| { | |
| "best_metric": 1.9584157466888428, | |
| "best_model_checkpoint": "pvt-small-224-ConcreteClassifier-PVT\\checkpoint-1927", | |
| "epoch": 1.0, | |
| "eval_steps": 500, | |
| "global_step": 1927, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.001, | |
| "loss": 2.7631, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.001, | |
| "loss": 2.0945, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.001, | |
| "loss": 1.9938, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.001, | |
| "loss": 1.9829, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.001, | |
| "loss": 2.0147, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.001, | |
| "loss": 2.096, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.001, | |
| "loss": 2.006, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.001, | |
| "loss": 1.9784, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.001, | |
| "loss": 1.9937, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.001, | |
| "loss": 1.9689, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.001, | |
| "loss": 2.0182, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.001, | |
| "loss": 1.9828, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.001, | |
| "loss": 1.9499, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.001, | |
| "loss": 1.9696, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.001, | |
| "loss": 1.9334, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.001, | |
| "loss": 1.9834, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.001, | |
| "loss": 1.9474, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.001, | |
| "loss": 1.9558, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.001, | |
| "loss": 1.9509, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.001, | |
| "loss": 1.9384, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.001, | |
| "loss": 1.9978, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.001, | |
| "loss": 1.9604, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.001, | |
| "loss": 1.971, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.001, | |
| "loss": 1.9705, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.001, | |
| "loss": 1.9544, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.001, | |
| "loss": 1.9343, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.001, | |
| "loss": 1.9263, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.001, | |
| "loss": 1.9695, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.001, | |
| "loss": 1.9348, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.001, | |
| "loss": 1.9506, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.001, | |
| "loss": 1.9976, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.001, | |
| "loss": 1.9527, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.001, | |
| "loss": 1.9526, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.001, | |
| "loss": 1.9623, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.001, | |
| "loss": 1.9455, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.001, | |
| "loss": 1.9286, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.001, | |
| "loss": 1.9476, | |
| "step": 370 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.001, | |
| "loss": 1.9484, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.001, | |
| "loss": 1.9564, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.001, | |
| "loss": 1.9471, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.001, | |
| "loss": 1.9434, | |
| "step": 410 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.001, | |
| "loss": 1.928, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.001, | |
| "loss": 1.9543, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.001, | |
| "loss": 1.8994, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.001, | |
| "loss": 2.029, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.001, | |
| "loss": 1.9449, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.001, | |
| "loss": 1.9539, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.001, | |
| "loss": 1.9541, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.001, | |
| "loss": 1.956, | |
| "step": 490 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.001, | |
| "loss": 1.9707, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.001, | |
| "loss": 1.9402, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.001, | |
| "loss": 1.9656, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.001, | |
| "loss": 1.9372, | |
| "step": 530 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.001, | |
| "loss": 1.955, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.001, | |
| "loss": 1.9288, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.001, | |
| "loss": 1.9574, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.001, | |
| "loss": 1.956, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.001, | |
| "loss": 1.9587, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.001, | |
| "loss": 1.9473, | |
| "step": 590 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.001, | |
| "loss": 1.9452, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.001, | |
| "loss": 1.9648, | |
| "step": 610 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.001, | |
| "loss": 1.9314, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.001, | |
| "loss": 1.961, | |
| "step": 630 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.001, | |
| "loss": 1.9386, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.001, | |
| "loss": 1.9546, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.001, | |
| "loss": 1.967, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.001, | |
| "loss": 1.9451, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.001, | |
| "loss": 1.9475, | |
| "step": 680 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.001, | |
| "loss": 1.9518, | |
| "step": 690 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.001, | |
| "loss": 1.9518, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.001, | |
| "loss": 1.9481, | |
| "step": 710 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.001, | |
| "loss": 1.9456, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.001, | |
| "loss": 1.9362, | |
| "step": 730 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.001, | |
| "loss": 1.9188, | |
| "step": 740 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.001, | |
| "loss": 2.0077, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.001, | |
| "loss": 1.9442, | |
| "step": 760 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.001, | |
| "loss": 1.9348, | |
| "step": 770 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.001, | |
| "loss": 1.9512, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.001, | |
| "loss": 1.9409, | |
| "step": 790 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.001, | |
| "loss": 1.9404, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.001, | |
| "loss": 1.9656, | |
| "step": 810 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.001, | |
| "loss": 1.948, | |
| "step": 820 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.001, | |
| "loss": 1.9471, | |
| "step": 830 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.001, | |
| "loss": 1.9453, | |
| "step": 840 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.001, | |
| "loss": 1.9386, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.001, | |
| "loss": 1.9568, | |
| "step": 860 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.001, | |
| "loss": 1.9498, | |
| "step": 870 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.001, | |
| "loss": 1.9621, | |
| "step": 880 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.001, | |
| "loss": 1.9444, | |
| "step": 890 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.001, | |
| "loss": 1.9515, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.001, | |
| "loss": 1.9419, | |
| "step": 910 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.001, | |
| "loss": 1.948, | |
| "step": 920 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.001, | |
| "loss": 1.9427, | |
| "step": 930 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.001, | |
| "loss": 1.9586, | |
| "step": 940 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.001, | |
| "loss": 1.9588, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.001, | |
| "loss": 1.9511, | |
| "step": 960 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.001, | |
| "loss": 1.9505, | |
| "step": 970 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.001, | |
| "loss": 1.9424, | |
| "step": 980 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.001, | |
| "loss": 1.9378, | |
| "step": 990 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.001, | |
| "loss": 1.9501, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.001, | |
| "loss": 1.9352, | |
| "step": 1010 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.001, | |
| "loss": 1.9338, | |
| "step": 1020 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.001, | |
| "loss": 1.9546, | |
| "step": 1030 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.001, | |
| "loss": 1.9436, | |
| "step": 1040 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.001, | |
| "loss": 1.9347, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.001, | |
| "loss": 1.9458, | |
| "step": 1060 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.001, | |
| "loss": 1.9389, | |
| "step": 1070 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.001, | |
| "loss": 1.9445, | |
| "step": 1080 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.001, | |
| "loss": 1.9594, | |
| "step": 1090 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.001, | |
| "loss": 1.9666, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.001, | |
| "loss": 1.9567, | |
| "step": 1110 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.001, | |
| "loss": 1.9527, | |
| "step": 1120 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.001, | |
| "loss": 1.9434, | |
| "step": 1130 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.001, | |
| "loss": 1.9252, | |
| "step": 1140 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.001, | |
| "loss": 1.9541, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.001, | |
| "loss": 1.9619, | |
| "step": 1160 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.001, | |
| "loss": 1.9531, | |
| "step": 1170 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.001, | |
| "loss": 1.9419, | |
| "step": 1180 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.001, | |
| "loss": 1.9521, | |
| "step": 1190 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.001, | |
| "loss": 1.9358, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.001, | |
| "loss": 1.9166, | |
| "step": 1210 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.001, | |
| "loss": 2.0, | |
| "step": 1220 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.001, | |
| "loss": 1.9459, | |
| "step": 1230 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.001, | |
| "loss": 1.9397, | |
| "step": 1240 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.001, | |
| "loss": 1.9722, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.001, | |
| "loss": 1.9704, | |
| "step": 1260 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.001, | |
| "loss": 1.957, | |
| "step": 1270 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.001, | |
| "loss": 1.9551, | |
| "step": 1280 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.001, | |
| "loss": 1.9534, | |
| "step": 1290 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.001, | |
| "loss": 1.9718, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.001, | |
| "loss": 1.9597, | |
| "step": 1310 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 0.001, | |
| "loss": 1.9527, | |
| "step": 1320 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 0.001, | |
| "loss": 1.9362, | |
| "step": 1330 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 0.001, | |
| "loss": 1.9636, | |
| "step": 1340 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 0.001, | |
| "loss": 1.95, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 0.001, | |
| "loss": 1.9551, | |
| "step": 1360 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 0.001, | |
| "loss": 2.0172, | |
| "step": 1370 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 0.001, | |
| "loss": 1.943, | |
| "step": 1380 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 0.001, | |
| "loss": 1.9486, | |
| "step": 1390 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 0.001, | |
| "loss": 1.9877, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 0.001, | |
| "loss": 1.976, | |
| "step": 1410 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 0.001, | |
| "loss": 1.9371, | |
| "step": 1420 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 0.001, | |
| "loss": 1.9797, | |
| "step": 1430 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 0.001, | |
| "loss": 1.9719, | |
| "step": 1440 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 0.001, | |
| "loss": 1.925, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 0.001, | |
| "loss": 1.9756, | |
| "step": 1460 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 0.001, | |
| "loss": 1.9553, | |
| "step": 1470 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 0.001, | |
| "loss": 1.949, | |
| "step": 1480 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 0.001, | |
| "loss": 1.9494, | |
| "step": 1490 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 0.001, | |
| "loss": 1.9454, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 0.001, | |
| "loss": 1.9347, | |
| "step": 1510 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 0.001, | |
| "loss": 1.9342, | |
| "step": 1520 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 0.001, | |
| "loss": 1.9762, | |
| "step": 1530 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 0.001, | |
| "loss": 1.9467, | |
| "step": 1540 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 0.001, | |
| "loss": 1.9493, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 0.001, | |
| "loss": 1.9309, | |
| "step": 1560 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 0.001, | |
| "loss": 1.9875, | |
| "step": 1570 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 0.001, | |
| "loss": 1.9379, | |
| "step": 1580 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 0.001, | |
| "loss": 1.9231, | |
| "step": 1590 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 0.001, | |
| "loss": 1.9163, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 0.001, | |
| "loss": 1.9894, | |
| "step": 1610 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 0.001, | |
| "loss": 1.9543, | |
| "step": 1620 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 0.001, | |
| "loss": 1.9237, | |
| "step": 1630 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 0.001, | |
| "loss": 2.1552, | |
| "step": 1640 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 0.001, | |
| "loss": 2.0399, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 0.001, | |
| "loss": 1.971, | |
| "step": 1660 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 0.001, | |
| "loss": 2.0006, | |
| "step": 1670 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 0.001, | |
| "loss": 1.9451, | |
| "step": 1680 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 0.001, | |
| "loss": 1.993, | |
| "step": 1690 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 0.001, | |
| "loss": 1.9586, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 0.001, | |
| "loss": 1.963, | |
| "step": 1710 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 0.001, | |
| "loss": 1.9641, | |
| "step": 1720 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 0.001, | |
| "loss": 1.9555, | |
| "step": 1730 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 0.001, | |
| "loss": 1.9381, | |
| "step": 1740 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 0.001, | |
| "loss": 1.934, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 0.001, | |
| "loss": 2.0062, | |
| "step": 1760 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 0.001, | |
| "loss": 1.9315, | |
| "step": 1770 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 0.001, | |
| "loss": 1.9626, | |
| "step": 1780 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 0.001, | |
| "loss": 1.9135, | |
| "step": 1790 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 0.001, | |
| "loss": 1.9756, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 0.001, | |
| "loss": 1.9628, | |
| "step": 1810 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 0.001, | |
| "loss": 1.952, | |
| "step": 1820 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 0.001, | |
| "loss": 1.9326, | |
| "step": 1830 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 0.001, | |
| "loss": 1.9552, | |
| "step": 1840 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 0.001, | |
| "loss": 1.9527, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 0.001, | |
| "loss": 1.938, | |
| "step": 1860 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 0.001, | |
| "loss": 1.9452, | |
| "step": 1870 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 0.001, | |
| "loss": 1.9515, | |
| "step": 1880 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 0.001, | |
| "loss": 1.9465, | |
| "step": 1890 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 0.001, | |
| "loss": 1.9524, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 0.001, | |
| "loss": 1.9271, | |
| "step": 1910 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 0.001, | |
| "loss": 1.981, | |
| "step": 1920 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": { | |
| "accuracy": 0.1556420233463035 | |
| }, | |
| "eval_f1": { | |
| "f1": 0.03848003848003848 | |
| }, | |
| "eval_loss": 1.9584157466888428, | |
| "eval_precision": { | |
| "precision": 0.022234574763757644 | |
| }, | |
| "eval_recall": { | |
| "recall": 0.14285714285714285 | |
| }, | |
| "eval_runtime": 201.5846, | |
| "eval_samples_per_second": 19.123, | |
| "eval_steps_per_second": 9.564, | |
| "step": 1927 | |
| } | |
| ], | |
| "logging_steps": 10, | |
| "max_steps": 57810, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 30, | |
| "save_steps": 500, | |
| "total_flos": 3.3382988667592704e+17, | |
| "train_batch_size": 2, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |