| { | |
| "best_global_step": null, | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 99.91201764057332, | |
| "eval_steps": 500, | |
| "global_step": 113300, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.9993384785005512, | |
| "grad_norm": 6.264042854309082, | |
| "learning_rate": 9.90035304501324e-06, | |
| "loss": 2.1264, | |
| "step": 1133 | |
| }, | |
| { | |
| "epoch": 1.9984564498346198, | |
| "grad_norm": 8.962105751037598, | |
| "learning_rate": 9.80035304501324e-06, | |
| "loss": 2.0771, | |
| "step": 2266 | |
| }, | |
| { | |
| "epoch": 2.997574421168688, | |
| "grad_norm": 13.9079008102417, | |
| "learning_rate": 9.70044130626655e-06, | |
| "loss": 2.0072, | |
| "step": 3399 | |
| }, | |
| { | |
| "epoch": 3.9966923925027564, | |
| "grad_norm": 9.212200164794922, | |
| "learning_rate": 9.60044130626655e-06, | |
| "loss": 1.9554, | |
| "step": 4532 | |
| }, | |
| { | |
| "epoch": 4.995810363836824, | |
| "grad_norm": 9.967486381530762, | |
| "learning_rate": 9.50052956751986e-06, | |
| "loss": 1.9156, | |
| "step": 5665 | |
| }, | |
| { | |
| "epoch": 5.9949283351708935, | |
| "grad_norm": 9.588685989379883, | |
| "learning_rate": 9.40061782877317e-06, | |
| "loss": 1.9031, | |
| "step": 6798 | |
| }, | |
| { | |
| "epoch": 6.994046306504961, | |
| "grad_norm": 5.2934889793396, | |
| "learning_rate": 9.30061782877317e-06, | |
| "loss": 1.8739, | |
| "step": 7931 | |
| }, | |
| { | |
| "epoch": 7.99316427783903, | |
| "grad_norm": 12.791667938232422, | |
| "learning_rate": 9.20070609002648e-06, | |
| "loss": 1.8476, | |
| "step": 9064 | |
| }, | |
| { | |
| "epoch": 8.992282249173098, | |
| "grad_norm": 5.592958927154541, | |
| "learning_rate": 9.100706090026479e-06, | |
| "loss": 1.8378, | |
| "step": 10197 | |
| }, | |
| { | |
| "epoch": 9.991400220507167, | |
| "grad_norm": 6.236687660217285, | |
| "learning_rate": 9.000706090026478e-06, | |
| "loss": 1.8141, | |
| "step": 11330 | |
| }, | |
| { | |
| "epoch": 10.990518191841234, | |
| "grad_norm": 14.459975242614746, | |
| "learning_rate": 8.90079435127979e-06, | |
| "loss": 1.8008, | |
| "step": 12463 | |
| }, | |
| { | |
| "epoch": 11.989636163175303, | |
| "grad_norm": 7.29625940322876, | |
| "learning_rate": 8.800794351279789e-06, | |
| "loss": 1.7965, | |
| "step": 13596 | |
| }, | |
| { | |
| "epoch": 12.98875413450937, | |
| "grad_norm": 7.927811622619629, | |
| "learning_rate": 8.7008826125331e-06, | |
| "loss": 1.7768, | |
| "step": 14729 | |
| }, | |
| { | |
| "epoch": 13.98787210584344, | |
| "grad_norm": 7.546788692474365, | |
| "learning_rate": 8.600882612533099e-06, | |
| "loss": 1.7735, | |
| "step": 15862 | |
| }, | |
| { | |
| "epoch": 14.986990077177508, | |
| "grad_norm": 2.396638870239258, | |
| "learning_rate": 8.500970873786408e-06, | |
| "loss": 1.7688, | |
| "step": 16995 | |
| }, | |
| { | |
| "epoch": 15.986108048511577, | |
| "grad_norm": 4.344698905944824, | |
| "learning_rate": 8.400970873786408e-06, | |
| "loss": 1.7521, | |
| "step": 18128 | |
| }, | |
| { | |
| "epoch": 16.985226019845644, | |
| "grad_norm": 3.3172178268432617, | |
| "learning_rate": 8.300970873786409e-06, | |
| "loss": 1.7524, | |
| "step": 19261 | |
| }, | |
| { | |
| "epoch": 17.984343991179713, | |
| "grad_norm": 4.233256816864014, | |
| "learning_rate": 8.200970873786408e-06, | |
| "loss": 1.7394, | |
| "step": 20394 | |
| }, | |
| { | |
| "epoch": 18.983461962513783, | |
| "grad_norm": 3.924440860748291, | |
| "learning_rate": 8.101147396293027e-06, | |
| "loss": 1.7355, | |
| "step": 21527 | |
| }, | |
| { | |
| "epoch": 19.982579933847852, | |
| "grad_norm": 8.452503204345703, | |
| "learning_rate": 8.001147396293027e-06, | |
| "loss": 1.7245, | |
| "step": 22660 | |
| }, | |
| { | |
| "epoch": 20.981697905181917, | |
| "grad_norm": 10.714750289916992, | |
| "learning_rate": 7.901147396293028e-06, | |
| "loss": 1.7205, | |
| "step": 23793 | |
| }, | |
| { | |
| "epoch": 21.980815876515987, | |
| "grad_norm": 12.158623695373535, | |
| "learning_rate": 7.801147396293029e-06, | |
| "loss": 1.7207, | |
| "step": 24926 | |
| }, | |
| { | |
| "epoch": 22.979933847850056, | |
| "grad_norm": 2.1460916996002197, | |
| "learning_rate": 7.701323918799648e-06, | |
| "loss": 1.716, | |
| "step": 26059 | |
| }, | |
| { | |
| "epoch": 23.979051819184125, | |
| "grad_norm": 2.6795060634613037, | |
| "learning_rate": 7.601323918799648e-06, | |
| "loss": 1.71, | |
| "step": 27192 | |
| }, | |
| { | |
| "epoch": 24.97816979051819, | |
| "grad_norm": 4.405874729156494, | |
| "learning_rate": 7.501412180052957e-06, | |
| "loss": 1.7143, | |
| "step": 28325 | |
| }, | |
| { | |
| "epoch": 25.97728776185226, | |
| "grad_norm": 2.3188281059265137, | |
| "learning_rate": 7.401412180052957e-06, | |
| "loss": 1.6879, | |
| "step": 29458 | |
| }, | |
| { | |
| "epoch": 26.97640573318633, | |
| "grad_norm": 4.247509002685547, | |
| "learning_rate": 7.301412180052957e-06, | |
| "loss": 1.7008, | |
| "step": 30591 | |
| }, | |
| { | |
| "epoch": 27.9755237045204, | |
| "grad_norm": 5.176025390625, | |
| "learning_rate": 7.2015004413062674e-06, | |
| "loss": 1.6904, | |
| "step": 31724 | |
| }, | |
| { | |
| "epoch": 28.974641675854464, | |
| "grad_norm": 3.4096691608428955, | |
| "learning_rate": 7.1015004413062675e-06, | |
| "loss": 1.6884, | |
| "step": 32857 | |
| }, | |
| { | |
| "epoch": 29.973759647188533, | |
| "grad_norm": 5.346144676208496, | |
| "learning_rate": 7.001588702559578e-06, | |
| "loss": 1.6874, | |
| "step": 33990 | |
| }, | |
| { | |
| "epoch": 30.972877618522602, | |
| "grad_norm": 4.809396743774414, | |
| "learning_rate": 6.901588702559577e-06, | |
| "loss": 1.6834, | |
| "step": 35123 | |
| }, | |
| { | |
| "epoch": 31.97199558985667, | |
| "grad_norm": 4.950044631958008, | |
| "learning_rate": 6.8016769638128865e-06, | |
| "loss": 1.6704, | |
| "step": 36256 | |
| }, | |
| { | |
| "epoch": 32.97111356119074, | |
| "grad_norm": 2.259593963623047, | |
| "learning_rate": 6.701676963812887e-06, | |
| "loss": 1.676, | |
| "step": 37389 | |
| }, | |
| { | |
| "epoch": 33.97023153252481, | |
| "grad_norm": 1.9857131242752075, | |
| "learning_rate": 6.601765225066197e-06, | |
| "loss": 1.6721, | |
| "step": 38522 | |
| }, | |
| { | |
| "epoch": 34.96934950385887, | |
| "grad_norm": 2.6523988246917725, | |
| "learning_rate": 6.501765225066197e-06, | |
| "loss": 1.6519, | |
| "step": 39655 | |
| }, | |
| { | |
| "epoch": 35.968467475192945, | |
| "grad_norm": 2.5178589820861816, | |
| "learning_rate": 6.401853486319506e-06, | |
| "loss": 1.6763, | |
| "step": 40788 | |
| }, | |
| { | |
| "epoch": 36.96758544652701, | |
| "grad_norm": 3.966003894805908, | |
| "learning_rate": 6.301941747572816e-06, | |
| "loss": 1.6729, | |
| "step": 41921 | |
| }, | |
| { | |
| "epoch": 37.96670341786108, | |
| "grad_norm": 2.874756336212158, | |
| "learning_rate": 6.201941747572816e-06, | |
| "loss": 1.6761, | |
| "step": 43054 | |
| }, | |
| { | |
| "epoch": 38.96582138919515, | |
| "grad_norm": 5.206564426422119, | |
| "learning_rate": 6.102030008826125e-06, | |
| "loss": 1.6731, | |
| "step": 44187 | |
| }, | |
| { | |
| "epoch": 39.964939360529215, | |
| "grad_norm": 3.3622446060180664, | |
| "learning_rate": 6.002030008826126e-06, | |
| "loss": 1.6649, | |
| "step": 45320 | |
| }, | |
| { | |
| "epoch": 40.96405733186329, | |
| "grad_norm": 2.243980884552002, | |
| "learning_rate": 5.902118270079436e-06, | |
| "loss": 1.6617, | |
| "step": 46453 | |
| }, | |
| { | |
| "epoch": 41.96317530319735, | |
| "grad_norm": 2.656219720840454, | |
| "learning_rate": 5.802118270079435e-06, | |
| "loss": 1.6539, | |
| "step": 47586 | |
| }, | |
| { | |
| "epoch": 42.96229327453142, | |
| "grad_norm": 2.1969571113586426, | |
| "learning_rate": 5.702206531332745e-06, | |
| "loss": 1.6507, | |
| "step": 48719 | |
| }, | |
| { | |
| "epoch": 43.96141124586549, | |
| "grad_norm": 2.570960760116577, | |
| "learning_rate": 5.602294792586055e-06, | |
| "loss": 1.6517, | |
| "step": 49852 | |
| }, | |
| { | |
| "epoch": 44.96052921719956, | |
| "grad_norm": 2.2769925594329834, | |
| "learning_rate": 5.502294792586055e-06, | |
| "loss": 1.6622, | |
| "step": 50985 | |
| }, | |
| { | |
| "epoch": 45.95964718853363, | |
| "grad_norm": 1.3121511936187744, | |
| "learning_rate": 5.402383053839365e-06, | |
| "loss": 1.6583, | |
| "step": 52118 | |
| }, | |
| { | |
| "epoch": 46.958765159867696, | |
| "grad_norm": 2.186378002166748, | |
| "learning_rate": 5.302383053839365e-06, | |
| "loss": 1.6579, | |
| "step": 53251 | |
| }, | |
| { | |
| "epoch": 47.95788313120176, | |
| "grad_norm": 1.257639765739441, | |
| "learning_rate": 5.202471315092674e-06, | |
| "loss": 1.653, | |
| "step": 54384 | |
| }, | |
| { | |
| "epoch": 48.957001102535834, | |
| "grad_norm": 2.570171594619751, | |
| "learning_rate": 5.102471315092676e-06, | |
| "loss": 1.6503, | |
| "step": 55517 | |
| }, | |
| { | |
| "epoch": 49.9561190738699, | |
| "grad_norm": 1.8046927452087402, | |
| "learning_rate": 5.002559576345984e-06, | |
| "loss": 1.6569, | |
| "step": 56650 | |
| }, | |
| { | |
| "epoch": 50.95523704520397, | |
| "grad_norm": 2.055699348449707, | |
| "learning_rate": 4.9026478375992945e-06, | |
| "loss": 1.649, | |
| "step": 57783 | |
| }, | |
| { | |
| "epoch": 51.95435501653804, | |
| "grad_norm": 1.5541220903396606, | |
| "learning_rate": 4.802647837599294e-06, | |
| "loss": 1.6463, | |
| "step": 58916 | |
| }, | |
| { | |
| "epoch": 52.953472987872104, | |
| "grad_norm": 1.3364524841308594, | |
| "learning_rate": 4.702736098852604e-06, | |
| "loss": 1.6406, | |
| "step": 60049 | |
| }, | |
| { | |
| "epoch": 53.95259095920618, | |
| "grad_norm": 1.8428481817245483, | |
| "learning_rate": 4.602736098852604e-06, | |
| "loss": 1.6443, | |
| "step": 61182 | |
| }, | |
| { | |
| "epoch": 54.95170893054024, | |
| "grad_norm": 1.3728575706481934, | |
| "learning_rate": 4.5028243601059144e-06, | |
| "loss": 1.6427, | |
| "step": 62315 | |
| }, | |
| { | |
| "epoch": 55.95082690187431, | |
| "grad_norm": 2.039762496948242, | |
| "learning_rate": 4.402824360105914e-06, | |
| "loss": 1.6523, | |
| "step": 63448 | |
| }, | |
| { | |
| "epoch": 56.94994487320838, | |
| "grad_norm": 0.4440517723560333, | |
| "learning_rate": 4.302912621359224e-06, | |
| "loss": 1.6346, | |
| "step": 64581 | |
| }, | |
| { | |
| "epoch": 57.949062844542446, | |
| "grad_norm": 2.8719191551208496, | |
| "learning_rate": 4.202912621359223e-06, | |
| "loss": 1.6481, | |
| "step": 65714 | |
| }, | |
| { | |
| "epoch": 58.94818081587652, | |
| "grad_norm": 2.116090774536133, | |
| "learning_rate": 4.1030008826125335e-06, | |
| "loss": 1.6403, | |
| "step": 66847 | |
| }, | |
| { | |
| "epoch": 59.947298787210585, | |
| "grad_norm": 2.5801730155944824, | |
| "learning_rate": 4.003089143865843e-06, | |
| "loss": 1.6445, | |
| "step": 67980 | |
| }, | |
| { | |
| "epoch": 60.94641675854465, | |
| "grad_norm": 1.5309211015701294, | |
| "learning_rate": 3.903089143865843e-06, | |
| "loss": 1.6495, | |
| "step": 69113 | |
| }, | |
| { | |
| "epoch": 61.94553472987872, | |
| "grad_norm": 2.1777145862579346, | |
| "learning_rate": 3.803177405119153e-06, | |
| "loss": 1.6477, | |
| "step": 70246 | |
| }, | |
| { | |
| "epoch": 62.94465270121279, | |
| "grad_norm": 1.003503441810608, | |
| "learning_rate": 3.7031774051191535e-06, | |
| "loss": 1.6387, | |
| "step": 71379 | |
| }, | |
| { | |
| "epoch": 63.943770672546854, | |
| "grad_norm": 2.015745162963867, | |
| "learning_rate": 3.603265666372463e-06, | |
| "loss": 1.6307, | |
| "step": 72512 | |
| }, | |
| { | |
| "epoch": 64.94288864388092, | |
| "grad_norm": 2.482320547103882, | |
| "learning_rate": 3.503265666372463e-06, | |
| "loss": 1.6402, | |
| "step": 73645 | |
| }, | |
| { | |
| "epoch": 65.94200661521499, | |
| "grad_norm": 2.404247999191284, | |
| "learning_rate": 3.403353927625773e-06, | |
| "loss": 1.6409, | |
| "step": 74778 | |
| }, | |
| { | |
| "epoch": 66.94112458654907, | |
| "grad_norm": 1.197210669517517, | |
| "learning_rate": 3.3034421888790823e-06, | |
| "loss": 1.6518, | |
| "step": 75911 | |
| }, | |
| { | |
| "epoch": 67.94024255788312, | |
| "grad_norm": 0.6208453178405762, | |
| "learning_rate": 3.2034421888790824e-06, | |
| "loss": 1.6461, | |
| "step": 77044 | |
| }, | |
| { | |
| "epoch": 68.9393605292172, | |
| "grad_norm": 1.967084527015686, | |
| "learning_rate": 3.1035304501323922e-06, | |
| "loss": 1.6393, | |
| "step": 78177 | |
| }, | |
| { | |
| "epoch": 69.93847850055127, | |
| "grad_norm": 1.0003052949905396, | |
| "learning_rate": 3.003530450132392e-06, | |
| "loss": 1.6259, | |
| "step": 79310 | |
| }, | |
| { | |
| "epoch": 70.93759647188534, | |
| "grad_norm": 0.6473856568336487, | |
| "learning_rate": 2.9036187113857018e-06, | |
| "loss": 1.6302, | |
| "step": 80443 | |
| }, | |
| { | |
| "epoch": 71.9367144432194, | |
| "grad_norm": 0.949741780757904, | |
| "learning_rate": 2.8036187113857015e-06, | |
| "loss": 1.6355, | |
| "step": 81576 | |
| }, | |
| { | |
| "epoch": 72.93583241455347, | |
| "grad_norm": 1.2901132106781006, | |
| "learning_rate": 2.7037069726390117e-06, | |
| "loss": 1.6335, | |
| "step": 82709 | |
| }, | |
| { | |
| "epoch": 73.93495038588755, | |
| "grad_norm": 1.8243396282196045, | |
| "learning_rate": 2.6037952338923216e-06, | |
| "loss": 1.6329, | |
| "step": 83842 | |
| }, | |
| { | |
| "epoch": 74.9340683572216, | |
| "grad_norm": 0.6137579679489136, | |
| "learning_rate": 2.5037952338923217e-06, | |
| "loss": 1.6421, | |
| "step": 84975 | |
| }, | |
| { | |
| "epoch": 75.93318632855568, | |
| "grad_norm": 1.270875096321106, | |
| "learning_rate": 2.403883495145631e-06, | |
| "loss": 1.628, | |
| "step": 86108 | |
| }, | |
| { | |
| "epoch": 76.93230429988975, | |
| "grad_norm": 0.9131399393081665, | |
| "learning_rate": 2.3038834951456313e-06, | |
| "loss": 1.6235, | |
| "step": 87241 | |
| }, | |
| { | |
| "epoch": 77.93142227122381, | |
| "grad_norm": 1.042668342590332, | |
| "learning_rate": 2.203971756398941e-06, | |
| "loss": 1.6323, | |
| "step": 88374 | |
| }, | |
| { | |
| "epoch": 78.93054024255788, | |
| "grad_norm": 2.174466609954834, | |
| "learning_rate": 2.103971756398941e-06, | |
| "loss": 1.6263, | |
| "step": 89507 | |
| }, | |
| { | |
| "epoch": 79.92965821389195, | |
| "grad_norm": 0.5852002501487732, | |
| "learning_rate": 2.004060017652251e-06, | |
| "loss": 1.6354, | |
| "step": 90640 | |
| }, | |
| { | |
| "epoch": 80.92877618522601, | |
| "grad_norm": 0.41708850860595703, | |
| "learning_rate": 1.9040600176522508e-06, | |
| "loss": 1.6177, | |
| "step": 91773 | |
| }, | |
| { | |
| "epoch": 81.92789415656009, | |
| "grad_norm": 1.2882851362228394, | |
| "learning_rate": 1.8041482789055606e-06, | |
| "loss": 1.6317, | |
| "step": 92906 | |
| }, | |
| { | |
| "epoch": 82.92701212789416, | |
| "grad_norm": 1.3567571640014648, | |
| "learning_rate": 1.7042365401588703e-06, | |
| "loss": 1.6336, | |
| "step": 94039 | |
| }, | |
| { | |
| "epoch": 83.92613009922823, | |
| "grad_norm": 1.7891902923583984, | |
| "learning_rate": 1.6042365401588702e-06, | |
| "loss": 1.6285, | |
| "step": 95172 | |
| }, | |
| { | |
| "epoch": 84.92524807056229, | |
| "grad_norm": 1.0167814493179321, | |
| "learning_rate": 1.5043248014121802e-06, | |
| "loss": 1.6233, | |
| "step": 96305 | |
| }, | |
| { | |
| "epoch": 85.92436604189636, | |
| "grad_norm": 0.6127368211746216, | |
| "learning_rate": 1.4043248014121801e-06, | |
| "loss": 1.6194, | |
| "step": 97438 | |
| }, | |
| { | |
| "epoch": 86.92348401323044, | |
| "grad_norm": 0.4271755516529083, | |
| "learning_rate": 1.30441306266549e-06, | |
| "loss": 1.6292, | |
| "step": 98571 | |
| }, | |
| { | |
| "epoch": 87.9226019845645, | |
| "grad_norm": 0.6342440843582153, | |
| "learning_rate": 1.20441306266549e-06, | |
| "loss": 1.6266, | |
| "step": 99704 | |
| }, | |
| { | |
| "epoch": 88.92171995589857, | |
| "grad_norm": 0.4804386794567108, | |
| "learning_rate": 1.1045013239187997e-06, | |
| "loss": 1.624, | |
| "step": 100837 | |
| }, | |
| { | |
| "epoch": 89.92083792723264, | |
| "grad_norm": 0.6205073595046997, | |
| "learning_rate": 1.0045895851721096e-06, | |
| "loss": 1.6225, | |
| "step": 101970 | |
| }, | |
| { | |
| "epoch": 90.9199558985667, | |
| "grad_norm": 0.5994776487350464, | |
| "learning_rate": 9.045895851721096e-07, | |
| "loss": 1.6296, | |
| "step": 103103 | |
| }, | |
| { | |
| "epoch": 91.91907386990077, | |
| "grad_norm": 1.454080581665039, | |
| "learning_rate": 8.046778464254193e-07, | |
| "loss": 1.6254, | |
| "step": 104236 | |
| }, | |
| { | |
| "epoch": 92.91819184123484, | |
| "grad_norm": 0.8811701536178589, | |
| "learning_rate": 7.046778464254193e-07, | |
| "loss": 1.6295, | |
| "step": 105369 | |
| }, | |
| { | |
| "epoch": 93.9173098125689, | |
| "grad_norm": 0.6332120895385742, | |
| "learning_rate": 6.047661076787291e-07, | |
| "loss": 1.6269, | |
| "step": 106502 | |
| }, | |
| { | |
| "epoch": 94.91642778390298, | |
| "grad_norm": 1.2232398986816406, | |
| "learning_rate": 5.047661076787291e-07, | |
| "loss": 1.6254, | |
| "step": 107635 | |
| }, | |
| { | |
| "epoch": 95.91554575523705, | |
| "grad_norm": 0.9554293751716614, | |
| "learning_rate": 4.048543689320389e-07, | |
| "loss": 1.6307, | |
| "step": 108768 | |
| }, | |
| { | |
| "epoch": 96.91466372657112, | |
| "grad_norm": 0.42345738410949707, | |
| "learning_rate": 3.0485436893203884e-07, | |
| "loss": 1.6314, | |
| "step": 109901 | |
| }, | |
| { | |
| "epoch": 97.91378169790518, | |
| "grad_norm": 0.4951033890247345, | |
| "learning_rate": 2.0494263018534864e-07, | |
| "loss": 1.6341, | |
| "step": 111034 | |
| }, | |
| { | |
| "epoch": 98.91289966923925, | |
| "grad_norm": 0.40377479791641235, | |
| "learning_rate": 1.0503089143865844e-07, | |
| "loss": 1.6174, | |
| "step": 112167 | |
| }, | |
| { | |
| "epoch": 99.91201764057332, | |
| "grad_norm": 0.5117936730384827, | |
| "learning_rate": 5.03089143865843e-09, | |
| "loss": 1.6313, | |
| "step": 113300 | |
| }, | |
| { | |
| "epoch": 99.91201764057332, | |
| "step": 113300, | |
| "total_flos": 5.364383961180322e+18, | |
| "train_loss": 1.6900030501417007, | |
| "train_runtime": 96835.149, | |
| "train_samples_per_second": 140.49, | |
| "train_steps_per_second": 1.17 | |
| } | |
| ], | |
| "logging_steps": 1133, | |
| "max_steps": 113300, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 100, | |
| "save_steps": 6800, | |
| "stateful_callbacks": { | |
| "TrainerControl": { | |
| "args": { | |
| "should_epoch_stop": false, | |
| "should_evaluate": false, | |
| "should_log": false, | |
| "should_save": true, | |
| "should_training_stop": true | |
| }, | |
| "attributes": {} | |
| } | |
| }, | |
| "total_flos": 5.364383961180322e+18, | |
| "train_batch_size": 30, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |