| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 7.230657989877079, | |
| "global_step": 5000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0002, | |
| "loss": 1.5637, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0002, | |
| "loss": 1.3117, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0002, | |
| "loss": 1.2738, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0002, | |
| "loss": 1.2728, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0002, | |
| "loss": 1.2202, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0002, | |
| "loss": 1.2241, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0002, | |
| "loss": 1.1647, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0002, | |
| "loss": 1.1426, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0002, | |
| "loss": 1.1432, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0002, | |
| "loss": 1.1319, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0002, | |
| "loss": 1.1882, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0002, | |
| "loss": 1.116, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0002, | |
| "loss": 1.1072, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0002, | |
| "loss": 1.1417, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0002, | |
| "loss": 1.1102, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0002, | |
| "loss": 1.1567, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0002, | |
| "loss": 1.1167, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0849, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0002, | |
| "loss": 1.1045, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0925, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0002, | |
| "loss": 1.1588, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0002, | |
| "loss": 1.1013, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0684, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0887, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0598, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0002, | |
| "loss": 1.153, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0932, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0747, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0729, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0683, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0002, | |
| "loss": 1.1444, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0992, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0648, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0701, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0628, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0002, | |
| "loss": 1.1322, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0736, | |
| "step": 370 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0485, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0571, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.0002, | |
| "loss": 1.076, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.0002, | |
| "loss": 1.1234, | |
| "step": 410 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0641, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0724, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0483, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.0002, | |
| "loss": 1.039, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.0002, | |
| "loss": 1.1064, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0551, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0224, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0001, | |
| "step": 490 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 0.0002, | |
| "loss": 0.96, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0563, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0359, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0219, | |
| "step": 530 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0062, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 0.0002, | |
| "loss": 0.948, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0781, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0279, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0209, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0166, | |
| "step": 590 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9426, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0778, | |
| "step": 610 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0314, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0092, | |
| "step": 630 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0022, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9352, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0628, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0157, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0169, | |
| "step": 680 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9608, | |
| "step": 690 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 0.0002, | |
| "loss": 1.033, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0205, | |
| "step": 710 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9868, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9838, | |
| "step": 730 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 0.0002, | |
| "loss": 0.925, | |
| "step": 740 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0044, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0188, | |
| "step": 760 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9815, | |
| "step": 770 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9745, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9249, | |
| "step": 790 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0042, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9997, | |
| "step": 810 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9889, | |
| "step": 820 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9679, | |
| "step": 830 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9192, | |
| "step": 840 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0285, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0116, | |
| "step": 860 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9794, | |
| "step": 870 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 0.0002, | |
| "loss": 0.974, | |
| "step": 880 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9272, | |
| "step": 890 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 0.0002, | |
| "loss": 1.016, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0145, | |
| "step": 910 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9856, | |
| "step": 920 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9517, | |
| "step": 930 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9321, | |
| "step": 940 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0238, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0162, | |
| "step": 960 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9882, | |
| "step": 970 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9704, | |
| "step": 980 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9221, | |
| "step": 990 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0302, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0084, | |
| "step": 1010 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9829, | |
| "step": 1020 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9673, | |
| "step": 1030 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9238, | |
| "step": 1040 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0006, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9912, | |
| "step": 1060 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 0.0002, | |
| "loss": 0.973, | |
| "step": 1070 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 0.0002, | |
| "loss": 0.961, | |
| "step": 1080 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9109, | |
| "step": 1090 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9946, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9957, | |
| "step": 1110 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9628, | |
| "step": 1120 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9551, | |
| "step": 1130 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9382, | |
| "step": 1140 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 0.0002, | |
| "loss": 1.0101, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 0.0002, | |
| "loss": 0.982, | |
| "step": 1160 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9751, | |
| "step": 1170 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9605, | |
| "step": 1180 | |
| }, | |
| { | |
| "epoch": 1.72, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9216, | |
| "step": 1190 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9855, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9885, | |
| "step": 1210 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9678, | |
| "step": 1220 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9642, | |
| "step": 1230 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9125, | |
| "step": 1240 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9917, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9832, | |
| "step": 1260 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9619, | |
| "step": 1270 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9523, | |
| "step": 1280 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9244, | |
| "step": 1290 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9929, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9795, | |
| "step": 1310 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9674, | |
| "step": 1320 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9671, | |
| "step": 1330 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9089, | |
| "step": 1340 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9829, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9712, | |
| "step": 1360 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9626, | |
| "step": 1370 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9273, | |
| "step": 1380 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 0.0002, | |
| "loss": 0.96, | |
| "step": 1390 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9706, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9352, | |
| "step": 1410 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9279, | |
| "step": 1420 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8945, | |
| "step": 1430 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9358, | |
| "step": 1440 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9542, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9328, | |
| "step": 1460 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 0.0002, | |
| "loss": 0.91, | |
| "step": 1470 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8871, | |
| "step": 1480 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9231, | |
| "step": 1490 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9682, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9527, | |
| "step": 1510 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9268, | |
| "step": 1520 | |
| }, | |
| { | |
| "epoch": 2.21, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8734, | |
| "step": 1530 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9363, | |
| "step": 1540 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9615, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9415, | |
| "step": 1560 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9177, | |
| "step": 1570 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9003, | |
| "step": 1580 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9628, | |
| "step": 1590 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9603, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9314, | |
| "step": 1610 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9305, | |
| "step": 1620 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8867, | |
| "step": 1630 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9304, | |
| "step": 1640 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 0.0002, | |
| "loss": 0.978, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9338, | |
| "step": 1660 | |
| }, | |
| { | |
| "epoch": 2.42, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9334, | |
| "step": 1670 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8967, | |
| "step": 1680 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9358, | |
| "step": 1690 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9557, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 2.47, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9332, | |
| "step": 1710 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9198, | |
| "step": 1720 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8794, | |
| "step": 1730 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9247, | |
| "step": 1740 | |
| }, | |
| { | |
| "epoch": 2.53, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9541, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9266, | |
| "step": 1760 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "learning_rate": 0.0002, | |
| "loss": 0.908, | |
| "step": 1770 | |
| }, | |
| { | |
| "epoch": 2.57, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8933, | |
| "step": 1780 | |
| }, | |
| { | |
| "epoch": 2.59, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9533, | |
| "step": 1790 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9567, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 2.62, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9514, | |
| "step": 1810 | |
| }, | |
| { | |
| "epoch": 2.63, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9099, | |
| "step": 1820 | |
| }, | |
| { | |
| "epoch": 2.65, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8934, | |
| "step": 1830 | |
| }, | |
| { | |
| "epoch": 2.66, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9207, | |
| "step": 1840 | |
| }, | |
| { | |
| "epoch": 2.68, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9655, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 2.69, | |
| "learning_rate": 0.0002, | |
| "loss": 0.934, | |
| "step": 1860 | |
| }, | |
| { | |
| "epoch": 2.7, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9299, | |
| "step": 1870 | |
| }, | |
| { | |
| "epoch": 2.72, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8832, | |
| "step": 1880 | |
| }, | |
| { | |
| "epoch": 2.73, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9434, | |
| "step": 1890 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9503, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 2.76, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9322, | |
| "step": 1910 | |
| }, | |
| { | |
| "epoch": 2.78, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9019, | |
| "step": 1920 | |
| }, | |
| { | |
| "epoch": 2.79, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8827, | |
| "step": 1930 | |
| }, | |
| { | |
| "epoch": 2.81, | |
| "learning_rate": 0.0002, | |
| "loss": 0.94, | |
| "step": 1940 | |
| }, | |
| { | |
| "epoch": 2.82, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9615, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 2.83, | |
| "learning_rate": 0.0002, | |
| "loss": 0.926, | |
| "step": 1960 | |
| }, | |
| { | |
| "epoch": 2.85, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9219, | |
| "step": 1970 | |
| }, | |
| { | |
| "epoch": 2.86, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8812, | |
| "step": 1980 | |
| }, | |
| { | |
| "epoch": 2.88, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9407, | |
| "step": 1990 | |
| }, | |
| { | |
| "epoch": 2.89, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9679, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 2.91, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9243, | |
| "step": 2010 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "learning_rate": 0.0002, | |
| "loss": 0.911, | |
| "step": 2020 | |
| }, | |
| { | |
| "epoch": 2.94, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8796, | |
| "step": 2030 | |
| }, | |
| { | |
| "epoch": 2.95, | |
| "learning_rate": 0.0002, | |
| "loss": 0.94, | |
| "step": 2040 | |
| }, | |
| { | |
| "epoch": 2.96, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9692, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 2.98, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9006, | |
| "step": 2060 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9048, | |
| "step": 2070 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8862, | |
| "step": 2080 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9506, | |
| "step": 2090 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 0.0002, | |
| "loss": 0.891, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8773, | |
| "step": 2110 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8564, | |
| "step": 2120 | |
| }, | |
| { | |
| "epoch": 3.08, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8797, | |
| "step": 2130 | |
| }, | |
| { | |
| "epoch": 3.09, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9263, | |
| "step": 2140 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8894, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 3.12, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8973, | |
| "step": 2160 | |
| }, | |
| { | |
| "epoch": 3.14, | |
| "learning_rate": 0.0002, | |
| "loss": 0.838, | |
| "step": 2170 | |
| }, | |
| { | |
| "epoch": 3.15, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8785, | |
| "step": 2180 | |
| }, | |
| { | |
| "epoch": 3.17, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9323, | |
| "step": 2190 | |
| }, | |
| { | |
| "epoch": 3.18, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8981, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 3.2, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8775, | |
| "step": 2210 | |
| }, | |
| { | |
| "epoch": 3.21, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8546, | |
| "step": 2220 | |
| }, | |
| { | |
| "epoch": 3.22, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8693, | |
| "step": 2230 | |
| }, | |
| { | |
| "epoch": 3.24, | |
| "learning_rate": 0.0002, | |
| "loss": 0.941, | |
| "step": 2240 | |
| }, | |
| { | |
| "epoch": 3.25, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9034, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 3.27, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8807, | |
| "step": 2260 | |
| }, | |
| { | |
| "epoch": 3.28, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8535, | |
| "step": 2270 | |
| }, | |
| { | |
| "epoch": 3.3, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8793, | |
| "step": 2280 | |
| }, | |
| { | |
| "epoch": 3.31, | |
| "learning_rate": 0.0002, | |
| "loss": 0.938, | |
| "step": 2290 | |
| }, | |
| { | |
| "epoch": 3.33, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9041, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 3.34, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8857, | |
| "step": 2310 | |
| }, | |
| { | |
| "epoch": 3.36, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8367, | |
| "step": 2320 | |
| }, | |
| { | |
| "epoch": 3.37, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8857, | |
| "step": 2330 | |
| }, | |
| { | |
| "epoch": 3.38, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9411, | |
| "step": 2340 | |
| }, | |
| { | |
| "epoch": 3.4, | |
| "learning_rate": 0.0002, | |
| "loss": 0.907, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 3.41, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8845, | |
| "step": 2360 | |
| }, | |
| { | |
| "epoch": 3.43, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8558, | |
| "step": 2370 | |
| }, | |
| { | |
| "epoch": 3.44, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8698, | |
| "step": 2380 | |
| }, | |
| { | |
| "epoch": 3.46, | |
| "learning_rate": 0.0002, | |
| "loss": 0.951, | |
| "step": 2390 | |
| }, | |
| { | |
| "epoch": 3.47, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8948, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 3.49, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8737, | |
| "step": 2410 | |
| }, | |
| { | |
| "epoch": 3.5, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8625, | |
| "step": 2420 | |
| }, | |
| { | |
| "epoch": 3.51, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8975, | |
| "step": 2430 | |
| }, | |
| { | |
| "epoch": 3.53, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9429, | |
| "step": 2440 | |
| }, | |
| { | |
| "epoch": 3.54, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9116, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 3.56, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8913, | |
| "step": 2460 | |
| }, | |
| { | |
| "epoch": 3.57, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8636, | |
| "step": 2470 | |
| }, | |
| { | |
| "epoch": 3.59, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8751, | |
| "step": 2480 | |
| }, | |
| { | |
| "epoch": 3.6, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9381, | |
| "step": 2490 | |
| }, | |
| { | |
| "epoch": 3.62, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9171, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 3.63, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8879, | |
| "step": 2510 | |
| }, | |
| { | |
| "epoch": 3.64, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8706, | |
| "step": 2520 | |
| }, | |
| { | |
| "epoch": 3.66, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8806, | |
| "step": 2530 | |
| }, | |
| { | |
| "epoch": 3.67, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9308, | |
| "step": 2540 | |
| }, | |
| { | |
| "epoch": 3.69, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8945, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 3.7, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8851, | |
| "step": 2560 | |
| }, | |
| { | |
| "epoch": 3.72, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8523, | |
| "step": 2570 | |
| }, | |
| { | |
| "epoch": 3.73, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8859, | |
| "step": 2580 | |
| }, | |
| { | |
| "epoch": 3.75, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9284, | |
| "step": 2590 | |
| }, | |
| { | |
| "epoch": 3.76, | |
| "learning_rate": 0.0002, | |
| "loss": 0.898, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 3.77, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8752, | |
| "step": 2610 | |
| }, | |
| { | |
| "epoch": 3.79, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8548, | |
| "step": 2620 | |
| }, | |
| { | |
| "epoch": 3.8, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8768, | |
| "step": 2630 | |
| }, | |
| { | |
| "epoch": 3.82, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9454, | |
| "step": 2640 | |
| }, | |
| { | |
| "epoch": 3.83, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8926, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 3.85, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8801, | |
| "step": 2660 | |
| }, | |
| { | |
| "epoch": 3.86, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8558, | |
| "step": 2670 | |
| }, | |
| { | |
| "epoch": 3.88, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8597, | |
| "step": 2680 | |
| }, | |
| { | |
| "epoch": 3.89, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9234, | |
| "step": 2690 | |
| }, | |
| { | |
| "epoch": 3.9, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8973, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 3.92, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8579, | |
| "step": 2710 | |
| }, | |
| { | |
| "epoch": 3.93, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8538, | |
| "step": 2720 | |
| }, | |
| { | |
| "epoch": 3.95, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8772, | |
| "step": 2730 | |
| }, | |
| { | |
| "epoch": 3.96, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9341, | |
| "step": 2740 | |
| }, | |
| { | |
| "epoch": 3.98, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8907, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 3.99, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8652, | |
| "step": 2760 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8527, | |
| "step": 2770 | |
| }, | |
| { | |
| "epoch": 4.02, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9176, | |
| "step": 2780 | |
| }, | |
| { | |
| "epoch": 4.03, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8707, | |
| "step": 2790 | |
| }, | |
| { | |
| "epoch": 4.05, | |
| "learning_rate": 0.0002, | |
| "loss": 0.849, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 4.06, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8208, | |
| "step": 2810 | |
| }, | |
| { | |
| "epoch": 4.08, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7985, | |
| "step": 2820 | |
| }, | |
| { | |
| "epoch": 4.09, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9227, | |
| "step": 2830 | |
| }, | |
| { | |
| "epoch": 4.11, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8879, | |
| "step": 2840 | |
| }, | |
| { | |
| "epoch": 4.12, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8378, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 4.14, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8242, | |
| "step": 2860 | |
| }, | |
| { | |
| "epoch": 4.15, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8257, | |
| "step": 2870 | |
| }, | |
| { | |
| "epoch": 4.16, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9259, | |
| "step": 2880 | |
| }, | |
| { | |
| "epoch": 4.18, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8598, | |
| "step": 2890 | |
| }, | |
| { | |
| "epoch": 4.19, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8437, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 4.21, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8156, | |
| "step": 2910 | |
| }, | |
| { | |
| "epoch": 4.22, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8178, | |
| "step": 2920 | |
| }, | |
| { | |
| "epoch": 4.24, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9162, | |
| "step": 2930 | |
| }, | |
| { | |
| "epoch": 4.25, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8678, | |
| "step": 2940 | |
| }, | |
| { | |
| "epoch": 4.27, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8499, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 4.28, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8106, | |
| "step": 2960 | |
| }, | |
| { | |
| "epoch": 4.3, | |
| "learning_rate": 0.0002, | |
| "loss": 0.821, | |
| "step": 2970 | |
| }, | |
| { | |
| "epoch": 4.31, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9278, | |
| "step": 2980 | |
| }, | |
| { | |
| "epoch": 4.32, | |
| "learning_rate": 0.0002, | |
| "loss": 0.865, | |
| "step": 2990 | |
| }, | |
| { | |
| "epoch": 4.34, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8549, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 4.35, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8248, | |
| "step": 3010 | |
| }, | |
| { | |
| "epoch": 4.37, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8262, | |
| "step": 3020 | |
| }, | |
| { | |
| "epoch": 4.38, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9076, | |
| "step": 3030 | |
| }, | |
| { | |
| "epoch": 4.4, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8706, | |
| "step": 3040 | |
| }, | |
| { | |
| "epoch": 4.41, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8503, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 4.43, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8351, | |
| "step": 3060 | |
| }, | |
| { | |
| "epoch": 4.44, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8267, | |
| "step": 3070 | |
| }, | |
| { | |
| "epoch": 4.45, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9008, | |
| "step": 3080 | |
| }, | |
| { | |
| "epoch": 4.47, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8624, | |
| "step": 3090 | |
| }, | |
| { | |
| "epoch": 4.48, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8703, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 4.5, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8284, | |
| "step": 3110 | |
| }, | |
| { | |
| "epoch": 4.51, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7899, | |
| "step": 3120 | |
| }, | |
| { | |
| "epoch": 4.53, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9195, | |
| "step": 3130 | |
| }, | |
| { | |
| "epoch": 4.54, | |
| "learning_rate": 0.0002, | |
| "loss": 0.881, | |
| "step": 3140 | |
| }, | |
| { | |
| "epoch": 4.56, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8647, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 4.57, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8115, | |
| "step": 3160 | |
| }, | |
| { | |
| "epoch": 4.58, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8282, | |
| "step": 3170 | |
| }, | |
| { | |
| "epoch": 4.6, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9087, | |
| "step": 3180 | |
| }, | |
| { | |
| "epoch": 4.61, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8789, | |
| "step": 3190 | |
| }, | |
| { | |
| "epoch": 4.63, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8359, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 4.64, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8223, | |
| "step": 3210 | |
| }, | |
| { | |
| "epoch": 4.66, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8178, | |
| "step": 3220 | |
| }, | |
| { | |
| "epoch": 4.67, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9243, | |
| "step": 3230 | |
| }, | |
| { | |
| "epoch": 4.69, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8732, | |
| "step": 3240 | |
| }, | |
| { | |
| "epoch": 4.7, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8427, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 4.71, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8358, | |
| "step": 3260 | |
| }, | |
| { | |
| "epoch": 4.73, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8308, | |
| "step": 3270 | |
| }, | |
| { | |
| "epoch": 4.74, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9175, | |
| "step": 3280 | |
| }, | |
| { | |
| "epoch": 4.76, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8806, | |
| "step": 3290 | |
| }, | |
| { | |
| "epoch": 4.77, | |
| "learning_rate": 0.0002, | |
| "loss": 0.848, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 4.79, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8159, | |
| "step": 3310 | |
| }, | |
| { | |
| "epoch": 4.8, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8202, | |
| "step": 3320 | |
| }, | |
| { | |
| "epoch": 4.82, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9166, | |
| "step": 3330 | |
| }, | |
| { | |
| "epoch": 4.83, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8756, | |
| "step": 3340 | |
| }, | |
| { | |
| "epoch": 4.84, | |
| "learning_rate": 0.0002, | |
| "loss": 0.842, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 4.86, | |
| "learning_rate": 0.0002, | |
| "loss": 0.831, | |
| "step": 3360 | |
| }, | |
| { | |
| "epoch": 4.87, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8228, | |
| "step": 3370 | |
| }, | |
| { | |
| "epoch": 4.89, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9109, | |
| "step": 3380 | |
| }, | |
| { | |
| "epoch": 4.9, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8746, | |
| "step": 3390 | |
| }, | |
| { | |
| "epoch": 4.92, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8605, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 4.93, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8263, | |
| "step": 3410 | |
| }, | |
| { | |
| "epoch": 4.95, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8009, | |
| "step": 3420 | |
| }, | |
| { | |
| "epoch": 4.96, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9056, | |
| "step": 3430 | |
| }, | |
| { | |
| "epoch": 4.97, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8647, | |
| "step": 3440 | |
| }, | |
| { | |
| "epoch": 4.99, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8481, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7768, | |
| "step": 3460 | |
| }, | |
| { | |
| "epoch": 5.02, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8872, | |
| "step": 3470 | |
| }, | |
| { | |
| "epoch": 5.03, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8517, | |
| "step": 3480 | |
| }, | |
| { | |
| "epoch": 5.05, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8096, | |
| "step": 3490 | |
| }, | |
| { | |
| "epoch": 5.06, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7853, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 5.08, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7469, | |
| "step": 3510 | |
| }, | |
| { | |
| "epoch": 5.09, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9045, | |
| "step": 3520 | |
| }, | |
| { | |
| "epoch": 5.1, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8387, | |
| "step": 3530 | |
| }, | |
| { | |
| "epoch": 5.12, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8039, | |
| "step": 3540 | |
| }, | |
| { | |
| "epoch": 5.13, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7629, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 5.15, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7336, | |
| "step": 3560 | |
| }, | |
| { | |
| "epoch": 5.16, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9018, | |
| "step": 3570 | |
| }, | |
| { | |
| "epoch": 5.18, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8511, | |
| "step": 3580 | |
| }, | |
| { | |
| "epoch": 5.19, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8043, | |
| "step": 3590 | |
| }, | |
| { | |
| "epoch": 5.21, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7877, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 5.22, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7438, | |
| "step": 3610 | |
| }, | |
| { | |
| "epoch": 5.23, | |
| "learning_rate": 0.0002, | |
| "loss": 0.903, | |
| "step": 3620 | |
| }, | |
| { | |
| "epoch": 5.25, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8562, | |
| "step": 3630 | |
| }, | |
| { | |
| "epoch": 5.26, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8198, | |
| "step": 3640 | |
| }, | |
| { | |
| "epoch": 5.28, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7953, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 5.29, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7606, | |
| "step": 3660 | |
| }, | |
| { | |
| "epoch": 5.31, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8996, | |
| "step": 3670 | |
| }, | |
| { | |
| "epoch": 5.32, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8519, | |
| "step": 3680 | |
| }, | |
| { | |
| "epoch": 5.34, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8307, | |
| "step": 3690 | |
| }, | |
| { | |
| "epoch": 5.35, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7875, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 5.37, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7508, | |
| "step": 3710 | |
| }, | |
| { | |
| "epoch": 5.38, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9087, | |
| "step": 3720 | |
| }, | |
| { | |
| "epoch": 5.39, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8564, | |
| "step": 3730 | |
| }, | |
| { | |
| "epoch": 5.41, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8265, | |
| "step": 3740 | |
| }, | |
| { | |
| "epoch": 5.42, | |
| "learning_rate": 0.0002, | |
| "loss": 0.794, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 5.44, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7502, | |
| "step": 3760 | |
| }, | |
| { | |
| "epoch": 5.45, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8966, | |
| "step": 3770 | |
| }, | |
| { | |
| "epoch": 5.47, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8521, | |
| "step": 3780 | |
| }, | |
| { | |
| "epoch": 5.48, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8195, | |
| "step": 3790 | |
| }, | |
| { | |
| "epoch": 5.5, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7985, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 5.51, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7676, | |
| "step": 3810 | |
| }, | |
| { | |
| "epoch": 5.52, | |
| "learning_rate": 0.0002, | |
| "loss": 0.9075, | |
| "step": 3820 | |
| }, | |
| { | |
| "epoch": 5.54, | |
| "learning_rate": 0.0002, | |
| "loss": 0.852, | |
| "step": 3830 | |
| }, | |
| { | |
| "epoch": 5.55, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8306, | |
| "step": 3840 | |
| }, | |
| { | |
| "epoch": 5.57, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7964, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 5.58, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7599, | |
| "step": 3860 | |
| }, | |
| { | |
| "epoch": 5.6, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8959, | |
| "step": 3870 | |
| }, | |
| { | |
| "epoch": 5.61, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8489, | |
| "step": 3880 | |
| }, | |
| { | |
| "epoch": 5.63, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8095, | |
| "step": 3890 | |
| }, | |
| { | |
| "epoch": 5.64, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7912, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 5.65, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7555, | |
| "step": 3910 | |
| }, | |
| { | |
| "epoch": 5.67, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8884, | |
| "step": 3920 | |
| }, | |
| { | |
| "epoch": 5.68, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8375, | |
| "step": 3930 | |
| }, | |
| { | |
| "epoch": 5.7, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8313, | |
| "step": 3940 | |
| }, | |
| { | |
| "epoch": 5.71, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7983, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 5.73, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7691, | |
| "step": 3960 | |
| }, | |
| { | |
| "epoch": 5.74, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8946, | |
| "step": 3970 | |
| }, | |
| { | |
| "epoch": 5.76, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8528, | |
| "step": 3980 | |
| }, | |
| { | |
| "epoch": 5.77, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8293, | |
| "step": 3990 | |
| }, | |
| { | |
| "epoch": 5.78, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7929, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 5.8, | |
| "learning_rate": 0.0002, | |
| "loss": 0.773, | |
| "step": 4010 | |
| }, | |
| { | |
| "epoch": 5.81, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8988, | |
| "step": 4020 | |
| }, | |
| { | |
| "epoch": 5.83, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8639, | |
| "step": 4030 | |
| }, | |
| { | |
| "epoch": 5.84, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8219, | |
| "step": 4040 | |
| }, | |
| { | |
| "epoch": 5.86, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8106, | |
| "step": 4050 | |
| }, | |
| { | |
| "epoch": 5.87, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7746, | |
| "step": 4060 | |
| }, | |
| { | |
| "epoch": 5.89, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8874, | |
| "step": 4070 | |
| }, | |
| { | |
| "epoch": 5.9, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8497, | |
| "step": 4080 | |
| }, | |
| { | |
| "epoch": 5.91, | |
| "learning_rate": 0.0002, | |
| "loss": 0.818, | |
| "step": 4090 | |
| }, | |
| { | |
| "epoch": 5.93, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7891, | |
| "step": 4100 | |
| }, | |
| { | |
| "epoch": 5.94, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7508, | |
| "step": 4110 | |
| }, | |
| { | |
| "epoch": 5.96, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8985, | |
| "step": 4120 | |
| }, | |
| { | |
| "epoch": 5.97, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8496, | |
| "step": 4130 | |
| }, | |
| { | |
| "epoch": 5.99, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8176, | |
| "step": 4140 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7464, | |
| "step": 4150 | |
| }, | |
| { | |
| "epoch": 6.02, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8725, | |
| "step": 4160 | |
| }, | |
| { | |
| "epoch": 6.03, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8238, | |
| "step": 4170 | |
| }, | |
| { | |
| "epoch": 6.04, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7955, | |
| "step": 4180 | |
| }, | |
| { | |
| "epoch": 6.06, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7304, | |
| "step": 4190 | |
| }, | |
| { | |
| "epoch": 6.07, | |
| "learning_rate": 0.0002, | |
| "loss": 0.6796, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 6.09, | |
| "learning_rate": 0.0002, | |
| "loss": 0.887, | |
| "step": 4210 | |
| }, | |
| { | |
| "epoch": 6.1, | |
| "learning_rate": 0.0002, | |
| "loss": 0.831, | |
| "step": 4220 | |
| }, | |
| { | |
| "epoch": 6.12, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7878, | |
| "step": 4230 | |
| }, | |
| { | |
| "epoch": 6.13, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7574, | |
| "step": 4240 | |
| }, | |
| { | |
| "epoch": 6.15, | |
| "learning_rate": 0.0002, | |
| "loss": 0.6826, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 6.16, | |
| "learning_rate": 0.0002, | |
| "loss": 0.881, | |
| "step": 4260 | |
| }, | |
| { | |
| "epoch": 6.17, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8245, | |
| "step": 4270 | |
| }, | |
| { | |
| "epoch": 6.19, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7827, | |
| "step": 4280 | |
| }, | |
| { | |
| "epoch": 6.2, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7556, | |
| "step": 4290 | |
| }, | |
| { | |
| "epoch": 6.22, | |
| "learning_rate": 0.0002, | |
| "loss": 0.6812, | |
| "step": 4300 | |
| }, | |
| { | |
| "epoch": 6.23, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8812, | |
| "step": 4310 | |
| }, | |
| { | |
| "epoch": 6.25, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8373, | |
| "step": 4320 | |
| }, | |
| { | |
| "epoch": 6.26, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7915, | |
| "step": 4330 | |
| }, | |
| { | |
| "epoch": 6.28, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7575, | |
| "step": 4340 | |
| }, | |
| { | |
| "epoch": 6.29, | |
| "learning_rate": 0.0002, | |
| "loss": 0.6715, | |
| "step": 4350 | |
| }, | |
| { | |
| "epoch": 6.31, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8731, | |
| "step": 4360 | |
| }, | |
| { | |
| "epoch": 6.32, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8362, | |
| "step": 4370 | |
| }, | |
| { | |
| "epoch": 6.33, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7917, | |
| "step": 4380 | |
| }, | |
| { | |
| "epoch": 6.35, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7803, | |
| "step": 4390 | |
| }, | |
| { | |
| "epoch": 6.36, | |
| "learning_rate": 0.0002, | |
| "loss": 0.6743, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 6.38, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8894, | |
| "step": 4410 | |
| }, | |
| { | |
| "epoch": 6.39, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8375, | |
| "step": 4420 | |
| }, | |
| { | |
| "epoch": 6.41, | |
| "learning_rate": 0.0002, | |
| "loss": 0.793, | |
| "step": 4430 | |
| }, | |
| { | |
| "epoch": 6.42, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7616, | |
| "step": 4440 | |
| }, | |
| { | |
| "epoch": 6.44, | |
| "learning_rate": 0.0002, | |
| "loss": 0.6821, | |
| "step": 4450 | |
| }, | |
| { | |
| "epoch": 6.45, | |
| "learning_rate": 0.0002, | |
| "loss": 0.884, | |
| "step": 4460 | |
| }, | |
| { | |
| "epoch": 6.46, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8264, | |
| "step": 4470 | |
| }, | |
| { | |
| "epoch": 6.48, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7921, | |
| "step": 4480 | |
| }, | |
| { | |
| "epoch": 6.49, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7699, | |
| "step": 4490 | |
| }, | |
| { | |
| "epoch": 6.51, | |
| "learning_rate": 0.0002, | |
| "loss": 0.6964, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 6.52, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8871, | |
| "step": 4510 | |
| }, | |
| { | |
| "epoch": 6.54, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8286, | |
| "step": 4520 | |
| }, | |
| { | |
| "epoch": 6.55, | |
| "learning_rate": 0.0002, | |
| "loss": 0.792, | |
| "step": 4530 | |
| }, | |
| { | |
| "epoch": 6.57, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7692, | |
| "step": 4540 | |
| }, | |
| { | |
| "epoch": 6.58, | |
| "learning_rate": 0.0002, | |
| "loss": 0.6903, | |
| "step": 4550 | |
| }, | |
| { | |
| "epoch": 6.59, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8976, | |
| "step": 4560 | |
| }, | |
| { | |
| "epoch": 6.61, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8361, | |
| "step": 4570 | |
| }, | |
| { | |
| "epoch": 6.62, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7947, | |
| "step": 4580 | |
| }, | |
| { | |
| "epoch": 6.64, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7715, | |
| "step": 4590 | |
| }, | |
| { | |
| "epoch": 6.65, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7046, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 6.67, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8952, | |
| "step": 4610 | |
| }, | |
| { | |
| "epoch": 6.68, | |
| "learning_rate": 0.0002, | |
| "loss": 0.848, | |
| "step": 4620 | |
| }, | |
| { | |
| "epoch": 6.7, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7885, | |
| "step": 4630 | |
| }, | |
| { | |
| "epoch": 6.71, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7753, | |
| "step": 4640 | |
| }, | |
| { | |
| "epoch": 6.72, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7011, | |
| "step": 4650 | |
| }, | |
| { | |
| "epoch": 6.74, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8948, | |
| "step": 4660 | |
| }, | |
| { | |
| "epoch": 6.75, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8186, | |
| "step": 4670 | |
| }, | |
| { | |
| "epoch": 6.77, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7991, | |
| "step": 4680 | |
| }, | |
| { | |
| "epoch": 6.78, | |
| "learning_rate": 0.0002, | |
| "loss": 0.756, | |
| "step": 4690 | |
| }, | |
| { | |
| "epoch": 6.8, | |
| "learning_rate": 0.0002, | |
| "loss": 0.6999, | |
| "step": 4700 | |
| }, | |
| { | |
| "epoch": 6.81, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8991, | |
| "step": 4710 | |
| }, | |
| { | |
| "epoch": 6.83, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8334, | |
| "step": 4720 | |
| }, | |
| { | |
| "epoch": 6.84, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7978, | |
| "step": 4730 | |
| }, | |
| { | |
| "epoch": 6.85, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7514, | |
| "step": 4740 | |
| }, | |
| { | |
| "epoch": 6.87, | |
| "learning_rate": 0.0002, | |
| "loss": 0.6968, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 6.88, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8783, | |
| "step": 4760 | |
| }, | |
| { | |
| "epoch": 6.9, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8291, | |
| "step": 4770 | |
| }, | |
| { | |
| "epoch": 6.91, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8059, | |
| "step": 4780 | |
| }, | |
| { | |
| "epoch": 6.93, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7764, | |
| "step": 4790 | |
| }, | |
| { | |
| "epoch": 6.94, | |
| "learning_rate": 0.0002, | |
| "loss": 0.6978, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 6.96, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8875, | |
| "step": 4810 | |
| }, | |
| { | |
| "epoch": 6.97, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8238, | |
| "step": 4820 | |
| }, | |
| { | |
| "epoch": 6.98, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7776, | |
| "step": 4830 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7001, | |
| "step": 4840 | |
| }, | |
| { | |
| "epoch": 7.01, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8558, | |
| "step": 4850 | |
| }, | |
| { | |
| "epoch": 7.03, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8014, | |
| "step": 4860 | |
| }, | |
| { | |
| "epoch": 7.04, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7671, | |
| "step": 4870 | |
| }, | |
| { | |
| "epoch": 7.06, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7089, | |
| "step": 4880 | |
| }, | |
| { | |
| "epoch": 7.07, | |
| "learning_rate": 0.0002, | |
| "loss": 0.6218, | |
| "step": 4890 | |
| }, | |
| { | |
| "epoch": 7.09, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8637, | |
| "step": 4900 | |
| }, | |
| { | |
| "epoch": 7.1, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8098, | |
| "step": 4910 | |
| }, | |
| { | |
| "epoch": 7.11, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7572, | |
| "step": 4920 | |
| }, | |
| { | |
| "epoch": 7.13, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7368, | |
| "step": 4930 | |
| }, | |
| { | |
| "epoch": 7.14, | |
| "learning_rate": 0.0002, | |
| "loss": 0.6251, | |
| "step": 4940 | |
| }, | |
| { | |
| "epoch": 7.16, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8433, | |
| "step": 4950 | |
| }, | |
| { | |
| "epoch": 7.17, | |
| "learning_rate": 0.0002, | |
| "loss": 0.8094, | |
| "step": 4960 | |
| }, | |
| { | |
| "epoch": 7.19, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7691, | |
| "step": 4970 | |
| }, | |
| { | |
| "epoch": 7.2, | |
| "learning_rate": 0.0002, | |
| "loss": 0.7159, | |
| "step": 4980 | |
| }, | |
| { | |
| "epoch": 7.22, | |
| "learning_rate": 0.0002, | |
| "loss": 0.6349, | |
| "step": 4990 | |
| }, | |
| { | |
| "epoch": 7.23, | |
| "learning_rate": 0.0002, | |
| "loss": 0.854, | |
| "step": 5000 | |
| } | |
| ], | |
| "max_steps": 5000, | |
| "num_train_epochs": 8, | |
| "total_flos": 1.0738578956363366e+18, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |