
+      "step": 457450
+    },
+    {
+      "epoch": 3.93853305785124,
+      "grad_norm": 3.0625,
+      "learning_rate": 2.1499681594438457e-05,
+      "loss": 4.4649,
+      "step": 457500
+    },
+    {
+      "epoch": 3.9389634986225897,
+      "grad_norm": 2.671875,
+      "learning_rate": 2.148290118247558e-05,
+      "loss": 4.1393,
+      "step": 457550
+    },
+    {
+      "epoch": 3.9393939393939394,
+      "grad_norm": 5.90625,
+      "learning_rate": 2.1466126533498788e-05,
+      "loss": 4.0418,
+      "step": 457600
+    },
+    {
+      "epoch": 3.9398243801652892,
+      "grad_norm": 2.15625,
+      "learning_rate": 2.1449357648739398e-05,
+      "loss": 4.2382,
+      "step": 457650
+    },
+    {
+      "epoch": 3.940254820936639,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.1432594529428087e-05,
+      "loss": 4.2597,
+      "step": 457700
+    },
+    {
+      "epoch": 3.940685261707989,
+      "grad_norm": 3.90625,
+      "learning_rate": 2.141583717679533e-05,
+      "loss": 4.3176,
+      "step": 457750
+    },
+    {
+      "epoch": 3.9411157024793386,
+      "grad_norm": 3.0,
+      "learning_rate": 2.1399085592071045e-05,
+      "loss": 4.5763,
+      "step": 457800
+    },
+    {
+      "epoch": 3.941546143250689,
+      "grad_norm": 4.34375,
+      "learning_rate": 2.1382339776484763e-05,
+      "loss": 4.7258,
+      "step": 457850
+    },
+    {
+      "epoch": 3.9419765840220387,
+      "grad_norm": 3.171875,
+      "learning_rate": 2.1365599731265595e-05,
+      "loss": 4.09,
+      "step": 457900
+    },
+    {
+      "epoch": 3.9424070247933884,
+      "grad_norm": 3.078125,
+      "learning_rate": 2.1348865457642207e-05,
+      "loss": 4.3835,
+      "step": 457950
+    },
+    {
+      "epoch": 3.9428374655647382,
+      "grad_norm": 2.40625,
+      "learning_rate": 2.133213695684286e-05,
+      "loss": 4.4301,
+      "step": 458000
+    },
+    {
+      "epoch": 3.943267906336088,
+      "grad_norm": 2.765625,
+      "learning_rate": 2.131541423009542e-05,
+      "loss": 4.5665,
+      "step": 458050
+    },
+    {
+      "epoch": 3.9436983471074383,
+      "grad_norm": 7.78125,
+      "learning_rate": 2.129869727862729e-05,
+      "loss": 4.2329,
+      "step": 458100
+    },
+    {
+      "epoch": 3.944128787878788,
+      "grad_norm": 2.75,
+      "learning_rate": 2.128198610366544e-05,
+      "loss": 4.1698,
+      "step": 458150
+    },
+    {
+      "epoch": 3.944559228650138,
+      "grad_norm": 2.34375,
+      "learning_rate": 2.1265280706436452e-05,
+      "loss": 4.2915,
+      "step": 458200
+    },
+    {
+      "epoch": 3.9449896694214877,
+      "grad_norm": 4.28125,
+      "learning_rate": 2.1248581088166453e-05,
+      "loss": 4.5038,
+      "step": 458250
+    },
+    {
+      "epoch": 3.9454201101928374,
+      "grad_norm": 2.328125,
+      "learning_rate": 2.1231887250081138e-05,
+      "loss": 4.23,
+      "step": 458300
+    },
+    {
+      "epoch": 3.9458505509641872,
+      "grad_norm": 3.3125,
+      "learning_rate": 2.1215199193405856e-05,
+      "loss": 4.443,
+      "step": 458350
+    },
+    {
+      "epoch": 3.946280991735537,
+      "grad_norm": 5.34375,
+      "learning_rate": 2.1198516919365428e-05,
+      "loss": 4.4481,
+      "step": 458400
+    },
+    {
+      "epoch": 3.946711432506887,
+      "grad_norm": 2.296875,
+      "learning_rate": 2.118184042918433e-05,
+      "loss": 4.4335,
+      "step": 458450
+    },
+    {
+      "epoch": 3.947141873278237,
+      "grad_norm": 3.5625,
+      "learning_rate": 2.1165169724086552e-05,
+      "loss": 4.648,
+      "step": 458500
+    },
+    {
+      "epoch": 3.947572314049587,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.1148504805295677e-05,
+      "loss": 4.4054,
+      "step": 458550
+    },
+    {
+      "epoch": 3.9480027548209367,
+      "grad_norm": 3.25,
+      "learning_rate": 2.113184567403492e-05,
+      "loss": 4.26,
+      "step": 458600
+    },
+    {
+      "epoch": 3.9484331955922864,
+      "grad_norm": 1.71875,
+      "learning_rate": 2.111519233152701e-05,
+      "loss": 4.0183,
+      "step": 458650
+    },
+    {
+      "epoch": 3.9488636363636362,
+      "grad_norm": 8.75,
+      "learning_rate": 2.1098544778994255e-05,
+      "loss": 4.4697,
+      "step": 458700
+    },
+    {
+      "epoch": 3.9492940771349865,
+      "grad_norm": 1.3671875,
+      "learning_rate": 2.1081903017658543e-05,
+      "loss": 4.1582,
+      "step": 458750
+    },
+    {
+      "epoch": 3.9497245179063363,
+      "grad_norm": 3.21875,
+      "learning_rate": 2.1065267048741365e-05,
+      "loss": 3.7962,
+      "step": 458800
+    },
+    {
+      "epoch": 3.950154958677686,
+      "grad_norm": 4.9375,
+      "learning_rate": 2.104863687346371e-05,
+      "loss": 4.2527,
+      "step": 458850
+    },
+    {
+      "epoch": 3.950585399449036,
+      "grad_norm": 4.03125,
+      "learning_rate": 2.1032012493046282e-05,
+      "loss": 4.2759,
+      "step": 458900
+    },
+    {
+      "epoch": 3.9510158402203857,
+      "grad_norm": 4.1875,
+      "learning_rate": 2.1015393908709258e-05,
+      "loss": 4.2437,
+      "step": 458950
+    },
+    {
+      "epoch": 3.9514462809917354,
+      "grad_norm": 4.21875,
+      "learning_rate": 2.0998781121672316e-05,
+      "loss": 4.3284,
+      "step": 459000
+    },
+    {
+      "epoch": 3.9514462809917354,
+      "eval_loss": 4.989922046661377,
+      "eval_runtime": 23.8491,
+      "eval_samples_per_second": 26.835,
+      "eval_steps_per_second": 13.418,
+      "eval_tts_loss": 7.461940387165521,
+      "step": 459000
+    },
+    {
+      "epoch": 3.9518767217630852,
+      "grad_norm": 3.515625,
+      "learning_rate": 2.0982174133154907e-05,
+      "loss": 4.2978,
+      "step": 459050
+    },
+    {
+      "epoch": 3.952307162534435,
+      "grad_norm": 0.80078125,
+      "learning_rate": 2.09655729443759e-05,
+      "loss": 4.4284,
+      "step": 459100
+    },
+    {
+      "epoch": 3.9527376033057853,
+      "grad_norm": 1.90625,
+      "learning_rate": 2.094897755655376e-05,
+      "loss": 4.2241,
+      "step": 459150
+    },
+    {
+      "epoch": 3.953168044077135,
+      "grad_norm": 3.03125,
+      "learning_rate": 2.093238797090664e-05,
+      "loss": 4.4189,
+      "step": 459200
+    },
+    {
+      "epoch": 3.953598484848485,
+      "grad_norm": 2.984375,
+      "learning_rate": 2.0915804188652077e-05,
+      "loss": 4.0136,
+      "step": 459250
+    },
+    {
+      "epoch": 3.9540289256198347,
+      "grad_norm": 5.875,
+      "learning_rate": 2.089922621100735e-05,
+      "loss": 4.0621,
+      "step": 459300
+    },
+    {
+      "epoch": 3.9544593663911844,
+      "grad_norm": 2.515625,
+      "learning_rate": 2.088265403918923e-05,
+      "loss": 4.2556,
+      "step": 459350
+    },
+    {
+      "epoch": 3.9548898071625347,
+      "grad_norm": 3.75,
+      "learning_rate": 2.086608767441409e-05,
+      "loss": 4.1928,
+      "step": 459400
+    },
+    {
+      "epoch": 3.9553202479338845,
+      "grad_norm": 3.6875,
+      "learning_rate": 2.0849527117897836e-05,
+      "loss": 4.7572,
+      "step": 459450
+    },
+    {
+      "epoch": 3.9557506887052343,
+      "grad_norm": 3.453125,
+      "learning_rate": 2.0832972370856007e-05,
+      "loss": 4.3422,
+      "step": 459500
+    },
+    {
+      "epoch": 3.956181129476584,
+      "grad_norm": 4.21875,
+      "learning_rate": 2.0816423434503652e-05,
+      "loss": 4.443,
+      "step": 459550
+    },
+    {
+      "epoch": 3.956611570247934,
+      "grad_norm": 2.953125,
+      "learning_rate": 2.0799880310055464e-05,
+      "loss": 4.3839,
+      "step": 459600
+    },
+    {
+      "epoch": 3.9570420110192837,
+      "grad_norm": 3.953125,
+      "learning_rate": 2.0783342998725665e-05,
+      "loss": 4.6666,
+      "step": 459650
+    },
+    {
+      "epoch": 3.9574724517906334,
+      "grad_norm": 3.515625,
+      "learning_rate": 2.0766811501728046e-05,
+      "loss": 4.3964,
+      "step": 459700
+    },
+    {
+      "epoch": 3.9579028925619832,
+      "grad_norm": 1.8671875,
+      "learning_rate": 2.0750285820275996e-05,
+      "loss": 4.1742,
+      "step": 459750
+    },
+    {
+      "epoch": 3.9583333333333335,
+      "grad_norm": 3.484375,
+      "learning_rate": 2.073376595558245e-05,
+      "loss": 4.2851,
+      "step": 459800
+    },
+    {
+      "epoch": 3.9587637741046833,
+      "grad_norm": 1.484375,
+      "learning_rate": 2.0717251908859903e-05,
+      "loss": 4.3001,
+      "step": 459850
+    },
+    {
+      "epoch": 3.959194214876033,
+      "grad_norm": 2.921875,
+      "learning_rate": 2.0700743681320523e-05,
+      "loss": 4.4291,
+      "step": 459900
+    },
+    {
+      "epoch": 3.959624655647383,
+      "grad_norm": 3.796875,
+      "learning_rate": 2.0684241274175942e-05,
+      "loss": 4.3904,
+      "step": 459950
+    },
+    {
+      "epoch": 3.9600550964187327,
+      "grad_norm": 4.0625,
+      "learning_rate": 2.0667744688637392e-05,
+      "loss": 4.5331,
+      "step": 460000
+    },
+    {
+      "epoch": 3.960485537190083,
+      "grad_norm": 1.8125,
+      "learning_rate": 2.0651253925915694e-05,
+      "loss": 4.2643,
+      "step": 460050
+    },
+    {
+      "epoch": 3.9609159779614327,
+      "grad_norm": 1.25,
+      "learning_rate": 2.0634768987221244e-05,
+      "loss": 4.6239,
+      "step": 460100
+    },
+    {
+      "epoch": 3.9613464187327825,
+      "grad_norm": 2.0625,
+      "learning_rate": 2.061828987376395e-05,
+      "loss": 4.4271,
+      "step": 460150
+    },
+    {
+      "epoch": 3.9617768595041323,
+      "grad_norm": 2.890625,
+      "learning_rate": 2.060181658675345e-05,
+      "loss": 4.3318,
+      "step": 460200
+    },
+    {
+      "epoch": 3.962207300275482,
+      "grad_norm": 1.75,
+      "learning_rate": 2.0585349127398724e-05,
+      "loss": 4.1366,
+      "step": 460250
+    },
+    {
+      "epoch": 3.962637741046832,
+      "grad_norm": 3.1875,
+      "learning_rate": 2.056888749690854e-05,
+      "loss": 4.5682,
+      "step": 460300
+    },
+    {
+      "epoch": 3.9630681818181817,
+      "grad_norm": 3.09375,
+      "learning_rate": 2.0552431696491104e-05,
+      "loss": 4.5183,
+      "step": 460350
+    },
+    {
+      "epoch": 3.9634986225895315,
+      "grad_norm": 2.515625,
+      "learning_rate": 2.0535981727354225e-05,
+      "loss": 4.1679,
+      "step": 460400
+    },
+    {
+      "epoch": 3.9639290633608817,
+      "grad_norm": 3.125,
+      "learning_rate": 2.051953759070534e-05,
+      "loss": 4.325,
+      "step": 460450
+    },
+    {
+      "epoch": 3.9643595041322315,
+      "grad_norm": 4.75,
+      "learning_rate": 2.0503099287751415e-05,
+      "loss": 4.0851,
+      "step": 460500
+    },
+    {
+      "epoch": 3.9647899449035813,
+      "grad_norm": 2.0625,
+      "learning_rate": 2.0486666819698908e-05,
+      "loss": 4.4,
+      "step": 460550
+    },
+    {
+      "epoch": 3.965220385674931,
+      "grad_norm": 4.90625,
+      "learning_rate": 2.047024018775401e-05,
+      "loss": 4.413,
+      "step": 460600
+    },
+    {
+      "epoch": 3.965650826446281,
+      "grad_norm": 2.5625,
+      "learning_rate": 2.0453819393122366e-05,
+      "loss": 4.2945,
+      "step": 460650
+    },
+    {
+      "epoch": 3.966081267217631,
+      "grad_norm": 1.5078125,
+      "learning_rate": 2.04374044370092e-05,
+      "loss": 4.3787,
+      "step": 460700
+    },
+    {
+      "epoch": 3.966511707988981,
+      "grad_norm": 2.234375,
+      "learning_rate": 2.042099532061943e-05,
+      "loss": 4.2596,
+      "step": 460750
+    },
+    {
+      "epoch": 3.9669421487603307,
+      "grad_norm": 4.3125,
+      "learning_rate": 2.040459204515732e-05,
+      "loss": 4.3229,
+      "step": 460800
+    },
+    {
+      "epoch": 3.9673725895316805,
+      "grad_norm": 1.7265625,
+      "learning_rate": 2.0388194611826927e-05,
+      "loss": 4.5099,
+      "step": 460850
+    },
+    {
+      "epoch": 3.9678030303030303,
+      "grad_norm": 1.9609375,
+      "learning_rate": 2.037180302183177e-05,
+      "loss": 3.9855,
+      "step": 460900
+    },
+    {
+      "epoch": 3.96823347107438,
+      "grad_norm": 2.390625,
+      "learning_rate": 2.0355417276374943e-05,
+      "loss": 4.3092,
+      "step": 460950
+    },
+    {
+      "epoch": 3.96866391184573,
+      "grad_norm": 4.84375,
+      "learning_rate": 2.033903737665912e-05,
+      "loss": 4.1461,
+      "step": 461000
+    },
+    {
+      "epoch": 3.9690943526170797,
+      "grad_norm": 3.59375,
+      "learning_rate": 2.032266332388657e-05,
+      "loss": 4.4845,
+      "step": 461050
+    },
+    {
+      "epoch": 3.96952479338843,
+      "grad_norm": 6.46875,
+      "learning_rate": 2.0306295119259066e-05,
+      "loss": 4.2084,
+      "step": 461100
+    },
+    {
+      "epoch": 3.9699552341597797,
+      "grad_norm": 2.625,
+      "learning_rate": 2.028993276397807e-05,
+      "loss": 4.5187,
+      "step": 461150
+    },
+    {
+      "epoch": 3.9703856749311295,
+      "grad_norm": 2.28125,
+      "learning_rate": 2.02735762592445e-05,
+      "loss": 4.0516,
+      "step": 461200
+    },
+    {
+      "epoch": 3.9708161157024793,
+      "grad_norm": 3.546875,
+      "learning_rate": 2.02572256062589e-05,
+      "loss": 4.0382,
+      "step": 461250
+    },
+    {
+      "epoch": 3.971246556473829,
+      "grad_norm": 2.9375,
+      "learning_rate": 2.0240880806221375e-05,
+      "loss": 4.4745,
+      "step": 461300
+    },
+    {
+      "epoch": 3.9716769972451793,
+      "grad_norm": 3.3125,
+      "learning_rate": 2.022454186033158e-05,
+      "loss": 4.272,
+      "step": 461350
+    },
+    {
+      "epoch": 3.972107438016529,
+      "grad_norm": 3.078125,
+      "learning_rate": 2.020820876978875e-05,
+      "loss": 4.1312,
+      "step": 461400
+    },
+    {
+      "epoch": 3.972537878787879,
+      "grad_norm": 4.84375,
+      "learning_rate": 2.019188153579176e-05,
+      "loss": 4.3924,
+      "step": 461450
+    },
+    {
+      "epoch": 3.9729683195592287,
+      "grad_norm": 3.875,
+      "learning_rate": 2.017556015953893e-05,
+      "loss": 4.3548,
+      "step": 461500
+    },
+    {
+      "epoch": 3.9733987603305785,
+      "grad_norm": 4.59375,
+      "learning_rate": 2.015924464222826e-05,
+      "loss": 4.5869,
+      "step": 461550
+    },
+    {
+      "epoch": 3.9738292011019283,
+      "grad_norm": 4.71875,
+      "learning_rate": 2.0142934985057237e-05,
+      "loss": 4.448,
+      "step": 461600
+    },
+    {
+      "epoch": 3.974259641873278,
+      "grad_norm": 3.140625,
+      "learning_rate": 2.0126631189222965e-05,
+      "loss": 4.0536,
+      "step": 461650
+    },
+    {
+      "epoch": 3.974690082644628,
+      "grad_norm": 3.09375,
+      "learning_rate": 2.0110333255922097e-05,
+      "loss": 4.1835,
+      "step": 461700
+    },
+    {
+      "epoch": 3.975120523415978,
+      "grad_norm": 7.3125,
+      "learning_rate": 2.009404118635093e-05,
+      "loss": 4.3824,
+      "step": 461750
+    },
+    {
+      "epoch": 3.975550964187328,
+      "grad_norm": 2.59375,
+      "learning_rate": 2.0077754981705165e-05,
+      "loss": 4.5331,
+      "step": 461800
+    },
+    {
+      "epoch": 3.9759814049586777,
+      "grad_norm": 2.890625,
+      "learning_rate": 2.0061474643180255e-05,
+      "loss": 4.3659,
+      "step": 461850
+    },
+    {
+      "epoch": 3.9764118457300275,
+      "grad_norm": 5.625,
+      "learning_rate": 2.0045200171971113e-05,
+      "loss": 4.4113,
+      "step": 461900
+    },
+    {
+      "epoch": 3.9768422865013773,
+      "grad_norm": 1.6953125,
+      "learning_rate": 2.0028931569272225e-05,
+      "loss": 4.5685,
+      "step": 461950
+    },
+    {
+      "epoch": 3.9772727272727275,
+      "grad_norm": 3.453125,
+      "learning_rate": 2.0012668836277748e-05,
+      "loss": 4.7479,
+      "step": 462000
+    },
+    {
+      "epoch": 3.9772727272727275,
+      "eval_loss": 4.989577293395996,
+      "eval_runtime": 24.1819,
+      "eval_samples_per_second": 26.466,
+      "eval_steps_per_second": 13.233,
+      "eval_tts_loss": 7.4560609404651474,
+      "step": 462000
+    },
+    {
+      "epoch": 3.9777031680440773,
+      "grad_norm": 3.734375,
+      "learning_rate": 1.999641197418126e-05,
+      "loss": 4.0643,
+      "step": 462050
+    },
+    {
+      "epoch": 3.978133608815427,
+      "grad_norm": 3.734375,
+      "learning_rate": 1.9980160984175977e-05,
+      "loss": 4.1987,
+      "step": 462100
+    },
+    {
+      "epoch": 3.978564049586777,
+      "grad_norm": 4.875,
+      "learning_rate": 1.996391586745473e-05,
+      "loss": 4.4418,
+      "step": 462150
+    },
+    {
+      "epoch": 3.9789944903581267,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.9947676625209855e-05,
+      "loss": 4.1893,
+      "step": 462200
+    },
+    {
+      "epoch": 3.9794249311294765,
+      "grad_norm": 3.84375,
+      "learning_rate": 1.9931443258633255e-05,
+      "loss": 4.4415,
+      "step": 462250
+    },
+    {
+      "epoch": 3.9798553719008263,
+      "grad_norm": 1.7109375,
+      "learning_rate": 1.9915215768916506e-05,
+      "loss": 4.2932,
+      "step": 462300
+    },
+    {
+      "epoch": 3.980285812672176,
+      "grad_norm": 3.96875,
+      "learning_rate": 1.9898994157250573e-05,
+      "loss": 4.2352,
+      "step": 462350
+    },
+    {
+      "epoch": 3.9807162534435263,
+      "grad_norm": 2.078125,
+      "learning_rate": 1.988277842482611e-05,
+      "loss": 4.3896,
+      "step": 462400
+    },
+    {
+      "epoch": 3.981146694214876,
+      "grad_norm": 2.6875,
+      "learning_rate": 1.9866568572833366e-05,
+      "loss": 4.3079,
+      "step": 462450
+    },
+    {
+      "epoch": 3.981577134986226,
+      "grad_norm": 5.71875,
+      "learning_rate": 1.9850364602462067e-05,
+      "loss": 4.3261,
+      "step": 462500
+    },
+    {
+      "epoch": 3.9820075757575757,
+      "grad_norm": 3.453125,
+      "learning_rate": 1.9834166514901565e-05,
+      "loss": 4.3331,
+      "step": 462550
+    },
+    {
+      "epoch": 3.9824380165289255,
+      "grad_norm": 3.125,
+      "learning_rate": 1.9817974311340757e-05,
+      "loss": 4.0866,
+      "step": 462600
+    },
+    {
+      "epoch": 3.9828684573002757,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.980178799296809e-05,
+      "loss": 4.253,
+      "step": 462650
+    },
+    {
+      "epoch": 3.9832988980716255,
+      "grad_norm": 3.375,
+      "learning_rate": 1.9785607560971664e-05,
+      "loss": 4.3319,
+      "step": 462700
+    },
+    {
+      "epoch": 3.9837293388429753,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.976943301653905e-05,
+      "loss": 4.4869,
+      "step": 462750
+    },
+    {
+      "epoch": 3.984159779614325,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.975326436085745e-05,
+      "loss": 4.4756,
+      "step": 462800
+    },
+    {
+      "epoch": 3.984590220385675,
+      "grad_norm": 1.4453125,
+      "learning_rate": 1.9737101595113572e-05,
+      "loss": 4.0252,
+      "step": 462850
+    },
+    {
+      "epoch": 3.9850206611570247,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.972094472049375e-05,
+      "loss": 4.3611,
+      "step": 462900
+    },
+    {
+      "epoch": 3.9854511019283745,
+      "grad_norm": 4.28125,
+      "learning_rate": 1.9704793738183846e-05,
+      "loss": 3.9808,
+      "step": 462950
+    },
+    {
+      "epoch": 3.9858815426997243,
+      "grad_norm": 3.578125,
+      "learning_rate": 1.9688648649369333e-05,
+      "loss": 4.3338,
+      "step": 463000
+    },
+    {
+      "epoch": 3.9863119834710745,
+      "grad_norm": 2.15625,
+      "learning_rate": 1.9672509455235255e-05,
+      "loss": 4.602,
+      "step": 463050
+    },
+    {
+      "epoch": 3.9867424242424243,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.9656376156966093e-05,
+      "loss": 4.7624,
+      "step": 463100
+    },
+    {
+      "epoch": 3.987172865013774,
+      "grad_norm": 2.796875,
+      "learning_rate": 1.9640248755746082e-05,
+      "loss": 4.2008,
+      "step": 463150
+    },
+    {
+      "epoch": 3.987603305785124,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.962412725275892e-05,
+      "loss": 4.4183,
+      "step": 463200
+    },
+    {
+      "epoch": 3.9880337465564737,
+      "grad_norm": 4.28125,
+      "learning_rate": 1.9608011649187853e-05,
+      "loss": 3.9213,
+      "step": 463250
+    },
+    {
+      "epoch": 3.988464187327824,
+      "grad_norm": 4.25,
+      "learning_rate": 1.9591901946215817e-05,
+      "loss": 4.0634,
+      "step": 463300
+    },
+    {
+      "epoch": 3.9888946280991737,
+      "grad_norm": 2.859375,
+      "learning_rate": 1.9575798145025127e-05,
+      "loss": 3.9527,
+      "step": 463350
+    },
+    {
+      "epoch": 3.9893250688705235,
+      "grad_norm": 3.375,
+      "learning_rate": 1.955970024679784e-05,
+      "loss": 4.4453,
+      "step": 463400
+    },
+    {
+      "epoch": 3.9897555096418733,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.9543608252715484e-05,
+      "loss": 4.7095,
+      "step": 463450
+    },
+    {
+      "epoch": 3.990185950413223,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.9527522163959144e-05,
+      "loss": 4.0063,
+      "step": 463500
+    },
+    {
+      "epoch": 3.990616391184573,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.951144198170959e-05,
+      "loss": 4.5552,
+      "step": 463550
+    },
+    {
+      "epoch": 3.9910468319559227,
+      "grad_norm": 3.0,
+      "learning_rate": 1.9495367707146994e-05,
+      "loss": 4.3213,
+      "step": 463600
+    },
+    {
+      "epoch": 3.9914772727272725,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.9479299341451174e-05,
+      "loss": 4.4042,
+      "step": 463650
+    },
+    {
+      "epoch": 3.9919077134986227,
+      "grad_norm": 2.875,
+      "learning_rate": 1.9463236885801573e-05,
+      "loss": 4.0999,
+      "step": 463700
+    },
+    {
+      "epoch": 3.9923381542699725,
+      "grad_norm": 2.875,
+      "learning_rate": 1.944718034137709e-05,
+      "loss": 4.734,
+      "step": 463750
+    },
+    {
+      "epoch": 3.9927685950413223,
+      "grad_norm": 2.25,
+      "learning_rate": 1.9431129709356267e-05,
+      "loss": 4.2549,
+      "step": 463800
+    },
+    {
+      "epoch": 3.993199035812672,
+      "grad_norm": 2.71875,
+      "learning_rate": 1.9415084990917166e-05,
+      "loss": 3.871,
+      "step": 463850
+    },
+    {
+      "epoch": 3.993629476584022,
+      "grad_norm": 4.84375,
+      "learning_rate": 1.9399046187237447e-05,
+      "loss": 4.1596,
+      "step": 463900
+    },
+    {
+      "epoch": 3.994059917355372,
+      "grad_norm": 2.578125,
+      "learning_rate": 1.9383013299494292e-05,
+      "loss": 4.3716,
+      "step": 463950
+    },
+    {
+      "epoch": 3.994490358126722,
+      "grad_norm": 5.65625,
+      "learning_rate": 1.9366986328864534e-05,
+      "loss": 4.2075,
+      "step": 464000
+    },
+    {
+      "epoch": 3.9949207988980717,
+      "grad_norm": 2.859375,
+      "learning_rate": 1.9350965276524503e-05,
+      "loss": 4.0072,
+      "step": 464050
+    },
+    {
+      "epoch": 3.9953512396694215,
+      "grad_norm": 4.125,
+      "learning_rate": 1.9334950143650086e-05,
+      "loss": 4.2279,
+      "step": 464100
+    },
+    {
+      "epoch": 3.9957816804407713,
+      "grad_norm": 5.15625,
+      "learning_rate": 1.9318940931416784e-05,
+      "loss": 4.3637,
+      "step": 464150
+    },
+    {
+      "epoch": 3.996212121212121,
+      "grad_norm": 2.625,
+      "learning_rate": 1.9302937640999595e-05,
+      "loss": 4.2214,
+      "step": 464200
+    },
+    {
+      "epoch": 3.996642561983471,
+      "grad_norm": 3.765625,
+      "learning_rate": 1.9286940273573196e-05,
+      "loss": 3.7824,
+      "step": 464250
+    },
+    {
+      "epoch": 3.9970730027548207,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.9270948830311707e-05,
+      "loss": 4.3371,
+      "step": 464300
+    },
+    {
+      "epoch": 3.997503443526171,
+      "grad_norm": 4.96875,
+      "learning_rate": 1.9254963312388895e-05,
+      "loss": 4.7692,
+      "step": 464350
+    },
+    {
+      "epoch": 3.9979338842975207,
+      "grad_norm": 3.609375,
+      "learning_rate": 1.9238983720978043e-05,
+      "loss": 4.0711,
+      "step": 464400
+    },
+    {
+      "epoch": 3.9983643250688705,
+      "grad_norm": 3.765625,
+      "learning_rate": 1.9223010057252024e-05,
+      "loss": 4.3957,
+      "step": 464450
+    },
+    {
+      "epoch": 3.9987947658402203,
+      "grad_norm": 3.453125,
+      "learning_rate": 1.920704232238324e-05,
+      "loss": 4.3544,
+      "step": 464500
+    },
+    {
+      "epoch": 3.99922520661157,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.9191080517543746e-05,
+      "loss": 4.6815,
+      "step": 464550
+    },
+    {
+      "epoch": 3.9996556473829203,
+      "grad_norm": 5.0625,
+      "learning_rate": 1.9175124643905107e-05,
+      "loss": 4.306,
+      "step": 464600
+    },
+    {
+      "epoch": 4.00008608815427,
+      "grad_norm": 2.359375,
+      "learning_rate": 1.915917470263837e-05,
+      "loss": 4.0637,
+      "step": 464650
+    },
+    {
+      "epoch": 4.00051652892562,
+      "grad_norm": 3.40625,
+      "learning_rate": 1.9143230694914292e-05,
+      "loss": 4.2511,
+      "step": 464700
+    },
+    {
+      "epoch": 4.00094696969697,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.9127292621903127e-05,
+      "loss": 4.1813,
+      "step": 464750
+    },
+    {
+      "epoch": 4.0013774104683195,
+      "grad_norm": 3.40625,
+      "learning_rate": 1.9111360484774644e-05,
+      "loss": 4.3493,
+      "step": 464800
+    },
+    {
+      "epoch": 4.001807851239669,
+      "grad_norm": 3.5,
+      "learning_rate": 1.909543428469832e-05,
+      "loss": 4.4086,
+      "step": 464850
+    },
+    {
+      "epoch": 4.002238292011019,
+      "grad_norm": 3.984375,
+      "learning_rate": 1.9079514022843002e-05,
+      "loss": 4.2353,
+      "step": 464900
+    },
+    {
+      "epoch": 4.002668732782369,
+      "grad_norm": 1.578125,
+      "learning_rate": 1.9063599700377265e-05,
+      "loss": 4.4949,
+      "step": 464950
+    },
+    {
+      "epoch": 4.003099173553719,
+      "grad_norm": 4.9375,
+      "learning_rate": 1.9047691318469185e-05,
+      "loss": 4.3271,
+      "step": 465000
+    },
+    {
+      "epoch": 4.003099173553719,
+      "eval_loss": 4.989533424377441,
+      "eval_runtime": 23.9105,
+      "eval_samples_per_second": 26.766,
+      "eval_steps_per_second": 13.383,
+      "eval_tts_loss": 7.465740052456312,
+      "step": 465000
+    },
+    {
+      "epoch": 4.0035296143250685,
+      "grad_norm": 6.53125,
+      "learning_rate": 1.903178887828636e-05,
+      "loss": 4.218,
+      "step": 465050
+    },
+    {
+      "epoch": 4.003960055096419,
+      "grad_norm": 2.15625,
+      "learning_rate": 1.901589238099607e-05,
+      "loss": 4.3344,
+      "step": 465100
+    },
+    {
+      "epoch": 4.004390495867769,
+      "grad_norm": 4.25,
+      "learning_rate": 1.900000182776501e-05,
+      "loss": 4.4278,
+      "step": 465150
+    },
+    {
+      "epoch": 4.004820936639119,
+      "grad_norm": 3.609375,
+      "learning_rate": 1.8984117219759513e-05,
+      "loss": 4.2559,
+      "step": 465200
+    },
+    {
+      "epoch": 4.0052513774104685,
+      "grad_norm": 4.5,
+      "learning_rate": 1.8968238558145533e-05,
+      "loss": 4.1277,
+      "step": 465250
+    },
+    {
+      "epoch": 4.005681818181818,
+      "grad_norm": 4.78125,
+      "learning_rate": 1.8952365844088504e-05,
+      "loss": 4.68,
+      "step": 465300
+    },
+    {
+      "epoch": 4.006112258953168,
+      "grad_norm": 3.828125,
+      "learning_rate": 1.8936499078753435e-05,
+      "loss": 4.2194,
+      "step": 465350
+    },
+    {
+      "epoch": 4.006542699724518,
+      "grad_norm": 4.53125,
+      "learning_rate": 1.8920638263304925e-05,
+      "loss": 4.0065,
+      "step": 465400
+    },
+    {
+      "epoch": 4.006973140495868,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.890478339890711e-05,
+      "loss": 4.3619,
+      "step": 465450
+    },
+    {
+      "epoch": 4.0074035812672175,
+      "grad_norm": 4.0625,
+      "learning_rate": 1.8888934486723696e-05,
+      "loss": 4.5314,
+      "step": 465500
+    },
+    {
+      "epoch": 4.007834022038567,
+      "grad_norm": 3.0,
+      "learning_rate": 1.8873091527917997e-05,
+      "loss": 4.3445,
+      "step": 465550
+    },
+    {
+      "epoch": 4.008264462809917,
+      "grad_norm": 2.390625,
+      "learning_rate": 1.8857254523652835e-05,
+      "loss": 4.4336,
+      "step": 465600
+    },
+    {
+      "epoch": 4.008694903581267,
+      "grad_norm": 2.484375,
+      "learning_rate": 1.8841423475090614e-05,
+      "loss": 4.507,
+      "step": 465650
+    },
+    {
+      "epoch": 4.009125344352617,
+      "grad_norm": 1.46875,
+      "learning_rate": 1.8825598383393285e-05,
+      "loss": 4.1518,
+      "step": 465700
+    },
+    {
+      "epoch": 4.009555785123967,
+      "grad_norm": 3.703125,
+      "learning_rate": 1.880977924972236e-05,
+      "loss": 4.3511,
+      "step": 465750
+    },
+    {
+      "epoch": 4.009986225895317,
+      "grad_norm": 3.6875,
+      "learning_rate": 1.8793966075238966e-05,
+      "loss": 4.4737,
+      "step": 465800
+    },
+    {
+      "epoch": 4.010416666666667,
+      "grad_norm": 5.875,
+      "learning_rate": 1.8778158861103755e-05,
+      "loss": 4.3586,
+      "step": 465850
+    },
+    {
+      "epoch": 4.010847107438017,
+      "grad_norm": 2.25,
+      "learning_rate": 1.8762357608476933e-05,
+      "loss": 4.0145,
+      "step": 465900
+    },
+    {
+      "epoch": 4.0112775482093666,
+      "grad_norm": 2.328125,
+      "learning_rate": 1.8746562318518267e-05,
+      "loss": 4.1285,
+      "step": 465950
+    },
+    {
+      "epoch": 4.011707988980716,
+      "grad_norm": 4.875,
+      "learning_rate": 1.8730772992387115e-05,
+      "loss": 4.1862,
+      "step": 466000
+    },
+    {
+      "epoch": 4.012138429752066,
+      "grad_norm": 3.6875,
+      "learning_rate": 1.871498963124233e-05,
+      "loss": 3.9745,
+      "step": 466050
+    },
+    {
+      "epoch": 4.012568870523416,
+      "grad_norm": 2.40625,
+      "learning_rate": 1.8699212236242458e-05,
+      "loss": 4.3076,
+      "step": 466100
+    },
+    {
+      "epoch": 4.012999311294766,
+      "grad_norm": 5.21875,
+      "learning_rate": 1.8683440808545506e-05,
+      "loss": 4.5248,
+      "step": 466150
+    },
+    {
+      "epoch": 4.0134297520661155,
+      "grad_norm": 3.921875,
+      "learning_rate": 1.8667675349308987e-05,
+      "loss": 4.3498,
+      "step": 466200
+    },
+    {
+      "epoch": 4.013860192837465,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.8651915859690128e-05,
+      "loss": 3.9982,
+      "step": 466250
+    },
+    {
+      "epoch": 4.014290633608815,
+      "grad_norm": 4.21875,
+      "learning_rate": 1.863616234084563e-05,
+      "loss": 4.78,
+      "step": 466300
+    },
+    {
+      "epoch": 4.014721074380165,
+      "grad_norm": 2.640625,
+      "learning_rate": 1.8620414793931727e-05,
+      "loss": 4.9355,
+      "step": 466350
+    },
+    {
+      "epoch": 4.015151515151516,
+      "grad_norm": 3.578125,
+      "learning_rate": 1.8604673220104328e-05,
+      "loss": 3.9323,
+      "step": 466400
+    },
+    {
+      "epoch": 4.015581955922865,
+      "grad_norm": 2.25,
+      "learning_rate": 1.8588937620518755e-05,
+      "loss": 4.6762,
+      "step": 466450
+    },
+    {
+      "epoch": 4.016012396694215,
+      "grad_norm": 3.375,
+      "learning_rate": 1.8573207996330012e-05,
+      "loss": 4.3824,
+      "step": 466500
+    },
+    {
+      "epoch": 4.016442837465565,
+      "grad_norm": 3.671875,
+      "learning_rate": 1.855748434869261e-05,
+      "loss": 4.2368,
+      "step": 466550
+    },
+    {
+      "epoch": 4.016873278236915,
+      "grad_norm": 3.5,
+      "learning_rate": 1.854176667876063e-05,
+      "loss": 4.5086,
+      "step": 466600
+    },
+    {
+      "epoch": 4.0173037190082646,
+      "grad_norm": 6.125,
+      "learning_rate": 1.8526054987687714e-05,
+      "loss": 4.1508,
+      "step": 466650
+    },
+    {
+      "epoch": 4.017734159779614,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.851034927662707e-05,
+      "loss": 4.2153,
+      "step": 466700
+    },
+    {
+      "epoch": 4.018164600550964,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.8494649546731434e-05,
+      "loss": 4.1755,
+      "step": 466750
+    },
+    {
+      "epoch": 4.018595041322314,
+      "grad_norm": 1.6484375,
+      "learning_rate": 1.8478955799153174e-05,
+      "loss": 4.0992,
+      "step": 466800
+    },
+    {
+      "epoch": 4.019025482093664,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.8463268035044168e-05,
+      "loss": 4.564,
+      "step": 466850
+    },
+    {
+      "epoch": 4.0194559228650135,
+      "grad_norm": 4.34375,
+      "learning_rate": 1.844758625555586e-05,
+      "loss": 4.1945,
+      "step": 466900
+    },
+    {
+      "epoch": 4.019886363636363,
+      "grad_norm": 2.984375,
+      "learning_rate": 1.8431910461839252e-05,
+      "loss": 4.3524,
+      "step": 466950
+    },
+    {
+      "epoch": 4.020316804407713,
+      "grad_norm": 3.9375,
+      "learning_rate": 1.8416240655044915e-05,
+      "loss": 4.2614,
+      "step": 467000
+    },
+    {
+      "epoch": 4.020747245179064,
+      "grad_norm": 3.40625,
+      "learning_rate": 1.840057683632297e-05,
+      "loss": 4.0348,
+      "step": 467050
+    },
+    {
+      "epoch": 4.021177685950414,
+      "grad_norm": 4.96875,
+      "learning_rate": 1.838491900682313e-05,
+      "loss": 4.5726,
+      "step": 467100
+    },
+    {
+      "epoch": 4.021608126721763,
+      "grad_norm": 4.53125,
+      "learning_rate": 1.8369267167694638e-05,
+      "loss": 4.1148,
+      "step": 467150
+    },
+    {
+      "epoch": 4.022038567493113,
+      "grad_norm": 3.8125,
+      "learning_rate": 1.835362132008631e-05,
+      "loss": 3.9623,
+      "step": 467200
+    },
+    {
+      "epoch": 4.022469008264463,
+      "grad_norm": 1.71875,
+      "learning_rate": 1.8337981465146504e-05,
+      "loss": 4.1407,
+      "step": 467250
+    },
+    {
+      "epoch": 4.022899449035813,
+      "grad_norm": 2.890625,
+      "learning_rate": 1.832234760402316e-05,
+      "loss": 4.5119,
+      "step": 467300
+    },
+    {
+      "epoch": 4.0233298898071626,
+      "grad_norm": 1.875,
+      "learning_rate": 1.8306719737863744e-05,
+      "loss": 4.2315,
+      "step": 467350
+    },
+    {
+      "epoch": 4.023760330578512,
+      "grad_norm": 3.3125,
+      "learning_rate": 1.8291097867815375e-05,
+      "loss": 4.6522,
+      "step": 467400
+    },
+    {
+      "epoch": 4.024190771349862,
+      "grad_norm": 4.09375,
+      "learning_rate": 1.8275481995024578e-05,
+      "loss": 4.2538,
+      "step": 467450
+    },
+    {
+      "epoch": 4.024621212121212,
+      "grad_norm": 2.453125,
+      "learning_rate": 1.82598721206376e-05,
+      "loss": 4.0893,
+      "step": 467500
+    },
+    {
+      "epoch": 4.025051652892562,
+      "grad_norm": 3.359375,
+      "learning_rate": 1.8244268245800135e-05,
+      "loss": 3.8504,
+      "step": 467550
+    },
+    {
+      "epoch": 4.0254820936639115,
+      "grad_norm": 1.953125,
+      "learning_rate": 1.8228670371657442e-05,
+      "loss": 4.1312,
+      "step": 467600
+    },
+    {
+      "epoch": 4.025912534435261,
+      "grad_norm": 5.0,
+      "learning_rate": 1.821307849935444e-05,
+      "loss": 4.145,
+      "step": 467650
+    },
+    {
+      "epoch": 4.026342975206612,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.819749263003554e-05,
+      "loss": 4.2722,
+      "step": 467700
+    },
+    {
+      "epoch": 4.026773415977962,
+      "grad_norm": 3.703125,
+      "learning_rate": 1.818191276484462e-05,
+      "loss": 4.7459,
+      "step": 467750
+    },
+    {
+      "epoch": 4.027203856749312,
+      "grad_norm": 5.3125,
+      "learning_rate": 1.8166338904925294e-05,
+      "loss": 4.6664,
+      "step": 467800
+    },
+    {
+      "epoch": 4.027634297520661,
+      "grad_norm": 2.78125,
+      "learning_rate": 1.8150771051420622e-05,
+      "loss": 4.482,
+      "step": 467850
+    },
+    {
+      "epoch": 4.028064738292011,
+      "grad_norm": 3.75,
+      "learning_rate": 1.8135209205473225e-05,
+      "loss": 4.5145,
+      "step": 467900
+    },
+    {
+      "epoch": 4.028495179063361,
+      "grad_norm": 1.609375,
+      "learning_rate": 1.8119653368225385e-05,
+      "loss": 4.4687,
+      "step": 467950
+    },
+    {
+      "epoch": 4.028925619834711,
+      "grad_norm": 5.875,
+      "learning_rate": 1.8104103540818773e-05,
+      "loss": 3.8254,
+      "step": 468000
+    },
+    {
+      "epoch": 4.028925619834711,
+      "eval_loss": 4.990596771240234,
+      "eval_runtime": 24.0267,
+      "eval_samples_per_second": 26.637,
+      "eval_steps_per_second": 13.319,
+      "eval_tts_loss": 7.479180711330167,
+      "step": 468000
+    },
+    {
+      "epoch": 4.0293560606060606,
+      "grad_norm": 2.703125,
+      "learning_rate": 1.8088559724394783e-05,
+      "loss": 4.6154,
+      "step": 468050
+    },
+    {
+      "epoch": 4.02978650137741,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.8073021920094267e-05,
+      "loss": 4.4653,
+      "step": 468100
+    },
+    {
+      "epoch": 4.03021694214876,
+      "grad_norm": 1.96875,
+      "learning_rate": 1.8057490129057674e-05,
+      "loss": 4.156,
+      "step": 468150
+    },
+    {
+      "epoch": 4.03064738292011,
+      "grad_norm": 3.4375,
+      "learning_rate": 1.8041964352425e-05,
+      "loss": 4.272,
+      "step": 468200
+    },
+    {
+      "epoch": 4.03107782369146,
+      "grad_norm": 2.78125,
+      "learning_rate": 1.8026444591335813e-05,
+      "loss": 4.5182,
+      "step": 468250
+    },
+    {
+      "epoch": 4.0315082644628095,
+      "grad_norm": 1.4921875,
+      "learning_rate": 1.8010930846929185e-05,
+      "loss": 4.3646,
+      "step": 468300
+    },
+    {
+      "epoch": 4.03193870523416,
+      "grad_norm": 3.609375,
+      "learning_rate": 1.7995423120343856e-05,
+      "loss": 4.3485,
+      "step": 468350
+    },
+    {
+      "epoch": 4.03236914600551,
+      "grad_norm": 1.4140625,
+      "learning_rate": 1.7979921412718048e-05,
+      "loss": 4.5566,
+      "step": 468400
+    },
+    {
+      "epoch": 4.03279958677686,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.7964425725189527e-05,
+      "loss": 4.4085,
+      "step": 468450
+    },
+    {
+      "epoch": 4.03323002754821,
+      "grad_norm": 8.25,
+      "learning_rate": 1.794893605889565e-05,
+      "loss": 4.7689,
+      "step": 468500
+    },
+    {
+      "epoch": 4.033660468319559,
+      "grad_norm": 4.28125,
+      "learning_rate": 1.793345241497333e-05,
+      "loss": 4.1303,
+      "step": 468550
+    },
+    {
+      "epoch": 4.034090909090909,
+      "grad_norm": 4.28125,
+      "learning_rate": 1.791797479455901e-05,
+      "loss": 4.1237,
+      "step": 468600
+    },
+    {
+      "epoch": 4.034521349862259,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.7902503198788755e-05,
+      "loss": 4.2392,
+      "step": 468650
+    },
+    {
+      "epoch": 4.034951790633609,
+      "grad_norm": 3.3125,
+      "learning_rate": 1.7887037628798132e-05,
+      "loss": 4.1755,
+      "step": 468700
+    },
+    {
+      "epoch": 4.0353822314049586,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.7871578085722285e-05,
+      "loss": 4.1829,
+      "step": 468750
+    },
+    {
+      "epoch": 4.035812672176308,
+      "grad_norm": 2.28125,
+      "learning_rate": 1.7856124570695888e-05,
+      "loss": 3.9362,
+      "step": 468800
+    },
+    {
+      "epoch": 4.036243112947658,
+      "grad_norm": 3.375,
+      "learning_rate": 1.784067708485322e-05,
+      "loss": 4.1969,
+      "step": 468850
+    },
+    {
+      "epoch": 4.036673553719008,
+      "grad_norm": 3.90625,
+      "learning_rate": 1.782523562932805e-05,
+      "loss": 4.0791,
+      "step": 468900
+    },
+    {
+      "epoch": 4.037103994490358,
+      "grad_norm": 3.78125,
+      "learning_rate": 1.7809800205253847e-05,
+      "loss": 4.5571,
+      "step": 468950
+    },
+    {
+      "epoch": 4.037534435261708,
+      "grad_norm": 2.484375,
+      "learning_rate": 1.7794370813763406e-05,
+      "loss": 3.9455,
+      "step": 469000
+    },
+    {
+      "epoch": 4.037964876033058,
+      "grad_norm": 4.21875,
+      "learning_rate": 1.7778947455989313e-05,
+      "loss": 4.0368,
+      "step": 469050
+    },
+    {
+      "epoch": 4.038395316804408,
+      "grad_norm": 3.75,
+      "learning_rate": 1.7763530133063566e-05,
+      "loss": 4.4333,
+      "step": 469100
+    },
+    {
+      "epoch": 4.038825757575758,
+      "grad_norm": 1.8515625,
+      "learning_rate": 1.7748118846117745e-05,
+      "loss": 4.5984,
+      "step": 469150
+    },
+    {
+      "epoch": 4.039256198347108,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.773271359628309e-05,
+      "loss": 4.1432,
+      "step": 469200
+    },
+    {
+      "epoch": 4.039686639118457,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.771731438469022e-05,
+      "loss": 4.0816,
+      "step": 469250
+    },
+    {
+      "epoch": 4.040117079889807,
+      "grad_norm": 5.1875,
+      "learning_rate": 1.7701921212469417e-05,
+      "loss": 3.943,
+      "step": 469300
+    },
+    {
+      "epoch": 4.040547520661157,
+      "grad_norm": 2.265625,
+      "learning_rate": 1.7686534080750562e-05,
+      "loss": 4.3908,
+      "step": 469350
+    },
+    {
+      "epoch": 4.040977961432507,
+      "grad_norm": 1.9375,
+      "learning_rate": 1.7671152990662986e-05,
+      "loss": 3.9232,
+      "step": 469400
+    },
+    {
+      "epoch": 4.0414084022038566,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.7655777943335638e-05,
+      "loss": 4.3269,
+      "step": 469450
+    },
+    {
+      "epoch": 4.041838842975206,
+      "grad_norm": 4.53125,
+      "learning_rate": 1.764040893989706e-05,
+      "loss": 4.583,
+      "step": 469500
+    },
+    {
+      "epoch": 4.042269283746556,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.762504598147524e-05,
+      "loss": 4.1703,
+      "step": 469550
+    },
+    {
+      "epoch": 4.042699724517906,
+      "grad_norm": 3.625,
+      "learning_rate": 1.760968906919779e-05,
+      "loss": 4.4236,
+      "step": 469600
+    },
+    {
+      "epoch": 4.043130165289257,
+      "grad_norm": 4.40625,
+      "learning_rate": 1.7594338204191918e-05,
+      "loss": 4.3865,
+      "step": 469650
+    },
+    {
+      "epoch": 4.043560606060606,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.757899338758432e-05,
+      "loss": 4.0767,
+      "step": 469700
+    },
+    {
+      "epoch": 4.043991046831956,
+      "grad_norm": 1.1640625,
+      "learning_rate": 1.756365462050128e-05,
+      "loss": 4.114,
+      "step": 469750
+    },
+    {
+      "epoch": 4.044421487603306,
+      "grad_norm": 1.3515625,
+      "learning_rate": 1.7548321904068622e-05,
+      "loss": 4.2026,
+      "step": 469800
+    },
+    {
+      "epoch": 4.044851928374656,
+      "grad_norm": 2.421875,
+      "learning_rate": 1.753299523941171e-05,
+      "loss": 4.3217,
+      "step": 469850
+    },
+    {
+      "epoch": 4.045282369146006,
+      "grad_norm": 3.546875,
+      "learning_rate": 1.7517674627655533e-05,
+      "loss": 4.171,
+      "step": 469900
+    },
+    {
+      "epoch": 4.045712809917355,
+      "grad_norm": 3.703125,
+      "learning_rate": 1.7502360069924583e-05,
+      "loss": 4.3414,
+      "step": 469950
+    },
+    {
+      "epoch": 4.046143250688705,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.7487051567342905e-05,
+      "loss": 4.2661,
+      "step": 470000
+    },
+    {
+      "epoch": 4.046573691460055,
+      "grad_norm": 3.953125,
+      "learning_rate": 1.7471749121034108e-05,
+      "loss": 4.5303,
+      "step": 470050
+    },
+    {
+      "epoch": 4.047004132231405,
+      "grad_norm": 6.5,
+      "learning_rate": 1.7456452732121364e-05,
+      "loss": 3.9519,
+      "step": 470100
+    },
+    {
+      "epoch": 4.0474345730027546,
+      "grad_norm": 2.5,
+      "learning_rate": 1.7441162401727362e-05,
+      "loss": 4.3258,
+      "step": 470150
+    },
+    {
+      "epoch": 4.047865013774104,
+      "grad_norm": 3.71875,
+      "learning_rate": 1.7425878130974436e-05,
+      "loss": 4.2178,
+      "step": 470200
+    },
+    {
+      "epoch": 4.048295454545454,
+      "grad_norm": 3.890625,
+      "learning_rate": 1.7410599920984395e-05,
+      "loss": 4.4268,
+      "step": 470250
+    },
+    {
+      "epoch": 4.048725895316805,
+      "grad_norm": 2.578125,
+      "learning_rate": 1.739532777287862e-05,
+      "loss": 4.0618,
+      "step": 470300
+    },
+    {
+      "epoch": 4.049156336088155,
+      "grad_norm": 5.5,
+      "learning_rate": 1.738006168777806e-05,
+      "loss": 4.2749,
+      "step": 470350
+    },
+    {
+      "epoch": 4.049586776859504,
+      "grad_norm": 2.734375,
+      "learning_rate": 1.736480166680322e-05,
+      "loss": 4.4164,
+      "step": 470400
+    },
+    {
+      "epoch": 4.050017217630854,
+      "grad_norm": 3.484375,
+      "learning_rate": 1.734954771107411e-05,
+      "loss": 4.2764,
+      "step": 470450
+    },
+    {
+      "epoch": 4.050447658402204,
+      "grad_norm": 5.34375,
+      "learning_rate": 1.733429982171042e-05,
+      "loss": 4.5184,
+      "step": 470500
+    },
+    {
+      "epoch": 4.050878099173554,
+      "grad_norm": 1.984375,
+      "learning_rate": 1.7319057999831213e-05,
+      "loss": 4.2812,
+      "step": 470550
+    },
+    {
+      "epoch": 4.051308539944904,
+      "grad_norm": 2.625,
+      "learning_rate": 1.7303822246555277e-05,
+      "loss": 4.2761,
+      "step": 470600
+    },
+    {
+      "epoch": 4.051738980716253,
+      "grad_norm": 2.078125,
+      "learning_rate": 1.7288592563000873e-05,
+      "loss": 4.4205,
+      "step": 470650
+    },
+    {
+      "epoch": 4.052169421487603,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.7273368950285785e-05,
+      "loss": 4.4832,
+      "step": 470700
+    },
+    {
+      "epoch": 4.052599862258953,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.725815140952748e-05,
+      "loss": 4.5096,
+      "step": 470750
+    },
+    {
+      "epoch": 4.053030303030303,
+      "grad_norm": 3.734375,
+      "learning_rate": 1.7242939941842816e-05,
+      "loss": 4.2955,
+      "step": 470800
+    },
+    {
+      "epoch": 4.0534607438016526,
+      "grad_norm": 3.125,
+      "learning_rate": 1.7227734548348273e-05,
+      "loss": 3.8665,
+      "step": 470850
+    },
+    {
+      "epoch": 4.053891184573002,
+      "grad_norm": 1.7734375,
+      "learning_rate": 1.7212535230159944e-05,
+      "loss": 4.5488,
+      "step": 470900
+    },
+    {
+      "epoch": 4.054321625344353,
+      "grad_norm": 3.234375,
+      "learning_rate": 1.7197341988393413e-05,
+      "loss": 4.3105,
+      "step": 470950
+    },
+    {
+      "epoch": 4.054752066115703,
+      "grad_norm": 3.8125,
+      "learning_rate": 1.718215482416383e-05,
+      "loss": 4.155,
+      "step": 471000
+    },
+    {
+      "epoch": 4.054752066115703,
+      "eval_loss": 4.990444183349609,
+      "eval_runtime": 24.202,
+      "eval_samples_per_second": 26.444,
+      "eval_steps_per_second": 13.222,
+      "eval_tts_loss": 7.478690195086511,
+      "step": 471000
+    },
+    {
+      "epoch": 4.055182506887053,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.716697373858589e-05,
+      "loss": 4.5042,
+      "step": 471050
+    },
+    {
+      "epoch": 4.055612947658402,
+      "grad_norm": 3.53125,
+      "learning_rate": 1.715179873277386e-05,
+      "loss": 4.3599,
+      "step": 471100
+    },
+    {
+      "epoch": 4.056043388429752,
+      "grad_norm": 2.875,
+      "learning_rate": 1.713662980784153e-05,
+      "loss": 4.3022,
+      "step": 471150
+    },
+    {
+      "epoch": 4.056473829201102,
+      "grad_norm": 2.75,
+      "learning_rate": 1.71214669649023e-05,
+      "loss": 4.16,
+      "step": 471200
+    },
+    {
+      "epoch": 4.056904269972452,
+      "grad_norm": 5.53125,
+      "learning_rate": 1.710631020506909e-05,
+      "loss": 4.3182,
+      "step": 471250
+    },
+    {
+      "epoch": 4.057334710743802,
+      "grad_norm": 3.84375,
+      "learning_rate": 1.7091159529454347e-05,
+      "loss": 4.296,
+      "step": 471300
+    },
+    {
+      "epoch": 4.057765151515151,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.7076014939170127e-05,
+      "loss": 4.2559,
+      "step": 471350
+    },
+    {
+      "epoch": 4.058195592286501,
+      "grad_norm": 3.765625,
+      "learning_rate": 1.7060876435327954e-05,
+      "loss": 4.0452,
+      "step": 471400
+    },
+    {
+      "epoch": 4.058626033057851,
+      "grad_norm": 4.1875,
+      "learning_rate": 1.7045744019039046e-05,
+      "loss": 4.3682,
+      "step": 471450
+    },
+    {
+      "epoch": 4.059056473829201,
+      "grad_norm": 4.03125,
+      "learning_rate": 1.703061769141403e-05,
+      "loss": 3.894,
+      "step": 471500
+    },
+    {
+      "epoch": 4.0594869146005506,
+      "grad_norm": 3.84375,
+      "learning_rate": 1.7015497453563167e-05,
+      "loss": 4.6356,
+      "step": 471550
+    },
+    {
+      "epoch": 4.059917355371901,
+      "grad_norm": 5.875,
+      "learning_rate": 1.7000383306596244e-05,
+      "loss": 4.1306,
+      "step": 471600
+    },
+    {
+      "epoch": 4.060347796143251,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.6985275251622612e-05,
+      "loss": 3.6677,
+      "step": 471650
+    },
+    {
+      "epoch": 4.060778236914601,
+      "grad_norm": 7.21875,
+      "learning_rate": 1.6970173289751145e-05,
+      "loss": 4.3977,
+      "step": 471700
+    },
+    {
+      "epoch": 4.061208677685951,
+      "grad_norm": 2.6875,
+      "learning_rate": 1.6955077422090328e-05,
+      "loss": 4.1676,
+      "step": 471750
+    },
+    {
+      "epoch": 4.0616391184573,
+      "grad_norm": 1.59375,
+      "learning_rate": 1.693998764974819e-05,
+      "loss": 4.1295,
+      "step": 471800
+    },
+    {
+      "epoch": 4.06206955922865,
+      "grad_norm": 3.34375,
+      "learning_rate": 1.6924903973832183e-05,
+      "loss": 4.3328,
+      "step": 471850
+    },
+    {
+      "epoch": 4.0625,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.6909826395449525e-05,
+      "loss": 4.5923,
+      "step": 471900
+    },
+    {
+      "epoch": 4.06293044077135,
+      "grad_norm": 4.9375,
+      "learning_rate": 1.6894754915706835e-05,
+      "loss": 4.7769,
+      "step": 471950
+    },
+    {
+      "epoch": 4.0633608815427,
+      "grad_norm": 4.0625,
+      "learning_rate": 1.6879689535710297e-05,
+      "loss": 4.2948,
+      "step": 472000
+    },
+    {
+      "epoch": 4.063791322314049,
+      "grad_norm": 3.59375,
+      "learning_rate": 1.686463025656576e-05,
+      "loss": 4.1265,
+      "step": 472050
+    },
+    {
+      "epoch": 4.064221763085399,
+      "grad_norm": 0.8125,
+      "learning_rate": 1.684957707937844e-05,
+      "loss": 4.1848,
+      "step": 472100
+    },
+    {
+      "epoch": 4.064652203856749,
+      "grad_norm": 2.59375,
+      "learning_rate": 1.683453000525329e-05,
+      "loss": 4.441,
+      "step": 472150
+    },
+    {
+      "epoch": 4.065082644628099,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.68194890352947e-05,
+      "loss": 4.8487,
+      "step": 472200
+    },
+    {
+      "epoch": 4.0655130853994494,
+      "grad_norm": 3.546875,
+      "learning_rate": 1.680445417060663e-05,
+      "loss": 4.3647,
+      "step": 472250
+    },
+    {
+      "epoch": 4.065943526170799,
+      "grad_norm": 3.59375,
+      "learning_rate": 1.6789425412292657e-05,
+      "loss": 4.3761,
+      "step": 472300
+    },
+    {
+      "epoch": 4.066373966942149,
+      "grad_norm": 3.296875,
+      "learning_rate": 1.677440276145582e-05,
+      "loss": 4.1368,
+      "step": 472350
+    },
+    {
+      "epoch": 4.066804407713499,
+      "grad_norm": 3.40625,
+      "learning_rate": 1.6759386219198726e-05,
+      "loss": 4.2209,
+      "step": 472400
+    },
+    {
+      "epoch": 4.067234848484849,
+      "grad_norm": 5.125,
+      "learning_rate": 1.67443757866236e-05,
+      "loss": 4.2474,
+      "step": 472450
+    },
+    {
+      "epoch": 4.067665289256198,
+      "grad_norm": 1.703125,
+      "learning_rate": 1.6729371464832176e-05,
+      "loss": 3.9704,
+      "step": 472500
+    },
+    {
+      "epoch": 4.068095730027548,
+      "grad_norm": 3.40625,
+      "learning_rate": 1.671437325492572e-05,
+      "loss": 4.3687,
+      "step": 472550
+    },
+    {
+      "epoch": 4.068526170798898,
+      "grad_norm": 2.703125,
+      "learning_rate": 1.6699381158005068e-05,
+      "loss": 4.474,
+      "step": 472600
+    },
+    {
+      "epoch": 4.068956611570248,
+      "grad_norm": 4.0,
+      "learning_rate": 1.668439517517062e-05,
+      "loss": 4.1668,
+      "step": 472650
+    },
+    {
+      "epoch": 4.069387052341598,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.6669415307522285e-05,
+      "loss": 4.2232,
+      "step": 472700
+    },
+    {
+      "epoch": 4.069817493112947,
+      "grad_norm": 2.03125,
+      "learning_rate": 1.6654441556159593e-05,
+      "loss": 4.3866,
+      "step": 472750
+    },
+    {
+      "epoch": 4.070247933884297,
+      "grad_norm": 7.5625,
+      "learning_rate": 1.663947392218157e-05,
+      "loss": 4.2389,
+      "step": 472800
+    },
+    {
+      "epoch": 4.070678374655647,
+      "grad_norm": 9.0,
+      "learning_rate": 1.662451240668681e-05,
+      "loss": 4.4785,
+      "step": 472850
+    },
+    {
+      "epoch": 4.071108815426998,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.6609557010773446e-05,
+      "loss": 4.1316,
+      "step": 472900
+    },
+    {
+      "epoch": 4.0715392561983474,
+      "grad_norm": 8.25,
+      "learning_rate": 1.659460773553917e-05,
+      "loss": 4.4142,
+      "step": 472950
+    },
+    {
+      "epoch": 4.071969696969697,
+      "grad_norm": 2.421875,
+      "learning_rate": 1.6579664582081246e-05,
+      "loss": 4.3803,
+      "step": 473000
+    },
+    {
+      "epoch": 4.072400137741047,
+      "grad_norm": 4.125,
+      "learning_rate": 1.6564727551496485e-05,
+      "loss": 4.1947,
+      "step": 473050
+    },
+    {
+      "epoch": 4.072830578512397,
+      "grad_norm": 4.5,
+      "learning_rate": 1.6549796644881156e-05,
+      "loss": 4.3879,
+      "step": 473100
+    },
+    {
+      "epoch": 4.073261019283747,
+      "grad_norm": 2.640625,
+      "learning_rate": 1.6534871863331224e-05,
+      "loss": 4.4808,
+      "step": 473150
+    },
+    {
+      "epoch": 4.073691460055096,
+      "grad_norm": 4.125,
+      "learning_rate": 1.651995320794213e-05,
+      "loss": 4.2729,
+      "step": 473200
+    },
+    {
+      "epoch": 4.074121900826446,
+      "grad_norm": 6.75,
+      "learning_rate": 1.6505040679808824e-05,
+      "loss": 4.5253,
+      "step": 473250
+    },
+    {
+      "epoch": 4.074552341597796,
+      "grad_norm": 2.09375,
+      "learning_rate": 1.6490134280025914e-05,
+      "loss": 3.7976,
+      "step": 473300
+    },
+    {
+      "epoch": 4.074982782369146,
+      "grad_norm": 3.015625,
+      "learning_rate": 1.64752340096875e-05,
+      "loss": 4.2162,
+      "step": 473350
+    },
+    {
+      "epoch": 4.075413223140496,
+      "grad_norm": 3.953125,
+      "learning_rate": 1.6460339869887152e-05,
+      "loss": 3.8813,
+      "step": 473400
+    },
+    {
+      "epoch": 4.075843663911845,
+      "grad_norm": 2.453125,
+      "learning_rate": 1.6445451861718143e-05,
+      "loss": 4.3434,
+      "step": 473450
+    },
+    {
+      "epoch": 4.076274104683195,
+      "grad_norm": 4.3125,
+      "learning_rate": 1.6430569986273194e-05,
+      "loss": 4.0991,
+      "step": 473500
+    },
+    {
+      "epoch": 4.076704545454546,
+      "grad_norm": 3.484375,
+      "learning_rate": 1.6415694244644575e-05,
+      "loss": 4.443,
+      "step": 473550
+    },
+    {
+      "epoch": 4.077134986225896,
+      "grad_norm": 3.875,
+      "learning_rate": 1.6400824637924216e-05,
+      "loss": 4.1666,
+      "step": 473600
+    },
+    {
+      "epoch": 4.0775654269972454,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.638596116720341e-05,
+      "loss": 4.3432,
+      "step": 473650
+    },
+    {
+      "epoch": 4.077995867768595,
+      "grad_norm": 5.9375,
+      "learning_rate": 1.637110383357319e-05,
+      "loss": 4.1673,
+      "step": 473700
+    },
+    {
+      "epoch": 4.078426308539945,
+      "grad_norm": 4.375,
+      "learning_rate": 1.6356252638123994e-05,
+      "loss": 4.4722,
+      "step": 473750
+    },
+    {
+      "epoch": 4.078856749311295,
+      "grad_norm": 3.59375,
+      "learning_rate": 1.63414075819459e-05,
+      "loss": 4.5102,
+      "step": 473800
+    },
+    {
+      "epoch": 4.079287190082645,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.6326568666128495e-05,
+      "loss": 4.3715,
+      "step": 473850
+    },
+    {
+      "epoch": 4.079717630853994,
+      "grad_norm": 3.53125,
+      "learning_rate": 1.6311735891760926e-05,
+      "loss": 4.0467,
+      "step": 473900
+    },
+    {
+      "epoch": 4.080148071625344,
+      "grad_norm": 7.71875,
+      "learning_rate": 1.6296909259931858e-05,
+      "loss": 4.4708,
+      "step": 473950
+    },
+    {
+      "epoch": 4.080578512396694,
+      "grad_norm": 0.71484375,
+      "learning_rate": 1.6282088771729575e-05,
+      "loss": 4.1761,
+      "step": 474000
+    },
+    {
+      "epoch": 4.080578512396694,
+      "eval_loss": 4.990285396575928,
+      "eval_runtime": 24.0607,
+      "eval_samples_per_second": 26.599,
+      "eval_steps_per_second": 13.3,
+      "eval_tts_loss": 7.4766609914280675,
+      "step": 474000
+    },
+    {
+      "epoch": 4.081008953168044,
+      "grad_norm": 2.09375,
+      "learning_rate": 1.6267274428241852e-05,
+      "loss": 4.3962,
+      "step": 474050
+    },
+    {
+      "epoch": 4.081439393939394,
+      "grad_norm": 1.0234375,
+      "learning_rate": 1.625246623055604e-05,
+      "loss": 4.3978,
+      "step": 474100
+    },
+    {
+      "epoch": 4.081869834710743,
+      "grad_norm": 4.1875,
+      "learning_rate": 1.6237664179759016e-05,
+      "loss": 4.2482,
+      "step": 474150
+    },
+    {
+      "epoch": 4.082300275482094,
+      "grad_norm": 3.0,
+      "learning_rate": 1.622286827693722e-05,
+      "loss": 4.098,
+      "step": 474200
+    },
+    {
+      "epoch": 4.082730716253444,
+      "grad_norm": 2.71875,
+      "learning_rate": 1.6208078523176617e-05,
+      "loss": 4.0942,
+      "step": 474250
+    },
+    {
+      "epoch": 4.083161157024794,
+      "grad_norm": 2.703125,
+      "learning_rate": 1.6193294919562786e-05,
+      "loss": 4.4851,
+      "step": 474300
+    },
+    {
+      "epoch": 4.0835915977961434,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.61785174671808e-05,
+      "loss": 4.2229,
+      "step": 474350
+    },
+    {
+      "epoch": 4.084022038567493,
+      "grad_norm": 3.296875,
+      "learning_rate": 1.616374616711528e-05,
+      "loss": 4.3572,
+      "step": 474400
+    },
+    {
+      "epoch": 4.084452479338843,
+      "grad_norm": 3.375,
+      "learning_rate": 1.6148981020450417e-05,
+      "loss": 4.332,
+      "step": 474450
+    },
+    {
+      "epoch": 4.084882920110193,
+      "grad_norm": 1.8828125,
+      "learning_rate": 1.613422202826992e-05,
+      "loss": 4.1032,
+      "step": 474500
+    },
+    {
+      "epoch": 4.085313360881543,
+      "grad_norm": 2.328125,
+      "learning_rate": 1.61194691916571e-05,
+      "loss": 4.1243,
+      "step": 474550
+    },
+    {
+      "epoch": 4.085743801652892,
+      "grad_norm": 3.296875,
+      "learning_rate": 1.610472251169479e-05,
+      "loss": 4.5435,
+      "step": 474600
+    },
+    {
+      "epoch": 4.086174242424242,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.6089981989465296e-05,
+      "loss": 4.3172,
+      "step": 474650
+    },
+    {
+      "epoch": 4.086604683195592,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.6075247626050617e-05,
+      "loss": 3.9243,
+      "step": 474700
+    },
+    {
+      "epoch": 4.087035123966942,
+      "grad_norm": 2.453125,
+      "learning_rate": 1.6060519422532182e-05,
+      "loss": 4.3399,
+      "step": 474750
+    },
+    {
+      "epoch": 4.087465564738292,
+      "grad_norm": 3.734375,
+      "learning_rate": 1.6045797379991e-05,
+      "loss": 3.9782,
+      "step": 474800
+    },
+    {
+      "epoch": 4.087896005509642,
+      "grad_norm": 2.875,
+      "learning_rate": 1.6031081499507717e-05,
+      "loss": 4.2701,
+      "step": 474850
+    },
+    {
+      "epoch": 4.088326446280992,
+      "grad_norm": 3.265625,
+      "learning_rate": 1.6016371782162355e-05,
+      "loss": 3.9421,
+      "step": 474900
+    },
+    {
+      "epoch": 4.088756887052342,
+      "grad_norm": 3.6875,
+      "learning_rate": 1.6001668229034584e-05,
+      "loss": 4.3552,
+      "step": 474950
+    },
+    {
+      "epoch": 4.089187327823692,
+      "grad_norm": 4.25,
+      "learning_rate": 1.5986970841203662e-05,
+      "loss": 4.0449,
+      "step": 475000
+    },
+    {
+      "epoch": 4.0896177685950414,
+      "grad_norm": 3.5,
+      "learning_rate": 1.597227961974833e-05,
+      "loss": 4.399,
+      "step": 475050
+    },
+    {
+      "epoch": 4.090048209366391,
+      "grad_norm": 2.359375,
+      "learning_rate": 1.5957594565746846e-05,
+      "loss": 4.2658,
+      "step": 475100
+    },
+    {
+      "epoch": 4.090478650137741,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.5942915680277158e-05,
+      "loss": 4.0079,
+      "step": 475150
+    },
+    {
+      "epoch": 4.090909090909091,
+      "grad_norm": 3.265625,
+      "learning_rate": 1.592824296441655e-05,
+      "loss": 4.3937,
+      "step": 475200
+    },
+    {
+      "epoch": 4.091339531680441,
+      "grad_norm": 3.625,
+      "learning_rate": 1.591357641924206e-05,
+      "loss": 4.4162,
+      "step": 475250
+    },
+    {
+      "epoch": 4.09176997245179,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.5898916045830136e-05,
+      "loss": 4.5583,
+      "step": 475300
+    },
+    {
+      "epoch": 4.09220041322314,
+      "grad_norm": 4.375,
+      "learning_rate": 1.588426184525682e-05,
+      "loss": 4.4763,
+      "step": 475350
+    },
+    {
+      "epoch": 4.09263085399449,
+      "grad_norm": 2.328125,
+      "learning_rate": 1.5869613818597716e-05,
+      "loss": 4.2226,
+      "step": 475400
+    },
+    {
+      "epoch": 4.09306129476584,
+      "grad_norm": 4.0,
+      "learning_rate": 1.5854971966927934e-05,
+      "loss": 4.3951,
+      "step": 475450
+    },
+    {
+      "epoch": 4.0934917355371905,
+      "grad_norm": 3.765625,
+      "learning_rate": 1.5840336291322156e-05,
+      "loss": 4.2528,
+      "step": 475500
+    },
+    {
+      "epoch": 4.09392217630854,
+      "grad_norm": 2.59375,
+      "learning_rate": 1.5825706792854633e-05,
+      "loss": 4.5994,
+      "step": 475550
+    },
+    {
+      "epoch": 4.09435261707989,
+      "grad_norm": 2.984375,
+      "learning_rate": 1.581108347259913e-05,
+      "loss": 4.1948,
+      "step": 475600
+    },
+    {
+      "epoch": 4.09478305785124,
+      "grad_norm": 3.453125,
+      "learning_rate": 1.5796466331628968e-05,
+      "loss": 4.2568,
+      "step": 475650
+    },
+    {
+      "epoch": 4.09521349862259,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.5781855371017008e-05,
+      "loss": 4.6507,
+      "step": 475700
+    },
+    {
+      "epoch": 4.0956439393939394,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.576725059183566e-05,
+      "loss": 4.0814,
+      "step": 475750
+    },
+    {
+      "epoch": 4.096074380165289,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.5752651995156863e-05,
+      "loss": 4.6558,
+      "step": 475800
+    },
+    {
+      "epoch": 4.096504820936639,
+      "grad_norm": 3.9375,
+      "learning_rate": 1.5738059582052166e-05,
+      "loss": 4.064,
+      "step": 475850
+    },
+    {
+      "epoch": 4.096935261707989,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.5723473353592612e-05,
+      "loss": 4.1358,
+      "step": 475900
+    },
+    {
+      "epoch": 4.097365702479339,
+      "grad_norm": 3.546875,
+      "learning_rate": 1.570889331084878e-05,
+      "loss": 4.6223,
+      "step": 475950
+    },
+    {
+      "epoch": 4.097796143250688,
+      "grad_norm": 6.46875,
+      "learning_rate": 1.5694319454890827e-05,
+      "loss": 4.4869,
+      "step": 476000
+    },
+    {
+      "epoch": 4.098226584022038,
+      "grad_norm": 3.5,
+      "learning_rate": 1.567975178678843e-05,
+      "loss": 4.2434,
+      "step": 476050
+    },
+    {
+      "epoch": 4.098657024793388,
+      "grad_norm": 3.34375,
+      "learning_rate": 1.5665190307610812e-05,
+      "loss": 3.9627,
+      "step": 476100
+    },
+    {
+      "epoch": 4.099087465564739,
+      "grad_norm": 5.375,
+      "learning_rate": 1.565063501842683e-05,
+      "loss": 4.4755,
+      "step": 476150
+    },
+    {
+      "epoch": 4.0995179063360885,
+      "grad_norm": 3.65625,
+      "learning_rate": 1.56360859203047e-05,
+      "loss": 4.613,
+      "step": 476200
+    },
+    {
+      "epoch": 4.099948347107438,
+      "grad_norm": 5.625,
+      "learning_rate": 1.5621543014312377e-05,
+      "loss": 4.2559,
+      "step": 476250
+    },
+    {
+      "epoch": 4.100378787878788,
+      "grad_norm": 2.359375,
+      "learning_rate": 1.5607006301517245e-05,
+      "loss": 4.3416,
+      "step": 476300
+    },
+    {
+      "epoch": 4.100809228650138,
+      "grad_norm": 2.359375,
+      "learning_rate": 1.559247578298626e-05,
+      "loss": 4.5287,
+      "step": 476350
+    },
+    {
+      "epoch": 4.101239669421488,
+      "grad_norm": 1.8359375,
+      "learning_rate": 1.557795145978599e-05,
+      "loss": 4.1401,
+      "step": 476400
+    },
+    {
+      "epoch": 4.1016701101928374,
+      "grad_norm": 3.671875,
+      "learning_rate": 1.5563433332982412e-05,
+      "loss": 4.5392,
+      "step": 476450
+    },
+    {
+      "epoch": 4.102100550964187,
+      "grad_norm": 4.125,
+      "learning_rate": 1.5548921403641137e-05,
+      "loss": 4.6783,
+      "step": 476500
+    },
+    {
+      "epoch": 4.102530991735537,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.5534415672827363e-05,
+      "loss": 4.5235,
+      "step": 476550
+    },
+    {
+      "epoch": 4.102961432506887,
+      "grad_norm": 2.296875,
+      "learning_rate": 1.551991614160574e-05,
+      "loss": 4.2722,
+      "step": 476600
+    },
+    {
+      "epoch": 4.103391873278237,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.5505422811040504e-05,
+      "loss": 4.4376,
+      "step": 476650
+    },
+    {
+      "epoch": 4.103822314049586,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.549093568219544e-05,
+      "loss": 4.2473,
+      "step": 476700
+    },
+    {
+      "epoch": 4.104252754820936,
+      "grad_norm": 5.1875,
+      "learning_rate": 1.5476454756133875e-05,
+      "loss": 4.3745,
+      "step": 476750
+    },
+    {
+      "epoch": 4.104683195592287,
+      "grad_norm": 2.15625,
+      "learning_rate": 1.5461980033918656e-05,
+      "loss": 4.4971,
+      "step": 476800
+    },
+    {
+      "epoch": 4.105113636363637,
+      "grad_norm": 6.28125,
+      "learning_rate": 1.5447511516612222e-05,
+      "loss": 4.1005,
+      "step": 476850
+    },
+    {
+      "epoch": 4.1055440771349865,
+      "grad_norm": 2.65625,
+      "learning_rate": 1.543304920527654e-05,
+      "loss": 3.9515,
+      "step": 476900
+    },
+    {
+      "epoch": 4.105974517906336,
+      "grad_norm": 3.34375,
+      "learning_rate": 1.541859310097309e-05,
+      "loss": 4.7117,
+      "step": 476950
+    },
+    {
+      "epoch": 4.106404958677686,
+      "grad_norm": 1.828125,
+      "learning_rate": 1.5404143204762922e-05,
+      "loss": 4.2185,
+      "step": 477000
+    },
+    {
+      "epoch": 4.106404958677686,
+      "eval_loss": 4.99039363861084,
+      "eval_runtime": 24.7587,
+      "eval_samples_per_second": 25.85,
+      "eval_steps_per_second": 12.925,
+      "eval_tts_loss": 7.485888761083619,
+      "step": 477000
+    },
+    {
+      "epoch": 4.106835399449036,
+      "grad_norm": 6.90625,
+      "learning_rate": 1.5389699517706602e-05,
+      "loss": 4.4078,
+      "step": 477050
+    },
+    {
+      "epoch": 4.107265840220386,
+      "grad_norm": 3.796875,
+      "learning_rate": 1.5375262040864323e-05,
+      "loss": 4.4557,
+      "step": 477100
+    },
+    {
+      "epoch": 4.1076962809917354,
+      "grad_norm": 2.046875,
+      "learning_rate": 1.5360830775295744e-05,
+      "loss": 4.2413,
+      "step": 477150
+    },
+    {
+      "epoch": 4.108126721763085,
+      "grad_norm": 3.359375,
+      "learning_rate": 1.5346405722060066e-05,
+      "loss": 4.5365,
+      "step": 477200
+    },
+    {
+      "epoch": 4.108557162534435,
+      "grad_norm": 4.28125,
+      "learning_rate": 1.5331986882216077e-05,
+      "loss": 4.1829,
+      "step": 477250
+    },
+    {
+      "epoch": 4.108987603305785,
+      "grad_norm": 2.515625,
+      "learning_rate": 1.5317574256822086e-05,
+      "loss": 4.3858,
+      "step": 477300
+    },
+    {
+      "epoch": 4.109418044077135,
+      "grad_norm": 5.78125,
+      "learning_rate": 1.5303167846935917e-05,
+      "loss": 4.3307,
+      "step": 477350
+    },
+    {
+      "epoch": 4.109848484848484,
+      "grad_norm": 3.125,
+      "learning_rate": 1.528876765361502e-05,
+      "loss": 4.0777,
+      "step": 477400
+    },
+    {
+      "epoch": 4.110278925619835,
+      "grad_norm": 3.6875,
+      "learning_rate": 1.527437367791632e-05,
+      "loss": 4.2276,
+      "step": 477450
+    },
+    {
+      "epoch": 4.110709366391185,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.5259985920896302e-05,
+      "loss": 4.0667,
+      "step": 477500
+    },
+    {
+      "epoch": 4.111139807162535,
+      "grad_norm": 3.765625,
+      "learning_rate": 1.5245604383610989e-05,
+      "loss": 4.4329,
+      "step": 477550
+    },
+    {
+      "epoch": 4.1115702479338845,
+      "grad_norm": 2.796875,
+      "learning_rate": 1.523122906711596e-05,
+      "loss": 4.1659,
+      "step": 477600
+    },
+    {
+      "epoch": 4.112000688705234,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.5216859972466314e-05,
+      "loss": 4.0445,
+      "step": 477650
+    },
+    {
+      "epoch": 4.112431129476584,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.5202497100716784e-05,
+      "loss": 4.7465,
+      "step": 477700
+    },
+    {
+      "epoch": 4.112861570247934,
+      "grad_norm": 3.890625,
+      "learning_rate": 1.5188140452921462e-05,
+      "loss": 4.2884,
+      "step": 477750
+    },
+    {
+      "epoch": 4.113292011019284,
+      "grad_norm": 2.140625,
+      "learning_rate": 1.5173790030134183e-05,
+      "loss": 4.2021,
+      "step": 477800
+    },
+    {
+      "epoch": 4.1137224517906334,
+      "grad_norm": 1.9375,
+      "learning_rate": 1.5159445833408203e-05,
+      "loss": 4.1972,
+      "step": 477850
+    },
+    {
+      "epoch": 4.114152892561983,
+      "grad_norm": 1.84375,
+      "learning_rate": 1.5145107863796337e-05,
+      "loss": 4.2857,
+      "step": 477900
+    },
+    {
+      "epoch": 4.114583333333333,
+      "grad_norm": 3.828125,
+      "learning_rate": 1.5130776122351031e-05,
+      "loss": 4.6469,
+      "step": 477950
+    },
+    {
+      "epoch": 4.115013774104683,
+      "grad_norm": 4.03125,
+      "learning_rate": 1.5116450610124134e-05,
+      "loss": 4.3185,
+      "step": 478000
+    },
+    {
+      "epoch": 4.115444214876033,
+      "grad_norm": 1.2734375,
+      "learning_rate": 1.5102131328167114e-05,
+      "loss": 4.2037,
+      "step": 478050
+    },
+    {
+      "epoch": 4.115874655647383,
+      "grad_norm": 3.375,
+      "learning_rate": 1.5087818277531007e-05,
+      "loss": 4.3135,
+      "step": 478100
+    },
+    {
+      "epoch": 4.116305096418733,
+      "grad_norm": 5.21875,
+      "learning_rate": 1.5073511459266344e-05,
+      "loss": 4.3542,
+      "step": 478150
+    },
+    {
+      "epoch": 4.116735537190083,
+      "grad_norm": 4.71875,
+      "learning_rate": 1.5059210874423213e-05,
+      "loss": 4.2297,
+      "step": 478200
+    },
+    {
+      "epoch": 4.117165977961433,
+      "grad_norm": 3.46875,
+      "learning_rate": 1.5044916524051256e-05,
+      "loss": 4.3947,
+      "step": 478250
+    },
+    {
+      "epoch": 4.1175964187327825,
+      "grad_norm": 5.0,
+      "learning_rate": 1.5030628409199643e-05,
+      "loss": 4.2481,
+      "step": 478300
+    },
+    {
+      "epoch": 4.118026859504132,
+      "grad_norm": 2.875,
+      "learning_rate": 1.5016346530917058e-05,
+      "loss": 4.3567,
+      "step": 478350
+    },
+    {
+      "epoch": 4.118457300275482,
+      "grad_norm": 2.984375,
+      "learning_rate": 1.5002070890251819e-05,
+      "loss": 4.5786,
+      "step": 478400
+    },
+    {
+      "epoch": 4.118887741046832,
+      "grad_norm": 2.125,
+      "learning_rate": 1.4987801488251695e-05,
+      "loss": 4.3133,
+      "step": 478450
+    },
+    {
+      "epoch": 4.119318181818182,
+      "grad_norm": 2.59375,
+      "learning_rate": 1.4973538325964042e-05,
+      "loss": 4.5338,
+      "step": 478500
+    },
+    {
+      "epoch": 4.1197486225895315,
+      "grad_norm": 3.84375,
+      "learning_rate": 1.4959281404435732e-05,
+      "loss": 4.4301,
+      "step": 478550
+    },
+    {
+      "epoch": 4.120179063360881,
+      "grad_norm": 5.9375,
+      "learning_rate": 1.4945030724713172e-05,
+      "loss": 4.2069,
+      "step": 478600
+    },
+    {
+      "epoch": 4.120609504132231,
+      "grad_norm": 3.34375,
+      "learning_rate": 1.493078628784238e-05,
+      "loss": 3.7939,
+      "step": 478650
+    },
+    {
+      "epoch": 4.121039944903581,
+      "grad_norm": 3.0,
+      "learning_rate": 1.4916548094868844e-05,
+      "loss": 4.4275,
+      "step": 478700
+    },
+    {
+      "epoch": 4.1214703856749315,
+      "grad_norm": 3.828125,
+      "learning_rate": 1.4902316146837613e-05,
+      "loss": 4.2769,
+      "step": 478750
+    },
+    {
+      "epoch": 4.121900826446281,
+      "grad_norm": 3.0,
+      "learning_rate": 1.4888090444793279e-05,
+      "loss": 4.2704,
+      "step": 478800
+    },
+    {
+      "epoch": 4.122331267217631,
+      "grad_norm": 3.640625,
+      "learning_rate": 1.4873870989779993e-05,
+      "loss": 4.0391,
+      "step": 478850
+    },
+    {
+      "epoch": 4.122761707988981,
+      "grad_norm": 2.984375,
+      "learning_rate": 1.4859657782841396e-05,
+      "loss": 4.0243,
+      "step": 478900
+    },
+    {
+      "epoch": 4.123192148760331,
+      "grad_norm": 1.5,
+      "learning_rate": 1.4845450825020757e-05,
+      "loss": 4.2848,
+      "step": 478950
+    },
+    {
+      "epoch": 4.1236225895316805,
+      "grad_norm": 2.046875,
+      "learning_rate": 1.483125011736084e-05,
+      "loss": 4.4641,
+      "step": 479000
+    },
+    {
+      "epoch": 4.12405303030303,
+      "grad_norm": 1.4375,
+      "learning_rate": 1.4817055660903867e-05,
+      "loss": 4.1726,
+      "step": 479050
+    },
+    {
+      "epoch": 4.12448347107438,
+      "grad_norm": 3.40625,
+      "learning_rate": 1.480286745669176e-05,
+      "loss": 4.4247,
+      "step": 479100
+    },
+    {
+      "epoch": 4.12491391184573,
+      "grad_norm": 1.1171875,
+      "learning_rate": 1.4788685505765875e-05,
+      "loss": 4.6344,
+      "step": 479150
+    },
+    {
+      "epoch": 4.12534435261708,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.4774509809167114e-05,
+      "loss": 4.4806,
+      "step": 479200
+    },
+    {
+      "epoch": 4.1257747933884295,
+      "grad_norm": 4.28125,
+      "learning_rate": 1.4760340367936009e-05,
+      "loss": 4.2063,
+      "step": 479250
+    },
+    {
+      "epoch": 4.126205234159779,
+      "grad_norm": 2.046875,
+      "learning_rate": 1.4746177183112475e-05,
+      "loss": 4.0203,
+      "step": 479300
+    },
+    {
+      "epoch": 4.12663567493113,
+      "grad_norm": 3.5,
+      "learning_rate": 1.4732020255736134e-05,
+      "loss": 4.6891,
+      "step": 479350
+    },
+    {
+      "epoch": 4.12706611570248,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.4717869586846056e-05,
+      "loss": 4.6682,
+      "step": 479400
+    },
+    {
+      "epoch": 4.1274965564738295,
+      "grad_norm": 1.8125,
+      "learning_rate": 1.470372517748082e-05,
+      "loss": 4.1853,
+      "step": 479450
+    },
+    {
+      "epoch": 4.127926997245179,
+      "grad_norm": 2.234375,
+      "learning_rate": 1.4689587028678698e-05,
+      "loss": 4.5874,
+      "step": 479500
+    },
+    {
+      "epoch": 4.128357438016529,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.4675455141477312e-05,
+      "loss": 4.1844,
+      "step": 479550
+    },
+    {
+      "epoch": 4.128787878787879,
+      "grad_norm": 5.15625,
+      "learning_rate": 1.4661329516913913e-05,
+      "loss": 4.3014,
+      "step": 479600
+    },
+    {
+      "epoch": 4.129218319559229,
+      "grad_norm": 3.015625,
+      "learning_rate": 1.4647210156025337e-05,
+      "loss": 4.2146,
+      "step": 479650
+    },
+    {
+      "epoch": 4.1296487603305785,
+      "grad_norm": 4.53125,
+      "learning_rate": 1.463309705984789e-05,
+      "loss": 4.2865,
+      "step": 479700
+    },
+    {
+      "epoch": 4.130079201101928,
+      "grad_norm": 3.265625,
+      "learning_rate": 1.461899022941745e-05,
+      "loss": 4.5302,
+      "step": 479750
+    },
+    {
+      "epoch": 4.130509641873278,
+      "grad_norm": 3.4375,
+      "learning_rate": 1.460488966576944e-05,
+      "loss": 4.5474,
+      "step": 479800
+    },
+    {
+      "epoch": 4.130940082644628,
+      "grad_norm": 2.890625,
+      "learning_rate": 1.459079536993878e-05,
+      "loss": 4.2235,
+      "step": 479850
+    },
+    {
+      "epoch": 4.131370523415978,
+      "grad_norm": 2.78125,
+      "learning_rate": 1.457670734295996e-05,
+      "loss": 4.3429,
+      "step": 479900
+    },
+    {
+      "epoch": 4.1318009641873275,
+      "grad_norm": 2.28125,
+      "learning_rate": 1.4562625585867052e-05,
+      "loss": 4.317,
+      "step": 479950
+    },
+    {
+      "epoch": 4.132231404958677,
+      "grad_norm": 2.359375,
+      "learning_rate": 1.4548550099693592e-05,
+      "loss": 4.3332,
+      "step": 480000
+    },
+    {
+      "epoch": 4.132231404958677,
+      "eval_loss": 4.990168571472168,
+      "eval_runtime": 23.9078,
+      "eval_samples_per_second": 26.769,
+      "eval_steps_per_second": 13.385,
+      "eval_tts_loss": 7.47751308258016,
+      "step": 480000
+    },
+    {
+      "epoch": 4.132661845730028,
+      "grad_norm": 4.6875,
+      "learning_rate": 1.4534480885472712e-05,
+      "loss": 4.2705,
+      "step": 480050
+    },
+    {
+      "epoch": 4.133092286501378,
+      "grad_norm": 2.015625,
+      "learning_rate": 1.4520417944237053e-05,
+      "loss": 4.1762,
+      "step": 480100
+    },
+    {
+      "epoch": 4.1335227272727275,
+      "grad_norm": 2.15625,
+      "learning_rate": 1.4506361277018777e-05,
+      "loss": 4.1225,
+      "step": 480150
+    },
+    {
+      "epoch": 4.133953168044077,
+      "grad_norm": 3.828125,
+      "learning_rate": 1.4492310884849657e-05,
+      "loss": 4.1823,
+      "step": 480200
+    },
+    {
+      "epoch": 4.134383608815427,
+      "grad_norm": 3.921875,
+      "learning_rate": 1.4478266768760972e-05,
+      "loss": 4.3453,
+      "step": 480250
+    },
+    {
+      "epoch": 4.134814049586777,
+      "grad_norm": 3.53125,
+      "learning_rate": 1.4464228929783453e-05,
+      "loss": 4.2912,
+      "step": 480300
+    },
+    {
+      "epoch": 4.135244490358127,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.4450197368947516e-05,
+      "loss": 4.2596,
+      "step": 480350
+    },
+    {
+      "epoch": 4.1356749311294765,
+      "grad_norm": 2.71875,
+      "learning_rate": 1.443617208728303e-05,
+      "loss": 4.5021,
+      "step": 480400
+    },
+    {
+      "epoch": 4.136105371900826,
+      "grad_norm": 2.296875,
+      "learning_rate": 1.4422153085819402e-05,
+      "loss": 4.2144,
+      "step": 480450
+    },
+    {
+      "epoch": 4.136535812672176,
+      "grad_norm": 2.5625,
+      "learning_rate": 1.440814036558562e-05,
+      "loss": 4.2672,
+      "step": 480500
+    },
+    {
+      "epoch": 4.136966253443526,
+      "grad_norm": 2.109375,
+      "learning_rate": 1.439413392761021e-05,
+      "loss": 4.4934,
+      "step": 480550
+    },
+    {
+      "epoch": 4.137396694214876,
+      "grad_norm": 2.5,
+      "learning_rate": 1.4380133772921134e-05,
+      "loss": 4.3124,
+      "step": 480600
+    },
+    {
+      "epoch": 4.137827134986226,
+      "grad_norm": 4.125,
+      "learning_rate": 1.4366139902546061e-05,
+      "loss": 4.3709,
+      "step": 480650
+    },
+    {
+      "epoch": 4.138257575757576,
+      "grad_norm": 3.0,
+      "learning_rate": 1.4352152317512057e-05,
+      "loss": 4.1758,
+      "step": 480700
+    },
+    {
+      "epoch": 4.138688016528926,
+      "grad_norm": 1.6171875,
+      "learning_rate": 1.433817101884578e-05,
+      "loss": 4.1738,
+      "step": 480750
+    },
+    {
+      "epoch": 4.139118457300276,
+      "grad_norm": 6.28125,
+      "learning_rate": 1.4324196007573487e-05,
+      "loss": 4.467,
+      "step": 480800
+    },
+    {
+      "epoch": 4.1395488980716255,
+      "grad_norm": 4.28125,
+      "learning_rate": 1.4310227284720823e-05,
+      "loss": 4.6783,
+      "step": 480850
+    },
+    {
+      "epoch": 4.139979338842975,
+      "grad_norm": 2.65625,
+      "learning_rate": 1.4296264851313146e-05,
+      "loss": 4.1553,
+      "step": 480900
+    },
+    {
+      "epoch": 4.140409779614325,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.4282308708375225e-05,
+      "loss": 4.4313,
+      "step": 480950
+    },
+    {
+      "epoch": 4.140840220385675,
+      "grad_norm": 4.8125,
+      "learning_rate": 1.4268358856931407e-05,
+      "loss": 4.198,
+      "step": 481000
+    },
+    {
+      "epoch": 4.141270661157025,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.4254415298005608e-05,
+      "loss": 4.0836,
+      "step": 481050
+    },
+    {
+      "epoch": 4.1417011019283745,
+      "grad_norm": 2.796875,
+      "learning_rate": 1.424047803262123e-05,
+      "loss": 3.8872,
+      "step": 481100
+    },
+    {
+      "epoch": 4.142131542699724,
+      "grad_norm": 3.78125,
+      "learning_rate": 1.4226547061801232e-05,
+      "loss": 4.2321,
+      "step": 481150
+    },
+    {
+      "epoch": 4.142561983471074,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.4212622386568153e-05,
+      "loss": 4.5142,
+      "step": 481200
+    },
+    {
+      "epoch": 4.142992424242424,
+      "grad_norm": 2.515625,
+      "learning_rate": 1.419870400794402e-05,
+      "loss": 4.2966,
+      "step": 481250
+    },
+    {
+      "epoch": 4.143422865013774,
+      "grad_norm": 3.484375,
+      "learning_rate": 1.4184791926950403e-05,
+      "loss": 4.3673,
+      "step": 481300
+    },
+    {
+      "epoch": 4.143853305785124,
+      "grad_norm": 1.125,
+      "learning_rate": 1.417088614460842e-05,
+      "loss": 4.1101,
+      "step": 481350
+    },
+    {
+      "epoch": 4.144283746556474,
+      "grad_norm": 3.75,
+      "learning_rate": 1.415698666193872e-05,
+      "loss": 4.4379,
+      "step": 481400
+    },
+    {
+      "epoch": 4.144714187327824,
+      "grad_norm": 5.03125,
+      "learning_rate": 1.41430934799615e-05,
+      "loss": 4.2962,
+      "step": 481450
+    },
+    {
+      "epoch": 4.145144628099174,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.41292065996965e-05,
+      "loss": 4.308,
+      "step": 481500
+    },
+    {
+      "epoch": 4.1455750688705235,
+      "grad_norm": 4.09375,
+      "learning_rate": 1.411532602216299e-05,
+      "loss": 4.4051,
+      "step": 481550
+    },
+    {
+      "epoch": 4.146005509641873,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.4101451748379757e-05,
+      "loss": 4.8246,
+      "step": 481600
+    },
+    {
+      "epoch": 4.146435950413223,
+      "grad_norm": 4.21875,
+      "learning_rate": 1.4087583779365143e-05,
+      "loss": 4.2333,
+      "step": 481650
+    },
+    {
+      "epoch": 4.146866391184573,
+      "grad_norm": 2.6875,
+      "learning_rate": 1.4073722116137022e-05,
+      "loss": 4.2659,
+      "step": 481700
+    },
+    {
+      "epoch": 4.147296831955923,
+      "grad_norm": 2.15625,
+      "learning_rate": 1.4059866759712847e-05,
+      "loss": 4.1725,
+      "step": 481750
+    },
+    {
+      "epoch": 4.1477272727272725,
+      "grad_norm": 2.390625,
+      "learning_rate": 1.404601771110956e-05,
+      "loss": 3.9544,
+      "step": 481800
+    },
+    {
+      "epoch": 4.148157713498622,
+      "grad_norm": 3.5,
+      "learning_rate": 1.4032174971343604e-05,
+      "loss": 4.4572,
+      "step": 481850
+    },
+    {
+      "epoch": 4.148588154269972,
+      "grad_norm": 4.9375,
+      "learning_rate": 1.4018338541431064e-05,
+      "loss": 4.2602,
+      "step": 481900
+    },
+    {
+      "epoch": 4.149018595041323,
+      "grad_norm": 4.03125,
+      "learning_rate": 1.4004508422387475e-05,
+      "loss": 4.3454,
+      "step": 481950
+    },
+    {
+      "epoch": 4.1494490358126725,
+      "grad_norm": 4.125,
+      "learning_rate": 1.3990684615227912e-05,
+      "loss": 4.4533,
+      "step": 482000
+    },
+    {
+      "epoch": 4.149879476584022,
+      "grad_norm": 2.703125,
+      "learning_rate": 1.3976867120967108e-05,
+      "loss": 4.576,
+      "step": 482050
+    },
+    {
+      "epoch": 4.150309917355372,
+      "grad_norm": 3.765625,
+      "learning_rate": 1.3963055940619141e-05,
+      "loss": 4.0701,
+      "step": 482100
+    },
+    {
+      "epoch": 4.150740358126722,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.3949251075197733e-05,
+      "loss": 4.2049,
+      "step": 482150
+    },
+    {
+      "epoch": 4.151170798898072,
+      "grad_norm": 4.46875,
+      "learning_rate": 1.3935452525716175e-05,
+      "loss": 4.3496,
+      "step": 482200
+    },
+    {
+      "epoch": 4.1516012396694215,
+      "grad_norm": 1.640625,
+      "learning_rate": 1.3921660293187222e-05,
+      "loss": 4.5934,
+      "step": 482250
+    },
+    {
+      "epoch": 4.152031680440771,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.3907874378623187e-05,
+      "loss": 4.7418,
+      "step": 482300
+    },
+    {
+      "epoch": 4.152462121212121,
+      "grad_norm": 2.328125,
+      "learning_rate": 1.3894094783035993e-05,
+      "loss": 3.9849,
+      "step": 482350
+    },
+    {
+      "epoch": 4.152892561983471,
+      "grad_norm": 2.640625,
+      "learning_rate": 1.3880321507436923e-05,
+      "loss": 4.3936,
+      "step": 482400
+    },
+    {
+      "epoch": 4.153323002754821,
+      "grad_norm": 2.15625,
+      "learning_rate": 1.3866554552836986e-05,
+      "loss": 3.9804,
+      "step": 482450
+    },
+    {
+      "epoch": 4.1537534435261705,
+      "grad_norm": 1.453125,
+      "learning_rate": 1.3852793920246631e-05,
+      "loss": 4.3132,
+      "step": 482500
+    },
+    {
+      "epoch": 4.15418388429752,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.3839039610675853e-05,
+      "loss": 4.3411,
+      "step": 482550
+    },
+    {
+      "epoch": 4.15461432506887,
+      "grad_norm": 2.53125,
+      "learning_rate": 1.3825291625134184e-05,
+      "loss": 4.0502,
+      "step": 482600
+    },
+    {
+      "epoch": 4.155044765840221,
+      "grad_norm": 3.40625,
+      "learning_rate": 1.38115499646307e-05,
+      "loss": 4.3411,
+      "step": 482650
+    },
+    {
+      "epoch": 4.1554752066115705,
+      "grad_norm": 5.65625,
+      "learning_rate": 1.3797814630173978e-05,
+      "loss": 4.5538,
+      "step": 482700
+    },
+    {
+      "epoch": 4.15590564738292,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.3784085622772224e-05,
+      "loss": 4.63,
+      "step": 482750
+    },
+    {
+      "epoch": 4.15633608815427,
+      "grad_norm": 3.578125,
+      "learning_rate": 1.377036294343309e-05,
+      "loss": 4.3575,
+      "step": 482800
+    },
+    {
+      "epoch": 4.15676652892562,
+      "grad_norm": 3.90625,
+      "learning_rate": 1.375664659316378e-05,
+      "loss": 4.3019,
+      "step": 482850
+    },
+    {
+      "epoch": 4.15719696969697,
+      "grad_norm": 2.40625,
+      "learning_rate": 1.3742936572971065e-05,
+      "loss": 4.2185,
+      "step": 482900
+    },
+    {
+      "epoch": 4.1576274104683195,
+      "grad_norm": 2.1875,
+      "learning_rate": 1.372923288386121e-05,
+      "loss": 4.8507,
+      "step": 482950
+    },
+    {
+      "epoch": 4.158057851239669,
+      "grad_norm": 2.703125,
+      "learning_rate": 1.371553552684003e-05,
+      "loss": 4.0664,
+      "step": 483000
+    },
+    {
+      "epoch": 4.158057851239669,
+      "eval_loss": 4.990303993225098,
+      "eval_runtime": 24.971,
+      "eval_samples_per_second": 25.63,
+      "eval_steps_per_second": 12.815,
+      "eval_tts_loss": 7.480562996970848,
+      "step": 483000
+    },
+    {
+      "epoch": 4.158488292011019,
+      "grad_norm": 3.828125,
+      "learning_rate": 1.3701844502912909e-05,
+      "loss": 4.4465,
+      "step": 483050
+    },
+    {
+      "epoch": 4.158918732782369,
+      "grad_norm": 3.828125,
+      "learning_rate": 1.3688159813084722e-05,
+      "loss": 4.5435,
+      "step": 483100
+    },
+    {
+      "epoch": 4.159349173553719,
+      "grad_norm": 3.359375,
+      "learning_rate": 1.3674481458359911e-05,
+      "loss": 4.33,
+      "step": 483150
+    },
+    {
+      "epoch": 4.1597796143250685,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.3660809439742428e-05,
+      "loss": 3.9643,
+      "step": 483200
+    },
+    {
+      "epoch": 4.160210055096419,
+      "grad_norm": 3.359375,
+      "learning_rate": 1.3647143758235759e-05,
+      "loss": 4.0003,
+      "step": 483250
+    },
+    {
+      "epoch": 4.160640495867769,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.3633484414842935e-05,
+      "loss": 4.4465,
+      "step": 483300
+    },
+    {
+      "epoch": 4.161070936639119,
+      "grad_norm": 1.6171875,
+      "learning_rate": 1.3619831410566564e-05,
+      "loss": 3.862,
+      "step": 483350
+    },
+    {
+      "epoch": 4.1615013774104685,
+      "grad_norm": 3.265625,
+      "learning_rate": 1.3606184746408679e-05,
+      "loss": 4.0066,
+      "step": 483400
+    },
+    {
+      "epoch": 4.161931818181818,
+      "grad_norm": 5.59375,
+      "learning_rate": 1.3592544423370978e-05,
+      "loss": 4.3242,
+      "step": 483450
+    },
+    {
+      "epoch": 4.162362258953168,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.3578910442454596e-05,
+      "loss": 4.3814,
+      "step": 483500
+    },
+    {
+      "epoch": 4.162792699724518,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.3565282804660229e-05,
+      "loss": 4.3254,
+      "step": 483550
+    },
+    {
+      "epoch": 4.163223140495868,
+      "grad_norm": 3.3125,
+      "learning_rate": 1.355166151098819e-05,
+      "loss": 4.5882,
+      "step": 483600
+    },
+    {
+      "epoch": 4.1636535812672175,
+      "grad_norm": 2.703125,
+      "learning_rate": 1.3538046562438167e-05,
+      "loss": 4.061,
+      "step": 483650
+    },
+    {
+      "epoch": 4.164084022038567,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.3524437960009473e-05,
+      "loss": 4.3966,
+      "step": 483700
+    },
+    {
+      "epoch": 4.164514462809917,
+      "grad_norm": 1.75,
+      "learning_rate": 1.3510835704700997e-05,
+      "loss": 4.697,
+      "step": 483750
+    },
+    {
+      "epoch": 4.164944903581267,
+      "grad_norm": 3.234375,
+      "learning_rate": 1.3497239797511097e-05,
+      "loss": 3.862,
+      "step": 483800
+    },
+    {
+      "epoch": 4.165375344352617,
+      "grad_norm": 3.859375,
+      "learning_rate": 1.3483650239437684e-05,
+      "loss": 4.3522,
+      "step": 483850
+    },
+    {
+      "epoch": 4.1658057851239665,
+      "grad_norm": 3.640625,
+      "learning_rate": 1.3470067031478195e-05,
+      "loss": 4.2348,
+      "step": 483900
+    },
+    {
+      "epoch": 4.166236225895317,
+      "grad_norm": 4.09375,
+      "learning_rate": 1.34564901746296e-05,
+      "loss": 4.3267,
+      "step": 483950
+    },
+    {
+      "epoch": 4.166666666666667,
+      "grad_norm": 2.796875,
+      "learning_rate": 1.3442919669888432e-05,
+      "loss": 3.9724,
+      "step": 484000
+    },
+    {
+      "epoch": 4.167097107438017,
+      "grad_norm": 4.03125,
+      "learning_rate": 1.342935551825073e-05,
+      "loss": 3.7752,
+      "step": 484050
+    },
+    {
+      "epoch": 4.1675275482093666,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.3415797720712076e-05,
+      "loss": 4.7967,
+      "step": 484100
+    },
+    {
+      "epoch": 4.167957988980716,
+      "grad_norm": 7.0,
+      "learning_rate": 1.3402246278267571e-05,
+      "loss": 4.2417,
+      "step": 484150
+    },
+    {
+      "epoch": 4.168388429752066,
+      "grad_norm": 4.6875,
+      "learning_rate": 1.338870119191188e-05,
+      "loss": 4.252,
+      "step": 484200
+    },
+    {
+      "epoch": 4.168818870523416,
+      "grad_norm": 2.046875,
+      "learning_rate": 1.3375162462639146e-05,
+      "loss": 4.4715,
+      "step": 484250
+    },
+    {
+      "epoch": 4.169249311294766,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.3361630091443122e-05,
+      "loss": 4.1569,
+      "step": 484300
+    },
+    {
+      "epoch": 4.1696797520661155,
+      "grad_norm": 3.4375,
+      "learning_rate": 1.3348104079317059e-05,
+      "loss": 4.0755,
+      "step": 484350
+    },
+    {
+      "epoch": 4.170110192837465,
+      "grad_norm": 4.28125,
+      "learning_rate": 1.3334584427253704e-05,
+      "loss": 4.3337,
+      "step": 484400
+    },
+    {
+      "epoch": 4.170540633608815,
+      "grad_norm": 1.8671875,
+      "learning_rate": 1.33210711362454e-05,
+      "loss": 4.2487,
+      "step": 484450
+    },
+    {
+      "epoch": 4.170971074380165,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.3307564207283974e-05,
+      "loss": 4.3619,
+      "step": 484500
+    },
+    {
+      "epoch": 4.171401515151516,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.3294063641360787e-05,
+      "loss": 4.2338,
+      "step": 484550
+    },
+    {
+      "epoch": 4.171831955922865,
+      "grad_norm": 3.578125,
+      "learning_rate": 1.3280569439466795e-05,
+      "loss": 4.2768,
+      "step": 484600
+    },
+    {
+      "epoch": 4.172262396694215,
+      "grad_norm": 4.65625,
+      "learning_rate": 1.3267081602592436e-05,
+      "loss": 4.048,
+      "step": 484650
+    },
+    {
+      "epoch": 4.172692837465565,
+      "grad_norm": 3.8125,
+      "learning_rate": 1.3253600131727672e-05,
+      "loss": 4.5198,
+      "step": 484700
+    },
+    {
+      "epoch": 4.173123278236915,
+      "grad_norm": 3.65625,
+      "learning_rate": 1.3240125027862016e-05,
+      "loss": 4.1501,
+      "step": 484750
+    },
+    {
+      "epoch": 4.1735537190082646,
+      "grad_norm": 5.3125,
+      "learning_rate": 1.3226656291984518e-05,
+      "loss": 4.1201,
+      "step": 484800
+    },
+    {
+      "epoch": 4.173984159779614,
+      "grad_norm": 3.75,
+      "learning_rate": 1.3213193925083721e-05,
+      "loss": 4.5307,
+      "step": 484850
+    },
+    {
+      "epoch": 4.174414600550964,
+      "grad_norm": 4.3125,
+      "learning_rate": 1.3199737928147815e-05,
+      "loss": 4.5901,
+      "step": 484900
+    },
+    {
+      "epoch": 4.174845041322314,
+      "grad_norm": 3.4375,
+      "learning_rate": 1.318628830216434e-05,
+      "loss": 4.4692,
+      "step": 484950
+    },
+    {
+      "epoch": 4.175275482093664,
+      "grad_norm": 1.78125,
+      "learning_rate": 1.3172845048120542e-05,
+      "loss": 4.0747,
+      "step": 485000
+    },
+    {
+      "epoch": 4.1757059228650135,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.3159408167003106e-05,
+      "loss": 4.7405,
+      "step": 485050
+    },
+    {
+      "epoch": 4.176136363636363,
+      "grad_norm": 1.7421875,
+      "learning_rate": 1.3145977659798247e-05,
+      "loss": 4.0364,
+      "step": 485100
+    },
+    {
+      "epoch": 4.176566804407713,
+      "grad_norm": 2.078125,
+      "learning_rate": 1.3132553527491797e-05,
+      "loss": 4.2683,
+      "step": 485150
+    },
+    {
+      "epoch": 4.176997245179064,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.3119135771069002e-05,
+      "loss": 4.0564,
+      "step": 485200
+    },
+    {
+      "epoch": 4.177427685950414,
+      "grad_norm": 1.2265625,
+      "learning_rate": 1.3105724391514696e-05,
+      "loss": 4.5556,
+      "step": 485250
+    },
+    {
+      "epoch": 4.177858126721763,
+      "grad_norm": 2.453125,
+      "learning_rate": 1.3092319389813268e-05,
+      "loss": 4.1933,
+      "step": 485300
+    },
+    {
+      "epoch": 4.178288567493113,
+      "grad_norm": 3.84375,
+      "learning_rate": 1.307892076694862e-05,
+      "loss": 4.4607,
+      "step": 485350
+    },
+    {
+      "epoch": 4.178719008264463,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.306552852390417e-05,
+      "loss": 4.5075,
+      "step": 485400
+    },
+    {
+      "epoch": 4.179149449035813,
+      "grad_norm": 2.6875,
+      "learning_rate": 1.305214266166288e-05,
+      "loss": 3.9504,
+      "step": 485450
+    },
+    {
+      "epoch": 4.1795798898071626,
+      "grad_norm": 3.875,
+      "learning_rate": 1.3038763181207248e-05,
+      "loss": 4.7029,
+      "step": 485500
+    },
+    {
+      "epoch": 4.180010330578512,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.3025390083519273e-05,
+      "loss": 4.0459,
+      "step": 485550
+    },
+    {
+      "epoch": 4.180440771349862,
+      "grad_norm": 7.59375,
+      "learning_rate": 1.3012023369580551e-05,
+      "loss": 4.2071,
+      "step": 485600
+    },
+    {
+      "epoch": 4.180871212121212,
+      "grad_norm": 3.9375,
+      "learning_rate": 1.299866304037216e-05,
+      "loss": 4.1605,
+      "step": 485650
+    },
+    {
+      "epoch": 4.181301652892562,
+      "grad_norm": 4.65625,
+      "learning_rate": 1.298530909687472e-05,
+      "loss": 4.1318,
+      "step": 485700
+    },
+    {
+      "epoch": 4.1817320936639115,
+      "grad_norm": 4.4375,
+      "learning_rate": 1.2971961540068367e-05,
+      "loss": 4.0893,
+      "step": 485750
+    },
+    {
+      "epoch": 4.182162534435261,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.2958620370932773e-05,
+      "loss": 4.3281,
+      "step": 485800
+    },
+    {
+      "epoch": 4.182592975206612,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.2945285590447188e-05,
+      "loss": 4.2348,
+      "step": 485850
+    },
+    {
+      "epoch": 4.183023415977962,
+      "grad_norm": 2.859375,
+      "learning_rate": 1.2931957199590361e-05,
+      "loss": 4.2268,
+      "step": 485900
+    },
+    {
+      "epoch": 4.183453856749312,
+      "grad_norm": 5.09375,
+      "learning_rate": 1.2918635199340501e-05,
+      "loss": 4.1926,
+      "step": 485950
+    },
+    {
+      "epoch": 4.183884297520661,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.290531959067548e-05,
+      "loss": 4.3034,
+      "step": 486000
+    },
+    {
+      "epoch": 4.183884297520661,
+      "eval_loss": 4.99005651473999,
+      "eval_runtime": 23.9809,
+      "eval_samples_per_second": 26.688,
+      "eval_steps_per_second": 13.344,
+      "eval_tts_loss": 7.475855944563067,
+      "step": 486000
+    },
+    {
+      "epoch": 4.184314738292011,
+      "grad_norm": 4.0,
+      "learning_rate": 1.2892010374572616e-05,
+      "loss": 4.6245,
+      "step": 486050
+    },
+    {
+      "epoch": 4.184745179063361,
+      "grad_norm": 1.953125,
+      "learning_rate": 1.287870755200874e-05,
+      "loss": 4.6211,
+      "step": 486100
+    },
+    {
+      "epoch": 4.185175619834711,
+      "grad_norm": 3.734375,
+      "learning_rate": 1.2865411123960313e-05,
+      "loss": 4.3865,
+      "step": 486150
+    },
+    {
+      "epoch": 4.1856060606060606,
+      "grad_norm": 3.46875,
+      "learning_rate": 1.2852121091403235e-05,
+      "loss": 4.2504,
+      "step": 486200
+    },
+    {
+      "epoch": 4.18603650137741,
+      "grad_norm": 3.609375,
+      "learning_rate": 1.2838837455312958e-05,
+      "loss": 4.6313,
+      "step": 486250
+    },
+    {
+      "epoch": 4.18646694214876,
+      "grad_norm": 5.4375,
+      "learning_rate": 1.2825560216664489e-05,
+      "loss": 4.4578,
+      "step": 486300
+    },
+    {
+      "epoch": 4.18689738292011,
+      "grad_norm": 1.9609375,
+      "learning_rate": 1.2812289376432329e-05,
+      "loss": 4.2719,
+      "step": 486350
+    },
+    {
+      "epoch": 4.18732782369146,
+      "grad_norm": 3.609375,
+      "learning_rate": 1.2799024935590531e-05,
+      "loss": 4.4011,
+      "step": 486400
+    },
+    {
+      "epoch": 4.1877582644628095,
+      "grad_norm": 3.53125,
+      "learning_rate": 1.2785766895112717e-05,
+      "loss": 4.1507,
+      "step": 486450
+    },
+    {
+      "epoch": 4.18818870523416,
+      "grad_norm": 3.515625,
+      "learning_rate": 1.2772515255971928e-05,
+      "loss": 4.0321,
+      "step": 486500
+    },
+    {
+      "epoch": 4.18861914600551,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.2759270019140867e-05,
+      "loss": 4.1574,
+      "step": 486550
+    },
+    {
+      "epoch": 4.18904958677686,
+      "grad_norm": 2.3125,
+      "learning_rate": 1.2746031185591677e-05,
+      "loss": 4.403,
+      "step": 486600
+    },
+    {
+      "epoch": 4.18948002754821,
+      "grad_norm": 1.578125,
+      "learning_rate": 1.273279875629605e-05,
+      "loss": 4.1491,
+      "step": 486650
+    },
+    {
+      "epoch": 4.189910468319559,
+      "grad_norm": 3.125,
+      "learning_rate": 1.2719572732225283e-05,
+      "loss": 4.235,
+      "step": 486700
+    },
+    {
+      "epoch": 4.190340909090909,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.2706353114350055e-05,
+      "loss": 4.3334,
+      "step": 486750
+    },
+    {
+      "epoch": 4.190771349862259,
+      "grad_norm": 3.8125,
+      "learning_rate": 1.2693139903640683e-05,
+      "loss": 4.5982,
+      "step": 486800
+    },
+    {
+      "epoch": 4.191201790633609,
+      "grad_norm": 5.3125,
+      "learning_rate": 1.267993310106701e-05,
+      "loss": 4.2806,
+      "step": 486850
+    },
+    {
+      "epoch": 4.1916322314049586,
+      "grad_norm": 3.453125,
+      "learning_rate": 1.2666732707598372e-05,
+      "loss": 4.1032,
+      "step": 486900
+    },
+    {
+      "epoch": 4.192062672176308,
+      "grad_norm": 4.78125,
+      "learning_rate": 1.265353872420365e-05,
+      "loss": 4.1981,
+      "step": 486950
+    },
+    {
+      "epoch": 4.192493112947658,
+      "grad_norm": 3.25,
+      "learning_rate": 1.264035115185126e-05,
+      "loss": 4.3563,
+      "step": 487000
+    },
+    {
+      "epoch": 4.192923553719008,
+      "grad_norm": 2.703125,
+      "learning_rate": 1.2627169991509136e-05,
+      "loss": 4.1019,
+      "step": 487050
+    },
+    {
+      "epoch": 4.193353994490358,
+      "grad_norm": 0.92578125,
+      "learning_rate": 1.2613995244144727e-05,
+      "loss": 4.1691,
+      "step": 487100
+    },
+    {
+      "epoch": 4.193784435261708,
+      "grad_norm": 3.40625,
+      "learning_rate": 1.2600826910725072e-05,
+      "loss": 4.3536,
+      "step": 487150
+    },
+    {
+      "epoch": 4.194214876033058,
+      "grad_norm": 3.125,
+      "learning_rate": 1.2587664992216674e-05,
+      "loss": 4.2473,
+      "step": 487200
+    },
+    {
+      "epoch": 4.194645316804408,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.2574509489585595e-05,
+      "loss": 4.6249,
+      "step": 487250
+    },
+    {
+      "epoch": 4.195075757575758,
+      "grad_norm": 2.890625,
+      "learning_rate": 1.2561360403797418e-05,
+      "loss": 4.3506,
+      "step": 487300
+    },
+    {
+      "epoch": 4.195506198347108,
+      "grad_norm": 2.21875,
+      "learning_rate": 1.2548217735817236e-05,
+      "loss": 4.4289,
+      "step": 487350
+    },
+    {
+      "epoch": 4.195936639118457,
+      "grad_norm": 3.40625,
+      "learning_rate": 1.2535081486609735e-05,
+      "loss": 4.3421,
+      "step": 487400
+    },
+    {
+      "epoch": 4.196367079889807,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.2521951657139086e-05,
+      "loss": 4.3659,
+      "step": 487450
+    },
+    {
+      "epoch": 4.196797520661157,
+      "grad_norm": 3.53125,
+      "learning_rate": 1.2508828248368942e-05,
+      "loss": 4.1716,
+      "step": 487500
+    },
+    {
+      "epoch": 4.197227961432507,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.2495711261262566e-05,
+      "loss": 4.5417,
+      "step": 487550
+    },
+    {
+      "epoch": 4.1976584022038566,
+      "grad_norm": 2.796875,
+      "learning_rate": 1.2482600696782731e-05,
+      "loss": 4.2121,
+      "step": 487600
+    },
+    {
+      "epoch": 4.198088842975206,
+      "grad_norm": 2.671875,
+      "learning_rate": 1.2469496555891669e-05,
+      "loss": 4.2866,
+      "step": 487650
+    },
+    {
+      "epoch": 4.198519283746556,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.2456398839551286e-05,
+      "loss": 3.9799,
+      "step": 487700
+    },
+    {
+      "epoch": 4.198949724517906,
+      "grad_norm": 4.125,
+      "learning_rate": 1.2443307548722848e-05,
+      "loss": 4.2645,
+      "step": 487750
+    },
+    {
+      "epoch": 4.199380165289257,
+      "grad_norm": 4.8125,
+      "learning_rate": 1.2430222684367233e-05,
+      "loss": 4.319,
+      "step": 487800
+    },
+    {
+      "epoch": 4.199810606060606,
+      "grad_norm": 4.40625,
+      "learning_rate": 1.2417144247444879e-05,
+      "loss": 4.1868,
+      "step": 487850
+    },
+    {
+      "epoch": 4.200241046831956,
+      "grad_norm": 3.296875,
+      "learning_rate": 1.240407223891571e-05,
+      "loss": 4.6528,
+      "step": 487900
+    },
+    {
+      "epoch": 4.200671487603306,
+      "grad_norm": 3.859375,
+      "learning_rate": 1.2391006659739135e-05,
+      "loss": 4.1551,
+      "step": 487950
+    },
+    {
+      "epoch": 4.201101928374656,
+      "grad_norm": 3.296875,
+      "learning_rate": 1.2377947510874233e-05,
+      "loss": 3.7787,
+      "step": 488000
+    },
+    {
+      "epoch": 4.201532369146006,
+      "grad_norm": 3.3125,
+      "learning_rate": 1.2364894793279402e-05,
+      "loss": 4.6407,
+      "step": 488050
+    },
+    {
+      "epoch": 4.201962809917355,
+      "grad_norm": 6.21875,
+      "learning_rate": 1.2351848507912767e-05,
+      "loss": 4.4595,
+      "step": 488100
+    },
+    {
+      "epoch": 4.202393250688705,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.2338808655731882e-05,
+      "loss": 4.3777,
+      "step": 488150
+    },
+    {
+      "epoch": 4.202823691460055,
+      "grad_norm": 3.828125,
+      "learning_rate": 1.2325775237693804e-05,
+      "loss": 3.9588,
+      "step": 488200
+    },
+    {
+      "epoch": 4.203254132231405,
+      "grad_norm": 4.40625,
+      "learning_rate": 1.2312748254755225e-05,
+      "loss": 3.8889,
+      "step": 488250
+    },
+    {
+      "epoch": 4.2036845730027546,
+      "grad_norm": 4.625,
+      "learning_rate": 1.2299727707872243e-05,
+      "loss": 4.2328,
+      "step": 488300
+    },
+    {
+      "epoch": 4.204115013774104,
+      "grad_norm": 3.46875,
+      "learning_rate": 1.2286713598000544e-05,
+      "loss": 4.4372,
+      "step": 488350
+    },
+    {
+      "epoch": 4.204545454545454,
+      "grad_norm": 3.34375,
+      "learning_rate": 1.2273705926095358e-05,
+      "loss": 4.3798,
+      "step": 488400
+    },
+    {
+      "epoch": 4.204975895316805,
+      "grad_norm": 7.90625,
+      "learning_rate": 1.2260704693111413e-05,
+      "loss": 4.2295,
+      "step": 488450
+    },
+    {
+      "epoch": 4.205406336088155,
+      "grad_norm": 2.28125,
+      "learning_rate": 1.2247709900002957e-05,
+      "loss": 3.9938,
+      "step": 488500
+    },
+    {
+      "epoch": 4.205836776859504,
+      "grad_norm": 2.28125,
+      "learning_rate": 1.2234721547723803e-05,
+      "loss": 4.1058,
+      "step": 488550
+    },
+    {
+      "epoch": 4.206267217630854,
+      "grad_norm": 1.53125,
+      "learning_rate": 1.2221739637227247e-05,
+      "loss": 4.3419,
+      "step": 488600
+    },
+    {
+      "epoch": 4.206697658402204,
+      "grad_norm": 2.46875,
+      "learning_rate": 1.2208764169466125e-05,
+      "loss": 4.1758,
+      "step": 488650
+    },
+    {
+      "epoch": 4.207128099173554,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.2195795145392852e-05,
+      "loss": 4.3812,
+      "step": 488700
+    },
+    {
+      "epoch": 4.207558539944904,
+      "grad_norm": 3.984375,
+      "learning_rate": 1.218283256595929e-05,
+      "loss": 4.1588,
+      "step": 488750
+    },
+    {
+      "epoch": 4.207988980716253,
+      "grad_norm": 3.0625,
+      "learning_rate": 1.2169876432116878e-05,
+      "loss": 4.2296,
+      "step": 488800
+    },
+    {
+      "epoch": 4.208419421487603,
+      "grad_norm": 4.1875,
+      "learning_rate": 1.2156926744816565e-05,
+      "loss": 4.2307,
+      "step": 488850
+    },
+    {
+      "epoch": 4.208849862258953,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.2143983505008815e-05,
+      "loss": 4.2124,
+      "step": 488900
+    },
+    {
+      "epoch": 4.209280303030303,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.2131046713643656e-05,
+      "loss": 4.29,
+      "step": 488950
+    },
+    {
+      "epoch": 4.2097107438016526,
+      "grad_norm": 8.125,
+      "learning_rate": 1.211811637167064e-05,
+      "loss": 3.9951,
+      "step": 489000
+    },
+    {
+      "epoch": 4.2097107438016526,
+      "eval_loss": 4.99033260345459,
+      "eval_runtime": 22.4777,
+      "eval_samples_per_second": 28.473,
+      "eval_steps_per_second": 14.236,
+      "eval_tts_loss": 7.48369155709714,
+      "step": 489000
+    },
+    {
+      "epoch": 4.210141184573002,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.2105192480038762e-05,
+      "loss": 4.3183,
+      "step": 489050
+    },
+    {
+      "epoch": 4.210571625344353,
+      "grad_norm": 3.75,
+      "learning_rate": 1.2092275039696666e-05,
+      "loss": 4.1129,
+      "step": 489100
+    },
+    {
+      "epoch": 4.211002066115703,
+      "grad_norm": 3.34375,
+      "learning_rate": 1.2079364051592446e-05,
+      "loss": 4.1003,
+      "step": 489150
+    },
+    {
+      "epoch": 4.211432506887053,
+      "grad_norm": 4.21875,
+      "learning_rate": 1.2066459516673711e-05,
+      "loss": 4.3532,
+      "step": 489200
+    },
+    {
+      "epoch": 4.211862947658402,
+      "grad_norm": 4.65625,
+      "learning_rate": 1.2053561435887694e-05,
+      "loss": 4.6279,
+      "step": 489250
+    },
+    {
+      "epoch": 4.212293388429752,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.2040669810181027e-05,
+      "loss": 4.5556,
+      "step": 489300
+    },
+    {
+      "epoch": 4.212723829201102,
+      "grad_norm": 3.609375,
+      "learning_rate": 1.2027784640499928e-05,
+      "loss": 4.3745,
+      "step": 489350
+    },
+    {
+      "epoch": 4.213154269972452,
+      "grad_norm": 2.625,
+      "learning_rate": 1.2014905927790176e-05,
+      "loss": 4.2876,
+      "step": 489400
+    },
+    {
+      "epoch": 4.213584710743802,
+      "grad_norm": 1.609375,
+      "learning_rate": 1.2002033672997026e-05,
+      "loss": 3.9901,
+      "step": 489450
+    },
+    {
+      "epoch": 4.214015151515151,
+      "grad_norm": 4.03125,
+      "learning_rate": 1.1989167877065276e-05,
+      "loss": 4.4062,
+      "step": 489500
+    },
+    {
+      "epoch": 4.214445592286501,
+      "grad_norm": 3.59375,
+      "learning_rate": 1.1976308540939241e-05,
+      "loss": 3.9376,
+      "step": 489550
+    },
+    {
+      "epoch": 4.214876033057851,
+      "grad_norm": 2.625,
+      "learning_rate": 1.196345566556275e-05,
+      "loss": 4.2519,
+      "step": 489600
+    },
+    {
+      "epoch": 4.215306473829201,
+      "grad_norm": 4.0,
+      "learning_rate": 1.195060925187922e-05,
+      "loss": 4.4874,
+      "step": 489650
+    },
+    {
+      "epoch": 4.2157369146005506,
+      "grad_norm": 4.28125,
+      "learning_rate": 1.1937769300831525e-05,
+      "loss": 4.5443,
+      "step": 489700
+    },
+    {
+      "epoch": 4.216167355371901,
+      "grad_norm": 5.71875,
+      "learning_rate": 1.1924935813362092e-05,
+      "loss": 4.2705,
+      "step": 489750
+    },
+    {
+      "epoch": 4.216597796143251,
+      "grad_norm": 1.671875,
+      "learning_rate": 1.191210879041288e-05,
+      "loss": 4.2869,
+      "step": 489800
+    },
+    {
+      "epoch": 4.217028236914601,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.1899288232925343e-05,
+      "loss": 4.2483,
+      "step": 489850
+    },
+    {
+      "epoch": 4.217458677685951,
+      "grad_norm": 5.1875,
+      "learning_rate": 1.1886474141840486e-05,
+      "loss": 4.6351,
+      "step": 489900
+    },
+    {
+      "epoch": 4.2178891184573,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.1873666518098857e-05,
+      "loss": 4.4643,
+      "step": 489950
+    },
+    {
+      "epoch": 4.21831955922865,
+      "grad_norm": 4.34375,
+      "learning_rate": 1.1860865362640505e-05,
+      "loss": 4.3536,
+      "step": 490000
+    },
+    {
+      "epoch": 4.21875,
+      "grad_norm": 2.890625,
+      "learning_rate": 1.1848070676404998e-05,
+      "loss": 4.6972,
+      "step": 490050
+    },
+    {
+      "epoch": 4.21918044077135,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.1835282460331443e-05,
+      "loss": 4.7093,
+      "step": 490100
+    },
+    {
+      "epoch": 4.2196108815427,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.1822500715358464e-05,
+      "loss": 4.4186,
+      "step": 490150
+    },
+    {
+      "epoch": 4.220041322314049,
+      "grad_norm": 2.734375,
+      "learning_rate": 1.180972544242418e-05,
+      "loss": 4.4695,
+      "step": 490200
+    },
+    {
+      "epoch": 4.220471763085399,
+      "grad_norm": 6.8125,
+      "learning_rate": 1.1796956642466339e-05,
+      "loss": 4.7324,
+      "step": 490250
+    },
+    {
+      "epoch": 4.220902203856749,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.1784194316422092e-05,
+      "loss": 3.6999,
+      "step": 490300
+    },
+    {
+      "epoch": 4.221332644628099,
+      "grad_norm": 3.0625,
+      "learning_rate": 1.177143846522819e-05,
+      "loss": 4.292,
+      "step": 490350
+    },
+    {
+      "epoch": 4.2217630853994494,
+      "grad_norm": 5.5,
+      "learning_rate": 1.1758689089820873e-05,
+      "loss": 4.0894,
+      "step": 490400
+    },
+    {
+      "epoch": 4.222193526170799,
+      "grad_norm": 3.875,
+      "learning_rate": 1.1745946191135903e-05,
+      "loss": 4.2006,
+      "step": 490450
+    },
+    {
+      "epoch": 4.222623966942149,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.1733209770108621e-05,
+      "loss": 4.4199,
+      "step": 490500
+    },
+    {
+      "epoch": 4.223054407713499,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.1720479827673847e-05,
+      "loss": 4.5137,
+      "step": 490550
+    },
+    {
+      "epoch": 4.223484848484849,
+      "grad_norm": 3.0625,
+      "learning_rate": 1.1707756364765877e-05,
+      "loss": 4.4945,
+      "step": 490600
+    },
+    {
+      "epoch": 4.223915289256198,
+      "grad_norm": 2.53125,
+      "learning_rate": 1.1695039382318652e-05,
+      "loss": 4.2744,
+      "step": 490650
+    },
+    {
+      "epoch": 4.224345730027548,
+      "grad_norm": 1.9140625,
+      "learning_rate": 1.1682328881265536e-05,
+      "loss": 4.1861,
+      "step": 490700
+    },
+    {
+      "epoch": 4.224776170798898,
+      "grad_norm": 3.890625,
+      "learning_rate": 1.166962486253943e-05,
+      "loss": 4.3473,
+      "step": 490750
+    },
+    {
+      "epoch": 4.225206611570248,
+      "grad_norm": 3.828125,
+      "learning_rate": 1.1656927327072864e-05,
+      "loss": 4.2845,
+      "step": 490800
+    },
+    {
+      "epoch": 4.225637052341598,
+      "grad_norm": 1.8125,
+      "learning_rate": 1.1644236275797737e-05,
+      "loss": 4.7198,
+      "step": 490850
+    },
+    {
+      "epoch": 4.226067493112947,
+      "grad_norm": 2.0625,
+      "learning_rate": 1.1631551709645538e-05,
+      "loss": 4.3826,
+      "step": 490900
+    },
+    {
+      "epoch": 4.226497933884297,
+      "grad_norm": 3.25,
+      "learning_rate": 1.1618873629547334e-05,
+      "loss": 4.6071,
+      "step": 490950
+    },
+    {
+      "epoch": 4.226928374655647,
+      "grad_norm": 4.53125,
+      "learning_rate": 1.1606202036433655e-05,
+      "loss": 4.3886,
+      "step": 491000
+    },
+    {
+      "epoch": 4.227358815426998,
+      "grad_norm": 4.28125,
+      "learning_rate": 1.1593536931234561e-05,
+      "loss": 4.6401,
+      "step": 491050
+    },
+    {
+      "epoch": 4.2277892561983474,
+      "grad_norm": 7.46875,
+      "learning_rate": 1.1580878314879639e-05,
+      "loss": 4.5303,
+      "step": 491100
+    },
+    {
+      "epoch": 4.228219696969697,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.1568226188298004e-05,
+      "loss": 4.3496,
+      "step": 491150
+    },
+    {
+      "epoch": 4.228650137741047,
+      "grad_norm": 2.890625,
+      "learning_rate": 1.155558055241831e-05,
+      "loss": 4.3177,
+      "step": 491200
+    },
+    {
+      "epoch": 4.229080578512397,
+      "grad_norm": 3.25,
+      "learning_rate": 1.1542941408168706e-05,
+      "loss": 4.6797,
+      "step": 491250
+    },
+    {
+      "epoch": 4.229511019283747,
+      "grad_norm": 3.515625,
+      "learning_rate": 1.1530308756476893e-05,
+      "loss": 4.4444,
+      "step": 491300
+    },
+    {
+      "epoch": 4.229941460055096,
+      "grad_norm": 4.25,
+      "learning_rate": 1.1517682598270063e-05,
+      "loss": 4.6788,
+      "step": 491350
+    },
+    {
+      "epoch": 4.230371900826446,
+      "grad_norm": 4.125,
+      "learning_rate": 1.1505062934474952e-05,
+      "loss": 4.2895,
+      "step": 491400
+    },
+    {
+      "epoch": 4.230802341597796,
+      "grad_norm": 6.25,
+      "learning_rate": 1.149244976601781e-05,
+      "loss": 4.0844,
+      "step": 491450
+    },
+    {
+      "epoch": 4.231232782369146,
+      "grad_norm": 3.84375,
+      "learning_rate": 1.1479843093824438e-05,
+      "loss": 3.9535,
+      "step": 491500
+    },
+    {
+      "epoch": 4.231663223140496,
+      "grad_norm": 2.890625,
+      "learning_rate": 1.1467242918820132e-05,
+      "loss": 4.4283,
+      "step": 491550
+    },
+    {
+      "epoch": 4.232093663911845,
+      "grad_norm": 3.4375,
+      "learning_rate": 1.1454649241929716e-05,
+      "loss": 4.6304,
+      "step": 491600
+    },
+    {
+      "epoch": 4.232524104683195,
+      "grad_norm": 3.046875,
+      "learning_rate": 1.144206206407753e-05,
+      "loss": 4.277,
+      "step": 491650
+    },
+    {
+      "epoch": 4.232954545454546,
+      "grad_norm": 1.640625,
+      "learning_rate": 1.1429481386187446e-05,
+      "loss": 4.1858,
+      "step": 491700
+    },
+    {
+      "epoch": 4.233384986225896,
+      "grad_norm": 2.78125,
+      "learning_rate": 1.1416907209182858e-05,
+      "loss": 4.4552,
+      "step": 491750
+    },
+    {
+      "epoch": 4.2338154269972454,
+      "grad_norm": 3.25,
+      "learning_rate": 1.1404339533986697e-05,
+      "loss": 4.4078,
+      "step": 491800
+    },
+    {
+      "epoch": 4.234245867768595,
+      "grad_norm": 3.484375,
+      "learning_rate": 1.1391778361521398e-05,
+      "loss": 4.3713,
+      "step": 491850
+    },
+    {
+      "epoch": 4.234676308539945,
+      "grad_norm": 2.796875,
+      "learning_rate": 1.1379223692708929e-05,
+      "loss": 4.0514,
+      "step": 491900
+    },
+    {
+      "epoch": 4.235106749311295,
+      "grad_norm": 2.046875,
+      "learning_rate": 1.136667552847076e-05,
+      "loss": 4.1963,
+      "step": 491950
+    },
+    {
+      "epoch": 4.235537190082645,
+      "grad_norm": 4.0625,
+      "learning_rate": 1.1354133869727912e-05,
+      "loss": 4.4756,
+      "step": 492000
+    },
+    {
+      "epoch": 4.235537190082645,
+      "eval_loss": 4.990378379821777,
+      "eval_runtime": 22.1678,
+      "eval_samples_per_second": 28.871,
+      "eval_steps_per_second": 14.435,
+      "eval_tts_loss": 7.481186896345774,
+      "step": 492000
+    },
+    {
+      "epoch": 4.235967630853994,
+      "grad_norm": 2.671875,
+      "learning_rate": 1.1341598717400892e-05,
+      "loss": 4.5046,
+      "step": 492050
+    },
+    {
+      "epoch": 4.236398071625344,
+      "grad_norm": 2.515625,
+      "learning_rate": 1.1329070072409808e-05,
+      "loss": 4.0987,
+      "step": 492100
+    },
+    {
+      "epoch": 4.236828512396694,
+      "grad_norm": 1.953125,
+      "learning_rate": 1.131654793567416e-05,
+      "loss": 4.2011,
+      "step": 492150
+    },
+    {
+      "epoch": 4.237258953168044,
+      "grad_norm": 4.09375,
+      "learning_rate": 1.130403230811311e-05,
+      "loss": 3.7776,
+      "step": 492200
+    },
+    {
+      "epoch": 4.237689393939394,
+      "grad_norm": 4.21875,
+      "learning_rate": 1.129152319064526e-05,
+      "loss": 4.0129,
+      "step": 492250
+    },
+    {
+      "epoch": 4.238119834710743,
+      "grad_norm": 2.609375,
+      "learning_rate": 1.1279020584188715e-05,
+      "loss": 4.2839,
+      "step": 492300
+    },
+    {
+      "epoch": 4.238550275482094,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.1266524489661212e-05,
+      "loss": 3.9612,
+      "step": 492350
+    },
+    {
+      "epoch": 4.238980716253444,
+      "grad_norm": 3.3125,
+      "learning_rate": 1.1254034907979882e-05,
+      "loss": 4.2487,
+      "step": 492400
+    },
+    {
+      "epoch": 4.239411157024794,
+      "grad_norm": 3.984375,
+      "learning_rate": 1.1241551840061426e-05,
+      "loss": 4.2714,
+      "step": 492450
+    },
+    {
+      "epoch": 4.2398415977961434,
+      "grad_norm": 5.21875,
+      "learning_rate": 1.1229075286822121e-05,
+      "loss": 4.0773,
+      "step": 492500
+    },
+    {
+      "epoch": 4.240272038567493,
+      "grad_norm": 3.140625,
+      "learning_rate": 1.1216605249177692e-05,
+      "loss": 3.8335,
+      "step": 492550
+    },
+    {
+      "epoch": 4.240702479338843,
+      "grad_norm": 2.890625,
+      "learning_rate": 1.1204141728043404e-05,
+      "loss": 4.1616,
+      "step": 492600
+    },
+    {
+      "epoch": 4.241132920110193,
+      "grad_norm": 3.265625,
+      "learning_rate": 1.1191684724334083e-05,
+      "loss": 4.4248,
+      "step": 492650
+    },
+    {
+      "epoch": 4.241563360881543,
+      "grad_norm": 1.6484375,
+      "learning_rate": 1.117923423896401e-05,
+      "loss": 4.0175,
+      "step": 492700
+    },
+    {
+      "epoch": 4.241993801652892,
+      "grad_norm": 2.3125,
+      "learning_rate": 1.1166790272847028e-05,
+      "loss": 4.2557,
+      "step": 492750
+    },
+    {
+      "epoch": 4.242424242424242,
+      "grad_norm": 4.15625,
+      "learning_rate": 1.1154352826896531e-05,
+      "loss": 4.7157,
+      "step": 492800
+    },
+    {
+      "epoch": 4.242854683195592,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.1141921902025387e-05,
+      "loss": 4.437,
+      "step": 492850
+    },
+    {
+      "epoch": 4.243285123966942,
+      "grad_norm": 3.671875,
+      "learning_rate": 1.112949749914598e-05,
+      "loss": 4.1353,
+      "step": 492900
+    },
+    {
+      "epoch": 4.243715564738292,
+      "grad_norm": 3.796875,
+      "learning_rate": 1.1117079619170256e-05,
+      "loss": 4.5801,
+      "step": 492950
+    },
+    {
+      "epoch": 4.244146005509642,
+      "grad_norm": 4.0625,
+      "learning_rate": 1.1104668263009622e-05,
+      "loss": 4.5561,
+      "step": 493000
+    },
+    {
+      "epoch": 4.244576446280992,
+      "grad_norm": 3.296875,
+      "learning_rate": 1.10922634315751e-05,
+      "loss": 4.7425,
+      "step": 493050
+    },
+    {
+      "epoch": 4.245006887052342,
+      "grad_norm": 3.921875,
+      "learning_rate": 1.1079865125777177e-05,
+      "loss": 4.5495,
+      "step": 493100
+    },
+    {
+      "epoch": 4.245437327823692,
+      "grad_norm": 2.90625,
+      "learning_rate": 1.1067473346525791e-05,
+      "loss": 3.8888,
+      "step": 493150
+    },
+    {
+      "epoch": 4.2458677685950414,
+      "grad_norm": 5.34375,
+      "learning_rate": 1.1055088094730538e-05,
+      "loss": 4.0591,
+      "step": 493200
+    },
+    {
+      "epoch": 4.246298209366391,
+      "grad_norm": 3.4375,
+      "learning_rate": 1.1042709371300453e-05,
+      "loss": 4.4691,
+      "step": 493250
+    },
+    {
+      "epoch": 4.246728650137741,
+      "grad_norm": 1.21875,
+      "learning_rate": 1.103033717714408e-05,
+      "loss": 4.0431,
+      "step": 493300
+    },
+    {
+      "epoch": 4.247159090909091,
+      "grad_norm": 3.09375,
+      "learning_rate": 1.1017971513169555e-05,
+      "loss": 4.3939,
+      "step": 493350
+    },
+    {
+      "epoch": 4.247589531680441,
+      "grad_norm": 4.8125,
+      "learning_rate": 1.1005612380284469e-05,
+      "loss": 4.3246,
+      "step": 493400
+    },
+    {
+      "epoch": 4.24801997245179,
+      "grad_norm": 3.0625,
+      "learning_rate": 1.099325977939597e-05,
+      "loss": 4.3051,
+      "step": 493450
+    },
+    {
+      "epoch": 4.24845041322314,
+      "grad_norm": 2.171875,
+      "learning_rate": 1.098091371141069e-05,
+      "loss": 4.3707,
+      "step": 493500
+    },
+    {
+      "epoch": 4.24888085399449,
+      "grad_norm": 3.15625,
+      "learning_rate": 1.0968574177234815e-05,
+      "loss": 4.4762,
+      "step": 493550
+    },
+    {
+      "epoch": 4.24931129476584,
+      "grad_norm": 5.5625,
+      "learning_rate": 1.0956241177774019e-05,
+      "loss": 4.3958,
+      "step": 493600
+    },
+    {
+      "epoch": 4.2497417355371905,
+      "grad_norm": 2.8125,
+      "learning_rate": 1.0943914713933578e-05,
+      "loss": 4.351,
+      "step": 493650
+    },
+    {
+      "epoch": 4.25017217630854,
+      "grad_norm": 3.234375,
+      "learning_rate": 1.0931594786618149e-05,
+      "loss": 4.0912,
+      "step": 493700
+    },
+    {
+      "epoch": 4.25060261707989,
+      "grad_norm": 4.9375,
+      "learning_rate": 1.091928139673204e-05,
+      "loss": 4.5503,
+      "step": 493750
+    },
+    {
+      "epoch": 4.25103305785124,
+      "grad_norm": 3.4375,
+      "learning_rate": 1.0906974545179027e-05,
+      "loss": 4.2763,
+      "step": 493800
+    },
+    {
+      "epoch": 4.25146349862259,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.0894674232862367e-05,
+      "loss": 4.3316,
+      "step": 493850
+    },
+    {
+      "epoch": 4.2518939393939394,
+      "grad_norm": 1.21875,
+      "learning_rate": 1.0882380460684938e-05,
+      "loss": 3.9106,
+      "step": 493900
+    },
+    {
+      "epoch": 4.252324380165289,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.0870093229549028e-05,
+      "loss": 4.0961,
+      "step": 493950
+    },
+    {
+      "epoch": 4.252754820936639,
+      "grad_norm": 4.8125,
+      "learning_rate": 1.0857812540356494e-05,
+      "loss": 4.435,
+      "step": 494000
+    },
+    {
+      "epoch": 4.253185261707989,
+      "grad_norm": 3.5,
+      "learning_rate": 1.0845538394008736e-05,
+      "loss": 4.5434,
+      "step": 494050
+    },
+    {
+      "epoch": 4.253615702479339,
+      "grad_norm": 4.0,
+      "learning_rate": 1.0833270791406635e-05,
+      "loss": 4.1598,
+      "step": 494100
+    },
+    {
+      "epoch": 4.254046143250688,
+      "grad_norm": 3.34375,
+      "learning_rate": 1.0821009733450616e-05,
+      "loss": 4.1035,
+      "step": 494150
+    },
+    {
+      "epoch": 4.254476584022038,
+      "grad_norm": 3.0,
+      "learning_rate": 1.0808755221040613e-05,
+      "loss": 4.4697,
+      "step": 494200
+    },
+    {
+      "epoch": 4.254907024793388,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.079650725507607e-05,
+      "loss": 4.1072,
+      "step": 494250
+    },
+    {
+      "epoch": 4.255337465564739,
+      "grad_norm": 2.734375,
+      "learning_rate": 1.0784265836455954e-05,
+      "loss": 4.5201,
+      "step": 494300
+    },
+    {
+      "epoch": 4.2557679063360885,
+      "grad_norm": 3.21875,
+      "learning_rate": 1.0772030966078795e-05,
+      "loss": 4.5165,
+      "step": 494350
+    },
+    {
+      "epoch": 4.256198347107438,
+      "grad_norm": 4.25,
+      "learning_rate": 1.0759802644842576e-05,
+      "loss": 4.6163,
+      "step": 494400
+    },
+    {
+      "epoch": 4.256628787878788,
+      "grad_norm": 1.46875,
+      "learning_rate": 1.0747580873644848e-05,
+      "loss": 4.406,
+      "step": 494450
+    },
+    {
+      "epoch": 4.257059228650138,
+      "grad_norm": 2.078125,
+      "learning_rate": 1.0735365653382645e-05,
+      "loss": 4.2456,
+      "step": 494500
+    },
+    {
+      "epoch": 4.257489669421488,
+      "grad_norm": 3.90625,
+      "learning_rate": 1.072315698495252e-05,
+      "loss": 4.0858,
+      "step": 494550
+    },
+    {
+      "epoch": 4.2579201101928374,
+      "grad_norm": 4.46875,
+      "learning_rate": 1.0710954869250622e-05,
+      "loss": 4.5307,
+      "step": 494600
+    },
+    {
+      "epoch": 4.258350550964187,
+      "grad_norm": 4.28125,
+      "learning_rate": 1.0698759307172534e-05,
+      "loss": 4.0033,
+      "step": 494650
+    },
+    {
+      "epoch": 4.258780991735537,
+      "grad_norm": 3.359375,
+      "learning_rate": 1.0686570299613341e-05,
+      "loss": 3.7,
+      "step": 494700
+    },
+    {
+      "epoch": 4.259211432506887,
+      "grad_norm": 4.75,
+      "learning_rate": 1.0674387847467748e-05,
+      "loss": 4.5149,
+      "step": 494750
+    },
+    {
+      "epoch": 4.259641873278237,
+      "grad_norm": 2.703125,
+      "learning_rate": 1.0662211951629896e-05,
+      "loss": 4.3212,
+      "step": 494800
+    },
+    {
+      "epoch": 4.260072314049586,
+      "grad_norm": 3.296875,
+      "learning_rate": 1.0650042612993449e-05,
+      "loss": 4.1083,
+      "step": 494850
+    },
+    {
+      "epoch": 4.260502754820937,
+      "grad_norm": 4.65625,
+      "learning_rate": 1.063787983245167e-05,
+      "loss": 4.3666,
+      "step": 494900
+    },
+    {
+      "epoch": 4.260933195592287,
+      "grad_norm": 2.546875,
+      "learning_rate": 1.0625723610897231e-05,
+      "loss": 4.482,
+      "step": 494950
+    },
+    {
+      "epoch": 4.261363636363637,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.0613573949222367e-05,
+      "loss": 4.7658,
+      "step": 495000
+    },
+    {
+      "epoch": 4.261363636363637,
+      "eval_loss": 4.990317344665527,
+      "eval_runtime": 22.5933,
+      "eval_samples_per_second": 28.327,
+      "eval_steps_per_second": 14.163,
+      "eval_tts_loss": 7.480617183203461,
+      "step": 495000
+    },
+    {
+      "epoch": 4.2617940771349865,
+      "grad_norm": 3.9375,
+      "learning_rate": 1.060143084831886e-05,
+      "loss": 4.3694,
+      "step": 495050
+    },
+    {
+      "epoch": 4.262224517906336,
+      "grad_norm": 3.765625,
+      "learning_rate": 1.0589294309078002e-05,
+      "loss": 3.9857,
+      "step": 495100
+    },
+    {
+      "epoch": 4.262654958677686,
+      "grad_norm": 1.7265625,
+      "learning_rate": 1.0577164332390532e-05,
+      "loss": 4.3198,
+      "step": 495150
+    },
+    {
+      "epoch": 4.263085399449036,
+      "grad_norm": 3.9375,
+      "learning_rate": 1.0565040919146851e-05,
+      "loss": 4.4039,
+      "step": 495200
+    },
+    {
+      "epoch": 4.263515840220386,
+      "grad_norm": 5.1875,
+      "learning_rate": 1.0552924070236703e-05,
+      "loss": 4.6953,
+      "step": 495250
+    },
+    {
+      "epoch": 4.2639462809917354,
+      "grad_norm": 5.40625,
+      "learning_rate": 1.0540813786549497e-05,
+      "loss": 4.1689,
+      "step": 495300
+    },
+    {
+      "epoch": 4.264376721763085,
+      "grad_norm": 5.34375,
+      "learning_rate": 1.052871006897409e-05,
+      "loss": 4.4301,
+      "step": 495350
+    },
+    {
+      "epoch": 4.264807162534435,
+      "grad_norm": 3.296875,
+      "learning_rate": 1.0516612918398827e-05,
+      "loss": 4.3754,
+      "step": 495400
+    },
+    {
+      "epoch": 4.265237603305785,
+      "grad_norm": 3.5,
+      "learning_rate": 1.0504522335711698e-05,
+      "loss": 4.3632,
+      "step": 495450
+    },
+    {
+      "epoch": 4.265668044077135,
+      "grad_norm": 4.71875,
+      "learning_rate": 1.0492438321800046e-05,
+      "loss": 4.1471,
+      "step": 495500
+    },
+    {
+      "epoch": 4.266098484848484,
+      "grad_norm": 3.796875,
+      "learning_rate": 1.0480360877550832e-05,
+      "loss": 4.4319,
+      "step": 495550
+    },
+    {
+      "epoch": 4.266528925619835,
+      "grad_norm": 2.1875,
+      "learning_rate": 1.0468290003850534e-05,
+      "loss": 4.1945,
+      "step": 495600
+    },
+    {
+      "epoch": 4.266959366391185,
+      "grad_norm": 5.9375,
+      "learning_rate": 1.045622570158512e-05,
+      "loss": 4.5434,
+      "step": 495650
+    },
+    {
+      "epoch": 4.267389807162535,
+      "grad_norm": 5.59375,
+      "learning_rate": 1.0444167971640085e-05,
+      "loss": 4.1925,
+      "step": 495700
+    },
+    {
+      "epoch": 4.2678202479338845,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.0432116814900439e-05,
+      "loss": 3.9483,
+      "step": 495750
+    },
+    {
+      "epoch": 4.268250688705234,
+      "grad_norm": 4.375,
+      "learning_rate": 1.04200722322507e-05,
+      "loss": 4.3413,
+      "step": 495800
+    },
+    {
+      "epoch": 4.268681129476584,
+      "grad_norm": 3.578125,
+      "learning_rate": 1.0408034224574903e-05,
+      "loss": 4.1517,
+      "step": 495850
+    },
+    {
+      "epoch": 4.269111570247934,
+      "grad_norm": 1.8515625,
+      "learning_rate": 1.0396002792756654e-05,
+      "loss": 4.3048,
+      "step": 495900
+    },
+    {
+      "epoch": 4.269542011019284,
+      "grad_norm": 1.921875,
+      "learning_rate": 1.0383977937679013e-05,
+      "loss": 3.8577,
+      "step": 495950
+    },
+    {
+      "epoch": 4.2699724517906334,
+      "grad_norm": 3.703125,
+      "learning_rate": 1.0371959660224574e-05,
+      "loss": 4.1211,
+      "step": 496000
+    },
+    {
+      "epoch": 4.270402892561983,
+      "grad_norm": 2.140625,
+      "learning_rate": 1.0359947961275463e-05,
+      "loss": 4.0847,
+      "step": 496050
+    },
+    {
+      "epoch": 4.270833333333333,
+      "grad_norm": 3.0,
+      "learning_rate": 1.0347942841713287e-05,
+      "loss": 4.2619,
+      "step": 496100
+    },
+    {
+      "epoch": 4.271263774104683,
+      "grad_norm": 5.71875,
+      "learning_rate": 1.0335944302419232e-05,
+      "loss": 4.6816,
+      "step": 496150
+    },
+    {
+      "epoch": 4.2716942148760335,
+      "grad_norm": 2.15625,
+      "learning_rate": 1.0323952344273956e-05,
+      "loss": 4.1899,
+      "step": 496200
+    },
+    {
+      "epoch": 4.272124655647383,
+      "grad_norm": 3.0625,
+      "learning_rate": 1.0311966968157604e-05,
+      "loss": 4.5647,
+      "step": 496250
+    },
+    {
+      "epoch": 4.272555096418733,
+      "grad_norm": 4.65625,
+      "learning_rate": 1.0299988174949937e-05,
+      "loss": 4.0187,
+      "step": 496300
+    },
+    {
+      "epoch": 4.272985537190083,
+      "grad_norm": 5.375,
+      "learning_rate": 1.0288015965530129e-05,
+      "loss": 4.3155,
+      "step": 496350
+    },
+    {
+      "epoch": 4.273415977961433,
+      "grad_norm": 3.5,
+      "learning_rate": 1.0276050340776921e-05,
+      "loss": 4.5485,
+      "step": 496400
+    },
+    {
+      "epoch": 4.2738464187327825,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.02640913015686e-05,
+      "loss": 4.3729,
+      "step": 496450
+    },
+    {
+      "epoch": 4.274276859504132,
+      "grad_norm": 2.5625,
+      "learning_rate": 1.0252138848782899e-05,
+      "loss": 4.3204,
+      "step": 496500
+    },
+    {
+      "epoch": 4.274707300275482,
+      "grad_norm": 1.890625,
+      "learning_rate": 1.0240192983297081e-05,
+      "loss": 4.1798,
+      "step": 496550
+    },
+    {
+      "epoch": 4.275137741046832,
+      "grad_norm": 3.828125,
+      "learning_rate": 1.0228253705988e-05,
+      "loss": 4.3045,
+      "step": 496600
+    },
+    {
+      "epoch": 4.275568181818182,
+      "grad_norm": 4.21875,
+      "learning_rate": 1.0216321017731945e-05,
+      "loss": 4.5311,
+      "step": 496650
+    },
+    {
+      "epoch": 4.2759986225895315,
+      "grad_norm": 4.21875,
+      "learning_rate": 1.0204394919404747e-05,
+      "loss": 4.5882,
+      "step": 496700
+    },
+    {
+      "epoch": 4.276429063360881,
+      "grad_norm": 1.5625,
+      "learning_rate": 1.0192475411881764e-05,
+      "loss": 4.1879,
+      "step": 496750
+    },
+    {
+      "epoch": 4.276859504132231,
+      "grad_norm": 4.59375,
+      "learning_rate": 1.0180562496037849e-05,
+      "loss": 4.5052,
+      "step": 496800
+    },
+    {
+      "epoch": 4.277289944903581,
+      "grad_norm": 4.5625,
+      "learning_rate": 1.0168656172747404e-05,
+      "loss": 4.51,
+      "step": 496850
+    },
+    {
+      "epoch": 4.2777203856749315,
+      "grad_norm": 4.40625,
+      "learning_rate": 1.015675644288432e-05,
+      "loss": 4.0977,
+      "step": 496900
+    },
+    {
+      "epoch": 4.278150826446281,
+      "grad_norm": 5.4375,
+      "learning_rate": 1.0144863307322017e-05,
+      "loss": 4.4966,
+      "step": 496950
+    },
+    {
+      "epoch": 4.278581267217631,
+      "grad_norm": 4.4375,
+      "learning_rate": 1.013297676693341e-05,
+      "loss": 4.6574,
+      "step": 497000
+    },
+    {
+      "epoch": 4.279011707988981,
+      "grad_norm": 1.71875,
+      "learning_rate": 1.0121096822590969e-05,
+      "loss": 4.5093,
+      "step": 497050
+    },
+    {
+      "epoch": 4.279442148760331,
+      "grad_norm": 4.9375,
+      "learning_rate": 1.0109223475166607e-05,
+      "loss": 4.3858,
+      "step": 497100
+    },
+    {
+      "epoch": 4.2798725895316805,
+      "grad_norm": 6.09375,
+      "learning_rate": 1.009735672553187e-05,
+      "loss": 4.2499,
+      "step": 497150
+    },
+    {
+      "epoch": 4.28030303030303,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.0085496574557708e-05,
+      "loss": 4.4491,
+      "step": 497200
+    },
+    {
+      "epoch": 4.28073347107438,
+      "grad_norm": 3.25,
+      "learning_rate": 1.0073643023114642e-05,
+      "loss": 4.2579,
+      "step": 497250
+    },
+    {
+      "epoch": 4.28116391184573,
+      "grad_norm": 2.796875,
+      "learning_rate": 1.0061796072072705e-05,
+      "loss": 4.5695,
+      "step": 497300
+    },
+    {
+      "epoch": 4.28159435261708,
+      "grad_norm": 1.9609375,
+      "learning_rate": 1.0049955722301419e-05,
+      "loss": 3.7701,
+      "step": 497350
+    },
+    {
+      "epoch": 4.2820247933884295,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.0038121974669834e-05,
+      "loss": 4.6326,
+      "step": 497400
+    },
+    {
+      "epoch": 4.282455234159779,
+      "grad_norm": 4.09375,
+      "learning_rate": 1.0026294830046556e-05,
+      "loss": 4.4966,
+      "step": 497450
+    },
+    {
+      "epoch": 4.28288567493113,
+      "grad_norm": 3.28125,
+      "learning_rate": 1.0014474289299657e-05,
+      "loss": 4.0949,
+      "step": 497500
+    },
+    {
+      "epoch": 4.28331611570248,
+      "grad_norm": 4.0625,
+      "learning_rate": 1.0002660353296745e-05,
+      "loss": 4.2209,
+      "step": 497550
+    },
+    {
+      "epoch": 4.2837465564738295,
+      "grad_norm": 2.75,
+      "learning_rate": 9.990853022904922e-06,
+      "loss": 4.1878,
+      "step": 497600
+    },
+    {
+      "epoch": 4.284176997245179,
+      "grad_norm": 2.640625,
+      "learning_rate": 9.979052298990798e-06,
+      "loss": 4.1017,
+      "step": 497650
+    },
+    {
+      "epoch": 4.284607438016529,
+      "grad_norm": 1.8671875,
+      "learning_rate": 9.96725818242058e-06,
+      "loss": 4.2861,
+      "step": 497700
+    },
+    {
+      "epoch": 4.285037878787879,
+      "grad_norm": 4.8125,
+      "learning_rate": 9.95547067405992e-06,
+      "loss": 4.4236,
+      "step": 497750
+    },
+    {
+      "epoch": 4.285468319559229,
+      "grad_norm": 3.375,
+      "learning_rate": 9.943689774773935e-06,
+      "loss": 4.0353,
+      "step": 497800
+    },
+    {
+      "epoch": 4.2858987603305785,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.931915485427379e-06,
+      "loss": 4.2317,
+      "step": 497850
+    },
+    {
+      "epoch": 4.286329201101928,
+      "grad_norm": 2.578125,
+      "learning_rate": 9.92014780688444e-06,
+      "loss": 4.4292,
+      "step": 497900
+    },
+    {
+      "epoch": 4.286759641873278,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.908386740008823e-06,
+      "loss": 4.6995,
+      "step": 497950
+    },
+    {
+      "epoch": 4.287190082644628,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.89663228566382e-06,
+      "loss": 4.3831,
+      "step": 498000
+    },
+    {
+      "epoch": 4.287190082644628,
+      "eval_loss": 4.990382194519043,
+      "eval_runtime": 23.7313,
+      "eval_samples_per_second": 26.969,
+      "eval_steps_per_second": 13.484,
+      "eval_tts_loss": 7.480655207959233,
+      "step": 498000
+    },
+    {
+      "epoch": 4.287620523415978,
+      "grad_norm": 1.9375,
+      "learning_rate": 9.884884444712128e-06,
+      "loss": 4.4527,
+      "step": 498050
+    },
+    {
+      "epoch": 4.2880509641873275,
+      "grad_norm": 2.625,
+      "learning_rate": 9.873143218016013e-06,
+      "loss": 4.3169,
+      "step": 498100
+    },
+    {
+      "epoch": 4.288481404958677,
+      "grad_norm": 6.15625,
+      "learning_rate": 9.861408606437283e-06,
+      "loss": 4.1037,
+      "step": 498150
+    },
+    {
+      "epoch": 4.288911845730028,
+      "grad_norm": 1.734375,
+      "learning_rate": 9.84968061083723e-06,
+      "loss": 4.3089,
+      "step": 498200
+    },
+    {
+      "epoch": 4.289342286501378,
+      "grad_norm": 1.2890625,
+      "learning_rate": 9.837959232076654e-06,
+      "loss": 4.4333,
+      "step": 498250
+    },
+    {
+      "epoch": 4.2897727272727275,
+      "grad_norm": 1.5234375,
+      "learning_rate": 9.826244471015877e-06,
+      "loss": 4.0469,
+      "step": 498300
+    },
+    {
+      "epoch": 4.290203168044077,
+      "grad_norm": 4.28125,
+      "learning_rate": 9.814536328514723e-06,
+      "loss": 4.1831,
+      "step": 498350
+    },
+    {
+      "epoch": 4.290633608815427,
+      "grad_norm": 5.3125,
+      "learning_rate": 9.802834805432593e-06,
+      "loss": 4.4076,
+      "step": 498400
+    },
+    {
+      "epoch": 4.291064049586777,
+      "grad_norm": 1.765625,
+      "learning_rate": 9.79113990262832e-06,
+      "loss": 4.666,
+      "step": 498450
+    },
+    {
+      "epoch": 4.291494490358127,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.77945162096029e-06,
+      "loss": 4.378,
+      "step": 498500
+    },
+    {
+      "epoch": 4.2919249311294765,
+      "grad_norm": 6.40625,
+      "learning_rate": 9.767769961286387e-06,
+      "loss": 4.38,
+      "step": 498550
+    },
+    {
+      "epoch": 4.292355371900826,
+      "grad_norm": 2.203125,
+      "learning_rate": 9.756094924464032e-06,
+      "loss": 4.2723,
+      "step": 498600
+    },
+    {
+      "epoch": 4.292785812672176,
+      "grad_norm": 4.25,
+      "learning_rate": 9.744426511350135e-06,
+      "loss": 4.2122,
+      "step": 498650
+    },
+    {
+      "epoch": 4.293216253443526,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.732764722801157e-06,
+      "loss": 4.5124,
+      "step": 498700
+    },
+    {
+      "epoch": 4.293646694214876,
+      "grad_norm": 5.4375,
+      "learning_rate": 9.721109559673026e-06,
+      "loss": 4.2672,
+      "step": 498750
+    },
+    {
+      "epoch": 4.294077134986226,
+      "grad_norm": 2.078125,
+      "learning_rate": 9.70946102282122e-06,
+      "loss": 4.1701,
+      "step": 498800
+    },
+    {
+      "epoch": 4.294507575757576,
+      "grad_norm": 4.875,
+      "learning_rate": 9.697819113100714e-06,
+      "loss": 3.9554,
+      "step": 498850
+    },
+    {
+      "epoch": 4.294938016528926,
+      "grad_norm": 5.1875,
+      "learning_rate": 9.686183831365991e-06,
+      "loss": 4.3497,
+      "step": 498900
+    },
+    {
+      "epoch": 4.295368457300276,
+      "grad_norm": 1.65625,
+      "learning_rate": 9.674555178471034e-06,
+      "loss": 4.0724,
+      "step": 498950
+    },
+    {
+      "epoch": 4.2957988980716255,
+      "grad_norm": 2.09375,
+      "learning_rate": 9.662933155269415e-06,
+      "loss": 4.3112,
+      "step": 499000
+    },
+    {
+      "epoch": 4.296229338842975,
+      "grad_norm": 3.8125,
+      "learning_rate": 9.651317762614132e-06,
+      "loss": 4.4373,
+      "step": 499050
+    },
+    {
+      "epoch": 4.296659779614325,
+      "grad_norm": 6.34375,
+      "learning_rate": 9.639709001357733e-06,
+      "loss": 4.3362,
+      "step": 499100
+    },
+    {
+      "epoch": 4.297090220385675,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.628106872352283e-06,
+      "loss": 3.9074,
+      "step": 499150
+    },
+    {
+      "epoch": 4.297520661157025,
+      "grad_norm": 4.46875,
+      "learning_rate": 9.616511376449355e-06,
+      "loss": 4.5121,
+      "step": 499200
+    },
+    {
+      "epoch": 4.2979511019283745,
+      "grad_norm": 2.75,
+      "learning_rate": 9.604922514499993e-06,
+      "loss": 4.1561,
+      "step": 499250
+    },
+    {
+      "epoch": 4.298381542699724,
+      "grad_norm": 3.359375,
+      "learning_rate": 9.59334028735488e-06,
+      "loss": 4.252,
+      "step": 499300
+    },
+    {
+      "epoch": 4.298811983471074,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.581764695864026e-06,
+      "loss": 4.5887,
+      "step": 499350
+    },
+    {
+      "epoch": 4.299242424242424,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.57019574087713e-06,
+      "loss": 4.2836,
+      "step": 499400
+    },
+    {
+      "epoch": 4.299672865013774,
+      "grad_norm": 2.28125,
+      "learning_rate": 9.55863342324329e-06,
+      "loss": 4.3688,
+      "step": 499450
+    },
+    {
+      "epoch": 4.300103305785124,
+      "grad_norm": 2.921875,
+      "learning_rate": 9.547077743811162e-06,
+      "loss": 4.263,
+      "step": 499500
+    },
+    {
+      "epoch": 4.300533746556474,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.535528703428942e-06,
+      "loss": 3.8768,
+      "step": 499550
+    },
+    {
+      "epoch": 4.300964187327824,
+      "grad_norm": 1.3359375,
+      "learning_rate": 9.523986302944265e-06,
+      "loss": 4.1092,
+      "step": 499600
+    },
+    {
+      "epoch": 4.301394628099174,
+      "grad_norm": 2.078125,
+      "learning_rate": 9.51245054320431e-06,
+      "loss": 4.65,
+      "step": 499650
+    },
+    {
+      "epoch": 4.3018250688705235,
+      "grad_norm": 5.78125,
+      "learning_rate": 9.500921425055809e-06,
+      "loss": 4.3907,
+      "step": 499700
+    },
+    {
+      "epoch": 4.302255509641873,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.489398949344963e-06,
+      "loss": 4.0317,
+      "step": 499750
+    },
+    {
+      "epoch": 4.302685950413223,
+      "grad_norm": 6.0625,
+      "learning_rate": 9.477883116917496e-06,
+      "loss": 4.2088,
+      "step": 499800
+    },
+    {
+      "epoch": 4.303116391184573,
+      "grad_norm": 5.4375,
+      "learning_rate": 9.466373928618643e-06,
+      "loss": 4.0614,
+      "step": 499850
+    },
+    {
+      "epoch": 4.303546831955923,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.45487138529314e-06,
+      "loss": 4.3172,
+      "step": 499900
+    },
+    {
+      "epoch": 4.3039772727272725,
+      "grad_norm": 2.96875,
+      "learning_rate": 9.443375487785278e-06,
+      "loss": 4.2179,
+      "step": 499950
+    },
+    {
+      "epoch": 4.304407713498622,
+      "grad_norm": 4.59375,
+      "learning_rate": 9.431886236938826e-06,
+      "loss": 4.1902,
+      "step": 500000
+    },
+    {
+      "epoch": 4.304838154269972,
+      "grad_norm": 3.90625,
+      "learning_rate": 9.420403633597052e-06,
+      "loss": 4.277,
+      "step": 500050
+    },
+    {
+      "epoch": 4.305268595041323,
+      "grad_norm": 5.125,
+      "learning_rate": 9.408927678602775e-06,
+      "loss": 4.1179,
+      "step": 500100
+    },
+    {
+      "epoch": 4.3056990358126725,
+      "grad_norm": 2.625,
+      "learning_rate": 9.397458372798285e-06,
+      "loss": 4.484,
+      "step": 500150
+    },
+    {
+      "epoch": 4.306129476584022,
+      "grad_norm": 2.328125,
+      "learning_rate": 9.385995717025408e-06,
+      "loss": 4.38,
+      "step": 500200
+    },
+    {
+      "epoch": 4.306559917355372,
+      "grad_norm": 2.390625,
+      "learning_rate": 9.374539712125484e-06,
+      "loss": 3.9738,
+      "step": 500250
+    },
+    {
+      "epoch": 4.306990358126722,
+      "grad_norm": 3.375,
+      "learning_rate": 9.363090358939398e-06,
+      "loss": 4.6081,
+      "step": 500300
+    },
+    {
+      "epoch": 4.307420798898072,
+      "grad_norm": 6.75,
+      "learning_rate": 9.351647658307416e-06,
+      "loss": 4.644,
+      "step": 500350
+    },
+    {
+      "epoch": 4.3078512396694215,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.340211611069493e-06,
+      "loss": 3.9021,
+      "step": 500400
+    },
+    {
+      "epoch": 4.308281680440771,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.328782218064968e-06,
+      "loss": 4.2323,
+      "step": 500450
+    },
+    {
+      "epoch": 4.308712121212121,
+      "grad_norm": 2.125,
+      "learning_rate": 9.317359480132726e-06,
+      "loss": 4.2335,
+      "step": 500500
+    },
+    {
+      "epoch": 4.309142561983471,
+      "grad_norm": 2.234375,
+      "learning_rate": 9.30594339811124e-06,
+      "loss": 4.1903,
+      "step": 500550
+    },
+    {
+      "epoch": 4.309573002754821,
+      "grad_norm": 2.703125,
+      "learning_rate": 9.294533972838327e-06,
+      "loss": 4.4114,
+      "step": 500600
+    },
+    {
+      "epoch": 4.3100034435261705,
+      "grad_norm": 3.3125,
+      "learning_rate": 9.283131205151474e-06,
+      "loss": 4.5617,
+      "step": 500650
+    },
+    {
+      "epoch": 4.31043388429752,
+      "grad_norm": 2.75,
+      "learning_rate": 9.271735095887623e-06,
+      "loss": 4.2109,
+      "step": 500700
+    },
+    {
+      "epoch": 4.31086432506887,
+      "grad_norm": 2.078125,
+      "learning_rate": 9.260345645883195e-06,
+      "loss": 4.5076,
+      "step": 500750
+    },
+    {
+      "epoch": 4.311294765840221,
+      "grad_norm": 5.21875,
+      "learning_rate": 9.248962855974153e-06,
+      "loss": 4.0628,
+      "step": 500800
+    },
+    {
+      "epoch": 4.3117252066115705,
+      "grad_norm": 6.03125,
+      "learning_rate": 9.237586726996028e-06,
+      "loss": 3.9696,
+      "step": 500850
+    },
+    {
+      "epoch": 4.31215564738292,
+      "grad_norm": 4.75,
+      "learning_rate": 9.226217259783709e-06,
+      "loss": 4.4315,
+      "step": 500900
+    },
+    {
+      "epoch": 4.31258608815427,
+      "grad_norm": 2.765625,
+      "learning_rate": 9.214854455171751e-06,
+      "loss": 4.5924,
+      "step": 500950
+    },
+    {
+      "epoch": 4.31301652892562,
+      "grad_norm": 2.625,
+      "learning_rate": 9.203498313994164e-06,
+      "loss": 4.3333,
+      "step": 501000
+    },
+    {
+      "epoch": 4.31301652892562,
+      "eval_loss": 4.990403175354004,
+      "eval_runtime": 24.1551,
+      "eval_samples_per_second": 26.495,
+      "eval_steps_per_second": 13.248,
+      "eval_tts_loss": 7.480433057473844,
+      "step": 501000
+    },
+    {
+      "epoch": 4.31344696969697,
+      "grad_norm": 2.578125,
+      "learning_rate": 9.192148837084414e-06,
+      "loss": 4.2487,
+      "step": 501050
+    },
+    {
+      "epoch": 4.3138774104683195,
+      "grad_norm": 2.578125,
+      "learning_rate": 9.1808060252756e-06,
+      "loss": 4.3544,
+      "step": 501100
+    },
+    {
+      "epoch": 4.314307851239669,
+      "grad_norm": 3.0625,
+      "learning_rate": 9.169469879400205e-06,
+      "loss": 4.3581,
+      "step": 501150
+    },
+    {
+      "epoch": 4.314738292011019,
+      "grad_norm": 2.71875,
+      "learning_rate": 9.158140400290282e-06,
+      "loss": 4.299,
+      "step": 501200
+    },
+    {
+      "epoch": 4.315168732782369,
+      "grad_norm": 1.984375,
+      "learning_rate": 9.146817588777413e-06,
+      "loss": 4.455,
+      "step": 501250
+    },
+    {
+      "epoch": 4.315599173553719,
+      "grad_norm": 2.90625,
+      "learning_rate": 9.135501445692662e-06,
+      "loss": 4.3407,
+      "step": 501300
+    },
+    {
+      "epoch": 4.3160296143250685,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.124191971866603e-06,
+      "loss": 4.2062,
+      "step": 501350
+    },
+    {
+      "epoch": 4.316460055096419,
+      "grad_norm": 2.4375,
+      "learning_rate": 9.112889168129323e-06,
+      "loss": 4.9257,
+      "step": 501400
+    },
+    {
+      "epoch": 4.316890495867769,
+      "grad_norm": 2.578125,
+      "learning_rate": 9.10159303531044e-06,
+      "loss": 4.5144,
+      "step": 501450
+    },
+    {
+      "epoch": 4.317320936639119,
+      "grad_norm": 3.59375,
+      "learning_rate": 9.090303574239034e-06,
+      "loss": 4.3472,
+      "step": 501500
+    },
+    {
+      "epoch": 4.3177513774104685,
+      "grad_norm": 4.5,
+      "learning_rate": 9.079020785743765e-06,
+      "loss": 4.2618,
+      "step": 501550
+    },
+    {
+      "epoch": 4.318181818181818,
+      "grad_norm": 3.6875,
+      "learning_rate": 9.067744670652744e-06,
+      "loss": 4.5121,
+      "step": 501600
+    },
+    {
+      "epoch": 4.318612258953168,
+      "grad_norm": 1.5,
+      "learning_rate": 9.056475229793627e-06,
+      "loss": 4.1174,
+      "step": 501650
+    },
+    {
+      "epoch": 4.319042699724518,
+      "grad_norm": 3.25,
+      "learning_rate": 9.045212463993546e-06,
+      "loss": 4.4511,
+      "step": 501700
+    },
+    {
+      "epoch": 4.319473140495868,
+      "grad_norm": 4.0625,
+      "learning_rate": 9.033956374079167e-06,
+      "loss": 4.0417,
+      "step": 501750
+    },
+    {
+      "epoch": 4.3199035812672175,
+      "grad_norm": 2.25,
+      "learning_rate": 9.022706960876692e-06,
+      "loss": 4.1759,
+      "step": 501800
+    },
+    {
+      "epoch": 4.320334022038567,
+      "grad_norm": 3.875,
+      "learning_rate": 9.011464225211797e-06,
+      "loss": 4.2091,
+      "step": 501850
+    },
+    {
+      "epoch": 4.320764462809917,
+      "grad_norm": 5.40625,
+      "learning_rate": 9.00022816790963e-06,
+      "loss": 4.0665,
+      "step": 501900
+    },
+    {
+      "epoch": 4.321194903581267,
+      "grad_norm": 3.828125,
+      "learning_rate": 8.988998789794934e-06,
+      "loss": 4.1778,
+      "step": 501950
+    },
+    {
+      "epoch": 4.321625344352617,
+      "grad_norm": 4.03125,
+      "learning_rate": 8.977776091691915e-06,
+      "loss": 4.1469,
+      "step": 502000
+    },
+    {
+      "epoch": 4.3220557851239665,
+      "grad_norm": 3.0,
+      "learning_rate": 8.96656007442428e-06,
+      "loss": 4.1407,
+      "step": 502050
+    },
+    {
+      "epoch": 4.322486225895317,
+      "grad_norm": 3.234375,
+      "learning_rate": 8.955350738815293e-06,
+      "loss": 4.3537,
+      "step": 502100
+    },
+    {
+      "epoch": 4.322916666666667,
+      "grad_norm": 2.46875,
+      "learning_rate": 8.944148085687665e-06,
+      "loss": 4.0011,
+      "step": 502150
+    },
+    {
+      "epoch": 4.323347107438017,
+      "grad_norm": 3.46875,
+      "learning_rate": 8.932952115863646e-06,
+      "loss": 4.2977,
+      "step": 502200
+    },
+    {
+      "epoch": 4.3237775482093666,
+      "grad_norm": 2.984375,
+      "learning_rate": 8.921762830165015e-06,
+      "loss": 4.244,
+      "step": 502250
+    },
+    {
+      "epoch": 4.324207988980716,
+      "grad_norm": 1.4453125,
+      "learning_rate": 8.910580229413035e-06,
+      "loss": 4.4479,
+      "step": 502300
+    },
+    {
+      "epoch": 4.324638429752066,
+      "grad_norm": 1.71875,
+      "learning_rate": 8.899404314428472e-06,
+      "loss": 4.1162,
+      "step": 502350
+    },
+    {
+      "epoch": 4.325068870523416,
+      "grad_norm": 2.359375,
+      "learning_rate": 8.888235086031637e-06,
+      "loss": 4.3571,
+      "step": 502400
+    },
+    {
+      "epoch": 4.325499311294766,
+      "grad_norm": 3.53125,
+      "learning_rate": 8.877072545042298e-06,
+      "loss": 4.4373,
+      "step": 502450
+    },
+    {
+      "epoch": 4.3259297520661155,
+      "grad_norm": 3.1875,
+      "learning_rate": 8.865916692279785e-06,
+      "loss": 4.0097,
+      "step": 502500
+    },
+    {
+      "epoch": 4.326360192837465,
+      "grad_norm": 4.625,
+      "learning_rate": 8.854767528562913e-06,
+      "loss": 4.1214,
+      "step": 502550
+    },
+    {
+      "epoch": 4.326790633608815,
+      "grad_norm": 2.796875,
+      "learning_rate": 8.843625054709981e-06,
+      "loss": 4.4159,
+      "step": 502600
+    },
+    {
+      "epoch": 4.327221074380165,
+      "grad_norm": 2.734375,
+      "learning_rate": 8.83248927153888e-06,
+      "loss": 4.5618,
+      "step": 502650
+    },
+    {
+      "epoch": 4.327651515151516,
+      "grad_norm": 4.21875,
+      "learning_rate": 8.8213601798669e-06,
+      "loss": 4.5627,
+      "step": 502700
+    },
+    {
+      "epoch": 4.328081955922865,
+      "grad_norm": 2.6875,
+      "learning_rate": 8.810237780510878e-06,
+      "loss": 4.3087,
+      "step": 502750
+    },
+    {
+      "epoch": 4.328512396694215,
+      "grad_norm": 3.8125,
+      "learning_rate": 8.799122074287224e-06,
+      "loss": 4.0904,
+      "step": 502800
+    },
+    {
+      "epoch": 4.328942837465565,
+      "grad_norm": 3.828125,
+      "learning_rate": 8.78801306201179e-06,
+      "loss": 4.023,
+      "step": 502850
+    },
+    {
+      "epoch": 4.329373278236915,
+      "grad_norm": 2.875,
+      "learning_rate": 8.776910744499956e-06,
+      "loss": 4.3179,
+      "step": 502900
+    },
+    {
+      "epoch": 4.3298037190082646,
+      "grad_norm": 3.1875,
+      "learning_rate": 8.76581512256659e-06,
+      "loss": 4.3492,
+      "step": 502950
+    },
+    {
+      "epoch": 4.330234159779614,
+      "grad_norm": 1.765625,
+      "learning_rate": 8.754726197026108e-06,
+      "loss": 4.1772,
+      "step": 503000
+    },
+    {
+      "epoch": 4.330664600550964,
+      "grad_norm": 3.0,
+      "learning_rate": 8.743643968692384e-06,
+      "loss": 4.321,
+      "step": 503050
+    },
+    {
+      "epoch": 4.331095041322314,
+      "grad_norm": 3.140625,
+      "learning_rate": 8.732568438378864e-06,
+      "loss": 4.4251,
+      "step": 503100
+    },
+    {
+      "epoch": 4.331525482093664,
+      "grad_norm": 2.53125,
+      "learning_rate": 8.721499606898453e-06,
+      "loss": 4.6072,
+      "step": 503150
+    },
+    {
+      "epoch": 4.3319559228650135,
+      "grad_norm": 4.21875,
+      "learning_rate": 8.710437475063582e-06,
+      "loss": 4.4484,
+      "step": 503200
+    },
+    {
+      "epoch": 4.332386363636363,
+      "grad_norm": 1.7265625,
+      "learning_rate": 8.699382043686177e-06,
+      "loss": 4.3609,
+      "step": 503250
+    },
+    {
+      "epoch": 4.332816804407713,
+      "grad_norm": 1.2890625,
+      "learning_rate": 8.68833331357769e-06,
+      "loss": 4.4118,
+      "step": 503300
+    },
+    {
+      "epoch": 4.333247245179063,
+      "grad_norm": 3.796875,
+      "learning_rate": 8.677291285549084e-06,
+      "loss": 4.3946,
+      "step": 503350
+    },
+    {
+      "epoch": 4.333677685950414,
+      "grad_norm": 2.75,
+      "learning_rate": 8.66625596041083e-06,
+      "loss": 4.3638,
+      "step": 503400
+    },
+    {
+      "epoch": 4.334108126721763,
+      "grad_norm": 4.3125,
+      "learning_rate": 8.655227338972848e-06,
+      "loss": 4.5647,
+      "step": 503450
+    },
+    {
+      "epoch": 4.334538567493113,
+      "grad_norm": 3.8125,
+      "learning_rate": 8.64420542204466e-06,
+      "loss": 3.8277,
+      "step": 503500
+    },
+    {
+      "epoch": 4.334969008264463,
+      "grad_norm": 1.3046875,
+      "learning_rate": 8.633190210435237e-06,
+      "loss": 4.1098,
+      "step": 503550
+    },
+    {
+      "epoch": 4.335399449035813,
+      "grad_norm": 3.640625,
+      "learning_rate": 8.622181704953047e-06,
+      "loss": 4.2239,
+      "step": 503600
+    },
+    {
+      "epoch": 4.3358298898071626,
+      "grad_norm": 2.625,
+      "learning_rate": 8.611179906406165e-06,
+      "loss": 4.2217,
+      "step": 503650
+    },
+    {
+      "epoch": 4.336260330578512,
+      "grad_norm": 2.609375,
+      "learning_rate": 8.60018481560202e-06,
+      "loss": 4.1941,
+      "step": 503700
+    },
+    {
+      "epoch": 4.336690771349862,
+      "grad_norm": 2.703125,
+      "learning_rate": 8.589196433347646e-06,
+      "loss": 4.3679,
+      "step": 503750
+    },
+    {
+      "epoch": 4.337121212121212,
+      "grad_norm": 1.484375,
+      "learning_rate": 8.578214760449599e-06,
+      "loss": 3.8968,
+      "step": 503800
+    },
+    {
+      "epoch": 4.337551652892562,
+      "grad_norm": 2.015625,
+      "learning_rate": 8.567239797713888e-06,
+      "loss": 4.3936,
+      "step": 503850
+    },
+    {
+      "epoch": 4.3379820936639115,
+      "grad_norm": 2.6875,
+      "learning_rate": 8.556271545946049e-06,
+      "loss": 4.4982,
+      "step": 503900
+    },
+    {
+      "epoch": 4.338412534435261,
+      "grad_norm": 3.59375,
+      "learning_rate": 8.545310005951135e-06,
+      "loss": 4.4685,
+      "step": 503950
+    },
+    {
+      "epoch": 4.338842975206612,
+      "grad_norm": 4.15625,
+      "learning_rate": 8.53435517853367e-06,
+      "loss": 4.1723,
+      "step": 504000
+    },
+    {
+      "epoch": 4.338842975206612,
+      "eval_loss": 4.990280628204346,
+      "eval_runtime": 24.3293,
+      "eval_samples_per_second": 26.306,
+      "eval_steps_per_second": 13.153,
+      "eval_tts_loss": 7.478574276202835,
+      "step": 504000
+    },
+    {
+      "epoch": 4.339273415977962,
+      "grad_norm": 3.5,
+      "learning_rate": 8.523407064497769e-06,
+      "loss": 4.7386,
+      "step": 504050
+    },
+    {
+      "epoch": 4.339703856749312,
+      "grad_norm": 4.1875,
+      "learning_rate": 8.512465664646973e-06,
+      "loss": 4.3512,
+      "step": 504100
+    },
+    {
+      "epoch": 4.340134297520661,
+      "grad_norm": 1.9453125,
+      "learning_rate": 8.501530979784356e-06,
+      "loss": 4.2346,
+      "step": 504150
+    },
+    {
+      "epoch": 4.340564738292011,
+      "grad_norm": 3.671875,
+      "learning_rate": 8.490603010712505e-06,
+      "loss": 4.8081,
+      "step": 504200
+    },
+    {
+      "epoch": 4.340995179063361,
+      "grad_norm": 4.25,
+      "learning_rate": 8.479681758233494e-06,
+      "loss": 4.301,
+      "step": 504250
+    },
+    {
+      "epoch": 4.341425619834711,
+      "grad_norm": 4.125,
+      "learning_rate": 8.468767223148921e-06,
+      "loss": 4.3199,
+      "step": 504300
+    },
+    {
+      "epoch": 4.3418560606060606,
+      "grad_norm": 3.75,
+      "learning_rate": 8.457859406259916e-06,
+      "loss": 3.8378,
+      "step": 504350
+    },
+    {
+      "epoch": 4.34228650137741,
+      "grad_norm": 1.3046875,
+      "learning_rate": 8.44695830836707e-06,
+      "loss": 3.7013,
+      "step": 504400
+    },
+    {
+      "epoch": 4.34271694214876,
+      "grad_norm": 1.9765625,
+      "learning_rate": 8.43606393027051e-06,
+      "loss": 4.3266,
+      "step": 504450
+    },
+    {
+      "epoch": 4.34314738292011,
+      "grad_norm": 4.125,
+      "learning_rate": 8.425176272769842e-06,
+      "loss": 4.3817,
+      "step": 504500
+    },
+    {
+      "epoch": 4.34357782369146,
+      "grad_norm": 3.265625,
+      "learning_rate": 8.414295336664214e-06,
+      "loss": 4.4488,
+      "step": 504550
+    },
+    {
+      "epoch": 4.3440082644628095,
+      "grad_norm": 6.34375,
+      "learning_rate": 8.403421122752242e-06,
+      "loss": 3.9959,
+      "step": 504600
+    },
+    {
+      "epoch": 4.344438705234159,
+      "grad_norm": 3.484375,
+      "learning_rate": 8.39255363183209e-06,
+      "loss": 4.2442,
+      "step": 504650
+    },
+    {
+      "epoch": 4.34486914600551,
+      "grad_norm": 2.484375,
+      "learning_rate": 8.381692864701419e-06,
+      "loss": 4.776,
+      "step": 504700
+    },
+    {
+      "epoch": 4.34529958677686,
+      "grad_norm": 3.375,
+      "learning_rate": 8.370838822157357e-06,
+      "loss": 4.2805,
+      "step": 504750
+    },
+    {
+      "epoch": 4.34573002754821,
+      "grad_norm": 5.1875,
+      "learning_rate": 8.359991504996578e-06,
+      "loss": 4.1398,
+      "step": 504800
+    },
+    {
+      "epoch": 4.346160468319559,
+      "grad_norm": 1.171875,
+      "learning_rate": 8.349150914015247e-06,
+      "loss": 4.4873,
+      "step": 504850
+    },
+    {
+      "epoch": 4.346590909090909,
+      "grad_norm": 3.21875,
+      "learning_rate": 8.338317050009058e-06,
+      "loss": 4.1217,
+      "step": 504900
+    },
+    {
+      "epoch": 4.347021349862259,
+      "grad_norm": 2.484375,
+      "learning_rate": 8.327489913773212e-06,
+      "loss": 4.4885,
+      "step": 504950
+    },
+    {
+      "epoch": 4.347451790633609,
+      "grad_norm": 2.828125,
+      "learning_rate": 8.316669506102314e-06,
+      "loss": 4.3804,
+      "step": 505000
+    },
+    {
+      "epoch": 4.3478822314049586,
+      "grad_norm": 5.0625,
+      "learning_rate": 8.305855827790642e-06,
+      "loss": 4.2444,
+      "step": 505050
+    },
+    {
+      "epoch": 4.348312672176308,
+      "grad_norm": 3.390625,
+      "learning_rate": 8.295048879631861e-06,
+      "loss": 4.4931,
+      "step": 505100
+    },
+    {
+      "epoch": 4.348743112947658,
+      "grad_norm": 5.75,
+      "learning_rate": 8.28424866241917e-06,
+      "loss": 4.6092,
+      "step": 505150
+    },
+    {
+      "epoch": 4.349173553719008,
+      "grad_norm": 3.359375,
+      "learning_rate": 8.273455176945333e-06,
+      "loss": 3.9988,
+      "step": 505200
+    },
+    {
+      "epoch": 4.349603994490358,
+      "grad_norm": 6.5,
+      "learning_rate": 8.262668424002517e-06,
+      "loss": 4.4437,
+      "step": 505250
+    },
+    {
+      "epoch": 4.350034435261708,
+      "grad_norm": 4.1875,
+      "learning_rate": 8.251888404382435e-06,
+      "loss": 4.4969,
+      "step": 505300
+    },
+    {
+      "epoch": 4.350464876033058,
+      "grad_norm": 2.5625,
+      "learning_rate": 8.241115118876364e-06,
+      "loss": 4.2591,
+      "step": 505350
+    },
+    {
+      "epoch": 4.350895316804408,
+      "grad_norm": 3.59375,
+      "learning_rate": 8.230348568275015e-06,
+      "loss": 4.0655,
+      "step": 505400
+    },
+    {
+      "epoch": 4.351325757575758,
+      "grad_norm": 2.46875,
+      "learning_rate": 8.219588753368634e-06,
+      "loss": 4.7343,
+      "step": 505450
+    },
+    {
+      "epoch": 4.351756198347108,
+      "grad_norm": 3.328125,
+      "learning_rate": 8.208835674946968e-06,
+      "loss": 4.2602,
+      "step": 505500
+    },
+    {
+      "epoch": 4.352186639118457,
+      "grad_norm": 6.59375,
+      "learning_rate": 8.198089333799264e-06,
+      "loss": 4.5904,
+      "step": 505550
+    },
+    {
+      "epoch": 4.352617079889807,
+      "grad_norm": 3.15625,
+      "learning_rate": 8.18734973071429e-06,
+      "loss": 4.3814,
+      "step": 505600
+    },
+    {
+      "epoch": 4.353047520661157,
+      "grad_norm": 3.796875,
+      "learning_rate": 8.176616866480313e-06,
+      "loss": 4.1741,
+      "step": 505650
+    },
+    {
+      "epoch": 4.353477961432507,
+      "grad_norm": 6.53125,
+      "learning_rate": 8.165890741885084e-06,
+      "loss": 4.6812,
+      "step": 505700
+    },
+    {
+      "epoch": 4.3539084022038566,
+      "grad_norm": 3.265625,
+      "learning_rate": 8.155171357715897e-06,
+      "loss": 4.4283,
+      "step": 505750
+    },
+    {
+      "epoch": 4.354338842975206,
+      "grad_norm": 5.9375,
+      "learning_rate": 8.144458714759528e-06,
+      "loss": 4.1386,
+      "step": 505800
+    },
+    {
+      "epoch": 4.354769283746556,
+      "grad_norm": 1.625,
+      "learning_rate": 8.13375281380223e-06,
+      "loss": 4.093,
+      "step": 505850
+    },
+    {
+      "epoch": 4.355199724517906,
+      "grad_norm": 3.5625,
+      "learning_rate": 8.123053655629842e-06,
+      "loss": 4.3499,
+      "step": 505900
+    },
+    {
+      "epoch": 4.355630165289256,
+      "grad_norm": 3.515625,
+      "learning_rate": 8.112361241027656e-06,
+      "loss": 4.5391,
+      "step": 505950
+    },
+    {
+      "epoch": 4.356060606060606,
+      "grad_norm": 2.78125,
+      "learning_rate": 8.10167557078042e-06,
+      "loss": 4.3041,
+      "step": 506000
+    },
+    {
+      "epoch": 4.356491046831956,
+      "grad_norm": 3.0625,
+      "learning_rate": 8.090996645672477e-06,
+      "loss": 3.7522,
+      "step": 506050
+    },
+    {
+      "epoch": 4.356921487603306,
+      "grad_norm": 4.84375,
+      "learning_rate": 8.080324466487643e-06,
+      "loss": 4.3118,
+      "step": 506100
+    },
+    {
+      "epoch": 4.357351928374656,
+      "grad_norm": 2.890625,
+      "learning_rate": 8.069659034009214e-06,
+      "loss": 4.3671,
+      "step": 506150
+    },
+    {
+      "epoch": 4.357782369146006,
+      "grad_norm": 3.828125,
+      "learning_rate": 8.05900034902003e-06,
+      "loss": 4.1839,
+      "step": 506200
+    },
+    {
+      "epoch": 4.358212809917355,
+      "grad_norm": 3.78125,
+      "learning_rate": 8.048348412302398e-06,
+      "loss": 4.3766,
+      "step": 506250
+    },
+    {
+      "epoch": 4.358643250688705,
+      "grad_norm": 2.84375,
+      "learning_rate": 8.03770322463816e-06,
+      "loss": 4.229,
+      "step": 506300
+    },
+    {
+      "epoch": 4.359073691460055,
+      "grad_norm": 2.109375,
+      "learning_rate": 8.027064786808635e-06,
+      "loss": 4.2059,
+      "step": 506350
+    },
+    {
+      "epoch": 4.359504132231405,
+      "grad_norm": 3.28125,
+      "learning_rate": 8.016433099594678e-06,
+      "loss": 4.2764,
+      "step": 506400
+    },
+    {
+      "epoch": 4.3599345730027546,
+      "grad_norm": 3.578125,
+      "learning_rate": 8.005808163776596e-06,
+      "loss": 4.1407,
+      "step": 506450
+    },
+    {
+      "epoch": 4.360365013774104,
+      "grad_norm": 0.78515625,
+      "learning_rate": 7.99518998013431e-06,
+      "loss": 4.2728,
+      "step": 506500
+    },
+    {
+      "epoch": 4.360795454545454,
+      "grad_norm": 4.0,
+      "learning_rate": 7.984578549447086e-06,
+      "loss": 4.5802,
+      "step": 506550
+    },
+    {
+      "epoch": 4.361225895316805,
+      "grad_norm": 5.125,
+      "learning_rate": 7.973973872493846e-06,
+      "loss": 4.2125,
+      "step": 506600
+    },
+    {
+      "epoch": 4.361656336088155,
+      "grad_norm": 4.09375,
+      "learning_rate": 7.96337595005291e-06,
+      "loss": 4.5784,
+      "step": 506650
+    },
+    {
+      "epoch": 4.362086776859504,
+      "grad_norm": 3.65625,
+      "learning_rate": 7.952784782902156e-06,
+      "loss": 4.2129,
+      "step": 506700
+    },
+    {
+      "epoch": 4.362517217630854,
+      "grad_norm": 4.625,
+      "learning_rate": 7.942200371818975e-06,
+      "loss": 3.9921,
+      "step": 506750
+    },
+    {
+      "epoch": 4.362947658402204,
+      "grad_norm": 3.46875,
+      "learning_rate": 7.931622717580211e-06,
+      "loss": 4.2852,
+      "step": 506800
+    },
+    {
+      "epoch": 4.363378099173554,
+      "grad_norm": 5.09375,
+      "learning_rate": 7.921051820962222e-06,
+      "loss": 4.3809,
+      "step": 506850
+    },
+    {
+      "epoch": 4.363808539944904,
+      "grad_norm": 4.5625,
+      "learning_rate": 7.910487682740942e-06,
+      "loss": 4.6271,
+      "step": 506900
+    },
+    {
+      "epoch": 4.364238980716253,
+      "grad_norm": 5.75,
+      "learning_rate": 7.899930303691728e-06,
+      "loss": 4.0671,
+      "step": 506950
+    },
+    {
+      "epoch": 4.364669421487603,
+      "grad_norm": 3.125,
+      "learning_rate": 7.889379684589482e-06,
+      "loss": 4.063,
+      "step": 507000
+    },
+    {
+      "epoch": 4.364669421487603,
+      "eval_loss": 4.990376949310303,
+      "eval_runtime": 24.8331,
+      "eval_samples_per_second": 25.772,
+      "eval_steps_per_second": 12.886,
+      "eval_tts_loss": 7.482996264546997,
+      "step": 507000
+    },
+    {
+      "epoch": 4.365099862258953,
+      "grad_norm": 2.59375,
+      "learning_rate": 7.878835826208575e-06,
+      "loss": 4.074,
+      "step": 507050
+    },
+    {
+      "epoch": 4.365530303030303,
+      "grad_norm": 3.71875,
+      "learning_rate": 7.868298729322898e-06,
+      "loss": 4.291,
+      "step": 507100
+    },
+    {
+      "epoch": 4.3659607438016526,
+      "grad_norm": 3.484375,
+      "learning_rate": 7.857768394705878e-06,
+      "loss": 4.2775,
+      "step": 507150
+    },
+    {
+      "epoch": 4.366391184573002,
+      "grad_norm": 4.28125,
+      "learning_rate": 7.847244823130418e-06,
+      "loss": 4.8031,
+      "step": 507200
+    },
+    {
+      "epoch": 4.366821625344353,
+      "grad_norm": 2.3125,
+      "learning_rate": 7.836728015368922e-06,
+      "loss": 4.7203,
+      "step": 507250
+    },
+    {
+      "epoch": 4.367252066115703,
+      "grad_norm": 7.46875,
+      "learning_rate": 7.826217972193285e-06,
+      "loss": 4.3567,
+      "step": 507300
+    },
+    {
+      "epoch": 4.367682506887053,
+      "grad_norm": 4.46875,
+      "learning_rate": 7.815714694374932e-06,
+      "loss": 4.5966,
+      "step": 507350
+    },
+    {
+      "epoch": 4.368112947658402,
+      "grad_norm": 4.25,
+      "learning_rate": 7.80521818268477e-06,
+      "loss": 4.1714,
+      "step": 507400
+    },
+    {
+      "epoch": 4.368543388429752,
+      "grad_norm": 4.875,
+      "learning_rate": 7.79472843789324e-06,
+      "loss": 4.0177,
+      "step": 507450
+    },
+    {
+      "epoch": 4.368973829201102,
+      "grad_norm": 2.109375,
+      "learning_rate": 7.78424546077028e-06,
+      "loss": 4.0039,
+      "step": 507500
+    },
+    {
+      "epoch": 4.369404269972452,
+      "grad_norm": 4.375,
+      "learning_rate": 7.77376925208526e-06,
+      "loss": 4.2144,
+      "step": 507550
+    },
+    {
+      "epoch": 4.369834710743802,
+      "grad_norm": 3.671875,
+      "learning_rate": 7.763299812607161e-06,
+      "loss": 4.1329,
+      "step": 507600
+    },
+    {
+      "epoch": 4.370265151515151,
+      "grad_norm": 3.15625,
+      "learning_rate": 7.752837143104408e-06,
+      "loss": 4.112,
+      "step": 507650
+    },
+    {
+      "epoch": 4.370695592286501,
+      "grad_norm": 4.78125,
+      "learning_rate": 7.742381244344921e-06,
+      "loss": 3.8912,
+      "step": 507700
+    },
+    {
+      "epoch": 4.371126033057851,
+      "grad_norm": 2.75,
+      "learning_rate": 7.731932117096185e-06,
+      "loss": 4.1174,
+      "step": 507750
+    },
+    {
+      "epoch": 4.371556473829201,
+      "grad_norm": 2.96875,
+      "learning_rate": 7.721489762125078e-06,
+      "loss": 4.2519,
+      "step": 507800
+    },
+    {
+      "epoch": 4.3719869146005506,
+      "grad_norm": 3.328125,
+      "learning_rate": 7.711054180198085e-06,
+      "loss": 4.6386,
+      "step": 507850
+    },
+    {
+      "epoch": 4.372417355371901,
+      "grad_norm": 3.34375,
+      "learning_rate": 7.70062537208116e-06,
+      "loss": 4.3162,
+      "step": 507900
+    },
+    {
+      "epoch": 4.372847796143251,
+      "grad_norm": 4.03125,
+      "learning_rate": 7.690203338539748e-06,
+      "loss": 4.3752,
+      "step": 507950
+    },
+    {
+      "epoch": 4.373278236914601,
+      "grad_norm": 1.9921875,
+      "learning_rate": 7.67978808033879e-06,
+      "loss": 4.3337,
+      "step": 508000
+    },
+    {
+      "epoch": 4.373708677685951,
+      "grad_norm": 3.3125,
+      "learning_rate": 7.669379598242776e-06,
+      "loss": 4.1397,
+      "step": 508050
+    },
+    {
+      "epoch": 4.3741391184573,
+      "grad_norm": 5.375,
+      "learning_rate": 7.658977893015618e-06,
+      "loss": 4.2318,
+      "step": 508100
+    },
+    {
+      "epoch": 4.37456955922865,
+      "grad_norm": 4.65625,
+      "learning_rate": 7.648582965420813e-06,
+      "loss": 4.4174,
+      "step": 508150
+    },
+    {
+      "epoch": 4.375,
+      "grad_norm": 4.65625,
+      "learning_rate": 7.638194816221334e-06,
+      "loss": 4.4217,
+      "step": 508200
+    },
+    {
+      "epoch": 4.37543044077135,
+      "grad_norm": 3.015625,
+      "learning_rate": 7.62781344617961e-06,
+      "loss": 4.1805,
+      "step": 508250
+    },
+    {
+      "epoch": 4.3758608815427,
+      "grad_norm": 2.71875,
+      "learning_rate": 7.617438856057668e-06,
+      "loss": 4.4549,
+      "step": 508300
+    },
+    {
+      "epoch": 4.376291322314049,
+      "grad_norm": 2.359375,
+      "learning_rate": 7.607071046616931e-06,
+      "loss": 4.2371,
+      "step": 508350
+    },
+    {
+      "epoch": 4.376721763085399,
+      "grad_norm": 3.953125,
+      "learning_rate": 7.596710018618369e-06,
+      "loss": 4.1101,
+      "step": 508400
+    },
+    {
+      "epoch": 4.377152203856749,
+      "grad_norm": 4.625,
+      "learning_rate": 7.586355772822506e-06,
+      "loss": 4.3206,
+      "step": 508450
+    },
+    {
+      "epoch": 4.377582644628099,
+      "grad_norm": 3.09375,
+      "learning_rate": 7.576008309989291e-06,
+      "loss": 4.4091,
+      "step": 508500
+    },
+    {
+      "epoch": 4.3780130853994494,
+      "grad_norm": 2.671875,
+      "learning_rate": 7.565667630878215e-06,
+      "loss": 4.2373,
+      "step": 508550
+    },
+    {
+      "epoch": 4.378443526170799,
+      "grad_norm": 1.5703125,
+      "learning_rate": 7.55533373624826e-06,
+      "loss": 4.4331,
+      "step": 508600
+    },
+    {
+      "epoch": 4.378873966942149,
+      "grad_norm": 4.5,
+      "learning_rate": 7.545006626857898e-06,
+      "loss": 4.6513,
+      "step": 508650
+    },
+    {
+      "epoch": 4.379304407713499,
+      "grad_norm": 3.515625,
+      "learning_rate": 7.534686303465122e-06,
+      "loss": 4.445,
+      "step": 508700
+    },
+    {
+      "epoch": 4.379734848484849,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.524372766827437e-06,
+      "loss": 4.3127,
+      "step": 508750
+    },
+    {
+      "epoch": 4.380165289256198,
+      "grad_norm": 2.765625,
+      "learning_rate": 7.5140660177018266e-06,
+      "loss": 4.6362,
+      "step": 508800
+    },
+    {
+      "epoch": 4.380595730027548,
+      "grad_norm": 1.8671875,
+      "learning_rate": 7.5037660568447745e-06,
+      "loss": 4.3959,
+      "step": 508850
+    },
+    {
+      "epoch": 4.381026170798898,
+      "grad_norm": 3.5625,
+      "learning_rate": 7.493472885012298e-06,
+      "loss": 4.13,
+      "step": 508900
+    },
+    {
+      "epoch": 4.381456611570248,
+      "grad_norm": 4.625,
+      "learning_rate": 7.483186502959849e-06,
+      "loss": 4.0516,
+      "step": 508950
+    },
+    {
+      "epoch": 4.381887052341598,
+      "grad_norm": 3.34375,
+      "learning_rate": 7.472906911442478e-06,
+      "loss": 4.4367,
+      "step": 509000
+    },
+    {
+      "epoch": 4.382317493112947,
+      "grad_norm": 2.53125,
+      "learning_rate": 7.462634111214684e-06,
+      "loss": 4.1198,
+      "step": 509050
+    },
+    {
+      "epoch": 4.382747933884297,
+      "grad_norm": 4.71875,
+      "learning_rate": 7.452368103030405e-06,
+      "loss": 4.4191,
+      "step": 509100
+    },
+    {
+      "epoch": 4.383178374655647,
+      "grad_norm": 2.53125,
+      "learning_rate": 7.442108887643206e-06,
+      "loss": 4.0874,
+      "step": 509150
+    },
+    {
+      "epoch": 4.383608815426998,
+      "grad_norm": 2.921875,
+      "learning_rate": 7.431856465806064e-06,
+      "loss": 4.0599,
+      "step": 509200
+    },
+    {
+      "epoch": 4.3840392561983474,
+      "grad_norm": 7.15625,
+      "learning_rate": 7.421610838271487e-06,
+      "loss": 4.2546,
+      "step": 509250
+    },
+    {
+      "epoch": 4.384469696969697,
+      "grad_norm": 2.5,
+      "learning_rate": 7.4113720057915174e-06,
+      "loss": 4.4595,
+      "step": 509300
+    },
+    {
+      "epoch": 4.384900137741047,
+      "grad_norm": 2.5625,
+      "learning_rate": 7.4011399691175985e-06,
+      "loss": 4.3373,
+      "step": 509350
+    },
+    {
+      "epoch": 4.385330578512397,
+      "grad_norm": 3.5625,
+      "learning_rate": 7.390914729000786e-06,
+      "loss": 4.3771,
+      "step": 509400
+    },
+    {
+      "epoch": 4.385761019283747,
+      "grad_norm": 4.59375,
+      "learning_rate": 7.38069628619158e-06,
+      "loss": 4.1849,
+      "step": 509450
+    },
+    {
+      "epoch": 4.386191460055096,
+      "grad_norm": 3.09375,
+      "learning_rate": 7.370484641440001e-06,
+      "loss": 4.3355,
+      "step": 509500
+    },
+    {
+      "epoch": 4.386621900826446,
+      "grad_norm": 3.078125,
+      "learning_rate": 7.360279795495539e-06,
+      "loss": 4.4696,
+      "step": 509550
+    },
+    {
+      "epoch": 4.387052341597796,
+      "grad_norm": 1.53125,
+      "learning_rate": 7.350081749107229e-06,
+      "loss": 4.197,
+      "step": 509600
+    },
+    {
+      "epoch": 4.387482782369146,
+      "grad_norm": 4.5,
+      "learning_rate": 7.339890503023561e-06,
+      "loss": 4.1877,
+      "step": 509650
+    },
+    {
+      "epoch": 4.387913223140496,
+      "grad_norm": 1.1796875,
+      "learning_rate": 7.32970605799258e-06,
+      "loss": 3.9881,
+      "step": 509700
+    },
+    {
+      "epoch": 4.388343663911845,
+      "grad_norm": 4.6875,
+      "learning_rate": 7.319528414761789e-06,
+      "loss": 4.2678,
+      "step": 509750
+    },
+    {
+      "epoch": 4.388774104683195,
+      "grad_norm": 5.3125,
+      "learning_rate": 7.309357574078191e-06,
+      "loss": 4.2218,
+      "step": 509800
+    },
+    {
+      "epoch": 4.389204545454546,
+      "grad_norm": 4.375,
+      "learning_rate": 7.299193536688343e-06,
+      "loss": 4.1335,
+      "step": 509850
+    },
+    {
+      "epoch": 4.389634986225896,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.2890363033382254e-06,
+      "loss": 4.0931,
+      "step": 509900
+    },
+    {
+      "epoch": 4.3900654269972454,
+      "grad_norm": 3.875,
+      "learning_rate": 7.278885874773356e-06,
+      "loss": 4.263,
+      "step": 509950
+    },
+    {
+      "epoch": 4.390495867768595,
+      "grad_norm": 4.90625,
+      "learning_rate": 7.268742251738792e-06,
+      "loss": 4.4627,
+      "step": 510000
+    },
+    {
+      "epoch": 4.390495867768595,
+      "eval_loss": 4.990321159362793,
+      "eval_runtime": 25.1056,
+      "eval_samples_per_second": 25.492,
+      "eval_steps_per_second": 12.746,
+      "eval_tts_loss": 7.480910927547624,
+      "step": 510000
+    },
+    {
+      "epoch": 4.390926308539945,
+      "grad_norm": 4.1875,
+      "learning_rate": 7.258605434979026e-06,
+      "loss": 4.5134,
+      "step": 510050
+    },
+    {
+      "epoch": 4.391356749311295,
+      "grad_norm": 3.859375,
+      "learning_rate": 7.2484754252380775e-06,
+      "loss": 4.4191,
+      "step": 510100
+    },
+    {
+      "epoch": 4.391787190082645,
+      "grad_norm": 4.46875,
+      "learning_rate": 7.23835222325947e-06,
+      "loss": 3.9403,
+      "step": 510150
+    },
+    {
+      "epoch": 4.392217630853994,
+      "grad_norm": 4.34375,
+      "learning_rate": 7.228235829786234e-06,
+      "loss": 4.4877,
+      "step": 510200
+    },
+    {
+      "epoch": 4.392648071625344,
+      "grad_norm": 3.453125,
+      "learning_rate": 7.2181262455608745e-06,
+      "loss": 3.9123,
+      "step": 510250
+    },
+    {
+      "epoch": 4.393078512396694,
+      "grad_norm": 3.859375,
+      "learning_rate": 7.208023471325431e-06,
+      "loss": 4.4258,
+      "step": 510300
+    },
+    {
+      "epoch": 4.393508953168044,
+      "grad_norm": 3.953125,
+      "learning_rate": 7.197927507821422e-06,
+      "loss": 4.3587,
+      "step": 510350
+    },
+    {
+      "epoch": 4.393939393939394,
+      "grad_norm": 4.9375,
+      "learning_rate": 7.187838355789855e-06,
+      "loss": 4.2157,
+      "step": 510400
+    },
+    {
+      "epoch": 4.394369834710744,
+      "grad_norm": 2.859375,
+      "learning_rate": 7.177756015971271e-06,
+      "loss": 4.2579,
+      "step": 510450
+    },
+    {
+      "epoch": 4.394800275482094,
+      "grad_norm": 2.203125,
+      "learning_rate": 7.1676804891056655e-06,
+      "loss": 4.2333,
+      "step": 510500
+    },
+    {
+      "epoch": 4.395230716253444,
+      "grad_norm": 3.109375,
+      "learning_rate": 7.157611775932593e-06,
+      "loss": 4.4928,
+      "step": 510550
+    },
+    {
+      "epoch": 4.395661157024794,
+      "grad_norm": 5.375,
+      "learning_rate": 7.147549877191084e-06,
+      "loss": 4.4959,
+      "step": 510600
+    },
+    {
+      "epoch": 4.3960915977961434,
+      "grad_norm": 3.75,
+      "learning_rate": 7.137494793619593e-06,
+      "loss": 4.1776,
+      "step": 510650
+    },
+    {
+      "epoch": 4.396522038567493,
+      "grad_norm": 3.40625,
+      "learning_rate": 7.127446525956205e-06,
+      "loss": 4.5399,
+      "step": 510700
+    },
+    {
+      "epoch": 4.396952479338843,
+      "grad_norm": 5.21875,
+      "learning_rate": 7.117405074938432e-06,
+      "loss": 4.5661,
+      "step": 510750
+    },
+    {
+      "epoch": 4.397382920110193,
+      "grad_norm": 3.5625,
+      "learning_rate": 7.107370441303252e-06,
+      "loss": 4.3027,
+      "step": 510800
+    },
+    {
+      "epoch": 4.397813360881543,
+      "grad_norm": 3.09375,
+      "learning_rate": 7.097342625787262e-06,
+      "loss": 4.3491,
+      "step": 510850
+    },
+    {
+      "epoch": 4.398243801652892,
+      "grad_norm": 5.46875,
+      "learning_rate": 7.08732162912642e-06,
+      "loss": 4.1872,
+      "step": 510900
+    },
+    {
+      "epoch": 4.398674242424242,
+      "grad_norm": 2.96875,
+      "learning_rate": 7.0773074520562476e-06,
+      "loss": 4.3698,
+      "step": 510950
+    },
+    {
+      "epoch": 4.399104683195592,
+      "grad_norm": 7.09375,
+      "learning_rate": 7.06730009531179e-06,
+      "loss": 4.5419,
+      "step": 511000
+    },
+    {
+      "epoch": 4.399535123966942,
+      "grad_norm": 3.546875,
+      "learning_rate": 7.057299559627562e-06,
+      "loss": 4.5538,
+      "step": 511050
+    },
+    {
+      "epoch": 4.399965564738292,
+      "grad_norm": 4.21875,
+      "learning_rate": 7.047305845737573e-06,
+      "loss": 4.0321,
+      "step": 511100
+    },
+    {
+      "epoch": 4.400396005509642,
+      "grad_norm": 3.453125,
+      "learning_rate": 7.037318954375338e-06,
+      "loss": 4.2875,
+      "step": 511150
+    },
+    {
+      "epoch": 4.400826446280992,
+      "grad_norm": 5.15625,
+      "learning_rate": 7.027338886273871e-06,
+      "loss": 4.3006,
+      "step": 511200
+    },
+    {
+      "epoch": 4.401256887052342,
+      "grad_norm": 3.265625,
+      "learning_rate": 7.017365642165696e-06,
+      "loss": 4.4986,
+      "step": 511250
+    },
+    {
+      "epoch": 4.401687327823692,
+      "grad_norm": 3.203125,
+      "learning_rate": 7.00739922278284e-06,
+      "loss": 4.2387,
+      "step": 511300
+    },
+    {
+      "epoch": 4.4021177685950414,
+      "grad_norm": 5.71875,
+      "learning_rate": 6.9974396288567945e-06,
+      "loss": 4.4865,
+      "step": 511350
+    },
+    {
+      "epoch": 4.402548209366391,
+      "grad_norm": 4.0625,
+      "learning_rate": 6.987486861118575e-06,
+      "loss": 4.08,
+      "step": 511400
+    },
+    {
+      "epoch": 4.402978650137741,
+      "grad_norm": 3.96875,
+      "learning_rate": 6.977540920298709e-06,
+      "loss": 4.1794,
+      "step": 511450
+    },
+    {
+      "epoch": 4.403409090909091,
+      "grad_norm": 3.65625,
+      "learning_rate": 6.967601807127177e-06,
+      "loss": 4.4802,
+      "step": 511500
+    },
+    {
+      "epoch": 4.403839531680441,
+      "grad_norm": 7.71875,
+      "learning_rate": 6.957669522333521e-06,
+      "loss": 4.2602,
+      "step": 511550
+    },
+    {
+      "epoch": 4.40426997245179,
+      "grad_norm": 1.8125,
+      "learning_rate": 6.947744066646744e-06,
+      "loss": 4.1937,
+      "step": 511600
+    },
+    {
+      "epoch": 4.40470041322314,
+      "grad_norm": 4.53125,
+      "learning_rate": 6.937825440795332e-06,
+      "loss": 4.2611,
+      "step": 511650
+    },
+    {
+      "epoch": 4.40513085399449,
+      "grad_norm": 2.8125,
+      "learning_rate": 6.927913645507311e-06,
+      "loss": 4.3111,
+      "step": 511700
+    },
+    {
+      "epoch": 4.405561294765841,
+      "grad_norm": 3.296875,
+      "learning_rate": 6.91800868151018e-06,
+      "loss": 4.3315,
+      "step": 511750
+    },
+    {
+      "epoch": 4.4059917355371905,
+      "grad_norm": 3.171875,
+      "learning_rate": 6.908110549530933e-06,
+      "loss": 4.5461,
+      "step": 511800
+    },
+    {
+      "epoch": 4.40642217630854,
+      "grad_norm": 3.046875,
+      "learning_rate": 6.898219250296078e-06,
+      "loss": 4.3747,
+      "step": 511850
+    },
+    {
+      "epoch": 4.40685261707989,
+      "grad_norm": 2.640625,
+      "learning_rate": 6.888334784531625e-06,
+      "loss": 4.4115,
+      "step": 511900
+    },
+    {
+      "epoch": 4.40728305785124,
+      "grad_norm": 5.125,
+      "learning_rate": 6.8784571529630696e-06,
+      "loss": 4.4447,
+      "step": 511950
+    },
+    {
+      "epoch": 4.40771349862259,
+      "grad_norm": 2.234375,
+      "learning_rate": 6.8685863563153985e-06,
+      "loss": 4.2815,
+      "step": 512000
+    },
+    {
+      "epoch": 4.4081439393939394,
+      "grad_norm": 4.0625,
+      "learning_rate": 6.858722395313089e-06,
+      "loss": 4.2215,
+      "step": 512050
+    },
+    {
+      "epoch": 4.408574380165289,
+      "grad_norm": 3.0,
+      "learning_rate": 6.848865270680183e-06,
+      "loss": 4.5206,
+      "step": 512100
+    },
+    {
+      "epoch": 4.409004820936639,
+      "grad_norm": 3.390625,
+      "learning_rate": 6.839014983140158e-06,
+      "loss": 4.0088,
+      "step": 512150
+    },
+    {
+      "epoch": 4.409435261707989,
+      "grad_norm": 3.0625,
+      "learning_rate": 6.8291715334159564e-06,
+      "loss": 4.0994,
+      "step": 512200
+    },
+    {
+      "epoch": 4.409865702479339,
+      "grad_norm": 2.328125,
+      "learning_rate": 6.819334922230125e-06,
+      "loss": 4.4212,
+      "step": 512250
+    },
+    {
+      "epoch": 4.410296143250688,
+      "grad_norm": 3.765625,
+      "learning_rate": 6.8095051503046185e-06,
+      "loss": 4.3757,
+      "step": 512300
+    },
+    {
+      "epoch": 4.410726584022038,
+      "grad_norm": 3.140625,
+      "learning_rate": 6.799682218360926e-06,
+      "loss": 4.3514,
+      "step": 512350
+    },
+    {
+      "epoch": 4.411157024793388,
+      "grad_norm": 2.9375,
+      "learning_rate": 6.789866127120048e-06,
+      "loss": 4.4842,
+      "step": 512400
+    },
+    {
+      "epoch": 4.411587465564739,
+      "grad_norm": 2.15625,
+      "learning_rate": 6.780056877302443e-06,
+      "loss": 4.2474,
+      "step": 512450
+    },
+    {
+      "epoch": 4.4120179063360885,
+      "grad_norm": 2.84375,
+      "learning_rate": 6.770254469628079e-06,
+      "loss": 4.1866,
+      "step": 512500
+    },
+    {
+      "epoch": 4.412448347107438,
+      "grad_norm": 5.25,
+      "learning_rate": 6.760458904816469e-06,
+      "loss": 4.5259,
+      "step": 512550
+    },
+    {
+      "epoch": 4.412878787878788,
+      "grad_norm": 4.25,
+      "learning_rate": 6.7506701835865605e-06,
+      "loss": 4.3188,
+      "step": 512600
+    },
+    {
+      "epoch": 4.413309228650138,
+      "grad_norm": 1.8515625,
+      "learning_rate": 6.740888306656823e-06,
+      "loss": 4.1352,
+      "step": 512650
+    },
+    {
+      "epoch": 4.413739669421488,
+      "grad_norm": 2.03125,
+      "learning_rate": 6.731113274745249e-06,
+      "loss": 4.028,
+      "step": 512700
+    },
+    {
+      "epoch": 4.4141701101928374,
+      "grad_norm": 5.25,
+      "learning_rate": 6.721345088569253e-06,
+      "loss": 4.2058,
+      "step": 512750
+    },
+    {
+      "epoch": 4.414600550964187,
+      "grad_norm": 3.90625,
+      "learning_rate": 6.7115837488458625e-06,
+      "loss": 4.5951,
+      "step": 512800
+    },
+    {
+      "epoch": 4.415030991735537,
+      "grad_norm": 3.328125,
+      "learning_rate": 6.701829256291503e-06,
+      "loss": 4.5285,
+      "step": 512850
+    },
+    {
+      "epoch": 4.415461432506887,
+      "grad_norm": 4.03125,
+      "learning_rate": 6.6920816116221365e-06,
+      "loss": 4.1794,
+      "step": 512900
+    },
+    {
+      "epoch": 4.415891873278237,
+      "grad_norm": 3.8125,
+      "learning_rate": 6.682340815553223e-06,
+      "loss": 4.1023,
+      "step": 512950
+    },
+    {
+      "epoch": 4.416322314049586,
+      "grad_norm": 4.0625,
+      "learning_rate": 6.672606868799725e-06,
+      "loss": 4.3001,
+      "step": 513000
+    },
+    {
+      "epoch": 4.416322314049586,
+      "eval_loss": 4.990197658538818,
+      "eval_runtime": 23.9965,
+      "eval_samples_per_second": 26.671,
+      "eval_steps_per_second": 13.335,
+      "eval_tts_loss": 7.4809472596331625,
+      "step": 513000
+    },
+    {
+      "epoch": 4.416752754820937,
+      "grad_norm": 3.109375,
+      "learning_rate": 6.6628797720760586e-06,
+      "loss": 4.5626,
+      "step": 513050
+    },
+    {
+      "epoch": 4.417183195592287,
+      "grad_norm": 2.171875,
+      "learning_rate": 6.65315952609622e-06,
+      "loss": 4.6766,
+      "step": 513100
+    },
+    {
+      "epoch": 4.417613636363637,
+      "grad_norm": 2.625,
+      "learning_rate": 6.643446131573638e-06,
+      "loss": 4.4865,
+      "step": 513150
+    },
+    {
+      "epoch": 4.4180440771349865,
+      "grad_norm": 2.546875,
+      "learning_rate": 6.633739589221221e-06,
+      "loss": 4.2308,
+      "step": 513200
+    },
+    {
+      "epoch": 4.418474517906336,
+      "grad_norm": 3.515625,
+      "learning_rate": 6.624039899751455e-06,
+      "loss": 4.2097,
+      "step": 513250
+    },
+    {
+      "epoch": 4.418904958677686,
+      "grad_norm": 2.109375,
+      "learning_rate": 6.614347063876259e-06,
+      "loss": 4.3102,
+      "step": 513300
+    },
+    {
+      "epoch": 4.419335399449036,
+      "grad_norm": 4.625,
+      "learning_rate": 6.60466108230704e-06,
+      "loss": 4.4591,
+      "step": 513350
+    },
+    {
+      "epoch": 4.419765840220386,
+      "grad_norm": 2.015625,
+      "learning_rate": 6.594981955754786e-06,
+      "loss": 4.2024,
+      "step": 513400
+    },
+    {
+      "epoch": 4.4201962809917354,
+      "grad_norm": 3.109375,
+      "learning_rate": 6.585309684929864e-06,
+      "loss": 4.412,
+      "step": 513450
+    },
+    {
+      "epoch": 4.420626721763085,
+      "grad_norm": 4.125,
+      "learning_rate": 6.575644270542236e-06,
+      "loss": 4.2866,
+      "step": 513500
+    },
+    {
+      "epoch": 4.421057162534435,
+      "grad_norm": 3.9375,
+      "learning_rate": 6.565985713301315e-06,
+      "loss": 4.4881,
+      "step": 513550
+    },
+    {
+      "epoch": 4.421487603305785,
+      "grad_norm": 3.78125,
+      "learning_rate": 6.556334013915988e-06,
+      "loss": 4.2589,
+      "step": 513600
+    },
+    {
+      "epoch": 4.421918044077135,
+      "grad_norm": 3.6875,
+      "learning_rate": 6.546689173094722e-06,
+      "loss": 4.1817,
+      "step": 513650
+    },
+    {
+      "epoch": 4.422348484848484,
+      "grad_norm": 1.6171875,
+      "learning_rate": 6.537051191545407e-06,
+      "loss": 4.0419,
+      "step": 513700
+    },
+    {
+      "epoch": 4.422778925619835,
+      "grad_norm": 4.4375,
+      "learning_rate": 6.5274200699754205e-06,
+      "loss": 4.0975,
+      "step": 513750
+    },
+    {
+      "epoch": 4.423209366391185,
+      "grad_norm": 2.765625,
+      "learning_rate": 6.517795809091698e-06,
+      "loss": 4.0693,
+      "step": 513800
+    },
+    {
+      "epoch": 4.423639807162535,
+      "grad_norm": 2.96875,
+      "learning_rate": 6.50817840960064e-06,
+      "loss": 3.9669,
+      "step": 513850
+    },
+    {
+      "epoch": 4.4240702479338845,
+      "grad_norm": 3.171875,
+      "learning_rate": 6.498567872208105e-06,
+      "loss": 4.1007,
+      "step": 513900
+    },
+    {
+      "epoch": 4.424500688705234,
+      "grad_norm": 3.34375,
+      "learning_rate": 6.488964197619563e-06,
+      "loss": 4.3688,
+      "step": 513950
+    },
+    {
+      "epoch": 4.424931129476584,
+      "grad_norm": 2.5,
+      "learning_rate": 6.479367386539825e-06,
+      "loss": 4.3292,
+      "step": 514000
+    },
+    {
+      "epoch": 4.425361570247934,
+      "grad_norm": 5.53125,
+      "learning_rate": 6.469777439673297e-06,
+      "loss": 4.3969,
+      "step": 514050
+    },
+    {
+      "epoch": 4.425792011019284,
+      "grad_norm": 3.4375,
+      "learning_rate": 6.460194357723881e-06,
+      "loss": 4.7134,
+      "step": 514100
+    },
+    {
+      "epoch": 4.4262224517906334,
+      "grad_norm": 2.71875,
+      "learning_rate": 6.450618141394949e-06,
+      "loss": 4.1469,
+      "step": 514150
+    },
+    {
+      "epoch": 4.426652892561983,
+      "grad_norm": 2.578125,
+      "learning_rate": 6.441048791389359e-06,
+      "loss": 4.4025,
+      "step": 514200
+    },
+    {
+      "epoch": 4.427083333333333,
+      "grad_norm": 3.8125,
+      "learning_rate": 6.431486308409496e-06,
+      "loss": 4.1695,
+      "step": 514250
+    },
+    {
+      "epoch": 4.427513774104683,
+      "grad_norm": 3.203125,
+      "learning_rate": 6.421930693157207e-06,
+      "loss": 4.3792,
+      "step": 514300
+    },
+    {
+      "epoch": 4.4279442148760335,
+      "grad_norm": 3.84375,
+      "learning_rate": 6.4123819463338785e-06,
+      "loss": 4.2667,
+      "step": 514350
+    },
+    {
+      "epoch": 4.428374655647383,
+      "grad_norm": 3.34375,
+      "learning_rate": 6.402840068640359e-06,
+      "loss": 3.8351,
+      "step": 514400
+    },
+    {
+      "epoch": 4.428805096418733,
+      "grad_norm": 3.109375,
+      "learning_rate": 6.393305060777e-06,
+      "loss": 4.3111,
+      "step": 514450
+    },
+    {
+      "epoch": 4.429235537190083,
+      "grad_norm": 6.0,
+      "learning_rate": 6.383776923443652e-06,
+      "loss": 4.532,
+      "step": 514500
+    },
+    {
+      "epoch": 4.429665977961433,
+      "grad_norm": 3.53125,
+      "learning_rate": 6.374255657339645e-06,
+      "loss": 4.8584,
+      "step": 514550
+    },
+    {
+      "epoch": 4.4300964187327825,
+      "grad_norm": 3.859375,
+      "learning_rate": 6.364741263163831e-06,
+      "loss": 4.0418,
+      "step": 514600
+    },
+    {
+      "epoch": 4.430526859504132,
+      "grad_norm": 5.40625,
+      "learning_rate": 6.355233741614552e-06,
+      "loss": 4.04,
+      "step": 514650
+    },
+    {
+      "epoch": 4.430957300275482,
+      "grad_norm": 1.4765625,
+      "learning_rate": 6.345733093389661e-06,
+      "loss": 4.2641,
+      "step": 514700
+    },
+    {
+      "epoch": 4.431387741046832,
+      "grad_norm": 4.71875,
+      "learning_rate": 6.33623931918641e-06,
+      "loss": 4.5212,
+      "step": 514750
+    },
+    {
+      "epoch": 4.431818181818182,
+      "grad_norm": 1.8359375,
+      "learning_rate": 6.326752419701698e-06,
+      "loss": 4.4224,
+      "step": 514800
+    },
+    {
+      "epoch": 4.4322486225895315,
+      "grad_norm": 4.03125,
+      "learning_rate": 6.3172723956318124e-06,
+      "loss": 4.221,
+      "step": 514850
+    },
+    {
+      "epoch": 4.432679063360881,
+      "grad_norm": 3.546875,
+      "learning_rate": 6.307799247672552e-06,
+      "loss": 4.101,
+      "step": 514900
+    },
+    {
+      "epoch": 4.433109504132231,
+      "grad_norm": 5.75,
+      "learning_rate": 6.298332976519283e-06,
+      "loss": 4.5273,
+      "step": 514950
+    },
+    {
+      "epoch": 4.433539944903581,
+      "grad_norm": 2.828125,
+      "learning_rate": 6.2888735828667275e-06,
+      "loss": 4.3606,
+      "step": 515000
+    },
+    {
+      "epoch": 4.4339703856749315,
+      "grad_norm": 1.15625,
+      "learning_rate": 6.279421067409252e-06,
+      "loss": 3.9342,
+      "step": 515050
+    },
+    {
+      "epoch": 4.434400826446281,
+      "grad_norm": 4.0625,
+      "learning_rate": 6.269975430840613e-06,
+      "loss": 4.5658,
+      "step": 515100
+    },
+    {
+      "epoch": 4.434831267217631,
+      "grad_norm": 4.34375,
+      "learning_rate": 6.260536673854123e-06,
+      "loss": 4.3632,
+      "step": 515150
+    },
+    {
+      "epoch": 4.435261707988981,
+      "grad_norm": 4.6875,
+      "learning_rate": 6.2511047971425485e-06,
+      "loss": 4.1757,
+      "step": 515200
+    },
+    {
+      "epoch": 4.435692148760331,
+      "grad_norm": 4.21875,
+      "learning_rate": 6.2416798013981815e-06,
+      "loss": 4.3564,
+      "step": 515250
+    },
+    {
+      "epoch": 4.4361225895316805,
+      "grad_norm": 1.703125,
+      "learning_rate": 6.232261687312768e-06,
+      "loss": 4.2246,
+      "step": 515300
+    },
+    {
+      "epoch": 4.43655303030303,
+      "grad_norm": 4.8125,
+      "learning_rate": 6.222850455577622e-06,
+      "loss": 4.1892,
+      "step": 515350
+    },
+    {
+      "epoch": 4.43698347107438,
+      "grad_norm": 6.65625,
+      "learning_rate": 6.213446106883481e-06,
+      "loss": 4.3059,
+      "step": 515400
+    },
+    {
+      "epoch": 4.43741391184573,
+      "grad_norm": 4.40625,
+      "learning_rate": 6.20404864192059e-06,
+      "loss": 4.1826,
+      "step": 515450
+    },
+    {
+      "epoch": 4.43784435261708,
+      "grad_norm": 2.609375,
+      "learning_rate": 6.1946580613787664e-06,
+      "loss": 4.3087,
+      "step": 515500
+    },
+    {
+      "epoch": 4.4382747933884295,
+      "grad_norm": 3.125,
+      "learning_rate": 6.185274365947191e-06,
+      "loss": 4.1832,
+      "step": 515550
+    },
+    {
+      "epoch": 4.438705234159779,
+      "grad_norm": 6.625,
+      "learning_rate": 6.175897556314603e-06,
+      "loss": 4.4377,
+      "step": 515600
+    },
+    {
+      "epoch": 4.43913567493113,
+      "grad_norm": 3.46875,
+      "learning_rate": 6.166527633169295e-06,
+      "loss": 4.245,
+      "step": 515650
+    },
+    {
+      "epoch": 4.43956611570248,
+      "grad_norm": 4.28125,
+      "learning_rate": 6.157164597198973e-06,
+      "loss": 3.8926,
+      "step": 515700
+    },
+    {
+      "epoch": 4.4399965564738295,
+      "grad_norm": 0.875,
+      "learning_rate": 6.147808449090853e-06,
+      "loss": 4.3383,
+      "step": 515750
+    },
+    {
+      "epoch": 4.440426997245179,
+      "grad_norm": 3.46875,
+      "learning_rate": 6.138459189531665e-06,
+      "loss": 4.3676,
+      "step": 515800
+    },
+    {
+      "epoch": 4.440857438016529,
+      "grad_norm": 4.625,
+      "learning_rate": 6.129116819207614e-06,
+      "loss": 4.2748,
+      "step": 515850
+    },
+    {
+      "epoch": 4.441287878787879,
+      "grad_norm": 2.546875,
+      "learning_rate": 6.1197813388044315e-06,
+      "loss": 4.4972,
+      "step": 515900
+    },
+    {
+      "epoch": 4.441718319559229,
+      "grad_norm": 3.671875,
+      "learning_rate": 6.110452749007311e-06,
+      "loss": 4.1191,
+      "step": 515950
+    },
+    {
+      "epoch": 4.4421487603305785,
+      "grad_norm": 1.9140625,
+      "learning_rate": 6.101131050500952e-06,
+      "loss": 4.3234,
+      "step": 516000
+    },
+    {
+      "epoch": 4.4421487603305785,
+      "eval_loss": 4.990259647369385,
+      "eval_runtime": 24.5592,
+      "eval_samples_per_second": 26.06,
+      "eval_steps_per_second": 13.03,
+      "eval_tts_loss": 7.480381709480102,
+      "step": 516000
+    },
+    {
+      "epoch": 4.442579201101928,
+      "grad_norm": 4.375,
+      "learning_rate": 6.09181624396955e-06,
+      "loss": 4.116,
+      "step": 516050
+    },
+    {
+      "epoch": 4.443009641873278,
+      "grad_norm": 4.46875,
+      "learning_rate": 6.082508330096781e-06,
+      "loss": 4.3434,
+      "step": 516100
+    },
+    {
+      "epoch": 4.443440082644628,
+      "grad_norm": 3.328125,
+      "learning_rate": 6.073207309565809e-06,
+      "loss": 4.6337,
+      "step": 516150
+    },
+    {
+      "epoch": 4.443870523415978,
+      "grad_norm": 4.59375,
+      "learning_rate": 6.0639131830593556e-06,
+      "loss": 4.404,
+      "step": 516200
+    },
+    {
+      "epoch": 4.4443009641873275,
+      "grad_norm": 4.25,
+      "learning_rate": 6.0546259512595845e-06,
+      "loss": 4.4904,
+      "step": 516250
+    },
+    {
+      "epoch": 4.444731404958677,
+      "grad_norm": 3.609375,
+      "learning_rate": 6.045345614848108e-06,
+      "loss": 4.1054,
+      "step": 516300
+    },
+    {
+      "epoch": 4.445161845730028,
+      "grad_norm": 2.71875,
+      "learning_rate": 6.036072174506124e-06,
+      "loss": 4.309,
+      "step": 516350
+    },
+    {
+      "epoch": 4.445592286501378,
+      "grad_norm": 2.265625,
+      "learning_rate": 6.026805630914279e-06,
+      "loss": 4.3595,
+      "step": 516400
+    },
+    {
+      "epoch": 4.4460227272727275,
+      "grad_norm": 2.375,
+      "learning_rate": 6.0175459847526925e-06,
+      "loss": 4.3329,
+      "step": 516450
+    },
+    {
+      "epoch": 4.446453168044077,
+      "grad_norm": 2.671875,
+      "learning_rate": 6.008293236701057e-06,
+      "loss": 4.5525,
+      "step": 516500
+    },
+    {
+      "epoch": 4.446883608815427,
+      "grad_norm": 2.484375,
+      "learning_rate": 5.999047387438439e-06,
+      "loss": 4.2963,
+      "step": 516550
+    },
+    {
+      "epoch": 4.447314049586777,
+      "grad_norm": 2.765625,
+      "learning_rate": 5.989808437643507e-06,
+      "loss": 4.2487,
+      "step": 516600
+    },
+    {
+      "epoch": 4.447744490358127,
+      "grad_norm": 2.125,
+      "learning_rate": 5.980576387994363e-06,
+      "loss": 4.5549,
+      "step": 516650
+    },
+    {
+      "epoch": 4.4481749311294765,
+      "grad_norm": 4.03125,
+      "learning_rate": 5.971351239168632e-06,
+      "loss": 4.1427,
+      "step": 516700
+    },
+    {
+      "epoch": 4.448605371900826,
+      "grad_norm": 3.96875,
+      "learning_rate": 5.962132991843405e-06,
+      "loss": 4.1719,
+      "step": 516750
+    },
+    {
+      "epoch": 4.449035812672176,
+      "grad_norm": 3.515625,
+      "learning_rate": 5.952921646695297e-06,
+      "loss": 3.8879,
+      "step": 516800
+    },
+    {
+      "epoch": 4.449466253443526,
+      "grad_norm": 2.453125,
+      "learning_rate": 5.943717204400379e-06,
+      "loss": 3.9986,
+      "step": 516850
+    },
+    {
+      "epoch": 4.449896694214876,
+      "grad_norm": 3.359375,
+      "learning_rate": 5.9345196656342525e-06,
+      "loss": 4.3833,
+      "step": 516900
+    },
+    {
+      "epoch": 4.450327134986226,
+      "grad_norm": 4.3125,
+      "learning_rate": 5.925329031072014e-06,
+      "loss": 4.1024,
+      "step": 516950
+    },
+    {
+      "epoch": 4.450757575757576,
+      "grad_norm": 2.90625,
+      "learning_rate": 5.9161453013881985e-06,
+      "loss": 4.5635,
+      "step": 517000
+    },
+    {
+      "epoch": 4.451188016528926,
+      "grad_norm": 0.8125,
+      "learning_rate": 5.906968477256902e-06,
+      "loss": 3.7934,
+      "step": 517050
+    },
+    {
+      "epoch": 4.451618457300276,
+      "grad_norm": 2.359375,
+      "learning_rate": 5.897798559351675e-06,
+      "loss": 4.347,
+      "step": 517100
+    },
+    {
+      "epoch": 4.4520488980716255,
+      "grad_norm": 5.28125,
+      "learning_rate": 5.888635548345545e-06,
+      "loss": 4.3105,
+      "step": 517150
+    },
+    {
+      "epoch": 4.452479338842975,
+      "grad_norm": 2.359375,
+      "learning_rate": 5.8794794449111065e-06,
+      "loss": 4.2003,
+      "step": 517200
+    },
+    {
+      "epoch": 4.452909779614325,
+      "grad_norm": 2.734375,
+      "learning_rate": 5.870330249720357e-06,
+      "loss": 4.359,
+      "step": 517250
+    },
+    {
+      "epoch": 4.453340220385675,
+      "grad_norm": 3.046875,
+      "learning_rate": 5.861187963444848e-06,
+      "loss": 4.2676,
+      "step": 517300
+    },
+    {
+      "epoch": 4.453770661157025,
+      "grad_norm": 4.6875,
+      "learning_rate": 5.852052586755597e-06,
+      "loss": 4.2297,
+      "step": 517350
+    },
+    {
+      "epoch": 4.4542011019283745,
+      "grad_norm": 5.03125,
+      "learning_rate": 5.842924120323123e-06,
+      "loss": 4.1846,
+      "step": 517400
+    },
+    {
+      "epoch": 4.454631542699724,
+      "grad_norm": 1.3203125,
+      "learning_rate": 5.833802564817425e-06,
+      "loss": 4.3206,
+      "step": 517450
+    },
+    {
+      "epoch": 4.455061983471074,
+      "grad_norm": 2.59375,
+      "learning_rate": 5.824687920908023e-06,
+      "loss": 4.6782,
+      "step": 517500
+    },
+    {
+      "epoch": 4.455492424242424,
+      "grad_norm": 2.875,
+      "learning_rate": 5.815580189263903e-06,
+      "loss": 4.3217,
+      "step": 517550
+    },
+    {
+      "epoch": 4.455922865013774,
+      "grad_norm": 3.15625,
+      "learning_rate": 5.806479370553553e-06,
+      "loss": 4.4238,
+      "step": 517600
+    },
+    {
+      "epoch": 4.456353305785124,
+      "grad_norm": 3.234375,
+      "learning_rate": 5.79738546544496e-06,
+      "loss": 4.302,
+      "step": 517650
+    },
+    {
+      "epoch": 4.456783746556474,
+      "grad_norm": 3.3125,
+      "learning_rate": 5.788298474605581e-06,
+      "loss": 4.2024,
+      "step": 517700
+    },
+    {
+      "epoch": 4.457214187327824,
+      "grad_norm": 3.671875,
+      "learning_rate": 5.779218398702402e-06,
+      "loss": 4.2905,
+      "step": 517750
+    },
+    {
+      "epoch": 4.457644628099174,
+      "grad_norm": 2.03125,
+      "learning_rate": 5.770145238401881e-06,
+      "loss": 4.1912,
+      "step": 517800
+    },
+    {
+      "epoch": 4.4580750688705235,
+      "grad_norm": 4.0,
+      "learning_rate": 5.7610789943699285e-06,
+      "loss": 4.4892,
+      "step": 517850
+    },
+    {
+      "epoch": 4.458505509641873,
+      "grad_norm": 5.28125,
+      "learning_rate": 5.752019667272046e-06,
+      "loss": 4.5276,
+      "step": 517900
+    },
+    {
+      "epoch": 4.458935950413223,
+      "grad_norm": 4.03125,
+      "learning_rate": 5.742967257773124e-06,
+      "loss": 4.5385,
+      "step": 517950
+    },
+    {
+      "epoch": 4.459366391184573,
+      "grad_norm": 2.59375,
+      "learning_rate": 5.733921766537598e-06,
+      "loss": 4.0946,
+      "step": 518000
+    },
+    {
+      "epoch": 4.459796831955923,
+      "grad_norm": 3.53125,
+      "learning_rate": 5.724883194229425e-06,
+      "loss": 4.5273,
+      "step": 518050
+    },
+    {
+      "epoch": 4.4602272727272725,
+      "grad_norm": 2.6875,
+      "learning_rate": 5.715851541511974e-06,
+      "loss": 4.1284,
+      "step": 518100
+    },
+    {
+      "epoch": 4.460657713498622,
+      "grad_norm": 2.90625,
+      "learning_rate": 5.706826809048149e-06,
+      "loss": 3.9941,
+      "step": 518150
+    },
+    {
+      "epoch": 4.461088154269972,
+      "grad_norm": 2.171875,
+      "learning_rate": 5.697808997500376e-06,
+      "loss": 4.1627,
+      "step": 518200
+    },
+    {
+      "epoch": 4.461518595041323,
+      "grad_norm": 3.046875,
+      "learning_rate": 5.688798107530513e-06,
+      "loss": 4.3969,
+      "step": 518250
+    },
+    {
+      "epoch": 4.4619490358126725,
+      "grad_norm": 2.96875,
+      "learning_rate": 5.679794139799965e-06,
+      "loss": 4.7208,
+      "step": 518300
+    },
+    {
+      "epoch": 4.462379476584022,
+      "grad_norm": 3.421875,
+      "learning_rate": 5.670797094969593e-06,
+      "loss": 4.33,
+      "step": 518350
+    },
+    {
+      "epoch": 4.462809917355372,
+      "grad_norm": 2.875,
+      "learning_rate": 5.661806973699735e-06,
+      "loss": 4.2921,
+      "step": 518400
+    },
+    {
+      "epoch": 4.463240358126722,
+      "grad_norm": 1.2734375,
+      "learning_rate": 5.652823776650295e-06,
+      "loss": 4.2804,
+      "step": 518450
+    },
+    {
+      "epoch": 4.463670798898072,
+      "grad_norm": 3.796875,
+      "learning_rate": 5.6438475044805925e-06,
+      "loss": 4.0299,
+      "step": 518500
+    },
+    {
+      "epoch": 4.4641012396694215,
+      "grad_norm": 1.2265625,
+      "learning_rate": 5.634878157849477e-06,
+      "loss": 4.5831,
+      "step": 518550
+    },
+    {
+      "epoch": 4.464531680440771,
+      "grad_norm": 3.859375,
+      "learning_rate": 5.625915737415266e-06,
+      "loss": 4.331,
+      "step": 518600
+    },
+    {
+      "epoch": 4.464962121212121,
+      "grad_norm": 2.65625,
+      "learning_rate": 5.616960243835789e-06,
+      "loss": 4.189,
+      "step": 518650
+    },
+    {
+      "epoch": 4.465392561983471,
+      "grad_norm": 3.15625,
+      "learning_rate": 5.608011677768343e-06,
+      "loss": 4.2639,
+      "step": 518700
+    },
+    {
+      "epoch": 4.465823002754821,
+      "grad_norm": 3.484375,
+      "learning_rate": 5.59907003986977e-06,
+      "loss": 4.5521,
+      "step": 518750
+    },
+    {
+      "epoch": 4.4662534435261705,
+      "grad_norm": 6.03125,
+      "learning_rate": 5.590135330796342e-06,
+      "loss": 4.2766,
+      "step": 518800
+    },
+    {
+      "epoch": 4.46668388429752,
+      "grad_norm": 1.34375,
+      "learning_rate": 5.58120755120386e-06,
+      "loss": 4.2728,
+      "step": 518850
+    },
+    {
+      "epoch": 4.46711432506887,
+      "grad_norm": 0.84375,
+      "learning_rate": 5.572286701747598e-06,
+      "loss": 4.0894,
+      "step": 518900
+    },
+    {
+      "epoch": 4.467544765840221,
+      "grad_norm": 4.625,
+      "learning_rate": 5.5633727830823215e-06,
+      "loss": 4.4455,
+      "step": 518950
+    },
+    {
+      "epoch": 4.4679752066115705,
+      "grad_norm": 2.09375,
+      "learning_rate": 5.554465795862274e-06,
+      "loss": 4.1913,
+      "step": 519000
+    },
+    {
+      "epoch": 4.4679752066115705,
+      "eval_loss": 4.990195274353027,
+      "eval_runtime": 25.0097,
+      "eval_samples_per_second": 25.59,
+      "eval_steps_per_second": 12.795,
+      "eval_tts_loss": 7.48109567921881,
+      "step": 519000
+    },
+    {
+      "epoch": 4.46840564738292,
+      "grad_norm": 3.171875,
+      "learning_rate": 5.545565740741254e-06,
+      "loss": 4.2293,
+      "step": 519050
+    },
+    {
+      "epoch": 4.46883608815427,
+      "grad_norm": 3.40625,
+      "learning_rate": 5.536672618372485e-06,
+      "loss": 4.7192,
+      "step": 519100
+    },
+    {
+      "epoch": 4.46926652892562,
+      "grad_norm": 4.59375,
+      "learning_rate": 5.52778642940871e-06,
+      "loss": 3.7329,
+      "step": 519150
+    },
+    {
+      "epoch": 4.46969696969697,
+      "grad_norm": 2.4375,
+      "learning_rate": 5.518907174502141e-06,
+      "loss": 4.2293,
+      "step": 519200
+    },
+    {
+      "epoch": 4.4701274104683195,
+      "grad_norm": 3.140625,
+      "learning_rate": 5.510034854304491e-06,
+      "loss": 4.3989,
+      "step": 519250
+    },
+    {
+      "epoch": 4.470557851239669,
+      "grad_norm": 3.453125,
+      "learning_rate": 5.501169469466994e-06,
+      "loss": 4.4038,
+      "step": 519300
+    },
+    {
+      "epoch": 4.470988292011019,
+      "grad_norm": 4.90625,
+      "learning_rate": 5.492311020640351e-06,
+      "loss": 4.3744,
+      "step": 519350
+    },
+    {
+      "epoch": 4.471418732782369,
+      "grad_norm": 2.5,
+      "learning_rate": 5.48345950847472e-06,
+      "loss": 4.3404,
+      "step": 519400
+    },
+    {
+      "epoch": 4.471849173553719,
+      "grad_norm": 2.8125,
+      "learning_rate": 5.474614933619815e-06,
+      "loss": 4.1944,
+      "step": 519450
+    },
+    {
+      "epoch": 4.4722796143250685,
+      "grad_norm": 3.59375,
+      "learning_rate": 5.465777296724794e-06,
+      "loss": 4.2664,
+      "step": 519500
+    },
+    {
+      "epoch": 4.472710055096419,
+      "grad_norm": 2.390625,
+      "learning_rate": 5.456946598438306e-06,
+      "loss": 3.9764,
+      "step": 519550
+    },
+    {
+      "epoch": 4.473140495867769,
+      "grad_norm": 6.875,
+      "learning_rate": 5.4481228394085646e-06,
+      "loss": 4.4265,
+      "step": 519600
+    },
+    {
+      "epoch": 4.473570936639119,
+      "grad_norm": 4.625,
+      "learning_rate": 5.439306020283141e-06,
+      "loss": 4.3629,
+      "step": 519650
+    },
+    {
+      "epoch": 4.4740013774104685,
+      "grad_norm": 3.328125,
+      "learning_rate": 5.430496141709207e-06,
+      "loss": 4.652,
+      "step": 519700
+    },
+    {
+      "epoch": 4.474431818181818,
+      "grad_norm": 3.640625,
+      "learning_rate": 5.421693204333389e-06,
+      "loss": 4.3145,
+      "step": 519750
+    },
+    {
+      "epoch": 4.474862258953168,
+      "grad_norm": 4.21875,
+      "learning_rate": 5.412897208801803e-06,
+      "loss": 4.3534,
+      "step": 519800
+    },
+    {
+      "epoch": 4.475292699724518,
+      "grad_norm": 2.796875,
+      "learning_rate": 5.404108155760046e-06,
+      "loss": 4.0033,
+      "step": 519850
+    },
+    {
+      "epoch": 4.475723140495868,
+      "grad_norm": 2.578125,
+      "learning_rate": 5.3953260458532215e-06,
+      "loss": 4.1615,
+      "step": 519900
+    },
+    {
+      "epoch": 4.4761535812672175,
+      "grad_norm": 6.5625,
+      "learning_rate": 5.386550879725915e-06,
+      "loss": 4.3982,
+      "step": 519950
+    },
+    {
+      "epoch": 4.476584022038567,
+      "grad_norm": 3.09375,
+      "learning_rate": 5.3777826580222126e-06,
+      "loss": 4.2962,
+      "step": 520000
+    },
+    {
+      "epoch": 4.477014462809917,
+      "grad_norm": 4.78125,
+      "learning_rate": 5.369021381385686e-06,
+      "loss": 4.2393,
+      "step": 520050
+    },
+    {
+      "epoch": 4.477444903581267,
+      "grad_norm": 2.59375,
+      "learning_rate": 5.360267050459378e-06,
+      "loss": 4.3622,
+      "step": 520100
+    },
+    {
+      "epoch": 4.477875344352617,
+      "grad_norm": 2.8125,
+      "learning_rate": 5.351519665885852e-06,
+      "loss": 4.4195,
+      "step": 520150
+    },
+    {
+      "epoch": 4.4783057851239665,
+      "grad_norm": 2.0,
+      "learning_rate": 5.342779228307138e-06,
+      "loss": 4.4245,
+      "step": 520200
+    },
+    {
+      "epoch": 4.478736225895317,
+      "grad_norm": 2.90625,
+      "learning_rate": 5.334045738364757e-06,
+      "loss": 4.5866,
+      "step": 520250
+    },
+    {
+      "epoch": 4.479166666666667,
+      "grad_norm": 1.0390625,
+      "learning_rate": 5.325319196699752e-06,
+      "loss": 4.5064,
+      "step": 520300
+    },
+    {
+      "epoch": 4.479597107438017,
+      "grad_norm": 3.140625,
+      "learning_rate": 5.316599603952632e-06,
+      "loss": 4.1147,
+      "step": 520350
+    },
+    {
+      "epoch": 4.4800275482093666,
+      "grad_norm": 3.390625,
+      "learning_rate": 5.307886960763353e-06,
+      "loss": 4.499,
+      "step": 520400
+    },
+    {
+      "epoch": 4.480457988980716,
+      "grad_norm": 4.6875,
+      "learning_rate": 5.299181267771458e-06,
+      "loss": 4.1368,
+      "step": 520450
+    },
+    {
+      "epoch": 4.480888429752066,
+      "grad_norm": 3.75,
+      "learning_rate": 5.290482525615903e-06,
+      "loss": 4.385,
+      "step": 520500
+    },
+    {
+      "epoch": 4.481318870523416,
+      "grad_norm": 10.0,
+      "learning_rate": 5.2817907349351435e-06,
+      "loss": 4.1886,
+      "step": 520550
+    },
+    {
+      "epoch": 4.481749311294766,
+      "grad_norm": 4.375,
+      "learning_rate": 5.273105896367181e-06,
+      "loss": 4.3442,
+      "step": 520600
+    },
+    {
+      "epoch": 4.4821797520661155,
+      "grad_norm": 1.9453125,
+      "learning_rate": 5.264428010549405e-06,
+      "loss": 3.8715,
+      "step": 520650
+    },
+    {
+      "epoch": 4.482610192837465,
+      "grad_norm": 4.9375,
+      "learning_rate": 5.255757078118795e-06,
+      "loss": 4.2651,
+      "step": 520700
+    },
+    {
+      "epoch": 4.483040633608815,
+      "grad_norm": 4.5,
+      "learning_rate": 5.247093099711775e-06,
+      "loss": 3.9816,
+      "step": 520750
+    },
+    {
+      "epoch": 4.483471074380165,
+      "grad_norm": 4.125,
+      "learning_rate": 5.238436075964237e-06,
+      "loss": 4.3572,
+      "step": 520800
+    },
+    {
+      "epoch": 4.483901515151516,
+      "grad_norm": 5.3125,
+      "learning_rate": 5.2297860075116164e-06,
+      "loss": 4.1667,
+      "step": 520850
+    },
+    {
+      "epoch": 4.484331955922865,
+      "grad_norm": 2.9375,
+      "learning_rate": 5.221142894988818e-06,
+      "loss": 3.9319,
+      "step": 520900
+    },
+    {
+      "epoch": 4.484762396694215,
+      "grad_norm": 3.609375,
+      "learning_rate": 5.212506739030187e-06,
+      "loss": 4.2496,
+      "step": 520950
+    },
+    {
+      "epoch": 4.485192837465565,
+      "grad_norm": 3.515625,
+      "learning_rate": 5.20387754026962e-06,
+      "loss": 4.4324,
+      "step": 521000
+    },
+    {
+      "epoch": 4.485623278236915,
+      "grad_norm": 3.515625,
+      "learning_rate": 5.195255299340496e-06,
+      "loss": 4.5426,
+      "step": 521050
+    },
+    {
+      "epoch": 4.4860537190082646,
+      "grad_norm": 2.875,
+      "learning_rate": 5.186640016875632e-06,
+      "loss": 4.3715,
+      "step": 521100
+    },
+    {
+      "epoch": 4.486484159779614,
+      "grad_norm": 3.0625,
+      "learning_rate": 5.1780316935074345e-06,
+      "loss": 3.9493,
+      "step": 521150
+    },
+    {
+      "epoch": 4.486914600550964,
+      "grad_norm": 4.65625,
+      "learning_rate": 5.169430329867675e-06,
+      "loss": 3.9616,
+      "step": 521200
+    },
+    {
+      "epoch": 4.487345041322314,
+      "grad_norm": 0.95703125,
+      "learning_rate": 5.1608359265876815e-06,
+      "loss": 4.6169,
+      "step": 521250
+    },
+    {
+      "epoch": 4.487775482093664,
+      "grad_norm": 2.03125,
+      "learning_rate": 5.152248484298294e-06,
+      "loss": 4.55,
+      "step": 521300
+    },
+    {
+      "epoch": 4.4882059228650135,
+      "grad_norm": 2.53125,
+      "learning_rate": 5.143668003629798e-06,
+      "loss": 4.5889,
+      "step": 521350
+    },
+    {
+      "epoch": 4.488636363636363,
+      "grad_norm": 3.203125,
+      "learning_rate": 5.135094485211977e-06,
+      "loss": 4.2215,
+      "step": 521400
+    },
+    {
+      "epoch": 4.489066804407713,
+      "grad_norm": 4.125,
+      "learning_rate": 5.126527929674119e-06,
+      "loss": 4.5214,
+      "step": 521450
+    },
+    {
+      "epoch": 4.489497245179063,
+      "grad_norm": 4.90625,
+      "learning_rate": 5.117968337644974e-06,
+      "loss": 4.1815,
+      "step": 521500
+    },
+    {
+      "epoch": 4.489927685950414,
+      "grad_norm": 2.765625,
+      "learning_rate": 5.109415709752818e-06,
+      "loss": 4.3825,
+      "step": 521550
+    },
+    {
+      "epoch": 4.490358126721763,
+      "grad_norm": 3.3125,
+      "learning_rate": 5.100870046625384e-06,
+      "loss": 4.3089,
+      "step": 521600
+    },
+    {
+      "epoch": 4.490788567493113,
+      "grad_norm": 3.796875,
+      "learning_rate": 5.092331348889912e-06,
+      "loss": 4.0753,
+      "step": 521650
+    },
+    {
+      "epoch": 4.491219008264463,
+      "grad_norm": 4.0625,
+      "learning_rate": 5.0837996171731235e-06,
+      "loss": 4.1591,
+      "step": 521700
+    },
+    {
+      "epoch": 4.491649449035813,
+      "grad_norm": 3.234375,
+      "learning_rate": 5.07527485210122e-06,
+      "loss": 4.533,
+      "step": 521750
+    },
+    {
+      "epoch": 4.4920798898071626,
+      "grad_norm": 3.34375,
+      "learning_rate": 5.066757054299898e-06,
+      "loss": 4.6478,
+      "step": 521800
+    },
+    {
+      "epoch": 4.492510330578512,
+      "grad_norm": 2.71875,
+      "learning_rate": 5.058246224394358e-06,
+      "loss": 4.3133,
+      "step": 521850
+    },
+    {
+      "epoch": 4.492940771349862,
+      "grad_norm": 3.03125,
+      "learning_rate": 5.049742363009291e-06,
+      "loss": 4.0999,
+      "step": 521900
+    },
+    {
+      "epoch": 4.493371212121212,
+      "grad_norm": 2.15625,
+      "learning_rate": 5.041245470768818e-06,
+      "loss": 4.4241,
+      "step": 521950
+    },
+    {
+      "epoch": 4.493801652892562,
+      "grad_norm": 5.40625,
+      "learning_rate": 5.0327555482966305e-06,
+      "loss": 4.5213,
+      "step": 522000
+    },
+    {
+      "epoch": 4.493801652892562,
+      "eval_loss": 4.990315914154053,
+      "eval_runtime": 24.6409,
+      "eval_samples_per_second": 25.973,
+      "eval_steps_per_second": 12.987,
+      "eval_tts_loss": 7.481205293532995,
+      "step": 522000
+    },
+    {
+      "epoch": 4.4942320936639115,
+      "grad_norm": 4.46875,
+      "learning_rate": 5.024272596215862e-06,
+      "loss": 4.48,
+      "step": 522050
+    },
+    {
+      "epoch": 4.494662534435261,
+      "grad_norm": 2.296875,
+      "learning_rate": 5.0157966151491265e-06,
+      "loss": 4.5593,
+      "step": 522100
+    },
+    {
+      "epoch": 4.495092975206612,
+      "grad_norm": 3.09375,
+      "learning_rate": 5.007327605718582e-06,
+      "loss": 4.3271,
+      "step": 522150
+    },
+    {
+      "epoch": 4.495523415977962,
+      "grad_norm": 3.40625,
+      "learning_rate": 4.998865568545774e-06,
+      "loss": 3.9936,
+      "step": 522200
+    },
+    {
+      "epoch": 4.495953856749312,
+      "grad_norm": 3.375,
+      "learning_rate": 4.990410504251863e-06,
+      "loss": 4.3553,
+      "step": 522250
+    },
+    {
+      "epoch": 4.496384297520661,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.981962413457386e-06,
+      "loss": 4.653,
+      "step": 522300
+    },
+    {
+      "epoch": 4.496814738292011,
+      "grad_norm": 3.953125,
+      "learning_rate": 4.973521296782435e-06,
+      "loss": 4.3257,
+      "step": 522350
+    },
+    {
+      "epoch": 4.497245179063361,
+      "grad_norm": 4.21875,
+      "learning_rate": 4.96508715484657e-06,
+      "loss": 4.2562,
+      "step": 522400
+    },
+    {
+      "epoch": 4.497675619834711,
+      "grad_norm": 3.0625,
+      "learning_rate": 4.956659988268841e-06,
+      "loss": 4.2002,
+      "step": 522450
+    },
+    {
+      "epoch": 4.4981060606060606,
+      "grad_norm": 3.4375,
+      "learning_rate": 4.948239797667753e-06,
+      "loss": 4.2381,
+      "step": 522500
+    },
+    {
+      "epoch": 4.49853650137741,
+      "grad_norm": 2.203125,
+      "learning_rate": 4.939826583661378e-06,
+      "loss": 4.2369,
+      "step": 522550
+    },
+    {
+      "epoch": 4.49896694214876,
+      "grad_norm": 6.96875,
+      "learning_rate": 4.931420346867188e-06,
+      "loss": 4.3375,
+      "step": 522600
+    },
+    {
+      "epoch": 4.49939738292011,
+      "grad_norm": 3.671875,
+      "learning_rate": 4.923021087902202e-06,
+      "loss": 4.5467,
+      "step": 522650
+    },
+    {
+      "epoch": 4.49982782369146,
+      "grad_norm": 3.109375,
+      "learning_rate": 4.914628807382926e-06,
+      "loss": 4.049,
+      "step": 522700
+    },
+    {
+      "epoch": 4.5002582644628095,
+      "grad_norm": 3.140625,
+      "learning_rate": 4.906243505925301e-06,
+      "loss": 4.6011,
+      "step": 522750
+    },
+    {
+      "epoch": 4.500688705234159,
+      "grad_norm": 2.3125,
+      "learning_rate": 4.89786518414479e-06,
+      "loss": 4.0835,
+      "step": 522800
+    },
+    {
+      "epoch": 4.50111914600551,
+      "grad_norm": 2.890625,
+      "learning_rate": 4.889493842656367e-06,
+      "loss": 4.1293,
+      "step": 522850
+    },
+    {
+      "epoch": 4.50154958677686,
+      "grad_norm": 5.6875,
+      "learning_rate": 4.881129482074465e-06,
+      "loss": 4.5677,
+      "step": 522900
+    },
+    {
+      "epoch": 4.50198002754821,
+      "grad_norm": 2.375,
+      "learning_rate": 4.872772103013001e-06,
+      "loss": 3.8966,
+      "step": 522950
+    },
+    {
+      "epoch": 4.502410468319559,
+      "grad_norm": 2.328125,
+      "learning_rate": 4.864421706085398e-06,
+      "loss": 4.9562,
+      "step": 523000
+    },
+    {
+      "epoch": 4.502840909090909,
+      "grad_norm": 3.125,
+      "learning_rate": 4.856078291904542e-06,
+      "loss": 4.3891,
+      "step": 523050
+    },
+    {
+      "epoch": 4.503271349862259,
+      "grad_norm": 2.9375,
+      "learning_rate": 4.8477418610828325e-06,
+      "loss": 4.5279,
+      "step": 523100
+    },
+    {
+      "epoch": 4.503701790633609,
+      "grad_norm": 3.3125,
+      "learning_rate": 4.839412414232158e-06,
+      "loss": 4.1265,
+      "step": 523150
+    },
+    {
+      "epoch": 4.5041322314049586,
+      "grad_norm": 2.890625,
+      "learning_rate": 4.831089951963863e-06,
+      "loss": 4.112,
+      "step": 523200
+    },
+    {
+      "epoch": 4.504562672176308,
+      "grad_norm": 3.21875,
+      "learning_rate": 4.822774474888803e-06,
+      "loss": 4.0946,
+      "step": 523250
+    },
+    {
+      "epoch": 4.504993112947658,
+      "grad_norm": 5.125,
+      "learning_rate": 4.814465983617322e-06,
+      "loss": 4.2552,
+      "step": 523300
+    },
+    {
+      "epoch": 4.505423553719008,
+      "grad_norm": 5.5,
+      "learning_rate": 4.806164478759223e-06,
+      "loss": 4.4739,
+      "step": 523350
+    },
+    {
+      "epoch": 4.505853994490359,
+      "grad_norm": 3.40625,
+      "learning_rate": 4.797869960923851e-06,
+      "loss": 4.2649,
+      "step": 523400
+    },
+    {
+      "epoch": 4.506284435261708,
+      "grad_norm": 3.859375,
+      "learning_rate": 4.7895824307200075e-06,
+      "loss": 4.1929,
+      "step": 523450
+    },
+    {
+      "epoch": 4.506714876033058,
+      "grad_norm": 2.8125,
+      "learning_rate": 4.78130188875594e-06,
+      "loss": 4.3796,
+      "step": 523500
+    },
+    {
+      "epoch": 4.507145316804408,
+      "grad_norm": 2.265625,
+      "learning_rate": 4.773028335639451e-06,
+      "loss": 4.2762,
+      "step": 523550
+    },
+    {
+      "epoch": 4.507575757575758,
+      "grad_norm": 1.5,
+      "learning_rate": 4.7647617719778e-06,
+      "loss": 4.6251,
+      "step": 523600
+    },
+    {
+      "epoch": 4.508006198347108,
+      "grad_norm": 4.75,
+      "learning_rate": 4.7565021983777105e-06,
+      "loss": 4.4283,
+      "step": 523650
+    },
+    {
+      "epoch": 4.508436639118457,
+      "grad_norm": 1.9765625,
+      "learning_rate": 4.748249615445477e-06,
+      "loss": 3.6475,
+      "step": 523700
+    },
+    {
+      "epoch": 4.508867079889807,
+      "grad_norm": 1.3984375,
+      "learning_rate": 4.740004023786748e-06,
+      "loss": 4.4065,
+      "step": 523750
+    },
+    {
+      "epoch": 4.509297520661157,
+      "grad_norm": 1.4765625,
+      "learning_rate": 4.731765424006773e-06,
+      "loss": 4.0777,
+      "step": 523800
+    },
+    {
+      "epoch": 4.509727961432507,
+      "grad_norm": 2.171875,
+      "learning_rate": 4.723533816710257e-06,
+      "loss": 4.337,
+      "step": 523850
+    },
+    {
+      "epoch": 4.5101584022038566,
+      "grad_norm": 2.046875,
+      "learning_rate": 4.71530920250135e-06,
+      "loss": 4.0588,
+      "step": 523900
+    },
+    {
+      "epoch": 4.510588842975206,
+      "grad_norm": 2.734375,
+      "learning_rate": 4.707091581983747e-06,
+      "loss": 4.8354,
+      "step": 523950
+    },
+    {
+      "epoch": 4.511019283746556,
+      "grad_norm": 2.640625,
+      "learning_rate": 4.698880955760598e-06,
+      "loss": 4.1805,
+      "step": 524000
+    },
+    {
+      "epoch": 4.511449724517906,
+      "grad_norm": 1.6015625,
+      "learning_rate": 4.690677324434511e-06,
+      "loss": 4.2731,
+      "step": 524050
+    },
+    {
+      "epoch": 4.511880165289256,
+      "grad_norm": 3.421875,
+      "learning_rate": 4.6824806886076715e-06,
+      "loss": 4.5869,
+      "step": 524100
+    },
+    {
+      "epoch": 4.512310606060606,
+      "grad_norm": 2.296875,
+      "learning_rate": 4.674291048881674e-06,
+      "loss": 4.1706,
+      "step": 524150
+    },
+    {
+      "epoch": 4.512741046831956,
+      "grad_norm": 2.1875,
+      "learning_rate": 4.666108405857606e-06,
+      "loss": 4.1391,
+      "step": 524200
+    },
+    {
+      "epoch": 4.513171487603306,
+      "grad_norm": 4.15625,
+      "learning_rate": 4.657932760136063e-06,
+      "loss": 4.2125,
+      "step": 524250
+    },
+    {
+      "epoch": 4.513601928374656,
+      "grad_norm": 2.953125,
+      "learning_rate": 4.649764112317123e-06,
+      "loss": 4.2078,
+      "step": 524300
+    },
+    {
+      "epoch": 4.514032369146006,
+      "grad_norm": 3.578125,
+      "learning_rate": 4.641602463000339e-06,
+      "loss": 4.1623,
+      "step": 524350
+    },
+    {
+      "epoch": 4.514462809917355,
+      "grad_norm": 3.75,
+      "learning_rate": 4.633447812784775e-06,
+      "loss": 4.2372,
+      "step": 524400
+    },
+    {
+      "epoch": 4.514893250688705,
+      "grad_norm": 3.0625,
+      "learning_rate": 4.625300162268953e-06,
+      "loss": 4.406,
+      "step": 524450
+    },
+    {
+      "epoch": 4.515323691460055,
+      "grad_norm": 6.5,
+      "learning_rate": 4.617159512050895e-06,
+      "loss": 4.2278,
+      "step": 524500
+    },
+    {
+      "epoch": 4.515754132231405,
+      "grad_norm": 3.703125,
+      "learning_rate": 4.6090258627281115e-06,
+      "loss": 4.6562,
+      "step": 524550
+    },
+    {
+      "epoch": 4.5161845730027546,
+      "grad_norm": 6.875,
+      "learning_rate": 4.600899214897581e-06,
+      "loss": 4.3026,
+      "step": 524600
+    },
+    {
+      "epoch": 4.516615013774104,
+      "grad_norm": 3.734375,
+      "learning_rate": 4.592779569155792e-06,
+      "loss": 4.4567,
+      "step": 524650
+    },
+    {
+      "epoch": 4.517045454545455,
+      "grad_norm": 2.65625,
+      "learning_rate": 4.584666926098713e-06,
+      "loss": 4.2037,
+      "step": 524700
+    },
+    {
+      "epoch": 4.517475895316805,
+      "grad_norm": 4.15625,
+      "learning_rate": 4.57656128632179e-06,
+      "loss": 4.385,
+      "step": 524750
+    },
+    {
+      "epoch": 4.517906336088155,
+      "grad_norm": 5.15625,
+      "learning_rate": 4.568462650419958e-06,
+      "loss": 4.381,
+      "step": 524800
+    },
+    {
+      "epoch": 4.518336776859504,
+      "grad_norm": 2.3125,
+      "learning_rate": 4.560371018987652e-06,
+      "loss": 4.2931,
+      "step": 524850
+    },
+    {
+      "epoch": 4.518767217630854,
+      "grad_norm": 2.984375,
+      "learning_rate": 4.552286392618743e-06,
+      "loss": 4.4834,
+      "step": 524900
+    },
+    {
+      "epoch": 4.519197658402204,
+      "grad_norm": 5.4375,
+      "learning_rate": 4.544208771906677e-06,
+      "loss": 4.1684,
+      "step": 524950
+    },
+    {
+      "epoch": 4.519628099173554,
+      "grad_norm": 3.484375,
+      "learning_rate": 4.536138157444314e-06,
+      "loss": 4.1747,
+      "step": 525000
+    },
+    {
+      "epoch": 4.519628099173554,
+      "eval_loss": 4.99021053314209,
+      "eval_runtime": 24.1038,
+      "eval_samples_per_second": 26.552,
+      "eval_steps_per_second": 13.276,
+      "eval_tts_loss": 7.480371839671666,
+      "step": 525000
+    },
+    {
+      "epoch": 4.520058539944904,
+      "grad_norm": 3.5,
+      "learning_rate": 4.528074549823991e-06,
+      "loss": 4.3075,
+      "step": 525050
+    },
+    {
+      "epoch": 4.520488980716253,
+      "grad_norm": 4.34375,
+      "learning_rate": 4.5200179496376e-06,
+      "loss": 4.2856,
+      "step": 525100
+    },
+    {
+      "epoch": 4.520919421487603,
+      "grad_norm": 2.375,
+      "learning_rate": 4.511968357476448e-06,
+      "loss": 4.3544,
+      "step": 525150
+    },
+    {
+      "epoch": 4.521349862258953,
+      "grad_norm": 1.953125,
+      "learning_rate": 4.5039257739313724e-06,
+      "loss": 4.3256,
+      "step": 525200
+    },
+    {
+      "epoch": 4.521780303030303,
+      "grad_norm": 1.1484375,
+      "learning_rate": 4.4958901995927004e-06,
+      "loss": 4.2391,
+      "step": 525250
+    },
+    {
+      "epoch": 4.5222107438016526,
+      "grad_norm": 4.875,
+      "learning_rate": 4.487861635050172e-06,
+      "loss": 4.1394,
+      "step": 525300
+    },
+    {
+      "epoch": 4.522641184573002,
+      "grad_norm": 3.671875,
+      "learning_rate": 4.479840080893127e-06,
+      "loss": 4.0337,
+      "step": 525350
+    },
+    {
+      "epoch": 4.523071625344352,
+      "grad_norm": 4.71875,
+      "learning_rate": 4.471825537710295e-06,
+      "loss": 4.1654,
+      "step": 525400
+    },
+    {
+      "epoch": 4.523502066115703,
+      "grad_norm": 3.03125,
+      "learning_rate": 4.463818006089937e-06,
+      "loss": 4.1178,
+      "step": 525450
+    },
+    {
+      "epoch": 4.523932506887053,
+      "grad_norm": 3.515625,
+      "learning_rate": 4.455817486619784e-06,
+      "loss": 4.0843,
+      "step": 525500
+    },
+    {
+      "epoch": 4.524362947658402,
+      "grad_norm": 2.09375,
+      "learning_rate": 4.447823979887056e-06,
+      "loss": 4.2138,
+      "step": 525550
+    },
+    {
+      "epoch": 4.524793388429752,
+      "grad_norm": 2.125,
+      "learning_rate": 4.439837486478448e-06,
+      "loss": 4.466,
+      "step": 525600
+    },
+    {
+      "epoch": 4.525223829201102,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.431858006980183e-06,
+      "loss": 4.3179,
+      "step": 525650
+    },
+    {
+      "epoch": 4.525654269972452,
+      "grad_norm": 7.375,
+      "learning_rate": 4.423885541977923e-06,
+      "loss": 4.0016,
+      "step": 525700
+    },
+    {
+      "epoch": 4.526084710743802,
+      "grad_norm": 5.0625,
+      "learning_rate": 4.415920092056824e-06,
+      "loss": 4.7607,
+      "step": 525750
+    },
+    {
+      "epoch": 4.526515151515151,
+      "grad_norm": 3.296875,
+      "learning_rate": 4.40796165780153e-06,
+      "loss": 4.3584,
+      "step": 525800
+    },
+    {
+      "epoch": 4.526945592286501,
+      "grad_norm": 2.390625,
+      "learning_rate": 4.400010239796193e-06,
+      "loss": 4.1827,
+      "step": 525850
+    },
+    {
+      "epoch": 4.527376033057851,
+      "grad_norm": 4.40625,
+      "learning_rate": 4.392065838624393e-06,
+      "loss": 4.6886,
+      "step": 525900
+    },
+    {
+      "epoch": 4.527806473829201,
+      "grad_norm": 4.15625,
+      "learning_rate": 4.384128454869263e-06,
+      "loss": 4.6378,
+      "step": 525950
+    },
+    {
+      "epoch": 4.528236914600551,
+      "grad_norm": 3.0625,
+      "learning_rate": 4.376198089113392e-06,
+      "loss": 4.4241,
+      "step": 526000
+    },
+    {
+      "epoch": 4.528667355371901,
+      "grad_norm": 3.140625,
+      "learning_rate": 4.3682747419388385e-06,
+      "loss": 4.1943,
+      "step": 526050
+    },
+    {
+      "epoch": 4.529097796143251,
+      "grad_norm": 4.84375,
+      "learning_rate": 4.360358413927157e-06,
+      "loss": 4.5571,
+      "step": 526100
+    },
+    {
+      "epoch": 4.529528236914601,
+      "grad_norm": 2.125,
+      "learning_rate": 4.3524491056593954e-06,
+      "loss": 4.146,
+      "step": 526150
+    },
+    {
+      "epoch": 4.529958677685951,
+      "grad_norm": 1.0390625,
+      "learning_rate": 4.344546817716067e-06,
+      "loss": 4.4104,
+      "step": 526200
+    },
+    {
+      "epoch": 4.5303891184573,
+      "grad_norm": 1.7265625,
+      "learning_rate": 4.336651550677217e-06,
+      "loss": 4.6521,
+      "step": 526250
+    },
+    {
+      "epoch": 4.53081955922865,
+      "grad_norm": 1.953125,
+      "learning_rate": 4.3287633051222965e-06,
+      "loss": 4.1393,
+      "step": 526300
+    },
+    {
+      "epoch": 4.53125,
+      "grad_norm": 3.5,
+      "learning_rate": 4.320882081630317e-06,
+      "loss": 4.0434,
+      "step": 526350
+    },
+    {
+      "epoch": 4.53168044077135,
+      "grad_norm": 3.546875,
+      "learning_rate": 4.31300788077974e-06,
+      "loss": 4.302,
+      "step": 526400
+    },
+    {
+      "epoch": 4.5321108815427,
+      "grad_norm": 3.53125,
+      "learning_rate": 4.305140703148491e-06,
+      "loss": 4.4763,
+      "step": 526450
+    },
+    {
+      "epoch": 4.532541322314049,
+      "grad_norm": 1.3515625,
+      "learning_rate": 4.2972805493140315e-06,
+      "loss": 4.5437,
+      "step": 526500
+    },
+    {
+      "epoch": 4.532971763085399,
+      "grad_norm": 3.15625,
+      "learning_rate": 4.289427419853287e-06,
+      "loss": 4.1178,
+      "step": 526550
+    },
+    {
+      "epoch": 4.533402203856749,
+      "grad_norm": 5.5,
+      "learning_rate": 4.28158131534262e-06,
+      "loss": 4.2925,
+      "step": 526600
+    },
+    {
+      "epoch": 4.533832644628099,
+      "grad_norm": 3.609375,
+      "learning_rate": 4.273742236357947e-06,
+      "loss": 4.2234,
+      "step": 526650
+    },
+    {
+      "epoch": 4.534263085399449,
+      "grad_norm": 2.53125,
+      "learning_rate": 4.265910183474642e-06,
+      "loss": 4.0474,
+      "step": 526700
+    },
+    {
+      "epoch": 4.534693526170799,
+      "grad_norm": 2.65625,
+      "learning_rate": 4.258085157267533e-06,
+      "loss": 4.4759,
+      "step": 526750
+    },
+    {
+      "epoch": 4.535123966942149,
+      "grad_norm": 3.4375,
+      "learning_rate": 4.250267158311005e-06,
+      "loss": 4.3425,
+      "step": 526800
+    },
+    {
+      "epoch": 4.535554407713499,
+      "grad_norm": 3.515625,
+      "learning_rate": 4.242456187178845e-06,
+      "loss": 4.5469,
+      "step": 526850
+    },
+    {
+      "epoch": 4.535984848484849,
+      "grad_norm": 2.796875,
+      "learning_rate": 4.234652244444359e-06,
+      "loss": 4.3049,
+      "step": 526900
+    },
+    {
+      "epoch": 4.536415289256198,
+      "grad_norm": 2.828125,
+      "learning_rate": 4.22685533068038e-06,
+      "loss": 4.1405,
+      "step": 526950
+    },
+    {
+      "epoch": 4.536845730027548,
+      "grad_norm": 3.171875,
+      "learning_rate": 4.219065446459136e-06,
+      "loss": 4.2781,
+      "step": 527000
+    },
+    {
+      "epoch": 4.537276170798898,
+      "grad_norm": 2.6875,
+      "learning_rate": 4.211282592352428e-06,
+      "loss": 4.5228,
+      "step": 527050
+    },
+    {
+      "epoch": 4.537706611570248,
+      "grad_norm": 3.09375,
+      "learning_rate": 4.203506768931476e-06,
+      "loss": 4.2064,
+      "step": 527100
+    },
+    {
+      "epoch": 4.538137052341598,
+      "grad_norm": 1.953125,
+      "learning_rate": 4.195737976767e-06,
+      "loss": 4.243,
+      "step": 527150
+    },
+    {
+      "epoch": 4.538567493112947,
+      "grad_norm": 3.3125,
+      "learning_rate": 4.1879762164292455e-06,
+      "loss": 4.3149,
+      "step": 527200
+    },
+    {
+      "epoch": 4.538997933884297,
+      "grad_norm": 3.03125,
+      "learning_rate": 4.1802214884878895e-06,
+      "loss": 4.4569,
+      "step": 527250
+    },
+    {
+      "epoch": 4.539428374655648,
+      "grad_norm": 4.71875,
+      "learning_rate": 4.172473793512122e-06,
+      "loss": 4.3919,
+      "step": 527300
+    },
+    {
+      "epoch": 4.539858815426998,
+      "grad_norm": 4.71875,
+      "learning_rate": 4.164733132070586e-06,
+      "loss": 4.926,
+      "step": 527350
+    },
+    {
+      "epoch": 4.5402892561983474,
+      "grad_norm": 2.859375,
+      "learning_rate": 4.156999504731451e-06,
+      "loss": 4.1797,
+      "step": 527400
+    },
+    {
+      "epoch": 4.540719696969697,
+      "grad_norm": 3.453125,
+      "learning_rate": 4.14927291206233e-06,
+      "loss": 4.5263,
+      "step": 527450
+    },
+    {
+      "epoch": 4.541150137741047,
+      "grad_norm": 4.34375,
+      "learning_rate": 4.141553354630367e-06,
+      "loss": 4.1936,
+      "step": 527500
+    },
+    {
+      "epoch": 4.541580578512397,
+      "grad_norm": 2.96875,
+      "learning_rate": 4.133840833002134e-06,
+      "loss": 3.9898,
+      "step": 527550
+    },
+    {
+      "epoch": 4.542011019283747,
+      "grad_norm": 1.8515625,
+      "learning_rate": 4.12613534774372e-06,
+      "loss": 4.4491,
+      "step": 527600
+    },
+    {
+      "epoch": 4.542441460055096,
+      "grad_norm": 2.390625,
+      "learning_rate": 4.118436899420697e-06,
+      "loss": 4.3879,
+      "step": 527650
+    },
+    {
+      "epoch": 4.542871900826446,
+      "grad_norm": 0.953125,
+      "learning_rate": 4.110745488598111e-06,
+      "loss": 4.3766,
+      "step": 527700
+    },
+    {
+      "epoch": 4.543302341597796,
+      "grad_norm": 3.03125,
+      "learning_rate": 4.10306111584049e-06,
+      "loss": 4.4666,
+      "step": 527750
+    },
+    {
+      "epoch": 4.543732782369146,
+      "grad_norm": 4.1875,
+      "learning_rate": 4.09538378171187e-06,
+      "loss": 4.2749,
+      "step": 527800
+    },
+    {
+      "epoch": 4.544163223140496,
+      "grad_norm": 4.59375,
+      "learning_rate": 4.087713486775713e-06,
+      "loss": 4.4894,
+      "step": 527850
+    },
+    {
+      "epoch": 4.544593663911845,
+      "grad_norm": 3.515625,
+      "learning_rate": 4.080050231595045e-06,
+      "loss": 4.5309,
+      "step": 527900
+    },
+    {
+      "epoch": 4.545024104683195,
+      "grad_norm": 2.640625,
+      "learning_rate": 4.072394016732306e-06,
+      "loss": 4.1269,
+      "step": 527950
+    },
+    {
+      "epoch": 4.545454545454545,
+      "grad_norm": 3.828125,
+      "learning_rate": 4.064744842749446e-06,
+      "loss": 4.1829,
+      "step": 528000
+    },
+    {
+      "epoch": 4.545454545454545,
+      "eval_loss": 4.990324974060059,
+      "eval_runtime": 24.3963,
+      "eval_samples_per_second": 26.234,
+      "eval_steps_per_second": 13.117,
+      "eval_tts_loss": 7.480024797597445,
+      "step": 528000
+    },
+    {
+      "epoch": 4.545884986225896,
+      "grad_norm": 3.4375,
+      "learning_rate": 4.0571027102079275e-06,
+      "loss": 4.5549,
+      "step": 528050
+    },
+    {
+      "epoch": 4.5463154269972454,
+      "grad_norm": 5.09375,
+      "learning_rate": 4.0494676196686236e-06,
+      "loss": 4.2084,
+      "step": 528100
+    },
+    {
+      "epoch": 4.546745867768595,
+      "grad_norm": 2.515625,
+      "learning_rate": 4.041839571691952e-06,
+      "loss": 4.2983,
+      "step": 528150
+    },
+    {
+      "epoch": 4.547176308539945,
+      "grad_norm": 3.359375,
+      "learning_rate": 4.0342185668377995e-06,
+      "loss": 4.6226,
+      "step": 528200
+    },
+    {
+      "epoch": 4.547606749311295,
+      "grad_norm": 1.6796875,
+      "learning_rate": 4.026604605665529e-06,
+      "loss": 4.3992,
+      "step": 528250
+    },
+    {
+      "epoch": 4.548037190082645,
+      "grad_norm": 3.140625,
+      "learning_rate": 4.018997688733983e-06,
+      "loss": 4.6786,
+      "step": 528300
+    },
+    {
+      "epoch": 4.548467630853994,
+      "grad_norm": 4.65625,
+      "learning_rate": 4.0113978166015255e-06,
+      "loss": 4.5706,
+      "step": 528350
+    },
+    {
+      "epoch": 4.548898071625344,
+      "grad_norm": 3.78125,
+      "learning_rate": 4.003804989825921e-06,
+      "loss": 3.9249,
+      "step": 528400
+    },
+    {
+      "epoch": 4.549328512396694,
+      "grad_norm": 2.8125,
+      "learning_rate": 3.99621920896448e-06,
+      "loss": 4.002,
+      "step": 528450
+    },
+    {
+      "epoch": 4.549758953168044,
+      "grad_norm": 3.375,
+      "learning_rate": 3.9886404745740125e-06,
+      "loss": 3.8839,
+      "step": 528500
+    },
+    {
+      "epoch": 4.550189393939394,
+      "grad_norm": 2.0625,
+      "learning_rate": 3.981068787210751e-06,
+      "loss": 4.2406,
+      "step": 528550
+    },
+    {
+      "epoch": 4.550619834710744,
+      "grad_norm": 2.53125,
+      "learning_rate": 3.973504147430451e-06,
+      "loss": 4.4745,
+      "step": 528600
+    },
+    {
+      "epoch": 4.551050275482094,
+      "grad_norm": 4.46875,
+      "learning_rate": 3.965946555788336e-06,
+      "loss": 4.4182,
+      "step": 528650
+    },
+    {
+      "epoch": 4.551480716253444,
+      "grad_norm": 2.203125,
+      "learning_rate": 3.958396012839116e-06,
+      "loss": 4.3205,
+      "step": 528700
+    },
+    {
+      "epoch": 4.551911157024794,
+      "grad_norm": 3.0625,
+      "learning_rate": 3.950852519137005e-06,
+      "loss": 4.4146,
+      "step": 528750
+    },
+    {
+      "epoch": 4.5523415977961434,
+      "grad_norm": 5.59375,
+      "learning_rate": 3.943316075235659e-06,
+      "loss": 4.5433,
+      "step": 528800
+    },
+    {
+      "epoch": 4.552772038567493,
+      "grad_norm": 3.734375,
+      "learning_rate": 3.935786681688247e-06,
+      "loss": 4.9725,
+      "step": 528850
+    },
+    {
+      "epoch": 4.553202479338843,
+      "grad_norm": 2.96875,
+      "learning_rate": 3.928264339047394e-06,
+      "loss": 4.2192,
+      "step": 528900
+    },
+    {
+      "epoch": 4.553632920110193,
+      "grad_norm": 3.59375,
+      "learning_rate": 3.920749047865247e-06,
+      "loss": 4.189,
+      "step": 528950
+    },
+    {
+      "epoch": 4.554063360881543,
+      "grad_norm": 1.9453125,
+      "learning_rate": 3.913240808693375e-06,
+      "loss": 4.0661,
+      "step": 529000
+    },
+    {
+      "epoch": 4.554493801652892,
+      "grad_norm": 5.53125,
+      "learning_rate": 3.905739622082916e-06,
+      "loss": 4.4467,
+      "step": 529050
+    },
+    {
+      "epoch": 4.554924242424242,
+      "grad_norm": 3.84375,
+      "learning_rate": 3.898245488584418e-06,
+      "loss": 4.4869,
+      "step": 529100
+    },
+    {
+      "epoch": 4.555354683195592,
+      "grad_norm": 3.859375,
+      "learning_rate": 3.890758408747919e-06,
+      "loss": 4.2002,
+      "step": 529150
+    },
+    {
+      "epoch": 4.555785123966942,
+      "grad_norm": 1.921875,
+      "learning_rate": 3.883278383122979e-06,
+      "loss": 4.3617,
+      "step": 529200
+    },
+    {
+      "epoch": 4.556215564738292,
+      "grad_norm": 5.125,
+      "learning_rate": 3.875805412258604e-06,
+      "loss": 3.9994,
+      "step": 529250
+    },
+    {
+      "epoch": 4.556646005509641,
+      "grad_norm": 2.015625,
+      "learning_rate": 3.868339496703277e-06,
+      "loss": 3.85,
+      "step": 529300
+    },
+    {
+      "epoch": 4.557076446280992,
+      "grad_norm": 6.59375,
+      "learning_rate": 3.860880637005027e-06,
+      "loss": 4.3406,
+      "step": 529350
+    },
+    {
+      "epoch": 4.557506887052342,
+      "grad_norm": 3.09375,
+      "learning_rate": 3.85342883371127e-06,
+      "loss": 4.1945,
+      "step": 529400
+    },
+    {
+      "epoch": 4.557937327823692,
+      "grad_norm": 4.71875,
+      "learning_rate": 3.84598408736897e-06,
+      "loss": 4.6027,
+      "step": 529450
+    },
+    {
+      "epoch": 4.5583677685950414,
+      "grad_norm": 3.296875,
+      "learning_rate": 3.838546398524567e-06,
+      "loss": 4.1592,
+      "step": 529500
+    },
+    {
+      "epoch": 4.558798209366391,
+      "grad_norm": 3.765625,
+      "learning_rate": 3.831115767723959e-06,
+      "loss": 4.2978,
+      "step": 529550
+    },
+    {
+      "epoch": 4.559228650137741,
+      "grad_norm": 4.1875,
+      "learning_rate": 3.823692195512552e-06,
+      "loss": 4.3131,
+      "step": 529600
+    },
+    {
+      "epoch": 4.559659090909091,
+      "grad_norm": 4.25,
+      "learning_rate": 3.8162756824351995e-06,
+      "loss": 4.3137,
+      "step": 529650
+    },
+    {
+      "epoch": 4.560089531680441,
+      "grad_norm": 2.625,
+      "learning_rate": 3.8088662290362563e-06,
+      "loss": 4.0229,
+      "step": 529700
+    },
+    {
+      "epoch": 4.56051997245179,
+      "grad_norm": 4.1875,
+      "learning_rate": 3.8014638358595865e-06,
+      "loss": 4.444,
+      "step": 529750
+    },
+    {
+      "epoch": 4.56095041322314,
+      "grad_norm": 2.84375,
+      "learning_rate": 3.794068503448489e-06,
+      "loss": 3.9654,
+      "step": 529800
+    },
+    {
+      "epoch": 4.56138085399449,
+      "grad_norm": 2.921875,
+      "learning_rate": 3.786680232345774e-06,
+      "loss": 4.5227,
+      "step": 529850
+    },
+    {
+      "epoch": 4.561811294765841,
+      "grad_norm": 3.828125,
+      "learning_rate": 3.7792990230937186e-06,
+      "loss": 4.4995,
+      "step": 529900
+    },
+    {
+      "epoch": 4.5622417355371905,
+      "grad_norm": 2.9375,
+      "learning_rate": 3.7719248762340897e-06,
+      "loss": 4.3621,
+      "step": 529950
+    },
+    {
+      "epoch": 4.56267217630854,
+      "grad_norm": 3.453125,
+      "learning_rate": 3.764557792308132e-06,
+      "loss": 4.357,
+      "step": 530000
+    },
+    {
+      "epoch": 4.56310261707989,
+      "grad_norm": 3.34375,
+      "learning_rate": 3.7571977718565797e-06,
+      "loss": 4.2724,
+      "step": 530050
+    },
+    {
+      "epoch": 4.56353305785124,
+      "grad_norm": 5.5,
+      "learning_rate": 3.7498448154196452e-06,
+      "loss": 4.7383,
+      "step": 530100
+    },
+    {
+      "epoch": 4.56396349862259,
+      "grad_norm": 3.3125,
+      "learning_rate": 3.7424989235370076e-06,
+      "loss": 4.4627,
+      "step": 530150
+    },
+    {
+      "epoch": 4.5643939393939394,
+      "grad_norm": 2.734375,
+      "learning_rate": 3.7351600967478474e-06,
+      "loss": 4.4491,
+      "step": 530200
+    },
+    {
+      "epoch": 4.564824380165289,
+      "grad_norm": 2.734375,
+      "learning_rate": 3.7278283355908107e-06,
+      "loss": 4.312,
+      "step": 530250
+    },
+    {
+      "epoch": 4.565254820936639,
+      "grad_norm": 2.21875,
+      "learning_rate": 3.720503640604045e-06,
+      "loss": 4.1889,
+      "step": 530300
+    },
+    {
+      "epoch": 4.565685261707989,
+      "grad_norm": 3.203125,
+      "learning_rate": 3.713186012325165e-06,
+      "loss": 4.3733,
+      "step": 530350
+    },
+    {
+      "epoch": 4.566115702479339,
+      "grad_norm": 3.359375,
+      "learning_rate": 3.705875451291263e-06,
+      "loss": 4.5894,
+      "step": 530400
+    },
+    {
+      "epoch": 4.566546143250688,
+      "grad_norm": 3.140625,
+      "learning_rate": 3.698571958038921e-06,
+      "loss": 4.1079,
+      "step": 530450
+    },
+    {
+      "epoch": 4.566976584022038,
+      "grad_norm": 3.609375,
+      "learning_rate": 3.6912755331041992e-06,
+      "loss": 4.0916,
+      "step": 530500
+    },
+    {
+      "epoch": 4.567407024793388,
+      "grad_norm": 2.765625,
+      "learning_rate": 3.6839861770226246e-06,
+      "loss": 4.2991,
+      "step": 530550
+    },
+    {
+      "epoch": 4.567837465564738,
+      "grad_norm": 3.171875,
+      "learning_rate": 3.676703890329247e-06,
+      "loss": 4.4796,
+      "step": 530600
+    },
+    {
+      "epoch": 4.5682679063360885,
+      "grad_norm": 6.34375,
+      "learning_rate": 3.6694286735585836e-06,
+      "loss": 4.0997,
+      "step": 530650
+    },
+    {
+      "epoch": 4.568698347107438,
+      "grad_norm": 2.265625,
+      "learning_rate": 3.6621605272445625e-06,
+      "loss": 4.3753,
+      "step": 530700
+    },
+    {
+      "epoch": 4.569128787878788,
+      "grad_norm": 4.84375,
+      "learning_rate": 3.6548994519206903e-06,
+      "loss": 4.5602,
+      "step": 530750
+    },
+    {
+      "epoch": 4.569559228650138,
+      "grad_norm": 3.6875,
+      "learning_rate": 3.6476454481199074e-06,
+      "loss": 4.3599,
+      "step": 530800
+    },
+    {
+      "epoch": 4.569989669421488,
+      "grad_norm": 2.875,
+      "learning_rate": 3.6403985163746325e-06,
+      "loss": 4.4335,
+      "step": 530850
+    },
+    {
+      "epoch": 4.5704201101928374,
+      "grad_norm": 2.796875,
+      "learning_rate": 3.633158657216806e-06,
+      "loss": 4.1776,
+      "step": 530900
+    },
+    {
+      "epoch": 4.570850550964187,
+      "grad_norm": 4.28125,
+      "learning_rate": 3.625925871177771e-06,
+      "loss": 4.2518,
+      "step": 530950
+    },
+    {
+      "epoch": 4.571280991735537,
+      "grad_norm": 6.28125,
+      "learning_rate": 3.6187001587884463e-06,
+      "loss": 4.5516,
+      "step": 531000
+    },
+    {
+      "epoch": 4.571280991735537,
+      "eval_loss": 4.990326404571533,
+      "eval_runtime": 25.3331,
+      "eval_samples_per_second": 25.263,
+      "eval_steps_per_second": 12.632,
+      "eval_tts_loss": 7.480250930217977,
+      "step": 531000
+    },
+    {
+      "epoch": 4.571711432506887,
+      "grad_norm": 3.625,
+      "learning_rate": 3.611481520579141e-06,
+      "loss": 4.7823,
+      "step": 531050
+    },
+    {
+      "epoch": 4.572141873278237,
+      "grad_norm": 4.875,
+      "learning_rate": 3.60426995707972e-06,
+      "loss": 4.4082,
+      "step": 531100
+    },
+    {
+      "epoch": 4.572572314049586,
+      "grad_norm": 3.203125,
+      "learning_rate": 3.5970654688194827e-06,
+      "loss": 4.1592,
+      "step": 531150
+    },
+    {
+      "epoch": 4.573002754820937,
+      "grad_norm": 4.4375,
+      "learning_rate": 3.5898680563272168e-06,
+      "loss": 4.378,
+      "step": 531200
+    },
+    {
+      "epoch": 4.573433195592287,
+      "grad_norm": 1.9296875,
+      "learning_rate": 3.5826777201312e-06,
+      "loss": 3.7464,
+      "step": 531250
+    },
+    {
+      "epoch": 4.573863636363637,
+      "grad_norm": 5.5,
+      "learning_rate": 3.5754944607591987e-06,
+      "loss": 4.4324,
+      "step": 531300
+    },
+    {
+      "epoch": 4.5742940771349865,
+      "grad_norm": 4.625,
+      "learning_rate": 3.5683182787384364e-06,
+      "loss": 4.1571,
+      "step": 531350
+    },
+    {
+      "epoch": 4.574724517906336,
+      "grad_norm": 2.953125,
+      "learning_rate": 3.561149174595646e-06,
+      "loss": 4.153,
+      "step": 531400
+    },
+    {
+      "epoch": 4.575154958677686,
+      "grad_norm": 3.640625,
+      "learning_rate": 3.5539871488570078e-06,
+      "loss": 4.4161,
+      "step": 531450
+    },
+    {
+      "epoch": 4.575585399449036,
+      "grad_norm": 5.0,
+      "learning_rate": 3.5468322020482e-06,
+      "loss": 4.055,
+      "step": 531500
+    },
+    {
+      "epoch": 4.576015840220386,
+      "grad_norm": 3.53125,
+      "learning_rate": 3.539684334694371e-06,
+      "loss": 4.0948,
+      "step": 531550
+    },
+    {
+      "epoch": 4.5764462809917354,
+      "grad_norm": 2.890625,
+      "learning_rate": 3.5325435473201884e-06,
+      "loss": 4.2058,
+      "step": 531600
+    },
+    {
+      "epoch": 4.576876721763085,
+      "grad_norm": 4.59375,
+      "learning_rate": 3.525409840449756e-06,
+      "loss": 4.8002,
+      "step": 531650
+    },
+    {
+      "epoch": 4.577307162534435,
+      "grad_norm": 3.390625,
+      "learning_rate": 3.518283214606666e-06,
+      "loss": 4.1627,
+      "step": 531700
+    },
+    {
+      "epoch": 4.577737603305785,
+      "grad_norm": 3.078125,
+      "learning_rate": 3.5111636703140104e-06,
+      "loss": 4.1937,
+      "step": 531750
+    },
+    {
+      "epoch": 4.578168044077135,
+      "grad_norm": 1.96875,
+      "learning_rate": 3.5040512080943278e-06,
+      "loss": 4.2847,
+      "step": 531800
+    },
+    {
+      "epoch": 4.578598484848484,
+      "grad_norm": 4.0625,
+      "learning_rate": 3.4969458284696664e-06,
+      "loss": 4.2713,
+      "step": 531850
+    },
+    {
+      "epoch": 4.579028925619834,
+      "grad_norm": 3.75,
+      "learning_rate": 3.4898475319615655e-06,
+      "loss": 4.2822,
+      "step": 531900
+    },
+    {
+      "epoch": 4.579459366391185,
+      "grad_norm": 5.875,
+      "learning_rate": 3.4827563190909963e-06,
+      "loss": 4.3862,
+      "step": 531950
+    },
+    {
+      "epoch": 4.579889807162535,
+      "grad_norm": 2.296875,
+      "learning_rate": 3.475672190378465e-06,
+      "loss": 4.3589,
+      "step": 532000
+    },
+    {
+      "epoch": 4.5803202479338845,
+      "grad_norm": 3.234375,
+      "learning_rate": 3.4685951463439116e-06,
+      "loss": 4.254,
+      "step": 532050
+    },
+    {
+      "epoch": 4.580750688705234,
+      "grad_norm": 2.5,
+      "learning_rate": 3.4615251875067644e-06,
+      "loss": 4.0744,
+      "step": 532100
+    },
+    {
+      "epoch": 4.581181129476584,
+      "grad_norm": 6.96875,
+      "learning_rate": 3.4544623143859756e-06,
+      "loss": 4.2891,
+      "step": 532150
+    },
+    {
+      "epoch": 4.581611570247934,
+      "grad_norm": 2.5625,
+      "learning_rate": 3.447406527499941e-06,
+      "loss": 4.3272,
+      "step": 532200
+    },
+    {
+      "epoch": 4.582042011019284,
+      "grad_norm": 3.234375,
+      "learning_rate": 3.440357827366514e-06,
+      "loss": 4.3425,
+      "step": 532250
+    },
+    {
+      "epoch": 4.5824724517906334,
+      "grad_norm": 5.15625,
+      "learning_rate": 3.4333162145030685e-06,
+      "loss": 4.5651,
+      "step": 532300
+    },
+    {
+      "epoch": 4.582902892561983,
+      "grad_norm": 6.125,
+      "learning_rate": 3.4262816894264473e-06,
+      "loss": 4.2723,
+      "step": 532350
+    },
+    {
+      "epoch": 4.583333333333333,
+      "grad_norm": 2.484375,
+      "learning_rate": 3.419254252652959e-06,
+      "loss": 3.9293,
+      "step": 532400
+    },
+    {
+      "epoch": 4.583763774104683,
+      "grad_norm": 2.203125,
+      "learning_rate": 3.4122339046984366e-06,
+      "loss": 4.4241,
+      "step": 532450
+    },
+    {
+      "epoch": 4.5841942148760335,
+      "grad_norm": 4.9375,
+      "learning_rate": 3.4052206460781e-06,
+      "loss": 4.3243,
+      "step": 532500
+    },
+    {
+      "epoch": 4.584624655647383,
+      "grad_norm": 1.6875,
+      "learning_rate": 3.3982144773067604e-06,
+      "loss": 4.1942,
+      "step": 532550
+    },
+    {
+      "epoch": 4.585055096418733,
+      "grad_norm": 2.890625,
+      "learning_rate": 3.3912153988986172e-06,
+      "loss": 4.5547,
+      "step": 532600
+    },
+    {
+      "epoch": 4.585485537190083,
+      "grad_norm": 5.5625,
+      "learning_rate": 3.3842234113674154e-06,
+      "loss": 4.2556,
+      "step": 532650
+    },
+    {
+      "epoch": 4.585915977961433,
+      "grad_norm": 5.09375,
+      "learning_rate": 3.3772385152263442e-06,
+      "loss": 4.3848,
+      "step": 532700
+    },
+    {
+      "epoch": 4.5863464187327825,
+      "grad_norm": 4.1875,
+      "learning_rate": 3.3702607109880713e-06,
+      "loss": 4.346,
+      "step": 532750
+    },
+    {
+      "epoch": 4.586776859504132,
+      "grad_norm": 3.078125,
+      "learning_rate": 3.3632899991647537e-06,
+      "loss": 4.4291,
+      "step": 532800
+    },
+    {
+      "epoch": 4.587207300275482,
+      "grad_norm": 4.53125,
+      "learning_rate": 3.356326380268038e-06,
+      "loss": 4.4228,
+      "step": 532850
+    },
+    {
+      "epoch": 4.587637741046832,
+      "grad_norm": 5.15625,
+      "learning_rate": 3.349369854809026e-06,
+      "loss": 4.3154,
+      "step": 532900
+    },
+    {
+      "epoch": 4.588068181818182,
+      "grad_norm": 2.984375,
+      "learning_rate": 3.3424204232983204e-06,
+      "loss": 3.9473,
+      "step": 532950
+    },
+    {
+      "epoch": 4.5884986225895315,
+      "grad_norm": 3.515625,
+      "learning_rate": 3.335478086245991e-06,
+      "loss": 4.0014,
+      "step": 533000
+    },
+    {
+      "epoch": 4.588929063360881,
+      "grad_norm": 6.5,
+      "learning_rate": 3.3285428441615973e-06,
+      "loss": 4.4765,
+      "step": 533050
+    },
+    {
+      "epoch": 4.589359504132231,
+      "grad_norm": 2.65625,
+      "learning_rate": 3.321614697554154e-06,
+      "loss": 4.3522,
+      "step": 533100
+    },
+    {
+      "epoch": 4.589789944903581,
+      "grad_norm": 3.625,
+      "learning_rate": 3.3146936469321875e-06,
+      "loss": 4.1383,
+      "step": 533150
+    },
+    {
+      "epoch": 4.5902203856749315,
+      "grad_norm": 3.875,
+      "learning_rate": 3.307779692803681e-06,
+      "loss": 4.6708,
+      "step": 533200
+    },
+    {
+      "epoch": 4.590650826446281,
+      "grad_norm": 3.625,
+      "learning_rate": 3.3008728356761164e-06,
+      "loss": 4.6503,
+      "step": 533250
+    },
+    {
+      "epoch": 4.591081267217631,
+      "grad_norm": 3.4375,
+      "learning_rate": 3.2939730760564224e-06,
+      "loss": 4.3016,
+      "step": 533300
+    },
+    {
+      "epoch": 4.591511707988981,
+      "grad_norm": 4.40625,
+      "learning_rate": 3.287080414451038e-06,
+      "loss": 4.2083,
+      "step": 533350
+    },
+    {
+      "epoch": 4.591942148760331,
+      "grad_norm": 3.4375,
+      "learning_rate": 3.2801948513658585e-06,
+      "loss": 4.3242,
+      "step": 533400
+    },
+    {
+      "epoch": 4.5923725895316805,
+      "grad_norm": 3.09375,
+      "learning_rate": 3.2733163873063023e-06,
+      "loss": 4.2407,
+      "step": 533450
+    },
+    {
+      "epoch": 4.59280303030303,
+      "grad_norm": 3.125,
+      "learning_rate": 3.2664450227771983e-06,
+      "loss": 4.5062,
+      "step": 533500
+    },
+    {
+      "epoch": 4.59323347107438,
+      "grad_norm": 3.421875,
+      "learning_rate": 3.2595807582828984e-06,
+      "loss": 4.3134,
+      "step": 533550
+    },
+    {
+      "epoch": 4.59366391184573,
+      "grad_norm": 6.65625,
+      "learning_rate": 3.252723594327234e-06,
+      "loss": 4.2919,
+      "step": 533600
+    },
+    {
+      "epoch": 4.59409435261708,
+      "grad_norm": 3.796875,
+      "learning_rate": 3.24587353141349e-06,
+      "loss": 4.4213,
+      "step": 533650
+    },
+    {
+      "epoch": 4.5945247933884295,
+      "grad_norm": 3.75,
+      "learning_rate": 3.239030570044466e-06,
+      "loss": 4.3427,
+      "step": 533700
+    },
+    {
+      "epoch": 4.594955234159779,
+      "grad_norm": 4.1875,
+      "learning_rate": 3.232194710722425e-06,
+      "loss": 4.2234,
+      "step": 533750
+    },
+    {
+      "epoch": 4.59538567493113,
+      "grad_norm": 2.046875,
+      "learning_rate": 3.2253659539490667e-06,
+      "loss": 3.8704,
+      "step": 533800
+    },
+    {
+      "epoch": 4.59581611570248,
+      "grad_norm": 5.53125,
+      "learning_rate": 3.2185443002256343e-06,
+      "loss": 4.2281,
+      "step": 533850
+    },
+    {
+      "epoch": 4.5962465564738295,
+      "grad_norm": 2.78125,
+      "learning_rate": 3.2117297500528276e-06,
+      "loss": 4.1042,
+      "step": 533900
+    },
+    {
+      "epoch": 4.596676997245179,
+      "grad_norm": 6.75,
+      "learning_rate": 3.2049223039308017e-06,
+      "loss": 3.956,
+      "step": 533950
+    },
+    {
+      "epoch": 4.597107438016529,
+      "grad_norm": 3.015625,
+      "learning_rate": 3.1981219623592463e-06,
+      "loss": 4.5374,
+      "step": 534000
+    },
+    {
+      "epoch": 4.597107438016529,
+      "eval_loss": 4.990235805511475,
+      "eval_runtime": 24.5558,
+      "eval_samples_per_second": 26.063,
+      "eval_steps_per_second": 13.032,
+      "eval_tts_loss": 7.479965384752732,
+      "step": 534000
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 580800,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 5,
+  "save_steps": 3000,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.925328502150398e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}