swinv2-base-patch4-window12-192-22k-ConcreteClassifier-PVT
/
tmp-checkpoint-7708
/trainer_state.json
| { | |
| "best_metric": 0.9808588624000549, | |
| "best_model_checkpoint": "swinv2-base-patch4-window12-192-22k-ConcreteClassifier-PVT\\checkpoint-1927", | |
| "epoch": 4.0, | |
| "eval_steps": 500, | |
| "global_step": 7708, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.7298045320878742e-06, | |
| "loss": 2.3066, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 3.4596090641757485e-06, | |
| "loss": 2.131, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 5.1894135962636225e-06, | |
| "loss": 2.0104, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 6.919218128351497e-06, | |
| "loss": 1.8938, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 8.649022660439371e-06, | |
| "loss": 1.7743, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.0378827192527245e-05, | |
| "loss": 1.73, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.2108631724615119e-05, | |
| "loss": 1.6552, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.3838436256702994e-05, | |
| "loss": 1.4119, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.5568240788790867e-05, | |
| "loss": 1.1716, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.7298045320878743e-05, | |
| "loss": 1.2431, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9027849852966615e-05, | |
| "loss": 1.1115, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 2.075765438505449e-05, | |
| "loss": 1.0365, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 2.2487458917142365e-05, | |
| "loss": 1.2179, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 2.4217263449230237e-05, | |
| "loss": 1.1192, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.5947067981318112e-05, | |
| "loss": 0.8697, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.7676872513405988e-05, | |
| "loss": 1.0286, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.940667704549386e-05, | |
| "loss": 0.9186, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 3.1136481577581735e-05, | |
| "loss": 1.0909, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 3.286628610966961e-05, | |
| "loss": 1.0519, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 3.4596090641757486e-05, | |
| "loss": 1.1043, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 3.632589517384536e-05, | |
| "loss": 0.8634, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 3.805569970593323e-05, | |
| "loss": 0.86, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 3.97855042380211e-05, | |
| "loss": 0.8325, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 4.151530877010898e-05, | |
| "loss": 1.1083, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 4.324511330219685e-05, | |
| "loss": 1.0759, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 4.497491783428473e-05, | |
| "loss": 0.9178, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 4.67047223663726e-05, | |
| "loss": 1.1132, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 4.8434526898460474e-05, | |
| "loss": 1.0484, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 5.0164331430548346e-05, | |
| "loss": 0.8629, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 5.1894135962636225e-05, | |
| "loss": 0.6426, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 5.36239404947241e-05, | |
| "loss": 1.0117, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 5.5353745026811976e-05, | |
| "loss": 0.9769, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 5.708354955889985e-05, | |
| "loss": 0.9459, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 5.881335409098772e-05, | |
| "loss": 1.3208, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 6.054315862307559e-05, | |
| "loss": 0.9631, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 6.227296315516347e-05, | |
| "loss": 1.3533, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 6.400276768725135e-05, | |
| "loss": 1.2228, | |
| "step": 370 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 6.573257221933921e-05, | |
| "loss": 0.9195, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 6.746237675142709e-05, | |
| "loss": 1.0466, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 6.919218128351497e-05, | |
| "loss": 0.6537, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 7.092198581560285e-05, | |
| "loss": 1.0766, | |
| "step": 410 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 7.265179034769071e-05, | |
| "loss": 1.3196, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 7.438159487977858e-05, | |
| "loss": 1.1902, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 7.611139941186646e-05, | |
| "loss": 0.8251, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 7.784120394395434e-05, | |
| "loss": 0.6196, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 7.95710084760422e-05, | |
| "loss": 1.1941, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 8.130081300813008e-05, | |
| "loss": 0.9509, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 8.303061754021796e-05, | |
| "loss": 1.0669, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 8.476042207230583e-05, | |
| "loss": 0.8493, | |
| "step": 490 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 8.64902266043937e-05, | |
| "loss": 1.0381, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 8.822003113648158e-05, | |
| "loss": 0.6853, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 8.994983566856946e-05, | |
| "loss": 1.0025, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 9.167964020065733e-05, | |
| "loss": 0.9758, | |
| "step": 530 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 9.34094447327452e-05, | |
| "loss": 1.2263, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 9.513924926483307e-05, | |
| "loss": 1.0245, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 9.686905379692095e-05, | |
| "loss": 0.9136, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 9.859885832900881e-05, | |
| "loss": 1.0586, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.00010032866286109669, | |
| "loss": 1.029, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00010205846739318457, | |
| "loss": 0.9973, | |
| "step": 590 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00010378827192527245, | |
| "loss": 0.7652, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00010551807645736032, | |
| "loss": 1.2198, | |
| "step": 610 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0001072478809894482, | |
| "loss": 1.3727, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00010897768552153607, | |
| "loss": 0.9157, | |
| "step": 630 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00011070749005362395, | |
| "loss": 1.4063, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00011243729458571182, | |
| "loss": 0.9697, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0001141670991177997, | |
| "loss": 1.0678, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00011589690364988756, | |
| "loss": 1.1373, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00011762670818197544, | |
| "loss": 1.1127, | |
| "step": 680 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0001193565127140633, | |
| "loss": 0.8828, | |
| "step": 690 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00012108631724615118, | |
| "loss": 1.0497, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00012281612177823906, | |
| "loss": 1.0252, | |
| "step": 710 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00012454592631032694, | |
| "loss": 1.0175, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00012627573084241482, | |
| "loss": 1.1303, | |
| "step": 730 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0001280055353745027, | |
| "loss": 0.8751, | |
| "step": 740 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00012973533990659055, | |
| "loss": 0.9646, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00013146514443867843, | |
| "loss": 1.0412, | |
| "step": 760 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0001331949489707663, | |
| "loss": 1.3243, | |
| "step": 770 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00013492475350285418, | |
| "loss": 1.0182, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00013665455803494206, | |
| "loss": 1.3814, | |
| "step": 790 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00013838436256702994, | |
| "loss": 0.9131, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00014011416709911782, | |
| "loss": 1.0305, | |
| "step": 810 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0001418439716312057, | |
| "loss": 0.9296, | |
| "step": 820 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00014357377616329355, | |
| "loss": 1.1304, | |
| "step": 830 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00014530358069538143, | |
| "loss": 0.964, | |
| "step": 840 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0001470333852274693, | |
| "loss": 0.9223, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00014876318975955716, | |
| "loss": 0.8419, | |
| "step": 860 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00015049299429164504, | |
| "loss": 0.7896, | |
| "step": 870 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00015222279882373292, | |
| "loss": 1.2753, | |
| "step": 880 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0001539526033558208, | |
| "loss": 1.01, | |
| "step": 890 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00015568240788790867, | |
| "loss": 0.8729, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00015741221241999653, | |
| "loss": 1.2337, | |
| "step": 910 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0001591420169520844, | |
| "loss": 1.0598, | |
| "step": 920 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00016087182148417228, | |
| "loss": 0.9216, | |
| "step": 930 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00016260162601626016, | |
| "loss": 1.2614, | |
| "step": 940 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00016433143054834804, | |
| "loss": 0.9602, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00016606123508043592, | |
| "loss": 1.1126, | |
| "step": 960 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.0001677910396125238, | |
| "loss": 0.7103, | |
| "step": 970 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00016952084414461165, | |
| "loss": 1.0462, | |
| "step": 980 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00017125064867669953, | |
| "loss": 0.9843, | |
| "step": 990 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0001729804532087874, | |
| "loss": 1.2419, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00017471025774087529, | |
| "loss": 1.3823, | |
| "step": 1010 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00017644006227296316, | |
| "loss": 0.9726, | |
| "step": 1020 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00017816986680505104, | |
| "loss": 1.1238, | |
| "step": 1030 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00017989967133713892, | |
| "loss": 0.934, | |
| "step": 1040 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.0001816294758692268, | |
| "loss": 1.5093, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00018335928040131465, | |
| "loss": 1.0234, | |
| "step": 1060 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00018508908493340253, | |
| "loss": 0.7998, | |
| "step": 1070 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.0001868188894654904, | |
| "loss": 1.0191, | |
| "step": 1080 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.0001885486939975783, | |
| "loss": 1.1171, | |
| "step": 1090 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00019027849852966614, | |
| "loss": 1.1417, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00019200830306175402, | |
| "loss": 0.9539, | |
| "step": 1110 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.0001937381075938419, | |
| "loss": 1.0955, | |
| "step": 1120 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00019546791212592978, | |
| "loss": 1.3748, | |
| "step": 1130 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00019719771665801763, | |
| "loss": 1.1126, | |
| "step": 1140 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.0001989275211901055, | |
| "loss": 1.0674, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00020065732572219339, | |
| "loss": 1.1236, | |
| "step": 1160 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00020238713025428126, | |
| "loss": 1.0824, | |
| "step": 1170 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00020411693478636914, | |
| "loss": 1.1181, | |
| "step": 1180 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00020584673931845702, | |
| "loss": 1.0881, | |
| "step": 1190 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.0002075765438505449, | |
| "loss": 1.1874, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00020930634838263278, | |
| "loss": 1.0377, | |
| "step": 1210 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00021103615291472063, | |
| "loss": 0.8027, | |
| "step": 1220 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.0002127659574468085, | |
| "loss": 0.8027, | |
| "step": 1230 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.0002144957619788964, | |
| "loss": 1.2303, | |
| "step": 1240 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00021622556651098427, | |
| "loss": 1.2914, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00021795537104307214, | |
| "loss": 1.2018, | |
| "step": 1260 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00021968517557516002, | |
| "loss": 0.994, | |
| "step": 1270 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.0002214149801072479, | |
| "loss": 1.1917, | |
| "step": 1280 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00022314478463933578, | |
| "loss": 0.9846, | |
| "step": 1290 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00022487458917142363, | |
| "loss": 1.1561, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.0002266043937035115, | |
| "loss": 1.3345, | |
| "step": 1310 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 0.0002283341982355994, | |
| "loss": 1.1663, | |
| "step": 1320 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 0.00023006400276768727, | |
| "loss": 1.0486, | |
| "step": 1330 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 0.00023179380729977512, | |
| "loss": 1.2478, | |
| "step": 1340 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 0.000233523611831863, | |
| "loss": 1.1334, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 0.00023525341636395088, | |
| "loss": 1.1248, | |
| "step": 1360 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 0.00023698322089603876, | |
| "loss": 1.2183, | |
| "step": 1370 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 0.0002387130254281266, | |
| "loss": 1.4618, | |
| "step": 1380 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 0.00024044282996021449, | |
| "loss": 1.5796, | |
| "step": 1390 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 0.00024217263449230237, | |
| "loss": 1.4877, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 0.00024390243902439024, | |
| "loss": 0.9612, | |
| "step": 1410 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 0.0002456322435564781, | |
| "loss": 1.3634, | |
| "step": 1420 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 0.000247362048088566, | |
| "loss": 1.1377, | |
| "step": 1430 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 0.0002490918526206539, | |
| "loss": 1.1365, | |
| "step": 1440 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 0.00025082165715274176, | |
| "loss": 1.2433, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 0.00025255146168482964, | |
| "loss": 1.2164, | |
| "step": 1460 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 0.0002542812662169175, | |
| "loss": 1.0668, | |
| "step": 1470 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 0.0002560110707490054, | |
| "loss": 1.174, | |
| "step": 1480 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 0.0002577408752810932, | |
| "loss": 1.1307, | |
| "step": 1490 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 0.0002594706798131811, | |
| "loss": 1.2406, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 0.000261200484345269, | |
| "loss": 1.2542, | |
| "step": 1510 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 0.00026293028887735685, | |
| "loss": 0.9126, | |
| "step": 1520 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 0.00026466009340944473, | |
| "loss": 1.1105, | |
| "step": 1530 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 0.0002663898979415326, | |
| "loss": 0.8491, | |
| "step": 1540 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 0.0002681197024736205, | |
| "loss": 1.4498, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 0.00026984950700570837, | |
| "loss": 1.3299, | |
| "step": 1560 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 0.00027157931153779625, | |
| "loss": 1.2222, | |
| "step": 1570 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 0.0002733091160698841, | |
| "loss": 0.9812, | |
| "step": 1580 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 0.000275038920601972, | |
| "loss": 1.2824, | |
| "step": 1590 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 0.0002767687251340599, | |
| "loss": 1.1105, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 0.00027849852966614776, | |
| "loss": 1.0978, | |
| "step": 1610 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 0.00028022833419823564, | |
| "loss": 1.1159, | |
| "step": 1620 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 0.0002819581387303235, | |
| "loss": 1.1441, | |
| "step": 1630 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 0.0002836879432624114, | |
| "loss": 1.2566, | |
| "step": 1640 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 0.0002854177477944992, | |
| "loss": 1.1327, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 0.0002871475523265871, | |
| "loss": 1.232, | |
| "step": 1660 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 0.000288877356858675, | |
| "loss": 1.0974, | |
| "step": 1670 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 0.00029060716139076286, | |
| "loss": 1.0083, | |
| "step": 1680 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 0.00029233696592285074, | |
| "loss": 1.5513, | |
| "step": 1690 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 0.0002940667704549386, | |
| "loss": 1.3731, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 0.00029579657498702644, | |
| "loss": 1.1877, | |
| "step": 1710 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 0.0002975263795191143, | |
| "loss": 1.199, | |
| "step": 1720 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 0.0002992561840512022, | |
| "loss": 1.4768, | |
| "step": 1730 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 0.0003009859885832901, | |
| "loss": 1.7352, | |
| "step": 1740 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 0.00030271579311537796, | |
| "loss": 1.3652, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 0.00030444559764746583, | |
| "loss": 1.1335, | |
| "step": 1760 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 0.0003061754021795537, | |
| "loss": 1.2259, | |
| "step": 1770 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 0.0003079052067116416, | |
| "loss": 1.3024, | |
| "step": 1780 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 0.00030963501124372947, | |
| "loss": 1.4433, | |
| "step": 1790 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 0.00031136481577581735, | |
| "loss": 1.1668, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 0.0003130946203079052, | |
| "loss": 1.3298, | |
| "step": 1810 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 0.00031482442483999305, | |
| "loss": 1.2601, | |
| "step": 1820 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 0.00031655422937208093, | |
| "loss": 1.1204, | |
| "step": 1830 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 0.0003182840339041688, | |
| "loss": 1.2142, | |
| "step": 1840 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 0.0003200138384362567, | |
| "loss": 0.93, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 0.00032174364296834457, | |
| "loss": 1.4335, | |
| "step": 1860 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 0.00032347344750043245, | |
| "loss": 1.3081, | |
| "step": 1870 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 0.0003252032520325203, | |
| "loss": 1.2692, | |
| "step": 1880 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 0.0003269330565646082, | |
| "loss": 1.2172, | |
| "step": 1890 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 0.0003286628610966961, | |
| "loss": 1.2254, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 0.00033039266562878396, | |
| "loss": 1.1448, | |
| "step": 1910 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 0.00033212247016087184, | |
| "loss": 1.2473, | |
| "step": 1920 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": { | |
| "accuracy": 0.5909208819714656 | |
| }, | |
| "eval_f1": { | |
| "f1": 0.5910435702573883 | |
| }, | |
| "eval_loss": 0.9808588624000549, | |
| "eval_precision": { | |
| "precision": 0.6661654194259725 | |
| }, | |
| "eval_recall": { | |
| "recall": 0.6016064778916844 | |
| }, | |
| "eval_runtime": 243.6104, | |
| "eval_samples_per_second": 15.824, | |
| "eval_steps_per_second": 7.914, | |
| "step": 1927 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 0.0003338522746929597, | |
| "loss": 1.1687, | |
| "step": 1930 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 0.0003355820792250476, | |
| "loss": 1.4294, | |
| "step": 1940 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 0.0003373118837571355, | |
| "loss": 1.2866, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 0.0003390416882892233, | |
| "loss": 1.0858, | |
| "step": 1960 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 0.0003407714928213112, | |
| "loss": 1.1465, | |
| "step": 1970 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 0.00034250129735339906, | |
| "loss": 1.2876, | |
| "step": 1980 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 0.00034423110188548694, | |
| "loss": 1.0266, | |
| "step": 1990 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 0.0003459609064175748, | |
| "loss": 1.1494, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 0.0003476907109496627, | |
| "loss": 1.5705, | |
| "step": 2010 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 0.00034942051548175057, | |
| "loss": 0.9916, | |
| "step": 2020 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 0.00035115032001383845, | |
| "loss": 1.2183, | |
| "step": 2030 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 0.00035288012454592633, | |
| "loss": 1.1186, | |
| "step": 2040 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 0.0003546099290780142, | |
| "loss": 1.1386, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 0.0003563397336101021, | |
| "loss": 1.413, | |
| "step": 2060 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 0.00035806953814218997, | |
| "loss": 1.4434, | |
| "step": 2070 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 0.00035979934267427784, | |
| "loss": 1.3144, | |
| "step": 2080 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 0.0003615291472063657, | |
| "loss": 1.4108, | |
| "step": 2090 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 0.0003632589517384536, | |
| "loss": 1.2305, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 0.0003649887562705415, | |
| "loss": 1.2335, | |
| "step": 2110 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 0.0003667185608026293, | |
| "loss": 1.3617, | |
| "step": 2120 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 0.0003684483653347172, | |
| "loss": 1.1457, | |
| "step": 2130 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 0.00037017816986680506, | |
| "loss": 1.0253, | |
| "step": 2140 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 0.00037190797439889294, | |
| "loss": 1.1486, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 0.0003736377789309808, | |
| "loss": 1.2603, | |
| "step": 2160 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 0.0003753675834630687, | |
| "loss": 1.25, | |
| "step": 2170 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 0.0003770973879951566, | |
| "loss": 1.5531, | |
| "step": 2180 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 0.0003788271925272444, | |
| "loss": 1.3442, | |
| "step": 2190 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 0.0003805569970593323, | |
| "loss": 1.1317, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 0.00038228680159142016, | |
| "loss": 1.3271, | |
| "step": 2210 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 0.00038401660612350804, | |
| "loss": 1.3165, | |
| "step": 2220 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 0.0003857464106555959, | |
| "loss": 1.2377, | |
| "step": 2230 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 0.0003874762151876838, | |
| "loss": 1.2677, | |
| "step": 2240 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 0.0003892060197197717, | |
| "loss": 1.2756, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 0.00039093582425185955, | |
| "loss": 1.0193, | |
| "step": 2260 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 0.00039266562878394743, | |
| "loss": 1.239, | |
| "step": 2270 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 0.00039439543331603526, | |
| "loss": 1.6044, | |
| "step": 2280 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 0.00039612523784812313, | |
| "loss": 1.2396, | |
| "step": 2290 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 0.000397855042380211, | |
| "loss": 1.0345, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 0.0003995848469122989, | |
| "loss": 1.3125, | |
| "step": 2310 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 0.00040131465144438677, | |
| "loss": 1.4382, | |
| "step": 2320 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 0.00040304445597647465, | |
| "loss": 1.1232, | |
| "step": 2330 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 0.00040477426050856253, | |
| "loss": 1.3415, | |
| "step": 2340 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 0.0004065040650406504, | |
| "loss": 1.3517, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 0.0004082338695727383, | |
| "loss": 1.1871, | |
| "step": 2360 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 0.00040996367410482616, | |
| "loss": 1.541, | |
| "step": 2370 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 0.00041169347863691404, | |
| "loss": 1.0871, | |
| "step": 2380 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 0.0004134232831690019, | |
| "loss": 1.0727, | |
| "step": 2390 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 0.0004151530877010898, | |
| "loss": 1.2664, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 0.0004168828922331777, | |
| "loss": 0.9518, | |
| "step": 2410 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 0.00041861269676526556, | |
| "loss": 1.5851, | |
| "step": 2420 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 0.0004203425012973534, | |
| "loss": 1.3433, | |
| "step": 2430 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 0.00042207230582944126, | |
| "loss": 1.1357, | |
| "step": 2440 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 0.00042380211036152914, | |
| "loss": 1.3124, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 0.000425531914893617, | |
| "loss": 1.3198, | |
| "step": 2460 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 0.0004272617194257049, | |
| "loss": 1.31, | |
| "step": 2470 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 0.0004289915239577928, | |
| "loss": 1.2563, | |
| "step": 2480 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 0.00043072132848988065, | |
| "loss": 1.2553, | |
| "step": 2490 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 0.00043245113302196853, | |
| "loss": 1.3314, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 0.0004341809375540564, | |
| "loss": 1.8077, | |
| "step": 2510 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 0.0004359107420861443, | |
| "loss": 1.7739, | |
| "step": 2520 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 0.00043764054661823217, | |
| "loss": 1.3835, | |
| "step": 2530 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 0.00043937035115032005, | |
| "loss": 1.3689, | |
| "step": 2540 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 0.0004411001556824079, | |
| "loss": 1.4259, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 0.0004428299602144958, | |
| "loss": 1.0997, | |
| "step": 2560 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 0.0004445597647465837, | |
| "loss": 0.944, | |
| "step": 2570 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 0.00044628956927867156, | |
| "loss": 1.431, | |
| "step": 2580 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 0.0004480193738107594, | |
| "loss": 1.492, | |
| "step": 2590 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 0.00044974917834284726, | |
| "loss": 1.0205, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 0.00045147898287493514, | |
| "loss": 1.2334, | |
| "step": 2610 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 0.000453208787407023, | |
| "loss": 1.0434, | |
| "step": 2620 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 0.0004549385919391109, | |
| "loss": 1.4912, | |
| "step": 2630 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 0.0004566683964711988, | |
| "loss": 1.2939, | |
| "step": 2640 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 0.00045839820100328666, | |
| "loss": 1.0389, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 0.00046012800553537454, | |
| "loss": 1.1925, | |
| "step": 2660 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 0.0004618578100674624, | |
| "loss": 1.5455, | |
| "step": 2670 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 0.00046358761459955024, | |
| "loss": 1.2916, | |
| "step": 2680 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 0.0004653174191316381, | |
| "loss": 1.253, | |
| "step": 2690 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 0.000467047223663726, | |
| "loss": 1.377, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 0.0004687770281958139, | |
| "loss": 1.4184, | |
| "step": 2710 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 0.00047050683272790175, | |
| "loss": 1.3838, | |
| "step": 2720 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 0.00047223663725998963, | |
| "loss": 1.1912, | |
| "step": 2730 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 0.0004739664417920775, | |
| "loss": 1.0304, | |
| "step": 2740 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 0.00047569624632416534, | |
| "loss": 1.3583, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 0.0004774260508562532, | |
| "loss": 1.3856, | |
| "step": 2760 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 0.0004791558553883411, | |
| "loss": 1.5254, | |
| "step": 2770 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 0.00048088565992042897, | |
| "loss": 1.2068, | |
| "step": 2780 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 0.00048261546445251685, | |
| "loss": 1.3795, | |
| "step": 2790 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 0.00048434526898460473, | |
| "loss": 1.5103, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 0.0004860750735166926, | |
| "loss": 1.6531, | |
| "step": 2810 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 0.0004878048780487805, | |
| "loss": 1.3675, | |
| "step": 2820 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 0.0004895346825808684, | |
| "loss": 1.6244, | |
| "step": 2830 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 0.0004912644871129562, | |
| "loss": 1.8395, | |
| "step": 2840 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 0.0004929942916450441, | |
| "loss": 1.6552, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 0.000494724096177132, | |
| "loss": 1.6432, | |
| "step": 2860 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 0.0004964539007092199, | |
| "loss": 1.4839, | |
| "step": 2870 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 0.0004981837052413078, | |
| "loss": 1.403, | |
| "step": 2880 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 0.0004999135097733956, | |
| "loss": 1.8624, | |
| "step": 2890 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 0.0005016433143054835, | |
| "loss": 1.4894, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 0.0005033731188375714, | |
| "loss": 2.1144, | |
| "step": 2910 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 0.0005051029233696593, | |
| "loss": 1.7605, | |
| "step": 2920 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 0.000506832727901747, | |
| "loss": 1.6352, | |
| "step": 2930 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 0.000508562532433835, | |
| "loss": 1.6826, | |
| "step": 2940 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 0.0005102923369659228, | |
| "loss": 1.8023, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 0.0005120221414980108, | |
| "loss": 1.6957, | |
| "step": 2960 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 0.0005137519460300986, | |
| "loss": 2.1667, | |
| "step": 2970 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 0.0005154817505621864, | |
| "loss": 2.1662, | |
| "step": 2980 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 0.0005172115550942743, | |
| "loss": 1.8724, | |
| "step": 2990 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 0.0005189413596263622, | |
| "loss": 1.7312, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 0.0005206711641584501, | |
| "loss": 2.0953, | |
| "step": 3010 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 0.000522400968690538, | |
| "loss": 1.8218, | |
| "step": 3020 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 0.0005241307732226258, | |
| "loss": 1.9732, | |
| "step": 3030 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 0.0005258605777547137, | |
| "loss": 2.229, | |
| "step": 3040 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 0.0005275903822868016, | |
| "loss": 1.9946, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 0.0005293201868188895, | |
| "loss": 2.162, | |
| "step": 3060 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 0.0005310499913509773, | |
| "loss": 2.0647, | |
| "step": 3070 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 0.0005327797958830652, | |
| "loss": 2.0291, | |
| "step": 3080 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 0.0005345096004151531, | |
| "loss": 2.0622, | |
| "step": 3090 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 0.000536239404947241, | |
| "loss": 1.9414, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 0.0005379692094793289, | |
| "loss": 2.0906, | |
| "step": 3110 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 0.0005396990140114167, | |
| "loss": 2.0018, | |
| "step": 3120 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 0.0005414288185435046, | |
| "loss": 1.9754, | |
| "step": 3130 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 0.0005431586230755925, | |
| "loss": 1.949, | |
| "step": 3140 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 0.0005448884276076803, | |
| "loss": 1.918, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 0.0005466182321397683, | |
| "loss": 2.1205, | |
| "step": 3160 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "learning_rate": 0.000548348036671856, | |
| "loss": 1.8933, | |
| "step": 3170 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "learning_rate": 0.000550077841203944, | |
| "loss": 1.9666, | |
| "step": 3180 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 0.0005518076457360318, | |
| "loss": 1.9286, | |
| "step": 3190 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 0.0005535374502681198, | |
| "loss": 1.9417, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 1.67, | |
| "learning_rate": 0.0005552672548002075, | |
| "loss": 1.8296, | |
| "step": 3210 | |
| }, | |
| { | |
| "epoch": 1.67, | |
| "learning_rate": 0.0005569970593322955, | |
| "loss": 1.9005, | |
| "step": 3220 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 0.0005587268638643833, | |
| "loss": 1.954, | |
| "step": 3230 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 0.0005604566683964713, | |
| "loss": 1.9252, | |
| "step": 3240 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 0.000562186472928559, | |
| "loss": 1.9827, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 0.000563916277460647, | |
| "loss": 1.7931, | |
| "step": 3260 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 0.0005656460819927348, | |
| "loss": 1.8842, | |
| "step": 3270 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 0.0005673758865248228, | |
| "loss": 1.8838, | |
| "step": 3280 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 0.0005691056910569106, | |
| "loss": 1.7902, | |
| "step": 3290 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 0.0005708354955889984, | |
| "loss": 1.9054, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 1.72, | |
| "learning_rate": 0.0005725653001210863, | |
| "loss": 1.972, | |
| "step": 3310 | |
| }, | |
| { | |
| "epoch": 1.72, | |
| "learning_rate": 0.0005742951046531742, | |
| "loss": 1.999, | |
| "step": 3320 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 0.0005760249091852621, | |
| "loss": 1.9566, | |
| "step": 3330 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 0.00057775471371735, | |
| "loss": 1.8273, | |
| "step": 3340 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 0.0005794845182494378, | |
| "loss": 1.9554, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 0.0005812143227815257, | |
| "loss": 2.0318, | |
| "step": 3360 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 0.0005829441273136136, | |
| "loss": 1.9838, | |
| "step": 3370 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 0.0005846739318457015, | |
| "loss": 1.9402, | |
| "step": 3380 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 0.0005864037363777894, | |
| "loss": 1.9904, | |
| "step": 3390 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 0.0005881335409098772, | |
| "loss": 1.9198, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 0.0005898633454419651, | |
| "loss": 1.9405, | |
| "step": 3410 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 0.0005915931499740529, | |
| "loss": 1.8445, | |
| "step": 3420 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 0.0005933229545061409, | |
| "loss": 1.9201, | |
| "step": 3430 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 0.0005950527590382286, | |
| "loss": 1.8237, | |
| "step": 3440 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 0.0005967825635703165, | |
| "loss": 1.7255, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 0.0005985123681024044, | |
| "loss": 1.844, | |
| "step": 3460 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 0.0006002421726344923, | |
| "loss": 2.0433, | |
| "step": 3470 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 0.0006019719771665802, | |
| "loss": 1.7753, | |
| "step": 3480 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 0.000603701781698668, | |
| "loss": 1.858, | |
| "step": 3490 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 0.0006054315862307559, | |
| "loss": 1.9348, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 0.0006071613907628438, | |
| "loss": 1.8196, | |
| "step": 3510 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 0.0006088911952949317, | |
| "loss": 2.0001, | |
| "step": 3520 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 0.0006106209998270195, | |
| "loss": 1.806, | |
| "step": 3530 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 0.0006123508043591074, | |
| "loss": 1.8369, | |
| "step": 3540 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 0.0006140806088911953, | |
| "loss": 1.8824, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 0.0006158104134232832, | |
| "loss": 1.8376, | |
| "step": 3560 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 0.0006175402179553711, | |
| "loss": 1.7554, | |
| "step": 3570 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 0.0006192700224874589, | |
| "loss": 1.835, | |
| "step": 3580 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 0.0006209998270195468, | |
| "loss": 1.7473, | |
| "step": 3590 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 0.0006227296315516347, | |
| "loss": 1.8827, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 0.0006244594360837226, | |
| "loss": 1.7536, | |
| "step": 3610 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 0.0006261892406158103, | |
| "loss": 1.8997, | |
| "step": 3620 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 0.0006279190451478983, | |
| "loss": 1.813, | |
| "step": 3630 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 0.0006296488496799861, | |
| "loss": 1.8474, | |
| "step": 3640 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 0.0006313786542120741, | |
| "loss": 1.8263, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 0.0006331084587441619, | |
| "loss": 1.8817, | |
| "step": 3660 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 0.0006348382632762499, | |
| "loss": 1.9347, | |
| "step": 3670 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 0.0006365680678083376, | |
| "loss": 1.786, | |
| "step": 3680 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 0.0006382978723404256, | |
| "loss": 2.197, | |
| "step": 3690 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 0.0006400276768725134, | |
| "loss": 1.9695, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 0.0006417574814046014, | |
| "loss": 1.8939, | |
| "step": 3710 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 0.0006434872859366891, | |
| "loss": 1.8662, | |
| "step": 3720 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 0.0006452170904687771, | |
| "loss": 1.8054, | |
| "step": 3730 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 0.0006469468950008649, | |
| "loss": 1.8351, | |
| "step": 3740 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 0.0006486766995329529, | |
| "loss": 1.9102, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 0.0006504065040650406, | |
| "loss": 1.8462, | |
| "step": 3760 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 0.0006521363085971285, | |
| "loss": 1.894, | |
| "step": 3770 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 0.0006538661131292164, | |
| "loss": 1.7559, | |
| "step": 3780 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 0.0006555959176613043, | |
| "loss": 1.8255, | |
| "step": 3790 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 0.0006573257221933922, | |
| "loss": 1.7118, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 0.00065905552672548, | |
| "loss": 1.8048, | |
| "step": 3810 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 0.0006607853312575679, | |
| "loss": 1.7759, | |
| "step": 3820 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 0.0006625151357896558, | |
| "loss": 1.7974, | |
| "step": 3830 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 0.0006642449403217437, | |
| "loss": 1.7673, | |
| "step": 3840 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 0.0006659747448538316, | |
| "loss": 1.7959, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": { | |
| "accuracy": 0.26848249027237353 | |
| }, | |
| "eval_f1": { | |
| "f1": 0.17635435227079216 | |
| }, | |
| "eval_loss": 1.8579649925231934, | |
| "eval_precision": { | |
| "precision": 0.15600300783137744 | |
| }, | |
| "eval_recall": { | |
| "recall": 0.2748312652287413 | |
| }, | |
| "eval_runtime": 115.151, | |
| "eval_samples_per_second": 33.478, | |
| "eval_steps_per_second": 16.743, | |
| "step": 3854 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 0.0006677045493859194, | |
| "loss": 1.9251, | |
| "step": 3860 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 0.0006694343539180073, | |
| "loss": 1.7743, | |
| "step": 3870 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 0.0006711641584500952, | |
| "loss": 1.8658, | |
| "step": 3880 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 0.000672893962982183, | |
| "loss": 1.8682, | |
| "step": 3890 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 0.000674623767514271, | |
| "loss": 1.9279, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 0.0006763535720463587, | |
| "loss": 2.0145, | |
| "step": 3910 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 0.0006780833765784466, | |
| "loss": 2.0323, | |
| "step": 3920 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 0.0006798131811105345, | |
| "loss": 1.9978, | |
| "step": 3930 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 0.0006815429856426224, | |
| "loss": 1.9258, | |
| "step": 3940 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 0.0006832727901747102, | |
| "loss": 1.9777, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 0.0006850025947067981, | |
| "loss": 1.9442, | |
| "step": 3960 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 0.000686732399238886, | |
| "loss": 1.9772, | |
| "step": 3970 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 0.0006884622037709739, | |
| "loss": 1.9368, | |
| "step": 3980 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 0.0006901920083030618, | |
| "loss": 1.8948, | |
| "step": 3990 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 0.0006919218128351496, | |
| "loss": 1.9485, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 0.0006936516173672375, | |
| "loss": 1.8472, | |
| "step": 4010 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 0.0006953814218993254, | |
| "loss": 1.7604, | |
| "step": 4020 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 0.0006971112264314133, | |
| "loss": 1.8371, | |
| "step": 4030 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 0.0006988410309635011, | |
| "loss": 1.7566, | |
| "step": 4040 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 0.000700570835495589, | |
| "loss": 1.9021, | |
| "step": 4050 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 0.0007023006400276769, | |
| "loss": 1.744, | |
| "step": 4060 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 0.0007040304445597648, | |
| "loss": 1.8639, | |
| "step": 4070 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 0.0007057602490918527, | |
| "loss": 1.857, | |
| "step": 4080 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 0.0007074900536239404, | |
| "loss": 1.8229, | |
| "step": 4090 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 0.0007092198581560284, | |
| "loss": 1.8733, | |
| "step": 4100 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 0.0007109496626881162, | |
| "loss": 1.8442, | |
| "step": 4110 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 0.0007126794672202042, | |
| "loss": 1.8209, | |
| "step": 4120 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 0.0007144092717522919, | |
| "loss": 1.7018, | |
| "step": 4130 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 0.0007161390762843799, | |
| "loss": 1.8013, | |
| "step": 4140 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 0.0007178688808164677, | |
| "loss": 1.9447, | |
| "step": 4150 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 0.0007195986853485557, | |
| "loss": 1.8768, | |
| "step": 4160 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 0.0007213284898806435, | |
| "loss": 1.822, | |
| "step": 4170 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 0.0007230582944127314, | |
| "loss": 1.77, | |
| "step": 4180 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 0.0007247880989448192, | |
| "loss": 1.8653, | |
| "step": 4190 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 0.0007265179034769072, | |
| "loss": 1.7695, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 0.000728247708008995, | |
| "loss": 1.8426, | |
| "step": 4210 | |
| }, | |
| { | |
| "epoch": 2.19, | |
| "learning_rate": 0.000729977512541083, | |
| "loss": 1.8379, | |
| "step": 4220 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 0.0007317073170731707, | |
| "loss": 1.8384, | |
| "step": 4230 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 0.0007334371216052586, | |
| "loss": 1.9245, | |
| "step": 4240 | |
| }, | |
| { | |
| "epoch": 2.21, | |
| "learning_rate": 0.0007351669261373465, | |
| "loss": 1.8017, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 2.21, | |
| "learning_rate": 0.0007368967306694344, | |
| "loss": 1.8151, | |
| "step": 4260 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 0.0007386265352015222, | |
| "loss": 1.8356, | |
| "step": 4270 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 0.0007403563397336101, | |
| "loss": 1.9064, | |
| "step": 4280 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 0.000742086144265698, | |
| "loss": 1.8325, | |
| "step": 4290 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 0.0007438159487977859, | |
| "loss": 1.8079, | |
| "step": 4300 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 0.0007455457533298738, | |
| "loss": 1.7581, | |
| "step": 4310 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 0.0007472755578619616, | |
| "loss": 1.9346, | |
| "step": 4320 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 0.0007490053623940495, | |
| "loss": 1.8706, | |
| "step": 4330 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 0.0007507351669261374, | |
| "loss": 1.8314, | |
| "step": 4340 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 0.0007524649714582253, | |
| "loss": 1.8881, | |
| "step": 4350 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 0.0007541947759903132, | |
| "loss": 1.8785, | |
| "step": 4360 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 0.000755924580522401, | |
| "loss": 1.7916, | |
| "step": 4370 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 0.0007576543850544888, | |
| "loss": 1.8089, | |
| "step": 4380 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 0.0007593841895865767, | |
| "loss": 1.7736, | |
| "step": 4390 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 0.0007611139941186646, | |
| "loss": 1.8184, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 0.0007628437986507524, | |
| "loss": 1.8084, | |
| "step": 4410 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 0.0007645736031828403, | |
| "loss": 2.0347, | |
| "step": 4420 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 0.0007663034077149282, | |
| "loss": 1.7984, | |
| "step": 4430 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 0.0007680332122470161, | |
| "loss": 1.7799, | |
| "step": 4440 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 0.000769763016779104, | |
| "loss": 1.7603, | |
| "step": 4450 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 0.0007714928213111918, | |
| "loss": 1.6663, | |
| "step": 4460 | |
| }, | |
| { | |
| "epoch": 2.32, | |
| "learning_rate": 0.0007732226258432797, | |
| "loss": 1.9232, | |
| "step": 4470 | |
| }, | |
| { | |
| "epoch": 2.32, | |
| "learning_rate": 0.0007749524303753676, | |
| "loss": 1.8086, | |
| "step": 4480 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 0.0007766822349074555, | |
| "loss": 1.7616, | |
| "step": 4490 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 0.0007784120394395433, | |
| "loss": 1.8018, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 0.0007801418439716312, | |
| "loss": 1.8413, | |
| "step": 4510 | |
| }, | |
| { | |
| "epoch": 2.35, | |
| "learning_rate": 0.0007818716485037191, | |
| "loss": 2.1126, | |
| "step": 4520 | |
| }, | |
| { | |
| "epoch": 2.35, | |
| "learning_rate": 0.000783601453035807, | |
| "loss": 1.9105, | |
| "step": 4530 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 0.0007853312575678949, | |
| "loss": 1.8485, | |
| "step": 4540 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 0.0007870610620999827, | |
| "loss": 1.8835, | |
| "step": 4550 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "learning_rate": 0.0007887908666320705, | |
| "loss": 1.7881, | |
| "step": 4560 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "learning_rate": 0.0007905206711641585, | |
| "loss": 1.7013, | |
| "step": 4570 | |
| }, | |
| { | |
| "epoch": 2.38, | |
| "learning_rate": 0.0007922504756962463, | |
| "loss": 1.7657, | |
| "step": 4580 | |
| }, | |
| { | |
| "epoch": 2.38, | |
| "learning_rate": 0.0007939802802283343, | |
| "loss": 1.9326, | |
| "step": 4590 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 0.000795710084760422, | |
| "loss": 1.6888, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 0.00079743988929251, | |
| "loss": 1.8114, | |
| "step": 4610 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 0.0007991696938245978, | |
| "loss": 1.9245, | |
| "step": 4620 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 0.0008008994983566858, | |
| "loss": 1.8287, | |
| "step": 4630 | |
| }, | |
| { | |
| "epoch": 2.41, | |
| "learning_rate": 0.0008026293028887735, | |
| "loss": 1.8763, | |
| "step": 4640 | |
| }, | |
| { | |
| "epoch": 2.41, | |
| "learning_rate": 0.0008043591074208615, | |
| "loss": 1.6896, | |
| "step": 4650 | |
| }, | |
| { | |
| "epoch": 2.42, | |
| "learning_rate": 0.0008060889119529493, | |
| "loss": 1.9996, | |
| "step": 4660 | |
| }, | |
| { | |
| "epoch": 2.42, | |
| "learning_rate": 0.0008078187164850373, | |
| "loss": 1.7345, | |
| "step": 4670 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 0.0008095485210171251, | |
| "loss": 1.7908, | |
| "step": 4680 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 0.000811278325549213, | |
| "loss": 1.8212, | |
| "step": 4690 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 0.0008130081300813008, | |
| "loss": 1.8434, | |
| "step": 4700 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 0.0008147379346133887, | |
| "loss": 1.9086, | |
| "step": 4710 | |
| }, | |
| { | |
| "epoch": 2.45, | |
| "learning_rate": 0.0008164677391454766, | |
| "loss": 1.9098, | |
| "step": 4720 | |
| }, | |
| { | |
| "epoch": 2.45, | |
| "learning_rate": 0.0008181975436775644, | |
| "loss": 1.7695, | |
| "step": 4730 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 0.0008199273482096523, | |
| "loss": 1.7415, | |
| "step": 4740 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 0.0008216571527417402, | |
| "loss": 1.8291, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 2.47, | |
| "learning_rate": 0.0008233869572738281, | |
| "loss": 1.7842, | |
| "step": 4760 | |
| }, | |
| { | |
| "epoch": 2.48, | |
| "learning_rate": 0.000825116761805916, | |
| "loss": 1.8745, | |
| "step": 4770 | |
| }, | |
| { | |
| "epoch": 2.48, | |
| "learning_rate": 0.0008268465663380038, | |
| "loss": 1.7075, | |
| "step": 4780 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "learning_rate": 0.0008285763708700917, | |
| "loss": 1.8133, | |
| "step": 4790 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "learning_rate": 0.0008303061754021796, | |
| "loss": 1.7318, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 0.0008320359799342675, | |
| "loss": 1.874, | |
| "step": 4810 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 0.0008337657844663554, | |
| "loss": 1.8895, | |
| "step": 4820 | |
| }, | |
| { | |
| "epoch": 2.51, | |
| "learning_rate": 0.0008354955889984432, | |
| "loss": 1.8485, | |
| "step": 4830 | |
| }, | |
| { | |
| "epoch": 2.51, | |
| "learning_rate": 0.0008372253935305311, | |
| "loss": 1.7643, | |
| "step": 4840 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "learning_rate": 0.000838955198062619, | |
| "loss": 1.8477, | |
| "step": 4850 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "learning_rate": 0.0008406850025947068, | |
| "loss": 1.8611, | |
| "step": 4860 | |
| }, | |
| { | |
| "epoch": 2.53, | |
| "learning_rate": 0.0008424148071267946, | |
| "loss": 1.8908, | |
| "step": 4870 | |
| }, | |
| { | |
| "epoch": 2.53, | |
| "learning_rate": 0.0008441446116588825, | |
| "loss": 1.8017, | |
| "step": 4880 | |
| }, | |
| { | |
| "epoch": 2.54, | |
| "learning_rate": 0.0008458744161909704, | |
| "loss": 1.9186, | |
| "step": 4890 | |
| }, | |
| { | |
| "epoch": 2.54, | |
| "learning_rate": 0.0008476042207230583, | |
| "loss": 1.8141, | |
| "step": 4900 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 0.0008493340252551462, | |
| "loss": 1.8298, | |
| "step": 4910 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 0.000851063829787234, | |
| "loss": 1.7711, | |
| "step": 4920 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "learning_rate": 0.0008527936343193219, | |
| "loss": 1.8265, | |
| "step": 4930 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "learning_rate": 0.0008545234388514098, | |
| "loss": 1.7649, | |
| "step": 4940 | |
| }, | |
| { | |
| "epoch": 2.57, | |
| "learning_rate": 0.0008562532433834977, | |
| "loss": 1.9159, | |
| "step": 4950 | |
| }, | |
| { | |
| "epoch": 2.57, | |
| "learning_rate": 0.0008579830479155855, | |
| "loss": 1.7968, | |
| "step": 4960 | |
| }, | |
| { | |
| "epoch": 2.58, | |
| "learning_rate": 0.0008597128524476734, | |
| "loss": 1.807, | |
| "step": 4970 | |
| }, | |
| { | |
| "epoch": 2.58, | |
| "learning_rate": 0.0008614426569797613, | |
| "loss": 1.7633, | |
| "step": 4980 | |
| }, | |
| { | |
| "epoch": 2.59, | |
| "learning_rate": 0.0008631724615118492, | |
| "loss": 1.8019, | |
| "step": 4990 | |
| }, | |
| { | |
| "epoch": 2.59, | |
| "learning_rate": 0.0008649022660439371, | |
| "loss": 1.8697, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "learning_rate": 0.0008666320705760249, | |
| "loss": 1.8348, | |
| "step": 5010 | |
| }, | |
| { | |
| "epoch": 2.61, | |
| "learning_rate": 0.0008683618751081128, | |
| "loss": 1.805, | |
| "step": 5020 | |
| }, | |
| { | |
| "epoch": 2.61, | |
| "learning_rate": 0.0008700916796402006, | |
| "loss": 1.7346, | |
| "step": 5030 | |
| }, | |
| { | |
| "epoch": 2.62, | |
| "learning_rate": 0.0008718214841722886, | |
| "loss": 1.9219, | |
| "step": 5040 | |
| }, | |
| { | |
| "epoch": 2.62, | |
| "learning_rate": 0.0008735512887043763, | |
| "loss": 2.0954, | |
| "step": 5050 | |
| }, | |
| { | |
| "epoch": 2.63, | |
| "learning_rate": 0.0008752810932364643, | |
| "loss": 1.9262, | |
| "step": 5060 | |
| }, | |
| { | |
| "epoch": 2.63, | |
| "learning_rate": 0.0008770108977685521, | |
| "loss": 1.8491, | |
| "step": 5070 | |
| }, | |
| { | |
| "epoch": 2.64, | |
| "learning_rate": 0.0008787407023006401, | |
| "loss": 2.0113, | |
| "step": 5080 | |
| }, | |
| { | |
| "epoch": 2.64, | |
| "learning_rate": 0.0008804705068327279, | |
| "loss": 1.9341, | |
| "step": 5090 | |
| }, | |
| { | |
| "epoch": 2.65, | |
| "learning_rate": 0.0008822003113648159, | |
| "loss": 1.8968, | |
| "step": 5100 | |
| }, | |
| { | |
| "epoch": 2.65, | |
| "learning_rate": 0.0008839301158969036, | |
| "loss": 2.0009, | |
| "step": 5110 | |
| }, | |
| { | |
| "epoch": 2.66, | |
| "learning_rate": 0.0008856599204289916, | |
| "loss": 1.9956, | |
| "step": 5120 | |
| }, | |
| { | |
| "epoch": 2.66, | |
| "learning_rate": 0.0008873897249610794, | |
| "loss": 1.9083, | |
| "step": 5130 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 0.0008891195294931674, | |
| "loss": 1.9274, | |
| "step": 5140 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 0.0008908493340252551, | |
| "loss": 1.8263, | |
| "step": 5150 | |
| }, | |
| { | |
| "epoch": 2.68, | |
| "learning_rate": 0.0008925791385573431, | |
| "loss": 1.8235, | |
| "step": 5160 | |
| }, | |
| { | |
| "epoch": 2.68, | |
| "learning_rate": 0.0008943089430894309, | |
| "loss": 1.8786, | |
| "step": 5170 | |
| }, | |
| { | |
| "epoch": 2.69, | |
| "learning_rate": 0.0008960387476215188, | |
| "loss": 1.8952, | |
| "step": 5180 | |
| }, | |
| { | |
| "epoch": 2.69, | |
| "learning_rate": 0.0008977685521536067, | |
| "loss": 1.7448, | |
| "step": 5190 | |
| }, | |
| { | |
| "epoch": 2.7, | |
| "learning_rate": 0.0008994983566856945, | |
| "loss": 1.9518, | |
| "step": 5200 | |
| }, | |
| { | |
| "epoch": 2.7, | |
| "learning_rate": 0.0009012281612177824, | |
| "loss": 1.8636, | |
| "step": 5210 | |
| }, | |
| { | |
| "epoch": 2.71, | |
| "learning_rate": 0.0009029579657498703, | |
| "loss": 1.9309, | |
| "step": 5220 | |
| }, | |
| { | |
| "epoch": 2.71, | |
| "learning_rate": 0.0009046877702819582, | |
| "loss": 1.8714, | |
| "step": 5230 | |
| }, | |
| { | |
| "epoch": 2.72, | |
| "learning_rate": 0.000906417574814046, | |
| "loss": 1.7624, | |
| "step": 5240 | |
| }, | |
| { | |
| "epoch": 2.72, | |
| "learning_rate": 0.0009081473793461339, | |
| "loss": 1.9421, | |
| "step": 5250 | |
| }, | |
| { | |
| "epoch": 2.73, | |
| "learning_rate": 0.0009098771838782218, | |
| "loss": 1.9204, | |
| "step": 5260 | |
| }, | |
| { | |
| "epoch": 2.73, | |
| "learning_rate": 0.0009116069884103097, | |
| "loss": 1.8561, | |
| "step": 5270 | |
| }, | |
| { | |
| "epoch": 2.74, | |
| "learning_rate": 0.0009133367929423976, | |
| "loss": 1.8766, | |
| "step": 5280 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 0.0009150665974744854, | |
| "loss": 1.8905, | |
| "step": 5290 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 0.0009167964020065733, | |
| "loss": 1.8466, | |
| "step": 5300 | |
| }, | |
| { | |
| "epoch": 2.76, | |
| "learning_rate": 0.0009185262065386612, | |
| "loss": 1.8735, | |
| "step": 5310 | |
| }, | |
| { | |
| "epoch": 2.76, | |
| "learning_rate": 0.0009202560110707491, | |
| "loss": 2.0758, | |
| "step": 5320 | |
| }, | |
| { | |
| "epoch": 2.77, | |
| "learning_rate": 0.0009219858156028368, | |
| "loss": 1.9582, | |
| "step": 5330 | |
| }, | |
| { | |
| "epoch": 2.77, | |
| "learning_rate": 0.0009237156201349248, | |
| "loss": 1.9317, | |
| "step": 5340 | |
| }, | |
| { | |
| "epoch": 2.78, | |
| "learning_rate": 0.0009254454246670126, | |
| "loss": 1.8719, | |
| "step": 5350 | |
| }, | |
| { | |
| "epoch": 2.78, | |
| "learning_rate": 0.0009271752291991005, | |
| "loss": 1.8377, | |
| "step": 5360 | |
| }, | |
| { | |
| "epoch": 2.79, | |
| "learning_rate": 0.0009289050337311884, | |
| "loss": 1.9785, | |
| "step": 5370 | |
| }, | |
| { | |
| "epoch": 2.79, | |
| "learning_rate": 0.0009306348382632762, | |
| "loss": 1.9259, | |
| "step": 5380 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "learning_rate": 0.0009323646427953641, | |
| "loss": 1.9097, | |
| "step": 5390 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "learning_rate": 0.000934094447327452, | |
| "loss": 1.9034, | |
| "step": 5400 | |
| }, | |
| { | |
| "epoch": 2.81, | |
| "learning_rate": 0.0009358242518595399, | |
| "loss": 1.8983, | |
| "step": 5410 | |
| }, | |
| { | |
| "epoch": 2.81, | |
| "learning_rate": 0.0009375540563916278, | |
| "loss": 1.8115, | |
| "step": 5420 | |
| }, | |
| { | |
| "epoch": 2.82, | |
| "learning_rate": 0.0009392838609237156, | |
| "loss": 1.8606, | |
| "step": 5430 | |
| }, | |
| { | |
| "epoch": 2.82, | |
| "learning_rate": 0.0009410136654558035, | |
| "loss": 1.8883, | |
| "step": 5440 | |
| }, | |
| { | |
| "epoch": 2.83, | |
| "learning_rate": 0.0009427434699878914, | |
| "loss": 1.7868, | |
| "step": 5450 | |
| }, | |
| { | |
| "epoch": 2.83, | |
| "learning_rate": 0.0009444732745199793, | |
| "loss": 1.9107, | |
| "step": 5460 | |
| }, | |
| { | |
| "epoch": 2.84, | |
| "learning_rate": 0.0009462030790520671, | |
| "loss": 1.9427, | |
| "step": 5470 | |
| }, | |
| { | |
| "epoch": 2.84, | |
| "learning_rate": 0.000947932883584155, | |
| "loss": 1.8159, | |
| "step": 5480 | |
| }, | |
| { | |
| "epoch": 2.85, | |
| "learning_rate": 0.0009496626881162429, | |
| "loss": 1.909, | |
| "step": 5490 | |
| }, | |
| { | |
| "epoch": 2.85, | |
| "learning_rate": 0.0009513924926483307, | |
| "loss": 1.8716, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 2.86, | |
| "learning_rate": 0.0009531222971804187, | |
| "loss": 1.9116, | |
| "step": 5510 | |
| }, | |
| { | |
| "epoch": 2.86, | |
| "learning_rate": 0.0009548521017125064, | |
| "loss": 1.8846, | |
| "step": 5520 | |
| }, | |
| { | |
| "epoch": 2.87, | |
| "learning_rate": 0.0009565819062445944, | |
| "loss": 1.8776, | |
| "step": 5530 | |
| }, | |
| { | |
| "epoch": 2.87, | |
| "learning_rate": 0.0009583117107766822, | |
| "loss": 1.828, | |
| "step": 5540 | |
| }, | |
| { | |
| "epoch": 2.88, | |
| "learning_rate": 0.0009600415153087702, | |
| "loss": 1.9392, | |
| "step": 5550 | |
| }, | |
| { | |
| "epoch": 2.89, | |
| "learning_rate": 0.0009617713198408579, | |
| "loss": 1.8823, | |
| "step": 5560 | |
| }, | |
| { | |
| "epoch": 2.89, | |
| "learning_rate": 0.0009635011243729459, | |
| "loss": 1.8782, | |
| "step": 5570 | |
| }, | |
| { | |
| "epoch": 2.9, | |
| "learning_rate": 0.0009652309289050337, | |
| "loss": 1.8898, | |
| "step": 5580 | |
| }, | |
| { | |
| "epoch": 2.9, | |
| "learning_rate": 0.0009669607334371217, | |
| "loss": 1.8107, | |
| "step": 5590 | |
| }, | |
| { | |
| "epoch": 2.91, | |
| "learning_rate": 0.0009686905379692095, | |
| "loss": 1.8026, | |
| "step": 5600 | |
| }, | |
| { | |
| "epoch": 2.91, | |
| "learning_rate": 0.0009704203425012974, | |
| "loss": 1.8616, | |
| "step": 5610 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "learning_rate": 0.0009721501470333852, | |
| "loss": 1.8443, | |
| "step": 5620 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "learning_rate": 0.0009738799515654732, | |
| "loss": 1.8465, | |
| "step": 5630 | |
| }, | |
| { | |
| "epoch": 2.93, | |
| "learning_rate": 0.000975609756097561, | |
| "loss": 1.8806, | |
| "step": 5640 | |
| }, | |
| { | |
| "epoch": 2.93, | |
| "learning_rate": 0.0009773395606296489, | |
| "loss": 1.8441, | |
| "step": 5650 | |
| }, | |
| { | |
| "epoch": 2.94, | |
| "learning_rate": 0.0009790693651617367, | |
| "loss": 1.8194, | |
| "step": 5660 | |
| }, | |
| { | |
| "epoch": 2.94, | |
| "learning_rate": 0.0009807991696938246, | |
| "loss": 1.9228, | |
| "step": 5670 | |
| }, | |
| { | |
| "epoch": 2.95, | |
| "learning_rate": 0.0009825289742259125, | |
| "loss": 1.792, | |
| "step": 5680 | |
| }, | |
| { | |
| "epoch": 2.95, | |
| "learning_rate": 0.0009842587787580004, | |
| "loss": 1.8211, | |
| "step": 5690 | |
| }, | |
| { | |
| "epoch": 2.96, | |
| "learning_rate": 0.0009859885832900882, | |
| "loss": 1.81, | |
| "step": 5700 | |
| }, | |
| { | |
| "epoch": 2.96, | |
| "learning_rate": 0.0009877183878221761, | |
| "loss": 1.8862, | |
| "step": 5710 | |
| }, | |
| { | |
| "epoch": 2.97, | |
| "learning_rate": 0.000989448192354264, | |
| "loss": 1.8373, | |
| "step": 5720 | |
| }, | |
| { | |
| "epoch": 2.97, | |
| "learning_rate": 0.0009911779968863519, | |
| "loss": 1.8639, | |
| "step": 5730 | |
| }, | |
| { | |
| "epoch": 2.98, | |
| "learning_rate": 0.0009929078014184398, | |
| "loss": 1.8718, | |
| "step": 5740 | |
| }, | |
| { | |
| "epoch": 2.98, | |
| "learning_rate": 0.0009946376059505276, | |
| "loss": 1.8258, | |
| "step": 5750 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "learning_rate": 0.0009963674104826155, | |
| "loss": 1.8564, | |
| "step": 5760 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "learning_rate": 0.0009980972150147034, | |
| "loss": 1.9462, | |
| "step": 5770 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 0.0009998270195467913, | |
| "loss": 1.8437, | |
| "step": 5780 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": { | |
| "accuracy": 0.1743190661478599 | |
| }, | |
| "eval_f1": { | |
| "f1": 0.1111482743173735 | |
| }, | |
| "eval_loss": 1.9242523908615112, | |
| "eval_precision": { | |
| "precision": 0.1516816437642854 | |
| }, | |
| "eval_recall": { | |
| "recall": 0.1887752678966561 | |
| }, | |
| "eval_runtime": 118.0197, | |
| "eval_samples_per_second": 32.664, | |
| "eval_steps_per_second": 16.336, | |
| "step": 5781 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 0.000999999926169841, | |
| "loss": 1.8482, | |
| "step": 5790 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 0.0009999996709545039, | |
| "loss": 1.8564, | |
| "step": 5800 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 0.0009999992334425987, | |
| "loss": 1.8474, | |
| "step": 5810 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 0.0009999986136342847, | |
| "loss": 1.8997, | |
| "step": 5820 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 0.000999997811529788, | |
| "loss": 1.8401, | |
| "step": 5830 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 0.000999996827129401, | |
| "loss": 1.8567, | |
| "step": 5840 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 0.0009999956604334825, | |
| "loss": 1.9804, | |
| "step": 5850 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 0.000999994311442458, | |
| "loss": 1.8427, | |
| "step": 5860 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 0.0009999927801568192, | |
| "loss": 1.8681, | |
| "step": 5870 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 0.0009999910665771247, | |
| "loss": 1.8561, | |
| "step": 5880 | |
| }, | |
| { | |
| "epoch": 3.06, | |
| "learning_rate": 0.000999989170703999, | |
| "loss": 1.9098, | |
| "step": 5890 | |
| }, | |
| { | |
| "epoch": 3.06, | |
| "learning_rate": 0.0009999870925381333, | |
| "loss": 1.935, | |
| "step": 5900 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "learning_rate": 0.0009999848320802853, | |
| "loss": 2.0045, | |
| "step": 5910 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "learning_rate": 0.0009999823893312794, | |
| "loss": 1.9466, | |
| "step": 5920 | |
| }, | |
| { | |
| "epoch": 3.08, | |
| "learning_rate": 0.0009999797642920058, | |
| "loss": 1.9047, | |
| "step": 5930 | |
| }, | |
| { | |
| "epoch": 3.08, | |
| "learning_rate": 0.0009999769569634218, | |
| "loss": 1.9901, | |
| "step": 5940 | |
| }, | |
| { | |
| "epoch": 3.09, | |
| "learning_rate": 0.000999973967346551, | |
| "loss": 1.9159, | |
| "step": 5950 | |
| }, | |
| { | |
| "epoch": 3.09, | |
| "learning_rate": 0.0009999707954424836, | |
| "loss": 1.9826, | |
| "step": 5960 | |
| }, | |
| { | |
| "epoch": 3.1, | |
| "learning_rate": 0.0009999674412523754, | |
| "loss": 1.9556, | |
| "step": 5970 | |
| }, | |
| { | |
| "epoch": 3.1, | |
| "learning_rate": 0.0009999639047774498, | |
| "loss": 1.9591, | |
| "step": 5980 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "learning_rate": 0.000999960186018996, | |
| "loss": 1.9551, | |
| "step": 5990 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "learning_rate": 0.0009999562849783699, | |
| "loss": 1.937, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 3.12, | |
| "learning_rate": 0.0009999522016569938, | |
| "loss": 1.9167, | |
| "step": 6010 | |
| }, | |
| { | |
| "epoch": 3.12, | |
| "learning_rate": 0.0009999479360563565, | |
| "loss": 1.9296, | |
| "step": 6020 | |
| }, | |
| { | |
| "epoch": 3.13, | |
| "learning_rate": 0.000999943488178013, | |
| "loss": 1.9602, | |
| "step": 6030 | |
| }, | |
| { | |
| "epoch": 3.13, | |
| "learning_rate": 0.0009999388580235853, | |
| "loss": 1.9229, | |
| "step": 6040 | |
| }, | |
| { | |
| "epoch": 3.14, | |
| "learning_rate": 0.0009999340455947613, | |
| "loss": 1.9508, | |
| "step": 6050 | |
| }, | |
| { | |
| "epoch": 3.14, | |
| "learning_rate": 0.0009999290508932957, | |
| "loss": 1.9451, | |
| "step": 6060 | |
| }, | |
| { | |
| "epoch": 3.15, | |
| "learning_rate": 0.0009999238739210093, | |
| "loss": 1.9379, | |
| "step": 6070 | |
| }, | |
| { | |
| "epoch": 3.16, | |
| "learning_rate": 0.00099991851467979, | |
| "loss": 1.9622, | |
| "step": 6080 | |
| }, | |
| { | |
| "epoch": 3.16, | |
| "learning_rate": 0.0009999129731715912, | |
| "loss": 1.9257, | |
| "step": 6090 | |
| }, | |
| { | |
| "epoch": 3.17, | |
| "learning_rate": 0.0009999072493984337, | |
| "loss": 1.936, | |
| "step": 6100 | |
| }, | |
| { | |
| "epoch": 3.17, | |
| "learning_rate": 0.0009999013433624041, | |
| "loss": 1.939, | |
| "step": 6110 | |
| }, | |
| { | |
| "epoch": 3.18, | |
| "learning_rate": 0.000999895255065656, | |
| "loss": 1.9445, | |
| "step": 6120 | |
| }, | |
| { | |
| "epoch": 3.18, | |
| "learning_rate": 0.0009998889845104092, | |
| "loss": 1.945, | |
| "step": 6130 | |
| }, | |
| { | |
| "epoch": 3.19, | |
| "learning_rate": 0.0009998825316989493, | |
| "loss": 1.989, | |
| "step": 6140 | |
| }, | |
| { | |
| "epoch": 3.19, | |
| "learning_rate": 0.0009998758966336297, | |
| "loss": 1.9469, | |
| "step": 6150 | |
| }, | |
| { | |
| "epoch": 3.2, | |
| "learning_rate": 0.0009998690793168688, | |
| "loss": 1.9297, | |
| "step": 6160 | |
| }, | |
| { | |
| "epoch": 3.2, | |
| "learning_rate": 0.0009998620797511525, | |
| "loss": 1.9461, | |
| "step": 6170 | |
| }, | |
| { | |
| "epoch": 3.21, | |
| "learning_rate": 0.0009998548979390331, | |
| "loss": 1.9486, | |
| "step": 6180 | |
| }, | |
| { | |
| "epoch": 3.21, | |
| "learning_rate": 0.0009998475338831286, | |
| "loss": 1.9473, | |
| "step": 6190 | |
| }, | |
| { | |
| "epoch": 3.22, | |
| "learning_rate": 0.000999839987586124, | |
| "loss": 1.9557, | |
| "step": 6200 | |
| }, | |
| { | |
| "epoch": 3.22, | |
| "learning_rate": 0.0009998322590507706, | |
| "loss": 1.9447, | |
| "step": 6210 | |
| }, | |
| { | |
| "epoch": 3.23, | |
| "learning_rate": 0.0009998243482798863, | |
| "loss": 1.9628, | |
| "step": 6220 | |
| }, | |
| { | |
| "epoch": 3.23, | |
| "learning_rate": 0.0009998162552763551, | |
| "loss": 1.9514, | |
| "step": 6230 | |
| }, | |
| { | |
| "epoch": 3.24, | |
| "learning_rate": 0.0009998079800431282, | |
| "loss": 1.9163, | |
| "step": 6240 | |
| }, | |
| { | |
| "epoch": 3.24, | |
| "learning_rate": 0.000999799522583222, | |
| "loss": 1.9278, | |
| "step": 6250 | |
| }, | |
| { | |
| "epoch": 3.25, | |
| "learning_rate": 0.00099979088289972, | |
| "loss": 1.9194, | |
| "step": 6260 | |
| }, | |
| { | |
| "epoch": 3.25, | |
| "learning_rate": 0.0009997820609957732, | |
| "loss": 1.9528, | |
| "step": 6270 | |
| }, | |
| { | |
| "epoch": 3.26, | |
| "learning_rate": 0.0009997730568745966, | |
| "loss": 1.9465, | |
| "step": 6280 | |
| }, | |
| { | |
| "epoch": 3.26, | |
| "learning_rate": 0.0009997638705394741, | |
| "loss": 1.9105, | |
| "step": 6290 | |
| }, | |
| { | |
| "epoch": 3.27, | |
| "learning_rate": 0.0009997545019937546, | |
| "loss": 1.9432, | |
| "step": 6300 | |
| }, | |
| { | |
| "epoch": 3.27, | |
| "learning_rate": 0.0009997449512408538, | |
| "loss": 1.9348, | |
| "step": 6310 | |
| }, | |
| { | |
| "epoch": 3.28, | |
| "learning_rate": 0.0009997352182842538, | |
| "loss": 1.9364, | |
| "step": 6320 | |
| }, | |
| { | |
| "epoch": 3.28, | |
| "learning_rate": 0.0009997253031275035, | |
| "loss": 1.9734, | |
| "step": 6330 | |
| }, | |
| { | |
| "epoch": 3.29, | |
| "learning_rate": 0.0009997152057742173, | |
| "loss": 1.9606, | |
| "step": 6340 | |
| }, | |
| { | |
| "epoch": 3.3, | |
| "learning_rate": 0.0009997049262280772, | |
| "loss": 1.9483, | |
| "step": 6350 | |
| }, | |
| { | |
| "epoch": 3.3, | |
| "learning_rate": 0.0009996944644928307, | |
| "loss": 1.9196, | |
| "step": 6360 | |
| }, | |
| { | |
| "epoch": 3.31, | |
| "learning_rate": 0.0009996838205722923, | |
| "loss": 1.9586, | |
| "step": 6370 | |
| }, | |
| { | |
| "epoch": 3.31, | |
| "learning_rate": 0.0009996729944703427, | |
| "loss": 1.9599, | |
| "step": 6380 | |
| }, | |
| { | |
| "epoch": 3.32, | |
| "learning_rate": 0.0009996619861909288, | |
| "loss": 1.9236, | |
| "step": 6390 | |
| }, | |
| { | |
| "epoch": 3.32, | |
| "learning_rate": 0.0009996507957380643, | |
| "loss": 1.9563, | |
| "step": 6400 | |
| }, | |
| { | |
| "epoch": 3.33, | |
| "learning_rate": 0.000999639423115829, | |
| "loss": 1.9324, | |
| "step": 6410 | |
| }, | |
| { | |
| "epoch": 3.33, | |
| "learning_rate": 0.0009996278683283696, | |
| "loss": 1.9386, | |
| "step": 6420 | |
| }, | |
| { | |
| "epoch": 3.34, | |
| "learning_rate": 0.0009996161313798985, | |
| "loss": 1.9506, | |
| "step": 6430 | |
| }, | |
| { | |
| "epoch": 3.34, | |
| "learning_rate": 0.0009996042122746955, | |
| "loss": 1.9184, | |
| "step": 6440 | |
| }, | |
| { | |
| "epoch": 3.35, | |
| "learning_rate": 0.0009995921110171058, | |
| "loss": 1.955, | |
| "step": 6450 | |
| }, | |
| { | |
| "epoch": 3.35, | |
| "learning_rate": 0.0009995798276115414, | |
| "loss": 1.9215, | |
| "step": 6460 | |
| }, | |
| { | |
| "epoch": 3.36, | |
| "learning_rate": 0.0009995673620624807, | |
| "loss": 1.9362, | |
| "step": 6470 | |
| }, | |
| { | |
| "epoch": 3.36, | |
| "learning_rate": 0.0009995547143744688, | |
| "loss": 1.939, | |
| "step": 6480 | |
| }, | |
| { | |
| "epoch": 3.37, | |
| "learning_rate": 0.000999541884552117, | |
| "loss": 1.8658, | |
| "step": 6490 | |
| }, | |
| { | |
| "epoch": 3.37, | |
| "learning_rate": 0.0009995288726001028, | |
| "loss": 1.9591, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 3.38, | |
| "learning_rate": 0.0009995156785231702, | |
| "loss": 1.9814, | |
| "step": 6510 | |
| }, | |
| { | |
| "epoch": 3.38, | |
| "learning_rate": 0.0009995023023261298, | |
| "loss": 1.9623, | |
| "step": 6520 | |
| }, | |
| { | |
| "epoch": 3.39, | |
| "learning_rate": 0.0009994887440138588, | |
| "loss": 1.9069, | |
| "step": 6530 | |
| }, | |
| { | |
| "epoch": 3.39, | |
| "learning_rate": 0.0009994750035912997, | |
| "loss": 1.8249, | |
| "step": 6540 | |
| }, | |
| { | |
| "epoch": 3.4, | |
| "learning_rate": 0.0009994610810634627, | |
| "loss": 1.8952, | |
| "step": 6550 | |
| }, | |
| { | |
| "epoch": 3.4, | |
| "learning_rate": 0.000999446976435424, | |
| "loss": 1.9932, | |
| "step": 6560 | |
| }, | |
| { | |
| "epoch": 3.41, | |
| "learning_rate": 0.0009994326897123256, | |
| "loss": 1.8771, | |
| "step": 6570 | |
| }, | |
| { | |
| "epoch": 3.41, | |
| "learning_rate": 0.0009994182208993764, | |
| "loss": 2.0021, | |
| "step": 6580 | |
| }, | |
| { | |
| "epoch": 3.42, | |
| "learning_rate": 0.000999403570001852, | |
| "loss": 1.9962, | |
| "step": 6590 | |
| }, | |
| { | |
| "epoch": 3.43, | |
| "learning_rate": 0.0009993887370250939, | |
| "loss": 1.9772, | |
| "step": 6600 | |
| }, | |
| { | |
| "epoch": 3.43, | |
| "learning_rate": 0.00099937372197451, | |
| "loss": 1.951, | |
| "step": 6610 | |
| }, | |
| { | |
| "epoch": 3.44, | |
| "learning_rate": 0.0009993585248555745, | |
| "loss": 1.9398, | |
| "step": 6620 | |
| }, | |
| { | |
| "epoch": 3.44, | |
| "learning_rate": 0.0009993431456738284, | |
| "loss": 1.9512, | |
| "step": 6630 | |
| }, | |
| { | |
| "epoch": 3.45, | |
| "learning_rate": 0.000999327584434879, | |
| "loss": 1.9954, | |
| "step": 6640 | |
| }, | |
| { | |
| "epoch": 3.45, | |
| "learning_rate": 0.0009993118411443995, | |
| "loss": 1.9426, | |
| "step": 6650 | |
| }, | |
| { | |
| "epoch": 3.46, | |
| "learning_rate": 0.00099929591580813, | |
| "loss": 1.9368, | |
| "step": 6660 | |
| }, | |
| { | |
| "epoch": 3.46, | |
| "learning_rate": 0.0009992798084318766, | |
| "loss": 1.9514, | |
| "step": 6670 | |
| }, | |
| { | |
| "epoch": 3.47, | |
| "learning_rate": 0.0009992635190215122, | |
| "loss": 1.9406, | |
| "step": 6680 | |
| }, | |
| { | |
| "epoch": 3.47, | |
| "learning_rate": 0.0009992470475829755, | |
| "loss": 1.9547, | |
| "step": 6690 | |
| }, | |
| { | |
| "epoch": 3.48, | |
| "learning_rate": 0.0009992303941222722, | |
| "loss": 1.9348, | |
| "step": 6700 | |
| }, | |
| { | |
| "epoch": 3.48, | |
| "learning_rate": 0.000999213558645474, | |
| "loss": 1.9717, | |
| "step": 6710 | |
| }, | |
| { | |
| "epoch": 3.49, | |
| "learning_rate": 0.0009991965411587187, | |
| "loss": 1.9505, | |
| "step": 6720 | |
| }, | |
| { | |
| "epoch": 3.49, | |
| "learning_rate": 0.000999179341668211, | |
| "loss": 1.925, | |
| "step": 6730 | |
| }, | |
| { | |
| "epoch": 3.5, | |
| "learning_rate": 0.0009991619601802217, | |
| "loss": 1.9599, | |
| "step": 6740 | |
| }, | |
| { | |
| "epoch": 3.5, | |
| "learning_rate": 0.000999144396701088, | |
| "loss": 1.9617, | |
| "step": 6750 | |
| }, | |
| { | |
| "epoch": 3.51, | |
| "learning_rate": 0.0009991266512372132, | |
| "loss": 1.9616, | |
| "step": 6760 | |
| }, | |
| { | |
| "epoch": 3.51, | |
| "learning_rate": 0.0009991087237950676, | |
| "loss": 1.9458, | |
| "step": 6770 | |
| }, | |
| { | |
| "epoch": 3.52, | |
| "learning_rate": 0.0009990906143811868, | |
| "loss": 1.9476, | |
| "step": 6780 | |
| }, | |
| { | |
| "epoch": 3.52, | |
| "learning_rate": 0.0009990723230021742, | |
| "loss": 1.919, | |
| "step": 6790 | |
| }, | |
| { | |
| "epoch": 3.53, | |
| "learning_rate": 0.000999053849664698, | |
| "loss": 1.9499, | |
| "step": 6800 | |
| }, | |
| { | |
| "epoch": 3.53, | |
| "learning_rate": 0.000999035194375494, | |
| "loss": 1.9444, | |
| "step": 6810 | |
| }, | |
| { | |
| "epoch": 3.54, | |
| "learning_rate": 0.0009990163571413633, | |
| "loss": 1.9498, | |
| "step": 6820 | |
| }, | |
| { | |
| "epoch": 3.54, | |
| "learning_rate": 0.0009989973379691739, | |
| "loss": 1.9729, | |
| "step": 6830 | |
| }, | |
| { | |
| "epoch": 3.55, | |
| "learning_rate": 0.0009989781368658604, | |
| "loss": 2.0205, | |
| "step": 6840 | |
| }, | |
| { | |
| "epoch": 3.55, | |
| "learning_rate": 0.0009989587538384232, | |
| "loss": 1.9512, | |
| "step": 6850 | |
| }, | |
| { | |
| "epoch": 3.56, | |
| "learning_rate": 0.0009989391888939295, | |
| "loss": 1.9329, | |
| "step": 6860 | |
| }, | |
| { | |
| "epoch": 3.57, | |
| "learning_rate": 0.000998919442039512, | |
| "loss": 1.9252, | |
| "step": 6870 | |
| }, | |
| { | |
| "epoch": 3.57, | |
| "learning_rate": 0.0009988995132823705, | |
| "loss": 1.9392, | |
| "step": 6880 | |
| }, | |
| { | |
| "epoch": 3.58, | |
| "learning_rate": 0.0009988794026297711, | |
| "loss": 1.9263, | |
| "step": 6890 | |
| }, | |
| { | |
| "epoch": 3.58, | |
| "learning_rate": 0.000998859110089046, | |
| "loss": 1.9332, | |
| "step": 6900 | |
| }, | |
| { | |
| "epoch": 3.59, | |
| "learning_rate": 0.0009988386356675932, | |
| "loss": 1.9138, | |
| "step": 6910 | |
| }, | |
| { | |
| "epoch": 3.59, | |
| "learning_rate": 0.000998817979372878, | |
| "loss": 1.9538, | |
| "step": 6920 | |
| }, | |
| { | |
| "epoch": 3.6, | |
| "learning_rate": 0.0009987971412124318, | |
| "loss": 1.9492, | |
| "step": 6930 | |
| }, | |
| { | |
| "epoch": 3.6, | |
| "learning_rate": 0.0009987761211938514, | |
| "loss": 1.9498, | |
| "step": 6940 | |
| }, | |
| { | |
| "epoch": 3.61, | |
| "learning_rate": 0.000998754919324801, | |
| "loss": 1.9517, | |
| "step": 6950 | |
| }, | |
| { | |
| "epoch": 3.61, | |
| "learning_rate": 0.0009987335356130107, | |
| "loss": 1.916, | |
| "step": 6960 | |
| }, | |
| { | |
| "epoch": 3.62, | |
| "learning_rate": 0.0009987119700662767, | |
| "loss": 1.9625, | |
| "step": 6970 | |
| }, | |
| { | |
| "epoch": 3.62, | |
| "learning_rate": 0.0009986902226924615, | |
| "loss": 1.9502, | |
| "step": 6980 | |
| }, | |
| { | |
| "epoch": 3.63, | |
| "learning_rate": 0.000998668293499494, | |
| "loss": 1.9368, | |
| "step": 6990 | |
| }, | |
| { | |
| "epoch": 3.63, | |
| "learning_rate": 0.0009986461824953701, | |
| "loss": 1.9518, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 3.64, | |
| "learning_rate": 0.0009986238896881508, | |
| "loss": 1.9244, | |
| "step": 7010 | |
| }, | |
| { | |
| "epoch": 3.64, | |
| "learning_rate": 0.0009986014150859637, | |
| "loss": 1.9348, | |
| "step": 7020 | |
| }, | |
| { | |
| "epoch": 3.65, | |
| "learning_rate": 0.0009985787586970034, | |
| "loss": 1.9646, | |
| "step": 7030 | |
| }, | |
| { | |
| "epoch": 3.65, | |
| "learning_rate": 0.0009985559205295301, | |
| "loss": 1.9216, | |
| "step": 7040 | |
| }, | |
| { | |
| "epoch": 3.66, | |
| "learning_rate": 0.0009985329005918703, | |
| "loss": 1.9065, | |
| "step": 7050 | |
| }, | |
| { | |
| "epoch": 3.66, | |
| "learning_rate": 0.0009985096988924169, | |
| "loss": 1.9527, | |
| "step": 7060 | |
| }, | |
| { | |
| "epoch": 3.67, | |
| "learning_rate": 0.0009984863154396293, | |
| "loss": 1.9143, | |
| "step": 7070 | |
| }, | |
| { | |
| "epoch": 3.67, | |
| "learning_rate": 0.0009984627502420329, | |
| "loss": 1.9082, | |
| "step": 7080 | |
| }, | |
| { | |
| "epoch": 3.68, | |
| "learning_rate": 0.0009984390033082194, | |
| "loss": 1.9009, | |
| "step": 7090 | |
| }, | |
| { | |
| "epoch": 3.68, | |
| "learning_rate": 0.0009984150746468465, | |
| "loss": 1.766, | |
| "step": 7100 | |
| }, | |
| { | |
| "epoch": 3.69, | |
| "learning_rate": 0.0009983909642666388, | |
| "loss": 1.8211, | |
| "step": 7110 | |
| }, | |
| { | |
| "epoch": 3.69, | |
| "learning_rate": 0.000998366672176387, | |
| "loss": 1.8446, | |
| "step": 7120 | |
| }, | |
| { | |
| "epoch": 3.7, | |
| "learning_rate": 0.000998342198384947, | |
| "loss": 1.8342, | |
| "step": 7130 | |
| }, | |
| { | |
| "epoch": 3.71, | |
| "learning_rate": 0.0009983175429012425, | |
| "loss": 1.8967, | |
| "step": 7140 | |
| }, | |
| { | |
| "epoch": 3.71, | |
| "learning_rate": 0.0009982927057342623, | |
| "loss": 1.8948, | |
| "step": 7150 | |
| }, | |
| { | |
| "epoch": 3.72, | |
| "learning_rate": 0.0009982676868930623, | |
| "loss": 1.9191, | |
| "step": 7160 | |
| }, | |
| { | |
| "epoch": 3.72, | |
| "learning_rate": 0.0009982424863867639, | |
| "loss": 1.9185, | |
| "step": 7170 | |
| }, | |
| { | |
| "epoch": 3.73, | |
| "learning_rate": 0.0009982171042245552, | |
| "loss": 1.8958, | |
| "step": 7180 | |
| }, | |
| { | |
| "epoch": 3.73, | |
| "learning_rate": 0.0009981915404156901, | |
| "loss": 1.9371, | |
| "step": 7190 | |
| }, | |
| { | |
| "epoch": 3.74, | |
| "learning_rate": 0.0009981657949694892, | |
| "loss": 1.8752, | |
| "step": 7200 | |
| }, | |
| { | |
| "epoch": 3.74, | |
| "learning_rate": 0.000998139867895339, | |
| "loss": 1.8434, | |
| "step": 7210 | |
| }, | |
| { | |
| "epoch": 3.75, | |
| "learning_rate": 0.0009981137592026926, | |
| "loss": 1.8518, | |
| "step": 7220 | |
| }, | |
| { | |
| "epoch": 3.75, | |
| "learning_rate": 0.0009980874689010687, | |
| "loss": 1.7907, | |
| "step": 7230 | |
| }, | |
| { | |
| "epoch": 3.76, | |
| "learning_rate": 0.000998060997000053, | |
| "loss": 1.8275, | |
| "step": 7240 | |
| }, | |
| { | |
| "epoch": 3.76, | |
| "learning_rate": 0.0009980343435092966, | |
| "loss": 1.9524, | |
| "step": 7250 | |
| }, | |
| { | |
| "epoch": 3.77, | |
| "learning_rate": 0.0009980075084385174, | |
| "loss": 1.8303, | |
| "step": 7260 | |
| }, | |
| { | |
| "epoch": 3.77, | |
| "learning_rate": 0.000997980491797499, | |
| "loss": 1.8564, | |
| "step": 7270 | |
| }, | |
| { | |
| "epoch": 3.78, | |
| "learning_rate": 0.000997953293596092, | |
| "loss": 1.8884, | |
| "step": 7280 | |
| }, | |
| { | |
| "epoch": 3.78, | |
| "learning_rate": 0.000997925913844212, | |
| "loss": 1.862, | |
| "step": 7290 | |
| }, | |
| { | |
| "epoch": 3.79, | |
| "learning_rate": 0.0009978983525518422, | |
| "loss": 1.8021, | |
| "step": 7300 | |
| }, | |
| { | |
| "epoch": 3.79, | |
| "learning_rate": 0.0009978706097290308, | |
| "loss": 1.9126, | |
| "step": 7310 | |
| }, | |
| { | |
| "epoch": 3.8, | |
| "learning_rate": 0.0009978426853858929, | |
| "loss": 1.9132, | |
| "step": 7320 | |
| }, | |
| { | |
| "epoch": 3.8, | |
| "learning_rate": 0.0009978145795326093, | |
| "loss": 1.8305, | |
| "step": 7330 | |
| }, | |
| { | |
| "epoch": 3.81, | |
| "learning_rate": 0.0009977862921794272, | |
| "loss": 1.871, | |
| "step": 7340 | |
| }, | |
| { | |
| "epoch": 3.81, | |
| "learning_rate": 0.0009977578233366603, | |
| "loss": 1.8928, | |
| "step": 7350 | |
| }, | |
| { | |
| "epoch": 3.82, | |
| "learning_rate": 0.0009977291730146876, | |
| "loss": 1.8817, | |
| "step": 7360 | |
| }, | |
| { | |
| "epoch": 3.82, | |
| "learning_rate": 0.0009977003412239556, | |
| "loss": 1.8307, | |
| "step": 7370 | |
| }, | |
| { | |
| "epoch": 3.83, | |
| "learning_rate": 0.0009976713279749754, | |
| "loss": 1.8866, | |
| "step": 7380 | |
| }, | |
| { | |
| "epoch": 3.83, | |
| "learning_rate": 0.0009976421332783256, | |
| "loss": 1.8442, | |
| "step": 7390 | |
| }, | |
| { | |
| "epoch": 3.84, | |
| "learning_rate": 0.00099761275714465, | |
| "loss": 1.8367, | |
| "step": 7400 | |
| }, | |
| { | |
| "epoch": 3.85, | |
| "learning_rate": 0.0009975831995846591, | |
| "loss": 1.8936, | |
| "step": 7410 | |
| }, | |
| { | |
| "epoch": 3.85, | |
| "learning_rate": 0.0009975534606091296, | |
| "loss": 1.807, | |
| "step": 7420 | |
| }, | |
| { | |
| "epoch": 3.86, | |
| "learning_rate": 0.0009975235402289039, | |
| "loss": 1.8231, | |
| "step": 7430 | |
| }, | |
| { | |
| "epoch": 3.86, | |
| "learning_rate": 0.0009974934384548906, | |
| "loss": 1.8122, | |
| "step": 7440 | |
| }, | |
| { | |
| "epoch": 3.87, | |
| "learning_rate": 0.000997463155298065, | |
| "loss": 1.8028, | |
| "step": 7450 | |
| }, | |
| { | |
| "epoch": 3.87, | |
| "learning_rate": 0.000997432690769468, | |
| "loss": 1.8177, | |
| "step": 7460 | |
| }, | |
| { | |
| "epoch": 3.88, | |
| "learning_rate": 0.0009974020448802066, | |
| "loss": 1.9069, | |
| "step": 7470 | |
| }, | |
| { | |
| "epoch": 3.88, | |
| "learning_rate": 0.0009973712176414543, | |
| "loss": 1.8245, | |
| "step": 7480 | |
| }, | |
| { | |
| "epoch": 3.89, | |
| "learning_rate": 0.0009973402090644504, | |
| "loss": 1.8876, | |
| "step": 7490 | |
| }, | |
| { | |
| "epoch": 3.89, | |
| "learning_rate": 0.0009973090191605005, | |
| "loss": 1.8971, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 3.9, | |
| "learning_rate": 0.000997277647940976, | |
| "loss": 1.7917, | |
| "step": 7510 | |
| }, | |
| { | |
| "epoch": 3.9, | |
| "learning_rate": 0.000997246095417315, | |
| "loss": 1.8608, | |
| "step": 7520 | |
| }, | |
| { | |
| "epoch": 3.91, | |
| "learning_rate": 0.000997214361601021, | |
| "loss": 1.7863, | |
| "step": 7530 | |
| }, | |
| { | |
| "epoch": 3.91, | |
| "learning_rate": 0.0009971824465036642, | |
| "loss": 1.8487, | |
| "step": 7540 | |
| }, | |
| { | |
| "epoch": 3.92, | |
| "learning_rate": 0.0009971503501368805, | |
| "loss": 1.8698, | |
| "step": 7550 | |
| }, | |
| { | |
| "epoch": 3.92, | |
| "learning_rate": 0.0009971180725123722, | |
| "loss": 1.8424, | |
| "step": 7560 | |
| }, | |
| { | |
| "epoch": 3.93, | |
| "learning_rate": 0.000997085613641907, | |
| "loss": 1.9235, | |
| "step": 7570 | |
| }, | |
| { | |
| "epoch": 3.93, | |
| "learning_rate": 0.0009970529735373198, | |
| "loss": 1.7802, | |
| "step": 7580 | |
| }, | |
| { | |
| "epoch": 3.94, | |
| "learning_rate": 0.0009970201522105106, | |
| "loss": 1.8938, | |
| "step": 7590 | |
| }, | |
| { | |
| "epoch": 3.94, | |
| "learning_rate": 0.000996987149673446, | |
| "loss": 1.7626, | |
| "step": 7600 | |
| }, | |
| { | |
| "epoch": 3.95, | |
| "learning_rate": 0.0009969539659381583, | |
| "loss": 1.8598, | |
| "step": 7610 | |
| }, | |
| { | |
| "epoch": 3.95, | |
| "learning_rate": 0.0009969206010167465, | |
| "loss": 1.8503, | |
| "step": 7620 | |
| }, | |
| { | |
| "epoch": 3.96, | |
| "learning_rate": 0.0009968870549213749, | |
| "loss": 1.9692, | |
| "step": 7630 | |
| }, | |
| { | |
| "epoch": 3.96, | |
| "learning_rate": 0.0009968533276642742, | |
| "loss": 1.8035, | |
| "step": 7640 | |
| }, | |
| { | |
| "epoch": 3.97, | |
| "learning_rate": 0.000996819419257741, | |
| "loss": 1.8945, | |
| "step": 7650 | |
| }, | |
| { | |
| "epoch": 3.98, | |
| "learning_rate": 0.0009967853297141384, | |
| "loss": 1.8432, | |
| "step": 7660 | |
| }, | |
| { | |
| "epoch": 3.98, | |
| "learning_rate": 0.0009967510590458953, | |
| "loss": 1.8781, | |
| "step": 7670 | |
| }, | |
| { | |
| "epoch": 3.99, | |
| "learning_rate": 0.000996716607265506, | |
| "loss": 1.8669, | |
| "step": 7680 | |
| }, | |
| { | |
| "epoch": 3.99, | |
| "learning_rate": 0.0009966819743855318, | |
| "loss": 1.8776, | |
| "step": 7690 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 0.0009966471604185997, | |
| "loss": 1.803, | |
| "step": 7700 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": { | |
| "accuracy": 0.1846952010376135 | |
| }, | |
| "eval_f1": { | |
| "f1": 0.1316379146314846 | |
| }, | |
| "eval_loss": 1.9003419876098633, | |
| "eval_precision": { | |
| "precision": 0.1460659091286986 | |
| }, | |
| "eval_recall": { | |
| "recall": 0.1928584239028949 | |
| }, | |
| "eval_runtime": 112.0693, | |
| "eval_samples_per_second": 34.398, | |
| "eval_steps_per_second": 17.204, | |
| "step": 7708 | |
| } | |
| ], | |
| "logging_steps": 10, | |
| "max_steps": 57810, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 30, | |
| "save_steps": 500, | |
| "total_flos": 3.555751357898293e+18, | |
| "train_batch_size": 2, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |