| { | |
| "best_metric": 0.5641044692944979, | |
| "best_model_checkpoint": "resnet-50-finetuned-dangerousSpiders/checkpoint-6728", | |
| "epoch": 1.9995542016494539, | |
| "eval_steps": 500, | |
| "global_step": 6728, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 7.429420505200595e-07, | |
| "loss": 5.405, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.485884101040119e-06, | |
| "loss": 5.4039, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 2.2288261515601783e-06, | |
| "loss": 5.4043, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 2.971768202080238e-06, | |
| "loss": 5.4038, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 3.7147102526002974e-06, | |
| "loss": 5.3921, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 4.457652303120357e-06, | |
| "loss": 5.3909, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 5.200594353640416e-06, | |
| "loss": 5.3826, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 5.943536404160476e-06, | |
| "loss": 5.3757, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 6.686478454680535e-06, | |
| "loss": 5.3772, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 7.429420505200595e-06, | |
| "loss": 5.3602, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 8.172362555720654e-06, | |
| "loss": 5.3515, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 8.915304606240713e-06, | |
| "loss": 5.3484, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.658246656760773e-06, | |
| "loss": 5.33, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.0401188707280832e-05, | |
| "loss": 5.3247, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.1144130757800892e-05, | |
| "loss": 5.3141, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.1887072808320951e-05, | |
| "loss": 5.2959, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.2630014858841011e-05, | |
| "loss": 5.2845, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.337295690936107e-05, | |
| "loss": 5.2776, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.411589895988113e-05, | |
| "loss": 5.2476, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.485884101040119e-05, | |
| "loss": 5.2416, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.560178306092125e-05, | |
| "loss": 5.218, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.6344725111441307e-05, | |
| "loss": 5.1922, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.708766716196137e-05, | |
| "loss": 5.1848, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.7830609212481426e-05, | |
| "loss": 5.1666, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.8573551263001488e-05, | |
| "loss": 5.1409, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9316493313521546e-05, | |
| "loss": 5.1223, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.0059435364041607e-05, | |
| "loss": 5.0851, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.0802377414561665e-05, | |
| "loss": 5.0707, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.1545319465081726e-05, | |
| "loss": 5.0444, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.2288261515601784e-05, | |
| "loss": 5.039, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.3031203566121842e-05, | |
| "loss": 4.9794, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 2.3774145616641903e-05, | |
| "loss": 4.9587, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 2.451708766716196e-05, | |
| "loss": 4.9096, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 2.5260029717682022e-05, | |
| "loss": 4.8677, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 2.6002971768202083e-05, | |
| "loss": 4.8764, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 2.674591381872214e-05, | |
| "loss": 4.8059, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 2.7488855869242203e-05, | |
| "loss": 4.8177, | |
| "step": 370 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 2.823179791976226e-05, | |
| "loss": 4.7389, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 2.8974739970282322e-05, | |
| "loss": 4.7242, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 2.971768202080238e-05, | |
| "loss": 4.7173, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 3.0460624071322434e-05, | |
| "loss": 4.6278, | |
| "step": 410 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 3.12035661218425e-05, | |
| "loss": 4.5485, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 3.1946508172362557e-05, | |
| "loss": 4.5207, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 3.2689450222882614e-05, | |
| "loss": 4.507, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 3.343239227340267e-05, | |
| "loss": 4.4682, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 3.417533432392274e-05, | |
| "loss": 4.4301, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 3.4918276374442795e-05, | |
| "loss": 4.4142, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 3.566121842496285e-05, | |
| "loss": 4.3468, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 3.640416047548291e-05, | |
| "loss": 4.3323, | |
| "step": 490 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 3.7147102526002975e-05, | |
| "loss": 4.2918, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 3.789004457652303e-05, | |
| "loss": 4.2163, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 3.863298662704309e-05, | |
| "loss": 4.2232, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 3.937592867756315e-05, | |
| "loss": 4.1624, | |
| "step": 530 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 4.0118870728083214e-05, | |
| "loss": 4.0693, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 4.086181277860327e-05, | |
| "loss": 4.0696, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 4.160475482912333e-05, | |
| "loss": 4.0959, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 4.234769687964339e-05, | |
| "loss": 4.0907, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 4.309063893016345e-05, | |
| "loss": 4.0332, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 4.383358098068351e-05, | |
| "loss": 4.0683, | |
| "step": 590 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 4.457652303120357e-05, | |
| "loss": 4.0444, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 4.5319465081723626e-05, | |
| "loss": 4.0296, | |
| "step": 610 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 4.6062407132243683e-05, | |
| "loss": 3.9832, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 4.680534918276375e-05, | |
| "loss": 3.926, | |
| "step": 630 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 4.7548291233283806e-05, | |
| "loss": 3.9267, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 4.8291233283803864e-05, | |
| "loss": 3.9513, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 4.903417533432392e-05, | |
| "loss": 3.8802, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 4.9777117384843986e-05, | |
| "loss": 3.9074, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 4.994219653179191e-05, | |
| "loss": 3.8442, | |
| "step": 680 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 4.985962014863749e-05, | |
| "loss": 3.8965, | |
| "step": 690 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 4.9777043765483074e-05, | |
| "loss": 3.826, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 4.9694467382328656e-05, | |
| "loss": 3.794, | |
| "step": 710 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 4.961189099917424e-05, | |
| "loss": 3.8224, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 4.952931461601982e-05, | |
| "loss": 3.7774, | |
| "step": 730 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 4.94467382328654e-05, | |
| "loss": 3.8306, | |
| "step": 740 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 4.9364161849710985e-05, | |
| "loss": 3.6768, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 4.928158546655657e-05, | |
| "loss": 3.6991, | |
| "step": 760 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 4.919900908340215e-05, | |
| "loss": 3.7721, | |
| "step": 770 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 4.911643270024773e-05, | |
| "loss": 3.7815, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 4.9033856317093314e-05, | |
| "loss": 3.907, | |
| "step": 790 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 4.8951279933938896e-05, | |
| "loss": 3.7194, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 4.886870355078448e-05, | |
| "loss": 3.6438, | |
| "step": 810 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 4.878612716763006e-05, | |
| "loss": 3.5943, | |
| "step": 820 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 4.870355078447564e-05, | |
| "loss": 3.6272, | |
| "step": 830 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 4.8620974401321225e-05, | |
| "loss": 3.6241, | |
| "step": 840 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 4.853839801816681e-05, | |
| "loss": 3.6919, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 4.845582163501239e-05, | |
| "loss": 3.5097, | |
| "step": 860 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 4.8373245251857965e-05, | |
| "loss": 3.7713, | |
| "step": 870 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 4.8290668868703554e-05, | |
| "loss": 3.7294, | |
| "step": 880 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 4.8208092485549136e-05, | |
| "loss": 3.5285, | |
| "step": 890 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 4.812551610239472e-05, | |
| "loss": 3.5451, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 4.80429397192403e-05, | |
| "loss": 3.492, | |
| "step": 910 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 4.796036333608588e-05, | |
| "loss": 3.621, | |
| "step": 920 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 4.7877786952931465e-05, | |
| "loss": 3.6066, | |
| "step": 930 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 4.779521056977704e-05, | |
| "loss": 3.6187, | |
| "step": 940 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 4.771263418662263e-05, | |
| "loss": 3.5503, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 4.763005780346821e-05, | |
| "loss": 3.5894, | |
| "step": 960 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 4.7547481420313794e-05, | |
| "loss": 3.4964, | |
| "step": 970 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 4.7464905037159377e-05, | |
| "loss": 3.3646, | |
| "step": 980 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 4.738232865400496e-05, | |
| "loss": 3.5244, | |
| "step": 990 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 4.729975227085054e-05, | |
| "loss": 3.5858, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 4.7217175887696117e-05, | |
| "loss": 3.4835, | |
| "step": 1010 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 4.7134599504541706e-05, | |
| "loss": 3.4223, | |
| "step": 1020 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 4.705202312138728e-05, | |
| "loss": 3.4574, | |
| "step": 1030 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 4.696944673823287e-05, | |
| "loss": 3.4394, | |
| "step": 1040 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 4.688687035507845e-05, | |
| "loss": 3.4179, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 4.680429397192403e-05, | |
| "loss": 3.4564, | |
| "step": 1060 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 4.672171758876962e-05, | |
| "loss": 3.349, | |
| "step": 1070 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 4.663914120561519e-05, | |
| "loss": 3.5325, | |
| "step": 1080 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 4.655656482246078e-05, | |
| "loss": 3.4651, | |
| "step": 1090 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 4.647398843930636e-05, | |
| "loss": 3.4254, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 4.6391412056151946e-05, | |
| "loss": 3.2702, | |
| "step": 1110 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 4.630883567299753e-05, | |
| "loss": 3.4774, | |
| "step": 1120 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 4.6226259289843103e-05, | |
| "loss": 3.2563, | |
| "step": 1130 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 4.614368290668869e-05, | |
| "loss": 3.3301, | |
| "step": 1140 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 4.606110652353427e-05, | |
| "loss": 3.2156, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 4.597853014037986e-05, | |
| "loss": 3.2333, | |
| "step": 1160 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 4.589595375722543e-05, | |
| "loss": 3.2974, | |
| "step": 1170 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 4.581337737407102e-05, | |
| "loss": 3.3184, | |
| "step": 1180 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 4.57308009909166e-05, | |
| "loss": 3.2369, | |
| "step": 1190 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 4.564822460776218e-05, | |
| "loss": 3.2378, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 4.556564822460777e-05, | |
| "loss": 3.1896, | |
| "step": 1210 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 4.5483071841453344e-05, | |
| "loss": 3.2404, | |
| "step": 1220 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 4.540049545829893e-05, | |
| "loss": 3.2338, | |
| "step": 1230 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 4.531791907514451e-05, | |
| "loss": 3.3576, | |
| "step": 1240 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 4.523534269199009e-05, | |
| "loss": 3.0155, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 4.515276630883567e-05, | |
| "loss": 3.0852, | |
| "step": 1260 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 4.5070189925681255e-05, | |
| "loss": 3.1457, | |
| "step": 1270 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 4.4987613542526844e-05, | |
| "loss": 3.1342, | |
| "step": 1280 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 4.490503715937242e-05, | |
| "loss": 3.1097, | |
| "step": 1290 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 4.482246077621801e-05, | |
| "loss": 3.2993, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 4.4739884393063584e-05, | |
| "loss": 3.1181, | |
| "step": 1310 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 4.4657308009909166e-05, | |
| "loss": 3.0694, | |
| "step": 1320 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 4.457473162675475e-05, | |
| "loss": 3.2075, | |
| "step": 1330 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 4.449215524360033e-05, | |
| "loss": 3.1879, | |
| "step": 1340 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 4.440957886044591e-05, | |
| "loss": 3.0836, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 4.4327002477291495e-05, | |
| "loss": 3.1561, | |
| "step": 1360 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 4.4244426094137084e-05, | |
| "loss": 3.1369, | |
| "step": 1370 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 4.416184971098266e-05, | |
| "loss": 2.9664, | |
| "step": 1380 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 4.407927332782824e-05, | |
| "loss": 3.0035, | |
| "step": 1390 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 4.3996696944673824e-05, | |
| "loss": 2.9866, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 4.3914120561519406e-05, | |
| "loss": 3.0382, | |
| "step": 1410 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 4.383154417836499e-05, | |
| "loss": 3.0908, | |
| "step": 1420 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 4.374896779521057e-05, | |
| "loss": 3.0488, | |
| "step": 1430 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 4.366639141205616e-05, | |
| "loss": 3.0779, | |
| "step": 1440 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 4.3583815028901735e-05, | |
| "loss": 2.9275, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 4.350123864574732e-05, | |
| "loss": 3.1103, | |
| "step": 1460 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 4.34186622625929e-05, | |
| "loss": 2.9926, | |
| "step": 1470 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 4.333608587943848e-05, | |
| "loss": 3.0584, | |
| "step": 1480 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 4.3253509496284064e-05, | |
| "loss": 3.0046, | |
| "step": 1490 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 4.3170933113129646e-05, | |
| "loss": 2.9953, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 4.308835672997523e-05, | |
| "loss": 3.0283, | |
| "step": 1510 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 4.300578034682081e-05, | |
| "loss": 2.9021, | |
| "step": 1520 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 4.292320396366639e-05, | |
| "loss": 3.1296, | |
| "step": 1530 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 4.2840627580511975e-05, | |
| "loss": 3.1279, | |
| "step": 1540 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 4.275805119735756e-05, | |
| "loss": 2.9625, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 4.267547481420314e-05, | |
| "loss": 2.9347, | |
| "step": 1560 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 4.259289843104872e-05, | |
| "loss": 3.0066, | |
| "step": 1570 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 4.2510322047894304e-05, | |
| "loss": 2.9883, | |
| "step": 1580 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 4.242774566473989e-05, | |
| "loss": 2.8869, | |
| "step": 1590 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 4.234516928158547e-05, | |
| "loss": 3.0961, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 4.226259289843105e-05, | |
| "loss": 2.9767, | |
| "step": 1610 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 4.2180016515276633e-05, | |
| "loss": 2.9925, | |
| "step": 1620 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 4.2097440132122216e-05, | |
| "loss": 2.8426, | |
| "step": 1630 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 4.20148637489678e-05, | |
| "loss": 2.9307, | |
| "step": 1640 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 4.193228736581338e-05, | |
| "loss": 2.739, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 4.184971098265896e-05, | |
| "loss": 2.7966, | |
| "step": 1660 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 4.1767134599504545e-05, | |
| "loss": 2.943, | |
| "step": 1670 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 4.168455821635013e-05, | |
| "loss": 2.9382, | |
| "step": 1680 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 4.160198183319571e-05, | |
| "loss": 2.823, | |
| "step": 1690 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.1519405450041285e-05, | |
| "loss": 2.9749, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.1436829066886874e-05, | |
| "loss": 2.9249, | |
| "step": 1710 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.1354252683732456e-05, | |
| "loss": 2.9081, | |
| "step": 1720 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.127167630057804e-05, | |
| "loss": 2.7706, | |
| "step": 1730 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.118909991742362e-05, | |
| "loss": 2.8004, | |
| "step": 1740 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.11065235342692e-05, | |
| "loss": 2.8529, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.1023947151114785e-05, | |
| "loss": 2.8549, | |
| "step": 1760 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.094137076796036e-05, | |
| "loss": 2.8035, | |
| "step": 1770 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.085879438480595e-05, | |
| "loss": 2.8468, | |
| "step": 1780 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.077621800165153e-05, | |
| "loss": 2.7341, | |
| "step": 1790 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.0693641618497114e-05, | |
| "loss": 2.9054, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.0611065235342696e-05, | |
| "loss": 2.8449, | |
| "step": 1810 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.052848885218828e-05, | |
| "loss": 2.8226, | |
| "step": 1820 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.044591246903386e-05, | |
| "loss": 2.8705, | |
| "step": 1830 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.0363336085879436e-05, | |
| "loss": 2.8243, | |
| "step": 1840 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.0280759702725025e-05, | |
| "loss": 2.7621, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.01981833195706e-05, | |
| "loss": 2.8886, | |
| "step": 1860 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 4.011560693641619e-05, | |
| "loss": 2.7429, | |
| "step": 1870 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 4.003303055326177e-05, | |
| "loss": 2.8031, | |
| "step": 1880 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 3.995045417010735e-05, | |
| "loss": 2.6617, | |
| "step": 1890 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 3.9867877786952936e-05, | |
| "loss": 2.9073, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 3.978530140379851e-05, | |
| "loss": 2.6695, | |
| "step": 1910 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 3.97027250206441e-05, | |
| "loss": 2.7126, | |
| "step": 1920 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 3.9620148637489676e-05, | |
| "loss": 2.6894, | |
| "step": 1930 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 3.9537572254335265e-05, | |
| "loss": 2.7675, | |
| "step": 1940 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 3.945499587118085e-05, | |
| "loss": 2.8897, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 3.937241948802642e-05, | |
| "loss": 2.7975, | |
| "step": 1960 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 3.928984310487201e-05, | |
| "loss": 2.5923, | |
| "step": 1970 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 3.920726672171759e-05, | |
| "loss": 2.7971, | |
| "step": 1980 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 3.9124690338563176e-05, | |
| "loss": 2.8001, | |
| "step": 1990 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 3.904211395540875e-05, | |
| "loss": 2.6936, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 3.895953757225434e-05, | |
| "loss": 2.7541, | |
| "step": 2010 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 3.8876961189099916e-05, | |
| "loss": 2.7161, | |
| "step": 2020 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 3.87943848059455e-05, | |
| "loss": 2.7168, | |
| "step": 2030 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.871180842279109e-05, | |
| "loss": 2.5064, | |
| "step": 2040 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.862923203963666e-05, | |
| "loss": 2.6671, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.854665565648225e-05, | |
| "loss": 2.7833, | |
| "step": 2060 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.846407927332783e-05, | |
| "loss": 2.7791, | |
| "step": 2070 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.838150289017341e-05, | |
| "loss": 2.7521, | |
| "step": 2080 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.829892650701899e-05, | |
| "loss": 2.7129, | |
| "step": 2090 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 3.8216350123864574e-05, | |
| "loss": 2.6336, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 3.813377374071016e-05, | |
| "loss": 2.7362, | |
| "step": 2110 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 3.805119735755574e-05, | |
| "loss": 2.8606, | |
| "step": 2120 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 3.796862097440133e-05, | |
| "loss": 2.5801, | |
| "step": 2130 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 3.78860445912469e-05, | |
| "loss": 2.7365, | |
| "step": 2140 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 3.7803468208092486e-05, | |
| "loss": 2.5675, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 3.772089182493807e-05, | |
| "loss": 2.7219, | |
| "step": 2160 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 3.763831544178365e-05, | |
| "loss": 2.5957, | |
| "step": 2170 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 3.755573905862924e-05, | |
| "loss": 2.7097, | |
| "step": 2180 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 3.7473162675474815e-05, | |
| "loss": 2.635, | |
| "step": 2190 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 3.7390586292320404e-05, | |
| "loss": 2.5425, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 3.730800990916598e-05, | |
| "loss": 2.6557, | |
| "step": 2210 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 3.722543352601156e-05, | |
| "loss": 2.5417, | |
| "step": 2220 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 3.7142857142857143e-05, | |
| "loss": 2.5551, | |
| "step": 2230 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 3.7060280759702726e-05, | |
| "loss": 2.4944, | |
| "step": 2240 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 3.697770437654831e-05, | |
| "loss": 2.7367, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 3.689512799339389e-05, | |
| "loss": 2.6595, | |
| "step": 2260 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 3.681255161023947e-05, | |
| "loss": 2.6913, | |
| "step": 2270 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 3.6729975227085055e-05, | |
| "loss": 2.6926, | |
| "step": 2280 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 3.664739884393064e-05, | |
| "loss": 2.5914, | |
| "step": 2290 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 3.656482246077622e-05, | |
| "loss": 2.6779, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 3.64822460776218e-05, | |
| "loss": 2.7525, | |
| "step": 2310 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 3.6399669694467384e-05, | |
| "loss": 2.5466, | |
| "step": 2320 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 3.6317093311312966e-05, | |
| "loss": 2.724, | |
| "step": 2330 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 3.623451692815855e-05, | |
| "loss": 2.6471, | |
| "step": 2340 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 3.615194054500413e-05, | |
| "loss": 2.5751, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 3.606936416184971e-05, | |
| "loss": 2.5643, | |
| "step": 2360 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 3.5986787778695295e-05, | |
| "loss": 2.6515, | |
| "step": 2370 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 3.590421139554088e-05, | |
| "loss": 2.7272, | |
| "step": 2380 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 3.582163501238646e-05, | |
| "loss": 2.6195, | |
| "step": 2390 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 3.573905862923204e-05, | |
| "loss": 2.6677, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.5656482246077624e-05, | |
| "loss": 2.5896, | |
| "step": 2410 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.5573905862923206e-05, | |
| "loss": 2.6583, | |
| "step": 2420 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.549132947976879e-05, | |
| "loss": 2.6532, | |
| "step": 2430 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.540875309661437e-05, | |
| "loss": 2.6657, | |
| "step": 2440 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.532617671345995e-05, | |
| "loss": 2.643, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.5243600330305535e-05, | |
| "loss": 2.6946, | |
| "step": 2460 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.516102394715112e-05, | |
| "loss": 2.532, | |
| "step": 2470 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.50784475639967e-05, | |
| "loss": 2.661, | |
| "step": 2480 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.499587118084228e-05, | |
| "loss": 2.7058, | |
| "step": 2490 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.4913294797687864e-05, | |
| "loss": 2.7606, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.4830718414533446e-05, | |
| "loss": 2.5542, | |
| "step": 2510 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.474814203137903e-05, | |
| "loss": 2.6459, | |
| "step": 2520 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.4665565648224604e-05, | |
| "loss": 2.6128, | |
| "step": 2530 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.458298926507019e-05, | |
| "loss": 2.4528, | |
| "step": 2540 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 3.4500412881915775e-05, | |
| "loss": 2.5791, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 3.441783649876136e-05, | |
| "loss": 2.6544, | |
| "step": 2560 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 3.433526011560694e-05, | |
| "loss": 2.4937, | |
| "step": 2570 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 3.425268373245252e-05, | |
| "loss": 2.5646, | |
| "step": 2580 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 3.4170107349298104e-05, | |
| "loss": 2.5199, | |
| "step": 2590 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 3.408753096614368e-05, | |
| "loss": 2.6409, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 3.400495458298927e-05, | |
| "loss": 2.4757, | |
| "step": 2610 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 3.392237819983485e-05, | |
| "loss": 2.495, | |
| "step": 2620 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 3.383980181668043e-05, | |
| "loss": 2.4249, | |
| "step": 2630 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 3.3757225433526015e-05, | |
| "loss": 2.4278, | |
| "step": 2640 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 3.367464905037159e-05, | |
| "loss": 2.5724, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 3.359207266721718e-05, | |
| "loss": 2.5795, | |
| "step": 2660 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 3.3509496284062755e-05, | |
| "loss": 2.512, | |
| "step": 2670 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 3.3426919900908344e-05, | |
| "loss": 2.5784, | |
| "step": 2680 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 3.334434351775392e-05, | |
| "loss": 2.5895, | |
| "step": 2690 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 3.326176713459951e-05, | |
| "loss": 2.44, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.317919075144509e-05, | |
| "loss": 2.407, | |
| "step": 2710 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.309661436829067e-05, | |
| "loss": 2.5788, | |
| "step": 2720 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.3014037985136256e-05, | |
| "loss": 2.6036, | |
| "step": 2730 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.293146160198183e-05, | |
| "loss": 2.5594, | |
| "step": 2740 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.284888521882742e-05, | |
| "loss": 2.5886, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.2766308835672996e-05, | |
| "loss": 2.4467, | |
| "step": 2760 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.2683732452518585e-05, | |
| "loss": 2.527, | |
| "step": 2770 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.260115606936417e-05, | |
| "loss": 2.3767, | |
| "step": 2780 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.251857968620974e-05, | |
| "loss": 2.6301, | |
| "step": 2790 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.243600330305533e-05, | |
| "loss": 2.5871, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 3.235342691990091e-05, | |
| "loss": 2.4975, | |
| "step": 2810 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 3.2270850536746496e-05, | |
| "loss": 2.5472, | |
| "step": 2820 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 3.218827415359207e-05, | |
| "loss": 2.4236, | |
| "step": 2830 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 3.2105697770437654e-05, | |
| "loss": 2.3969, | |
| "step": 2840 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 3.202312138728324e-05, | |
| "loss": 2.4679, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 3.194054500412882e-05, | |
| "loss": 2.7221, | |
| "step": 2860 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 3.185796862097441e-05, | |
| "loss": 2.4721, | |
| "step": 2870 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 3.177539223781998e-05, | |
| "loss": 2.4576, | |
| "step": 2880 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 3.169281585466557e-05, | |
| "loss": 2.5149, | |
| "step": 2890 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 3.161023947151115e-05, | |
| "loss": 2.4146, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 3.152766308835673e-05, | |
| "loss": 2.4355, | |
| "step": 2910 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 3.144508670520231e-05, | |
| "loss": 2.474, | |
| "step": 2920 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 3.1362510322047894e-05, | |
| "loss": 2.3925, | |
| "step": 2930 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 3.127993393889348e-05, | |
| "loss": 2.321, | |
| "step": 2940 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 3.119735755573906e-05, | |
| "loss": 2.5465, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 3.111478117258465e-05, | |
| "loss": 2.4741, | |
| "step": 2960 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 3.103220478943022e-05, | |
| "loss": 2.6036, | |
| "step": 2970 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 3.0949628406275805e-05, | |
| "loss": 2.2974, | |
| "step": 2980 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 3.086705202312139e-05, | |
| "loss": 2.3861, | |
| "step": 2990 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 3.078447563996697e-05, | |
| "loss": 2.3389, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 3.070189925681256e-05, | |
| "loss": 2.4809, | |
| "step": 3010 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 3.0619322873658134e-05, | |
| "loss": 2.4182, | |
| "step": 3020 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 3.0536746490503716e-05, | |
| "loss": 2.3364, | |
| "step": 3030 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 3.0454170107349302e-05, | |
| "loss": 2.5519, | |
| "step": 3040 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 3.037159372419488e-05, | |
| "loss": 2.5753, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 3.0289017341040466e-05, | |
| "loss": 2.5368, | |
| "step": 3060 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 3.0206440957886045e-05, | |
| "loss": 2.5524, | |
| "step": 3070 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.012386457473163e-05, | |
| "loss": 2.3608, | |
| "step": 3080 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.004128819157721e-05, | |
| "loss": 2.4974, | |
| "step": 3090 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 2.9958711808422792e-05, | |
| "loss": 2.391, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 2.9876135425268374e-05, | |
| "loss": 2.3334, | |
| "step": 3110 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.9793559042113956e-05, | |
| "loss": 2.5954, | |
| "step": 3120 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.9710982658959542e-05, | |
| "loss": 2.324, | |
| "step": 3130 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.962840627580512e-05, | |
| "loss": 2.4917, | |
| "step": 3140 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.9545829892650706e-05, | |
| "loss": 2.4194, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.9463253509496285e-05, | |
| "loss": 2.5282, | |
| "step": 3160 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.9380677126341864e-05, | |
| "loss": 2.4709, | |
| "step": 3170 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 2.929810074318745e-05, | |
| "loss": 2.5022, | |
| "step": 3180 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 2.9215524360033032e-05, | |
| "loss": 2.6882, | |
| "step": 3190 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 2.9132947976878618e-05, | |
| "loss": 2.3625, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 2.9050371593724197e-05, | |
| "loss": 2.4684, | |
| "step": 3210 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 2.8967795210569775e-05, | |
| "loss": 2.4225, | |
| "step": 3220 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 2.888521882741536e-05, | |
| "loss": 2.4635, | |
| "step": 3230 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 2.880264244426094e-05, | |
| "loss": 2.4034, | |
| "step": 3240 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 2.8720066061106526e-05, | |
| "loss": 2.4898, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 2.8637489677952108e-05, | |
| "loss": 2.4432, | |
| "step": 3260 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 2.855491329479769e-05, | |
| "loss": 2.3896, | |
| "step": 3270 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 2.8472336911643272e-05, | |
| "loss": 2.3928, | |
| "step": 3280 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.838976052848885e-05, | |
| "loss": 2.456, | |
| "step": 3290 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.8307184145334437e-05, | |
| "loss": 2.3562, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.8224607762180016e-05, | |
| "loss": 2.3572, | |
| "step": 3310 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 2.81420313790256e-05, | |
| "loss": 2.4603, | |
| "step": 3320 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 2.805945499587118e-05, | |
| "loss": 2.5909, | |
| "step": 3330 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 2.7976878612716766e-05, | |
| "loss": 2.3332, | |
| "step": 3340 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 2.7894302229562348e-05, | |
| "loss": 2.3701, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 2.7811725846407927e-05, | |
| "loss": 2.5677, | |
| "step": 3360 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.5328231229334621, | |
| "eval_f1": 0.05253328324856182, | |
| "eval_loss": 2.0752506256103516, | |
| "eval_precision": 0.0867161409969943, | |
| "eval_recall": 0.06132451140338959, | |
| "eval_runtime": 171.6239, | |
| "eval_samples_per_second": 156.837, | |
| "eval_steps_per_second": 9.806, | |
| "step": 3364 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 2.7729149463253512e-05, | |
| "loss": 2.458, | |
| "step": 3370 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 2.764657308009909e-05, | |
| "loss": 2.3922, | |
| "step": 3380 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 2.7563996696944677e-05, | |
| "loss": 2.3989, | |
| "step": 3390 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 2.7481420313790256e-05, | |
| "loss": 2.5164, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 2.7398843930635838e-05, | |
| "loss": 2.3984, | |
| "step": 3410 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 2.7316267547481424e-05, | |
| "loss": 2.2326, | |
| "step": 3420 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 2.7233691164327003e-05, | |
| "loss": 2.515, | |
| "step": 3430 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 2.7151114781172588e-05, | |
| "loss": 2.4275, | |
| "step": 3440 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 2.7068538398018167e-05, | |
| "loss": 2.3412, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 2.6985962014863753e-05, | |
| "loss": 2.4333, | |
| "step": 3460 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 2.690338563170933e-05, | |
| "loss": 2.4109, | |
| "step": 3470 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 2.6820809248554914e-05, | |
| "loss": 2.4189, | |
| "step": 3480 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 2.6738232865400496e-05, | |
| "loss": 2.4701, | |
| "step": 3490 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 2.6655656482246078e-05, | |
| "loss": 2.346, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 2.6573080099091664e-05, | |
| "loss": 2.3714, | |
| "step": 3510 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 2.6490503715937243e-05, | |
| "loss": 2.3199, | |
| "step": 3520 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 2.640792733278283e-05, | |
| "loss": 2.6262, | |
| "step": 3530 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 2.6325350949628407e-05, | |
| "loss": 2.3353, | |
| "step": 3540 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 2.6242774566473986e-05, | |
| "loss": 2.3815, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 2.616019818331957e-05, | |
| "loss": 2.4503, | |
| "step": 3560 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 2.6077621800165154e-05, | |
| "loss": 2.2936, | |
| "step": 3570 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 2.599504541701074e-05, | |
| "loss": 2.2848, | |
| "step": 3580 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 2.591246903385632e-05, | |
| "loss": 2.5068, | |
| "step": 3590 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 2.5829892650701897e-05, | |
| "loss": 2.2433, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 2.5747316267547483e-05, | |
| "loss": 2.4485, | |
| "step": 3610 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 2.5664739884393062e-05, | |
| "loss": 2.3825, | |
| "step": 3620 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 2.5582163501238647e-05, | |
| "loss": 2.299, | |
| "step": 3630 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 2.549958711808423e-05, | |
| "loss": 2.3278, | |
| "step": 3640 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 2.5417010734929815e-05, | |
| "loss": 2.2822, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 2.5334434351775394e-05, | |
| "loss": 2.3737, | |
| "step": 3660 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 2.5251857968620973e-05, | |
| "loss": 2.4658, | |
| "step": 3670 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 2.516928158546656e-05, | |
| "loss": 2.3204, | |
| "step": 3680 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 2.5086705202312137e-05, | |
| "loss": 2.2318, | |
| "step": 3690 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 2.5004128819157723e-05, | |
| "loss": 2.3069, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 2.4921552436003305e-05, | |
| "loss": 2.1357, | |
| "step": 3710 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 2.4838976052848888e-05, | |
| "loss": 2.2658, | |
| "step": 3720 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 2.475639966969447e-05, | |
| "loss": 2.4089, | |
| "step": 3730 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 2.4673823286540052e-05, | |
| "loss": 2.2985, | |
| "step": 3740 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 2.4591246903385634e-05, | |
| "loss": 2.3225, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 2.4508670520231213e-05, | |
| "loss": 2.2832, | |
| "step": 3760 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 2.4426094137076795e-05, | |
| "loss": 2.414, | |
| "step": 3770 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 2.4343517753922378e-05, | |
| "loss": 2.2114, | |
| "step": 3780 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 2.4260941370767963e-05, | |
| "loss": 2.258, | |
| "step": 3790 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 2.4178364987613546e-05, | |
| "loss": 2.3341, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 2.4095788604459128e-05, | |
| "loss": 2.2882, | |
| "step": 3810 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 2.4013212221304707e-05, | |
| "loss": 2.3976, | |
| "step": 3820 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 2.393063583815029e-05, | |
| "loss": 2.3602, | |
| "step": 3830 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 2.384805945499587e-05, | |
| "loss": 2.3432, | |
| "step": 3840 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 2.3765483071841453e-05, | |
| "loss": 2.4601, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 2.3682906688687036e-05, | |
| "loss": 2.3406, | |
| "step": 3860 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 2.360033030553262e-05, | |
| "loss": 2.3154, | |
| "step": 3870 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 2.3517753922378203e-05, | |
| "loss": 2.5264, | |
| "step": 3880 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 2.3435177539223782e-05, | |
| "loss": 2.4018, | |
| "step": 3890 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 2.3352601156069365e-05, | |
| "loss": 2.4116, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 2.3270024772914947e-05, | |
| "loss": 2.2176, | |
| "step": 3910 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 2.318744838976053e-05, | |
| "loss": 2.3973, | |
| "step": 3920 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 2.310487200660611e-05, | |
| "loss": 2.3694, | |
| "step": 3930 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 2.3022295623451694e-05, | |
| "loss": 2.3191, | |
| "step": 3940 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 2.2939719240297276e-05, | |
| "loss": 2.2085, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 2.2857142857142858e-05, | |
| "loss": 2.1206, | |
| "step": 3960 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 2.277456647398844e-05, | |
| "loss": 2.2114, | |
| "step": 3970 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 2.2691990090834023e-05, | |
| "loss": 2.342, | |
| "step": 3980 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 2.2609413707679605e-05, | |
| "loss": 2.4578, | |
| "step": 3990 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 2.2526837324525187e-05, | |
| "loss": 2.4828, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 2.244426094137077e-05, | |
| "loss": 2.2953, | |
| "step": 4010 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 2.236168455821635e-05, | |
| "loss": 2.3039, | |
| "step": 4020 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 2.2279108175061934e-05, | |
| "loss": 2.4023, | |
| "step": 4030 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 2.2196531791907516e-05, | |
| "loss": 2.3618, | |
| "step": 4040 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 2.2113955408753098e-05, | |
| "loss": 2.2806, | |
| "step": 4050 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 2.203137902559868e-05, | |
| "loss": 2.352, | |
| "step": 4060 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 2.1948802642444263e-05, | |
| "loss": 2.2781, | |
| "step": 4070 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 2.186622625928984e-05, | |
| "loss": 2.3673, | |
| "step": 4080 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 2.1783649876135427e-05, | |
| "loss": 2.1819, | |
| "step": 4090 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 2.170107349298101e-05, | |
| "loss": 2.3615, | |
| "step": 4100 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 2.1618497109826592e-05, | |
| "loss": 2.5218, | |
| "step": 4110 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 2.1535920726672174e-05, | |
| "loss": 2.342, | |
| "step": 4120 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 2.1453344343517756e-05, | |
| "loss": 2.384, | |
| "step": 4130 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 2.1370767960363335e-05, | |
| "loss": 2.2196, | |
| "step": 4140 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 2.1288191577208917e-05, | |
| "loss": 2.2029, | |
| "step": 4150 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 2.12056151940545e-05, | |
| "loss": 2.2433, | |
| "step": 4160 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 2.1123038810900085e-05, | |
| "loss": 2.4107, | |
| "step": 4170 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 2.1040462427745667e-05, | |
| "loss": 2.4475, | |
| "step": 4180 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 2.095788604459125e-05, | |
| "loss": 2.4814, | |
| "step": 4190 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 2.087530966143683e-05, | |
| "loss": 2.3149, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 2.079273327828241e-05, | |
| "loss": 2.2011, | |
| "step": 4210 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 2.0710156895127993e-05, | |
| "loss": 2.3177, | |
| "step": 4220 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 2.0627580511973575e-05, | |
| "loss": 2.2045, | |
| "step": 4230 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 2.0545004128819157e-05, | |
| "loss": 2.353, | |
| "step": 4240 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 2.0462427745664743e-05, | |
| "loss": 2.2882, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 2.0379851362510325e-05, | |
| "loss": 2.2864, | |
| "step": 4260 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 2.0297274979355904e-05, | |
| "loss": 2.3833, | |
| "step": 4270 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 2.0214698596201486e-05, | |
| "loss": 2.3435, | |
| "step": 4280 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 2.013212221304707e-05, | |
| "loss": 2.2024, | |
| "step": 4290 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 2.004954582989265e-05, | |
| "loss": 2.4308, | |
| "step": 4300 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 1.9966969446738233e-05, | |
| "loss": 2.3202, | |
| "step": 4310 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 1.988439306358382e-05, | |
| "loss": 2.3706, | |
| "step": 4320 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 1.9801816680429398e-05, | |
| "loss": 2.3166, | |
| "step": 4330 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 1.971924029727498e-05, | |
| "loss": 2.2627, | |
| "step": 4340 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 1.9636663914120562e-05, | |
| "loss": 2.2601, | |
| "step": 4350 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 1.9554087530966144e-05, | |
| "loss": 2.2989, | |
| "step": 4360 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 1.9471511147811727e-05, | |
| "loss": 2.0918, | |
| "step": 4370 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 1.938893476465731e-05, | |
| "loss": 2.3354, | |
| "step": 4380 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 1.930635838150289e-05, | |
| "loss": 2.4271, | |
| "step": 4390 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 1.9223781998348473e-05, | |
| "loss": 2.2184, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 1.9141205615194056e-05, | |
| "loss": 2.2901, | |
| "step": 4410 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 1.9058629232039638e-05, | |
| "loss": 2.3239, | |
| "step": 4420 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 1.897605284888522e-05, | |
| "loss": 2.1658, | |
| "step": 4430 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 1.8893476465730802e-05, | |
| "loss": 2.2772, | |
| "step": 4440 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 1.8810900082576385e-05, | |
| "loss": 2.2348, | |
| "step": 4450 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 1.8728323699421967e-05, | |
| "loss": 2.3802, | |
| "step": 4460 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 1.864574731626755e-05, | |
| "loss": 2.2576, | |
| "step": 4470 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 1.856317093311313e-05, | |
| "loss": 2.3755, | |
| "step": 4480 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 1.8480594549958714e-05, | |
| "loss": 2.2188, | |
| "step": 4490 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 1.8398018166804296e-05, | |
| "loss": 2.1154, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 1.8315441783649878e-05, | |
| "loss": 2.1984, | |
| "step": 4510 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 1.8232865400495457e-05, | |
| "loss": 2.1404, | |
| "step": 4520 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 1.815028901734104e-05, | |
| "loss": 2.3174, | |
| "step": 4530 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 1.8067712634186625e-05, | |
| "loss": 2.3589, | |
| "step": 4540 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 1.7985136251032207e-05, | |
| "loss": 2.1742, | |
| "step": 4550 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 1.790255986787779e-05, | |
| "loss": 2.1817, | |
| "step": 4560 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 1.781998348472337e-05, | |
| "loss": 2.2441, | |
| "step": 4570 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 1.773740710156895e-05, | |
| "loss": 2.3572, | |
| "step": 4580 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 1.7654830718414533e-05, | |
| "loss": 2.2951, | |
| "step": 4590 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 1.7572254335260115e-05, | |
| "loss": 2.4111, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 1.7489677952105697e-05, | |
| "loss": 2.3184, | |
| "step": 4610 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 1.7407101568951283e-05, | |
| "loss": 2.3761, | |
| "step": 4620 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 1.7324525185796865e-05, | |
| "loss": 2.3063, | |
| "step": 4630 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 1.7241948802642447e-05, | |
| "loss": 2.2103, | |
| "step": 4640 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 1.7159372419488026e-05, | |
| "loss": 2.1707, | |
| "step": 4650 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 1.707679603633361e-05, | |
| "loss": 2.293, | |
| "step": 4660 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 1.699421965317919e-05, | |
| "loss": 2.3733, | |
| "step": 4670 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 1.6911643270024773e-05, | |
| "loss": 2.1965, | |
| "step": 4680 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 1.6829066886870355e-05, | |
| "loss": 2.3514, | |
| "step": 4690 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 1.674649050371594e-05, | |
| "loss": 2.3012, | |
| "step": 4700 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 1.666391412056152e-05, | |
| "loss": 2.2125, | |
| "step": 4710 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 1.6581337737407102e-05, | |
| "loss": 2.4134, | |
| "step": 4720 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 1.6498761354252684e-05, | |
| "loss": 2.309, | |
| "step": 4730 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 1.6416184971098266e-05, | |
| "loss": 2.3972, | |
| "step": 4740 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 1.633360858794385e-05, | |
| "loss": 2.2777, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 1.625103220478943e-05, | |
| "loss": 2.118, | |
| "step": 4760 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 1.6168455821635013e-05, | |
| "loss": 2.1258, | |
| "step": 4770 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 1.6085879438480595e-05, | |
| "loss": 2.2947, | |
| "step": 4780 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 1.6003303055326177e-05, | |
| "loss": 2.2381, | |
| "step": 4790 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 1.592072667217176e-05, | |
| "loss": 2.3334, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 1.5838150289017342e-05, | |
| "loss": 2.2339, | |
| "step": 4810 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 1.5755573905862924e-05, | |
| "loss": 2.3019, | |
| "step": 4820 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 1.5672997522708506e-05, | |
| "loss": 2.353, | |
| "step": 4830 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 1.559042113955409e-05, | |
| "loss": 2.3393, | |
| "step": 4840 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 1.550784475639967e-05, | |
| "loss": 2.3197, | |
| "step": 4850 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 1.5425268373245253e-05, | |
| "loss": 2.2364, | |
| "step": 4860 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 1.5342691990090835e-05, | |
| "loss": 2.2193, | |
| "step": 4870 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 1.5260115606936418e-05, | |
| "loss": 2.1397, | |
| "step": 4880 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 1.5177539223782e-05, | |
| "loss": 2.2839, | |
| "step": 4890 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 1.509496284062758e-05, | |
| "loss": 2.2511, | |
| "step": 4900 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 1.5012386457473163e-05, | |
| "loss": 2.2102, | |
| "step": 4910 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 1.4929810074318745e-05, | |
| "loss": 2.1302, | |
| "step": 4920 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 1.4847233691164327e-05, | |
| "loss": 2.1151, | |
| "step": 4930 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 1.4764657308009911e-05, | |
| "loss": 2.3325, | |
| "step": 4940 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 1.4682080924855493e-05, | |
| "loss": 2.3439, | |
| "step": 4950 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 1.4599504541701072e-05, | |
| "loss": 2.2699, | |
| "step": 4960 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 1.4516928158546656e-05, | |
| "loss": 2.4093, | |
| "step": 4970 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 1.4434351775392238e-05, | |
| "loss": 2.3453, | |
| "step": 4980 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 1.435177539223782e-05, | |
| "loss": 2.2587, | |
| "step": 4990 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 1.4269199009083403e-05, | |
| "loss": 2.3447, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 1.4186622625928985e-05, | |
| "loss": 2.3058, | |
| "step": 5010 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 1.4104046242774569e-05, | |
| "loss": 2.3383, | |
| "step": 5020 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 1.4021469859620148e-05, | |
| "loss": 2.2672, | |
| "step": 5030 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 1.393889347646573e-05, | |
| "loss": 2.3583, | |
| "step": 5040 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 1.3856317093311314e-05, | |
| "loss": 2.2302, | |
| "step": 5050 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 1.3773740710156896e-05, | |
| "loss": 2.3005, | |
| "step": 5060 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 1.3691164327002479e-05, | |
| "loss": 2.2938, | |
| "step": 5070 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 1.360858794384806e-05, | |
| "loss": 2.3045, | |
| "step": 5080 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 1.3526011560693641e-05, | |
| "loss": 2.208, | |
| "step": 5090 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 1.3443435177539224e-05, | |
| "loss": 2.3105, | |
| "step": 5100 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 1.3360858794384806e-05, | |
| "loss": 2.2803, | |
| "step": 5110 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 1.3278282411230388e-05, | |
| "loss": 2.2378, | |
| "step": 5120 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 1.3195706028075972e-05, | |
| "loss": 2.2361, | |
| "step": 5130 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 1.3113129644921554e-05, | |
| "loss": 2.2824, | |
| "step": 5140 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 1.3030553261767133e-05, | |
| "loss": 2.2924, | |
| "step": 5150 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 1.2947976878612717e-05, | |
| "loss": 2.3041, | |
| "step": 5160 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 1.28654004954583e-05, | |
| "loss": 2.2128, | |
| "step": 5170 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 1.2782824112303882e-05, | |
| "loss": 2.2239, | |
| "step": 5180 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 1.2700247729149464e-05, | |
| "loss": 2.2884, | |
| "step": 5190 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 1.2617671345995046e-05, | |
| "loss": 2.3694, | |
| "step": 5200 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 1.253509496284063e-05, | |
| "loss": 2.2848, | |
| "step": 5210 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 1.245251857968621e-05, | |
| "loss": 2.2381, | |
| "step": 5220 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 1.2369942196531791e-05, | |
| "loss": 2.3515, | |
| "step": 5230 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 1.2287365813377375e-05, | |
| "loss": 2.3403, | |
| "step": 5240 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 1.2204789430222957e-05, | |
| "loss": 2.2665, | |
| "step": 5250 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 1.212221304706854e-05, | |
| "loss": 2.2445, | |
| "step": 5260 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 1.203963666391412e-05, | |
| "loss": 2.343, | |
| "step": 5270 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 1.1957060280759704e-05, | |
| "loss": 2.2634, | |
| "step": 5280 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 1.1874483897605286e-05, | |
| "loss": 2.3119, | |
| "step": 5290 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 1.1791907514450867e-05, | |
| "loss": 2.3031, | |
| "step": 5300 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 1.1709331131296449e-05, | |
| "loss": 2.2753, | |
| "step": 5310 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 1.1626754748142033e-05, | |
| "loss": 2.1869, | |
| "step": 5320 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 1.1544178364987614e-05, | |
| "loss": 2.2223, | |
| "step": 5330 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 1.1461601981833196e-05, | |
| "loss": 2.2915, | |
| "step": 5340 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 1.1379025598678778e-05, | |
| "loss": 2.2428, | |
| "step": 5350 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 1.129644921552436e-05, | |
| "loss": 2.3804, | |
| "step": 5360 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 1.1213872832369943e-05, | |
| "loss": 2.3171, | |
| "step": 5370 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 1.1131296449215525e-05, | |
| "loss": 2.438, | |
| "step": 5380 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 1.1048720066061107e-05, | |
| "loss": 2.3282, | |
| "step": 5390 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 1.096614368290669e-05, | |
| "loss": 2.514, | |
| "step": 5400 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 1.0883567299752271e-05, | |
| "loss": 2.1497, | |
| "step": 5410 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 1.0800990916597854e-05, | |
| "loss": 2.3027, | |
| "step": 5420 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 1.0718414533443436e-05, | |
| "loss": 2.2309, | |
| "step": 5430 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 1.0635838150289018e-05, | |
| "loss": 2.2857, | |
| "step": 5440 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 1.05532617671346e-05, | |
| "loss": 2.2155, | |
| "step": 5450 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 1.0470685383980183e-05, | |
| "loss": 2.1981, | |
| "step": 5460 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 1.0388109000825765e-05, | |
| "loss": 2.2261, | |
| "step": 5470 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 1.0305532617671347e-05, | |
| "loss": 2.1544, | |
| "step": 5480 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 1.0222956234516928e-05, | |
| "loss": 2.3149, | |
| "step": 5490 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 1.0140379851362512e-05, | |
| "loss": 2.2896, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 1.0057803468208094e-05, | |
| "loss": 2.2211, | |
| "step": 5510 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 9.975227085053674e-06, | |
| "loss": 2.1627, | |
| "step": 5520 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 9.892650701899257e-06, | |
| "loss": 2.2883, | |
| "step": 5530 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "learning_rate": 9.81007431874484e-06, | |
| "loss": 2.2187, | |
| "step": 5540 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "learning_rate": 9.727497935590421e-06, | |
| "loss": 2.1557, | |
| "step": 5550 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "learning_rate": 9.644921552436003e-06, | |
| "loss": 2.3091, | |
| "step": 5560 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 9.562345169281586e-06, | |
| "loss": 2.3263, | |
| "step": 5570 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 9.479768786127168e-06, | |
| "loss": 2.2828, | |
| "step": 5580 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 9.39719240297275e-06, | |
| "loss": 2.2468, | |
| "step": 5590 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 9.314616019818332e-06, | |
| "loss": 2.1766, | |
| "step": 5600 | |
| }, | |
| { | |
| "epoch": 1.67, | |
| "learning_rate": 9.232039636663915e-06, | |
| "loss": 2.2523, | |
| "step": 5610 | |
| }, | |
| { | |
| "epoch": 1.67, | |
| "learning_rate": 9.149463253509497e-06, | |
| "loss": 2.1417, | |
| "step": 5620 | |
| }, | |
| { | |
| "epoch": 1.67, | |
| "learning_rate": 9.066886870355079e-06, | |
| "loss": 2.13, | |
| "step": 5630 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 8.984310487200661e-06, | |
| "loss": 2.1691, | |
| "step": 5640 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 8.901734104046244e-06, | |
| "loss": 2.2776, | |
| "step": 5650 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 8.819157720891826e-06, | |
| "loss": 2.2718, | |
| "step": 5660 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 8.736581337737408e-06, | |
| "loss": 2.2006, | |
| "step": 5670 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 8.654004954582989e-06, | |
| "loss": 2.2386, | |
| "step": 5680 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 8.571428571428573e-06, | |
| "loss": 2.2667, | |
| "step": 5690 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 8.488852188274155e-06, | |
| "loss": 2.1965, | |
| "step": 5700 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 8.406275805119735e-06, | |
| "loss": 2.1731, | |
| "step": 5710 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 8.323699421965318e-06, | |
| "loss": 2.1508, | |
| "step": 5720 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 8.241123038810902e-06, | |
| "loss": 2.1833, | |
| "step": 5730 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 8.158546655656482e-06, | |
| "loss": 2.1511, | |
| "step": 5740 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 8.075970272502064e-06, | |
| "loss": 2.3589, | |
| "step": 5750 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 7.993393889347647e-06, | |
| "loss": 2.281, | |
| "step": 5760 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 7.910817506193229e-06, | |
| "loss": 2.2117, | |
| "step": 5770 | |
| }, | |
| { | |
| "epoch": 1.72, | |
| "learning_rate": 7.828241123038811e-06, | |
| "loss": 2.2086, | |
| "step": 5780 | |
| }, | |
| { | |
| "epoch": 1.72, | |
| "learning_rate": 7.745664739884393e-06, | |
| "loss": 2.214, | |
| "step": 5790 | |
| }, | |
| { | |
| "epoch": 1.72, | |
| "learning_rate": 7.663088356729976e-06, | |
| "loss": 2.223, | |
| "step": 5800 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 7.580511973575558e-06, | |
| "loss": 2.2489, | |
| "step": 5810 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 7.49793559042114e-06, | |
| "loss": 2.3186, | |
| "step": 5820 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 7.415359207266722e-06, | |
| "loss": 2.1881, | |
| "step": 5830 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 7.332782824112304e-06, | |
| "loss": 2.203, | |
| "step": 5840 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 7.250206440957887e-06, | |
| "loss": 2.2772, | |
| "step": 5850 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 7.167630057803469e-06, | |
| "loss": 2.1845, | |
| "step": 5860 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 7.0850536746490505e-06, | |
| "loss": 2.1627, | |
| "step": 5870 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 7.002477291494633e-06, | |
| "loss": 2.3377, | |
| "step": 5880 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 6.919900908340216e-06, | |
| "loss": 2.1565, | |
| "step": 5890 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 6.837324525185796e-06, | |
| "loss": 2.2705, | |
| "step": 5900 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 6.7547481420313794e-06, | |
| "loss": 2.1833, | |
| "step": 5910 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 6.672171758876962e-06, | |
| "loss": 2.3386, | |
| "step": 5920 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 6.589595375722543e-06, | |
| "loss": 2.2546, | |
| "step": 5930 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 6.507018992568126e-06, | |
| "loss": 2.2714, | |
| "step": 5940 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 6.424442609413708e-06, | |
| "loss": 2.243, | |
| "step": 5950 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 6.34186622625929e-06, | |
| "loss": 2.2826, | |
| "step": 5960 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 6.259289843104872e-06, | |
| "loss": 2.1877, | |
| "step": 5970 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 6.176713459950455e-06, | |
| "loss": 2.142, | |
| "step": 5980 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 6.0941370767960365e-06, | |
| "loss": 2.1688, | |
| "step": 5990 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 6.011560693641619e-06, | |
| "loss": 2.1811, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 5.928984310487201e-06, | |
| "loss": 2.2439, | |
| "step": 6010 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 5.846407927332783e-06, | |
| "loss": 2.1966, | |
| "step": 6020 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 5.763831544178365e-06, | |
| "loss": 2.2321, | |
| "step": 6030 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 5.681255161023948e-06, | |
| "loss": 2.2609, | |
| "step": 6040 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 5.598678777869529e-06, | |
| "loss": 2.3517, | |
| "step": 6050 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 5.516102394715112e-06, | |
| "loss": 2.2532, | |
| "step": 6060 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 5.433526011560694e-06, | |
| "loss": 2.3205, | |
| "step": 6070 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 5.350949628406276e-06, | |
| "loss": 2.116, | |
| "step": 6080 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 5.268373245251858e-06, | |
| "loss": 2.2879, | |
| "step": 6090 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 5.18579686209744e-06, | |
| "loss": 2.2843, | |
| "step": 6100 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 5.103220478943023e-06, | |
| "loss": 2.2843, | |
| "step": 6110 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 5.020644095788605e-06, | |
| "loss": 2.1346, | |
| "step": 6120 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 4.938067712634187e-06, | |
| "loss": 2.2408, | |
| "step": 6130 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 4.8554913294797685e-06, | |
| "loss": 2.2361, | |
| "step": 6140 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 4.772914946325352e-06, | |
| "loss": 2.3251, | |
| "step": 6150 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 4.690338563170933e-06, | |
| "loss": 2.3783, | |
| "step": 6160 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 4.607762180016516e-06, | |
| "loss": 2.2199, | |
| "step": 6170 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 4.5251857968620975e-06, | |
| "loss": 2.3079, | |
| "step": 6180 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 4.44260941370768e-06, | |
| "loss": 2.3608, | |
| "step": 6190 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 4.360033030553262e-06, | |
| "loss": 2.3257, | |
| "step": 6200 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 4.277456647398844e-06, | |
| "loss": 2.1902, | |
| "step": 6210 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 4.194880264244426e-06, | |
| "loss": 2.309, | |
| "step": 6220 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 4.112303881090009e-06, | |
| "loss": 2.2283, | |
| "step": 6230 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 4.02972749793559e-06, | |
| "loss": 2.3299, | |
| "step": 6240 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 3.947151114781173e-06, | |
| "loss": 2.2564, | |
| "step": 6250 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 3.8645747316267546e-06, | |
| "loss": 2.1607, | |
| "step": 6260 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 3.781998348472337e-06, | |
| "loss": 2.1553, | |
| "step": 6270 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 3.6994219653179195e-06, | |
| "loss": 2.2015, | |
| "step": 6280 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 3.6168455821635013e-06, | |
| "loss": 2.2779, | |
| "step": 6290 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 3.534269199009083e-06, | |
| "loss": 2.1265, | |
| "step": 6300 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 3.4516928158546658e-06, | |
| "loss": 2.1531, | |
| "step": 6310 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 3.3691164327002476e-06, | |
| "loss": 2.1807, | |
| "step": 6320 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 3.2865400495458303e-06, | |
| "loss": 2.2759, | |
| "step": 6330 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 3.203963666391412e-06, | |
| "loss": 2.1689, | |
| "step": 6340 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 3.1213872832369943e-06, | |
| "loss": 2.1759, | |
| "step": 6350 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 3.0388109000825766e-06, | |
| "loss": 2.2255, | |
| "step": 6360 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 2.956234516928159e-06, | |
| "loss": 2.1884, | |
| "step": 6370 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 2.873658133773741e-06, | |
| "loss": 2.2668, | |
| "step": 6380 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 2.791081750619323e-06, | |
| "loss": 2.3213, | |
| "step": 6390 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 2.708505367464905e-06, | |
| "loss": 2.1226, | |
| "step": 6400 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 2.6259289843104874e-06, | |
| "loss": 2.2776, | |
| "step": 6410 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 2.5433526011560696e-06, | |
| "loss": 2.303, | |
| "step": 6420 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 2.4607762180016514e-06, | |
| "loss": 2.3322, | |
| "step": 6430 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 2.3781998348472337e-06, | |
| "loss": 2.2012, | |
| "step": 6440 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 2.295623451692816e-06, | |
| "loss": 2.3653, | |
| "step": 6450 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 2.213047068538398e-06, | |
| "loss": 2.1154, | |
| "step": 6460 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 2.13047068538398e-06, | |
| "loss": 2.4081, | |
| "step": 6470 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 2.0478943022295622e-06, | |
| "loss": 2.1207, | |
| "step": 6480 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 1.9653179190751445e-06, | |
| "loss": 2.2064, | |
| "step": 6490 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 1.882741535920727e-06, | |
| "loss": 2.2633, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 1.8001651527663087e-06, | |
| "loss": 2.3085, | |
| "step": 6510 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 1.717588769611891e-06, | |
| "loss": 2.2987, | |
| "step": 6520 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 1.6350123864574732e-06, | |
| "loss": 2.2352, | |
| "step": 6530 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 1.5524360033030555e-06, | |
| "loss": 2.1567, | |
| "step": 6540 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 1.4698596201486377e-06, | |
| "loss": 2.2328, | |
| "step": 6550 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 1.3872832369942197e-06, | |
| "loss": 2.0991, | |
| "step": 6560 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 1.3047068538398018e-06, | |
| "loss": 2.2181, | |
| "step": 6570 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 1.222130470685384e-06, | |
| "loss": 2.2654, | |
| "step": 6580 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 1.139554087530966e-06, | |
| "loss": 2.2258, | |
| "step": 6590 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 1.0569777043765483e-06, | |
| "loss": 2.1085, | |
| "step": 6600 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 9.744013212221305e-07, | |
| "loss": 2.2112, | |
| "step": 6610 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 8.918249380677127e-07, | |
| "loss": 2.2201, | |
| "step": 6620 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 8.092485549132948e-07, | |
| "loss": 2.1453, | |
| "step": 6630 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 7.266721717588771e-07, | |
| "loss": 2.2454, | |
| "step": 6640 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 6.440957886044592e-07, | |
| "loss": 2.1955, | |
| "step": 6650 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 5.615194054500413e-07, | |
| "loss": 2.3055, | |
| "step": 6660 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 4.789430222956235e-07, | |
| "loss": 2.1449, | |
| "step": 6670 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 3.963666391412056e-07, | |
| "loss": 2.2541, | |
| "step": 6680 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 3.137902559867878e-07, | |
| "loss": 2.2948, | |
| "step": 6690 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 2.3121387283236997e-07, | |
| "loss": 2.1985, | |
| "step": 6700 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 1.486374896779521e-07, | |
| "loss": 2.1649, | |
| "step": 6710 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 6.606110652353427e-08, | |
| "loss": 2.1896, | |
| "step": 6720 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.5641044692944979, | |
| "eval_f1": 0.07339856199148617, | |
| "eval_loss": 1.8686530590057373, | |
| "eval_precision": 0.11176558876610455, | |
| "eval_recall": 0.08071017936635987, | |
| "eval_runtime": 119.898, | |
| "eval_samples_per_second": 224.499, | |
| "eval_steps_per_second": 14.037, | |
| "step": 6728 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "step": 6728, | |
| "total_flos": 9.319747123440876e+18, | |
| "train_loss": 2.763173609369575, | |
| "train_runtime": 6283.8661, | |
| "train_samples_per_second": 68.536, | |
| "train_steps_per_second": 1.071 | |
| } | |
| ], | |
| "logging_steps": 10, | |
| "max_steps": 6728, | |
| "num_train_epochs": 2, | |
| "save_steps": 500, | |
| "total_flos": 9.319747123440876e+18, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |