| { | |
| "best_metric": 0.9607948442534908, | |
| "best_model_checkpoint": "convnext-tiny-224-finetuned-aiornot/checkpoint-1569", | |
| "epoch": 2.9957040572792364, | |
| "global_step": 1569, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 3.1847133757961785e-06, | |
| "loss": 0.6826, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 6.369426751592357e-06, | |
| "loss": 0.6715, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.554140127388536e-06, | |
| "loss": 0.6717, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.2738853503184714e-05, | |
| "loss": 0.6628, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.592356687898089e-05, | |
| "loss": 0.6325, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.910828025477707e-05, | |
| "loss": 0.636, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 2.229299363057325e-05, | |
| "loss": 0.5867, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 2.5477707006369428e-05, | |
| "loss": 0.5556, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 2.8662420382165606e-05, | |
| "loss": 0.5039, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 3.184713375796178e-05, | |
| "loss": 0.464, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 3.503184713375796e-05, | |
| "loss": 0.3867, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 3.821656050955414e-05, | |
| "loss": 0.3865, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 4.1401273885350325e-05, | |
| "loss": 0.3504, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 4.45859872611465e-05, | |
| "loss": 0.2962, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 4.777070063694268e-05, | |
| "loss": 0.2928, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 4.989376770538244e-05, | |
| "loss": 0.278, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 4.953966005665723e-05, | |
| "loss": 0.2251, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 4.918555240793201e-05, | |
| "loss": 0.2488, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 4.88314447592068e-05, | |
| "loss": 0.2488, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 4.847733711048159e-05, | |
| "loss": 0.2952, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 4.812322946175637e-05, | |
| "loss": 0.2654, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 4.776912181303116e-05, | |
| "loss": 0.1734, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 4.741501416430595e-05, | |
| "loss": 0.172, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 4.706090651558074e-05, | |
| "loss": 0.2531, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 4.670679886685553e-05, | |
| "loss": 0.2789, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 4.635269121813031e-05, | |
| "loss": 0.2007, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 4.59985835694051e-05, | |
| "loss": 0.2171, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.564447592067989e-05, | |
| "loss": 0.2264, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.5290368271954675e-05, | |
| "loss": 0.248, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 4.4936260623229464e-05, | |
| "loss": 0.1643, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 4.458215297450425e-05, | |
| "loss": 0.2226, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 4.4228045325779036e-05, | |
| "loss": 0.2129, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 4.3873937677053826e-05, | |
| "loss": 0.2213, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 4.351983002832861e-05, | |
| "loss": 0.1747, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 4.3165722379603405e-05, | |
| "loss": 0.2705, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.2811614730878195e-05, | |
| "loss": 0.1971, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.245750708215298e-05, | |
| "loss": 0.2161, | |
| "step": 370 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 4.210339943342777e-05, | |
| "loss": 0.133, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 4.174929178470255e-05, | |
| "loss": 0.1454, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 4.139518413597734e-05, | |
| "loss": 0.1347, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 4.104107648725213e-05, | |
| "loss": 0.1519, | |
| "step": 410 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.068696883852691e-05, | |
| "loss": 0.1675, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 4.03328611898017e-05, | |
| "loss": 0.1809, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 3.997875354107649e-05, | |
| "loss": 0.1558, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 3.962464589235127e-05, | |
| "loss": 0.1858, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 3.927053824362606e-05, | |
| "loss": 0.1394, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 3.891643059490085e-05, | |
| "loss": 0.1956, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.856232294617564e-05, | |
| "loss": 0.157, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 3.820821529745043e-05, | |
| "loss": 0.1587, | |
| "step": 490 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 3.7854107648725214e-05, | |
| "loss": 0.1685, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 3.7500000000000003e-05, | |
| "loss": 0.1056, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 3.714589235127479e-05, | |
| "loss": 0.2188, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.9430719656283566, | |
| "eval_loss": 0.14191117882728577, | |
| "eval_runtime": 24.6183, | |
| "eval_samples_per_second": 75.635, | |
| "eval_steps_per_second": 9.465, | |
| "step": 523 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 3.6791784702549576e-05, | |
| "loss": 0.1677, | |
| "step": 530 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 3.6437677053824365e-05, | |
| "loss": 0.1708, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 3.6083569405099155e-05, | |
| "loss": 0.1175, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 3.572946175637394e-05, | |
| "loss": 0.1241, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 3.537535410764873e-05, | |
| "loss": 0.136, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 3.502124645892351e-05, | |
| "loss": 0.1484, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 3.46671388101983e-05, | |
| "loss": 0.1319, | |
| "step": 590 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 3.431303116147309e-05, | |
| "loss": 0.1259, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 3.395892351274788e-05, | |
| "loss": 0.1002, | |
| "step": 610 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 3.360481586402267e-05, | |
| "loss": 0.1399, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 3.325070821529746e-05, | |
| "loss": 0.1688, | |
| "step": 630 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 3.289660056657224e-05, | |
| "loss": 0.1578, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 3.254249291784703e-05, | |
| "loss": 0.1494, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 3.218838526912181e-05, | |
| "loss": 0.1125, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 3.18342776203966e-05, | |
| "loss": 0.1223, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 3.148016997167139e-05, | |
| "loss": 0.1331, | |
| "step": 680 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 3.1126062322946174e-05, | |
| "loss": 0.1109, | |
| "step": 690 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 3.0771954674220964e-05, | |
| "loss": 0.1703, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 3.0417847025495753e-05, | |
| "loss": 0.1578, | |
| "step": 710 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 3.006373937677054e-05, | |
| "loss": 0.1431, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 2.970963172804533e-05, | |
| "loss": 0.1723, | |
| "step": 730 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 2.9355524079320112e-05, | |
| "loss": 0.154, | |
| "step": 740 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 2.90014164305949e-05, | |
| "loss": 0.1036, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 2.864730878186969e-05, | |
| "loss": 0.0967, | |
| "step": 760 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 2.8293201133144477e-05, | |
| "loss": 0.1329, | |
| "step": 770 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 2.7939093484419266e-05, | |
| "loss": 0.1127, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 2.7584985835694056e-05, | |
| "loss": 0.1035, | |
| "step": 790 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 2.723087818696884e-05, | |
| "loss": 0.1385, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 2.6876770538243628e-05, | |
| "loss": 0.1368, | |
| "step": 810 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 2.652266288951841e-05, | |
| "loss": 0.1094, | |
| "step": 820 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 2.61685552407932e-05, | |
| "loss": 0.0838, | |
| "step": 830 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 2.5814447592067993e-05, | |
| "loss": 0.1057, | |
| "step": 840 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 2.5460339943342776e-05, | |
| "loss": 0.1281, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 2.5106232294617566e-05, | |
| "loss": 0.1404, | |
| "step": 860 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 2.4752124645892352e-05, | |
| "loss": 0.1737, | |
| "step": 870 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 2.4398016997167138e-05, | |
| "loss": 0.1201, | |
| "step": 880 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 2.4043909348441927e-05, | |
| "loss": 0.0907, | |
| "step": 890 | |
| }, | |
| { | |
| "epoch": 1.72, | |
| "learning_rate": 2.3689801699716717e-05, | |
| "loss": 0.1083, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 2.3335694050991503e-05, | |
| "loss": 0.1465, | |
| "step": 910 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 2.298158640226629e-05, | |
| "loss": 0.1056, | |
| "step": 920 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 2.262747875354108e-05, | |
| "loss": 0.0841, | |
| "step": 930 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 2.2273371104815865e-05, | |
| "loss": 0.0632, | |
| "step": 940 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 2.191926345609065e-05, | |
| "loss": 0.136, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 2.156515580736544e-05, | |
| "loss": 0.0835, | |
| "step": 960 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 2.121104815864023e-05, | |
| "loss": 0.0859, | |
| "step": 970 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 2.0856940509915016e-05, | |
| "loss": 0.1056, | |
| "step": 980 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 2.0502832861189802e-05, | |
| "loss": 0.137, | |
| "step": 990 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 2.014872521246459e-05, | |
| "loss": 0.1185, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 1.9794617563739378e-05, | |
| "loss": 0.1349, | |
| "step": 1010 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 1.9440509915014164e-05, | |
| "loss": 0.1017, | |
| "step": 1020 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 1.9086402266288954e-05, | |
| "loss": 0.1012, | |
| "step": 1030 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 1.873229461756374e-05, | |
| "loss": 0.0957, | |
| "step": 1040 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.9591836734693877, | |
| "eval_loss": 0.10178032517433167, | |
| "eval_runtime": 24.8731, | |
| "eval_samples_per_second": 74.86, | |
| "eval_steps_per_second": 9.368, | |
| "step": 1047 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 1.837818696883853e-05, | |
| "loss": 0.0973, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 1.8024079320113316e-05, | |
| "loss": 0.134, | |
| "step": 1060 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 1.76699716713881e-05, | |
| "loss": 0.1119, | |
| "step": 1070 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 1.7315864022662888e-05, | |
| "loss": 0.1115, | |
| "step": 1080 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 1.6961756373937677e-05, | |
| "loss": 0.0794, | |
| "step": 1090 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 1.6607648725212467e-05, | |
| "loss": 0.0566, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 1.6253541076487253e-05, | |
| "loss": 0.1112, | |
| "step": 1110 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 1.589943342776204e-05, | |
| "loss": 0.0727, | |
| "step": 1120 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 1.554532577903683e-05, | |
| "loss": 0.0896, | |
| "step": 1130 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 1.5191218130311616e-05, | |
| "loss": 0.1158, | |
| "step": 1140 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 1.4837110481586403e-05, | |
| "loss": 0.0713, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 2.21, | |
| "learning_rate": 1.4483002832861189e-05, | |
| "loss": 0.0826, | |
| "step": 1160 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 1.4128895184135978e-05, | |
| "loss": 0.0819, | |
| "step": 1170 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 1.3774787535410766e-05, | |
| "loss": 0.131, | |
| "step": 1180 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 1.3420679886685552e-05, | |
| "loss": 0.1084, | |
| "step": 1190 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 1.306657223796034e-05, | |
| "loss": 0.1509, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 1.271246458923513e-05, | |
| "loss": 0.1293, | |
| "step": 1210 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 1.2358356940509916e-05, | |
| "loss": 0.1098, | |
| "step": 1220 | |
| }, | |
| { | |
| "epoch": 2.35, | |
| "learning_rate": 1.2004249291784702e-05, | |
| "loss": 0.0677, | |
| "step": 1230 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "learning_rate": 1.1650141643059491e-05, | |
| "loss": 0.1223, | |
| "step": 1240 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 1.1296033994334278e-05, | |
| "loss": 0.1121, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 2.41, | |
| "learning_rate": 1.0941926345609065e-05, | |
| "loss": 0.113, | |
| "step": 1260 | |
| }, | |
| { | |
| "epoch": 2.42, | |
| "learning_rate": 1.0587818696883853e-05, | |
| "loss": 0.1379, | |
| "step": 1270 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 1.0233711048158641e-05, | |
| "loss": 0.0928, | |
| "step": 1280 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 9.879603399433427e-06, | |
| "loss": 0.0977, | |
| "step": 1290 | |
| }, | |
| { | |
| "epoch": 2.48, | |
| "learning_rate": 9.525495750708217e-06, | |
| "loss": 0.1089, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 9.171388101983003e-06, | |
| "loss": 0.0295, | |
| "step": 1310 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "learning_rate": 8.81728045325779e-06, | |
| "loss": 0.1063, | |
| "step": 1320 | |
| }, | |
| { | |
| "epoch": 2.54, | |
| "learning_rate": 8.463172804532578e-06, | |
| "loss": 0.1034, | |
| "step": 1330 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "learning_rate": 8.109065155807366e-06, | |
| "loss": 0.0782, | |
| "step": 1340 | |
| }, | |
| { | |
| "epoch": 2.58, | |
| "learning_rate": 7.754957507082152e-06, | |
| "loss": 0.1049, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "learning_rate": 7.400849858356941e-06, | |
| "loss": 0.1219, | |
| "step": 1360 | |
| }, | |
| { | |
| "epoch": 2.62, | |
| "learning_rate": 7.046742209631728e-06, | |
| "loss": 0.0677, | |
| "step": 1370 | |
| }, | |
| { | |
| "epoch": 2.63, | |
| "learning_rate": 6.692634560906516e-06, | |
| "loss": 0.1263, | |
| "step": 1380 | |
| }, | |
| { | |
| "epoch": 2.65, | |
| "learning_rate": 6.338526912181303e-06, | |
| "loss": 0.0728, | |
| "step": 1390 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 5.984419263456091e-06, | |
| "loss": 0.1082, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 2.69, | |
| "learning_rate": 5.630311614730879e-06, | |
| "loss": 0.115, | |
| "step": 1410 | |
| }, | |
| { | |
| "epoch": 2.71, | |
| "learning_rate": 5.276203966005666e-06, | |
| "loss": 0.0729, | |
| "step": 1420 | |
| }, | |
| { | |
| "epoch": 2.73, | |
| "learning_rate": 4.922096317280453e-06, | |
| "loss": 0.0887, | |
| "step": 1430 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 4.567988668555241e-06, | |
| "loss": 0.0933, | |
| "step": 1440 | |
| }, | |
| { | |
| "epoch": 2.77, | |
| "learning_rate": 4.213881019830029e-06, | |
| "loss": 0.0709, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 2.79, | |
| "learning_rate": 3.859773371104816e-06, | |
| "loss": 0.0599, | |
| "step": 1460 | |
| }, | |
| { | |
| "epoch": 2.81, | |
| "learning_rate": 3.505665722379604e-06, | |
| "loss": 0.0673, | |
| "step": 1470 | |
| }, | |
| { | |
| "epoch": 2.83, | |
| "learning_rate": 3.1515580736543913e-06, | |
| "loss": 0.0469, | |
| "step": 1480 | |
| }, | |
| { | |
| "epoch": 2.84, | |
| "learning_rate": 2.7974504249291787e-06, | |
| "loss": 0.0559, | |
| "step": 1490 | |
| }, | |
| { | |
| "epoch": 2.86, | |
| "learning_rate": 2.443342776203966e-06, | |
| "loss": 0.14, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 2.88, | |
| "learning_rate": 2.089235127478754e-06, | |
| "loss": 0.1153, | |
| "step": 1510 | |
| }, | |
| { | |
| "epoch": 2.9, | |
| "learning_rate": 1.7351274787535413e-06, | |
| "loss": 0.1182, | |
| "step": 1520 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "learning_rate": 1.3810198300283285e-06, | |
| "loss": 0.0636, | |
| "step": 1530 | |
| }, | |
| { | |
| "epoch": 2.94, | |
| "learning_rate": 1.0269121813031164e-06, | |
| "loss": 0.1232, | |
| "step": 1540 | |
| }, | |
| { | |
| "epoch": 2.96, | |
| "learning_rate": 6.728045325779038e-07, | |
| "loss": 0.0924, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 2.98, | |
| "learning_rate": 3.186968838526912e-07, | |
| "loss": 0.0847, | |
| "step": 1560 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.9607948442534908, | |
| "eval_loss": 0.10509096086025238, | |
| "eval_runtime": 23.1726, | |
| "eval_samples_per_second": 80.354, | |
| "eval_steps_per_second": 10.055, | |
| "step": 1569 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "step": 1569, | |
| "total_flos": 1.2614074495819776e+18, | |
| "train_loss": 0.17057580525059088, | |
| "train_runtime": 1281.8017, | |
| "train_samples_per_second": 39.217, | |
| "train_steps_per_second": 1.224 | |
| } | |
| ], | |
| "max_steps": 1569, | |
| "num_train_epochs": 3, | |
| "total_flos": 1.2614074495819776e+18, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |