| { | |
| "best_metric": 0.26776641607284546, | |
| "best_model_checkpoint": "food-recognition\\checkpoint-1500", | |
| "epoch": 4.0, | |
| "global_step": 1876, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00019893390191897656, | |
| "loss": 2.1647, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0001978678038379531, | |
| "loss": 1.7478, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.00019680170575692965, | |
| "loss": 1.4997, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0001957356076759062, | |
| "loss": 1.1421, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.00019466950959488275, | |
| "loss": 1.0115, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.00019360341151385927, | |
| "loss": 0.8962, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.00019253731343283584, | |
| "loss": 0.774, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0001914712153518124, | |
| "loss": 0.8508, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0001904051172707889, | |
| "loss": 0.8852, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00018933901918976546, | |
| "loss": 0.6346, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "eval_accuracy": 0.8136, | |
| "eval_loss": 0.6600381135940552, | |
| "eval_runtime": 910.9195, | |
| "eval_samples_per_second": 2.744, | |
| "eval_steps_per_second": 0.344, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00018827292110874203, | |
| "loss": 0.7183, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00018720682302771858, | |
| "loss": 0.7806, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0001861407249466951, | |
| "loss": 0.7702, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.00018507462686567165, | |
| "loss": 0.8338, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0001840085287846482, | |
| "loss": 0.732, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00018294243070362474, | |
| "loss": 0.6746, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0001818763326226013, | |
| "loss": 0.5369, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00018081023454157784, | |
| "loss": 0.6082, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00017974413646055436, | |
| "loss": 0.5756, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00017867803837953093, | |
| "loss": 0.4988, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "eval_accuracy": 0.8212, | |
| "eval_loss": 0.5744591951370239, | |
| "eval_runtime": 901.9851, | |
| "eval_samples_per_second": 2.772, | |
| "eval_steps_per_second": 0.347, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00017761194029850748, | |
| "loss": 0.672, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00017654584221748402, | |
| "loss": 0.6575, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00017547974413646055, | |
| "loss": 0.5028, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00017441364605543712, | |
| "loss": 0.5451, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00017334754797441367, | |
| "loss": 0.6499, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.0001722814498933902, | |
| "loss": 0.7127, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00017121535181236673, | |
| "loss": 0.538, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00017014925373134328, | |
| "loss": 0.6385, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00016908315565031986, | |
| "loss": 0.4637, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00016801705756929638, | |
| "loss": 0.4619, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "eval_accuracy": 0.8852, | |
| "eval_loss": 0.3846997916698456, | |
| "eval_runtime": 886.5096, | |
| "eval_samples_per_second": 2.82, | |
| "eval_steps_per_second": 0.353, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00016695095948827292, | |
| "loss": 0.6006, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00016588486140724947, | |
| "loss": 0.4678, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 0.00016481876332622602, | |
| "loss": 0.5459, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 0.00016375266524520257, | |
| "loss": 0.4216, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 0.00016268656716417911, | |
| "loss": 0.6596, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 0.00016162046908315566, | |
| "loss": 0.6491, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 0.0001605543710021322, | |
| "loss": 0.5269, | |
| "step": 370 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 0.00015948827292110876, | |
| "loss": 0.441, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 0.0001584221748400853, | |
| "loss": 0.4428, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 0.00015735607675906182, | |
| "loss": 0.4205, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "eval_accuracy": 0.864, | |
| "eval_loss": 0.43827134370803833, | |
| "eval_runtime": 905.4707, | |
| "eval_samples_per_second": 2.761, | |
| "eval_steps_per_second": 0.346, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 0.0001562899786780384, | |
| "loss": 0.6422, | |
| "step": 410 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 0.00015522388059701495, | |
| "loss": 0.4558, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 0.0001541577825159915, | |
| "loss": 0.4575, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 0.000153091684434968, | |
| "loss": 0.4454, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 0.00015202558635394456, | |
| "loss": 0.4917, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 0.00015095948827292113, | |
| "loss": 0.5301, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 0.00014989339019189766, | |
| "loss": 0.3294, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 0.0001488272921108742, | |
| "loss": 0.2967, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 0.00014776119402985075, | |
| "loss": 0.3406, | |
| "step": 490 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 0.0001466950959488273, | |
| "loss": 0.287, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "eval_accuracy": 0.8992, | |
| "eval_loss": 0.3214336931705475, | |
| "eval_runtime": 905.5537, | |
| "eval_samples_per_second": 2.761, | |
| "eval_steps_per_second": 0.346, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 0.00014562899786780384, | |
| "loss": 0.3278, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 0.0001445628997867804, | |
| "loss": 0.2406, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 0.00014349680170575694, | |
| "loss": 0.2839, | |
| "step": 530 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 0.0001424307036247335, | |
| "loss": 0.2843, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 0.00014136460554371003, | |
| "loss": 0.2085, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 0.00014029850746268658, | |
| "loss": 0.2465, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 0.0001392324093816631, | |
| "loss": 0.2444, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 0.00013816631130063965, | |
| "loss": 0.3261, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 0.00013710021321961622, | |
| "loss": 0.1768, | |
| "step": 590 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 0.00013603411513859277, | |
| "loss": 0.2963, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "eval_accuracy": 0.8796, | |
| "eval_loss": 0.35504186153411865, | |
| "eval_runtime": 901.3195, | |
| "eval_samples_per_second": 2.774, | |
| "eval_steps_per_second": 0.347, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 0.0001349680170575693, | |
| "loss": 0.3221, | |
| "step": 610 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 0.00013390191897654584, | |
| "loss": 0.3559, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 0.0001328358208955224, | |
| "loss": 0.2589, | |
| "step": 630 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 0.00013176972281449893, | |
| "loss": 0.317, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 0.00013070362473347548, | |
| "loss": 0.2702, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 0.00012963752665245203, | |
| "loss": 0.2416, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 0.00012857142857142858, | |
| "loss": 0.3619, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 0.00012750533049040512, | |
| "loss": 0.1777, | |
| "step": 680 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 0.00012643923240938167, | |
| "loss": 0.3028, | |
| "step": 690 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 0.00012537313432835822, | |
| "loss": 0.2908, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "eval_accuracy": 0.9064, | |
| "eval_loss": 0.3182031810283661, | |
| "eval_runtime": 904.1865, | |
| "eval_samples_per_second": 2.765, | |
| "eval_steps_per_second": 0.346, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 0.00012430703624733474, | |
| "loss": 0.236, | |
| "step": 710 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 0.0001232409381663113, | |
| "loss": 0.1903, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 0.00012217484008528786, | |
| "loss": 0.2509, | |
| "step": 730 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 0.00012110874200426441, | |
| "loss": 0.3468, | |
| "step": 740 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 0.00012004264392324094, | |
| "loss": 0.4192, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 0.00011897654584221749, | |
| "loss": 0.2604, | |
| "step": 760 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 0.00011791044776119405, | |
| "loss": 0.2523, | |
| "step": 770 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 0.00011684434968017057, | |
| "loss": 0.169, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 0.00011577825159914713, | |
| "loss": 0.3698, | |
| "step": 790 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 0.00011471215351812368, | |
| "loss": 0.3791, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "eval_accuracy": 0.9012, | |
| "eval_loss": 0.30747249722480774, | |
| "eval_runtime": 905.3757, | |
| "eval_samples_per_second": 2.761, | |
| "eval_steps_per_second": 0.346, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 0.00011364605543710023, | |
| "loss": 0.4315, | |
| "step": 810 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 0.00011257995735607676, | |
| "loss": 0.2439, | |
| "step": 820 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 0.0001115138592750533, | |
| "loss": 0.1582, | |
| "step": 830 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 0.00011044776119402987, | |
| "loss": 0.1514, | |
| "step": 840 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 0.00010938166311300639, | |
| "loss": 0.302, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 0.00010831556503198295, | |
| "loss": 0.1793, | |
| "step": 860 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 0.0001072494669509595, | |
| "loss": 0.2064, | |
| "step": 870 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 0.00010618336886993603, | |
| "loss": 0.1789, | |
| "step": 880 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 0.00010511727078891258, | |
| "loss": 0.2323, | |
| "step": 890 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 0.00010405117270788914, | |
| "loss": 0.1399, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "eval_accuracy": 0.9108, | |
| "eval_loss": 0.31127122044563293, | |
| "eval_runtime": 903.7995, | |
| "eval_samples_per_second": 2.766, | |
| "eval_steps_per_second": 0.346, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 0.00010298507462686569, | |
| "loss": 0.2957, | |
| "step": 910 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 0.00010191897654584222, | |
| "loss": 0.2673, | |
| "step": 920 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 0.00010085287846481877, | |
| "loss": 0.2396, | |
| "step": 930 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 9.978678038379531e-05, | |
| "loss": 0.1854, | |
| "step": 940 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 9.872068230277186e-05, | |
| "loss": 0.0911, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 9.76545842217484e-05, | |
| "loss": 0.0954, | |
| "step": 960 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 9.658848614072496e-05, | |
| "loss": 0.0634, | |
| "step": 970 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 9.552238805970149e-05, | |
| "loss": 0.1001, | |
| "step": 980 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 9.445628997867805e-05, | |
| "loss": 0.076, | |
| "step": 990 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 9.339019189765458e-05, | |
| "loss": 0.0501, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "eval_accuracy": 0.9004, | |
| "eval_loss": 0.33236250281333923, | |
| "eval_runtime": 901.7876, | |
| "eval_samples_per_second": 2.772, | |
| "eval_steps_per_second": 0.347, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 9.232409381663113e-05, | |
| "loss": 0.1246, | |
| "step": 1010 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 9.125799573560768e-05, | |
| "loss": 0.1432, | |
| "step": 1020 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 9.019189765458423e-05, | |
| "loss": 0.1998, | |
| "step": 1030 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 8.912579957356077e-05, | |
| "loss": 0.1226, | |
| "step": 1040 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 8.805970149253732e-05, | |
| "loss": 0.0544, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 8.699360341151387e-05, | |
| "loss": 0.0842, | |
| "step": 1060 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 8.59275053304904e-05, | |
| "loss": 0.063, | |
| "step": 1070 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 8.486140724946695e-05, | |
| "loss": 0.0682, | |
| "step": 1080 | |
| }, | |
| { | |
| "epoch": 2.32, | |
| "learning_rate": 8.37953091684435e-05, | |
| "loss": 0.0897, | |
| "step": 1090 | |
| }, | |
| { | |
| "epoch": 2.35, | |
| "learning_rate": 8.272921108742005e-05, | |
| "loss": 0.0873, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 2.35, | |
| "eval_accuracy": 0.9196, | |
| "eval_loss": 0.28453874588012695, | |
| "eval_runtime": 903.3406, | |
| "eval_samples_per_second": 2.768, | |
| "eval_steps_per_second": 0.346, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "learning_rate": 8.166311300639659e-05, | |
| "loss": 0.076, | |
| "step": 1110 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 8.059701492537314e-05, | |
| "loss": 0.0576, | |
| "step": 1120 | |
| }, | |
| { | |
| "epoch": 2.41, | |
| "learning_rate": 7.953091684434967e-05, | |
| "loss": 0.0622, | |
| "step": 1130 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 7.846481876332623e-05, | |
| "loss": 0.0799, | |
| "step": 1140 | |
| }, | |
| { | |
| "epoch": 2.45, | |
| "learning_rate": 7.739872068230277e-05, | |
| "loss": 0.1253, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 2.47, | |
| "learning_rate": 7.633262260127933e-05, | |
| "loss": 0.0698, | |
| "step": 1160 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "learning_rate": 7.526652452025586e-05, | |
| "loss": 0.1532, | |
| "step": 1170 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "learning_rate": 7.420042643923242e-05, | |
| "loss": 0.1728, | |
| "step": 1180 | |
| }, | |
| { | |
| "epoch": 2.54, | |
| "learning_rate": 7.313432835820896e-05, | |
| "loss": 0.0871, | |
| "step": 1190 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "learning_rate": 7.20682302771855e-05, | |
| "loss": 0.0809, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "eval_accuracy": 0.9104, | |
| "eval_loss": 0.3186958134174347, | |
| "eval_runtime": 903.5081, | |
| "eval_samples_per_second": 2.767, | |
| "eval_steps_per_second": 0.346, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 2.58, | |
| "learning_rate": 7.100213219616205e-05, | |
| "loss": 0.1217, | |
| "step": 1210 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "learning_rate": 6.993603411513859e-05, | |
| "loss": 0.0469, | |
| "step": 1220 | |
| }, | |
| { | |
| "epoch": 2.62, | |
| "learning_rate": 6.886993603411515e-05, | |
| "loss": 0.0659, | |
| "step": 1230 | |
| }, | |
| { | |
| "epoch": 2.64, | |
| "learning_rate": 6.780383795309168e-05, | |
| "loss": 0.1019, | |
| "step": 1240 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 6.673773987206824e-05, | |
| "loss": 0.046, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 2.69, | |
| "learning_rate": 6.567164179104478e-05, | |
| "loss": 0.0203, | |
| "step": 1260 | |
| }, | |
| { | |
| "epoch": 2.71, | |
| "learning_rate": 6.460554371002132e-05, | |
| "loss": 0.0483, | |
| "step": 1270 | |
| }, | |
| { | |
| "epoch": 2.73, | |
| "learning_rate": 6.353944562899787e-05, | |
| "loss": 0.0785, | |
| "step": 1280 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 6.247334754797442e-05, | |
| "loss": 0.0358, | |
| "step": 1290 | |
| }, | |
| { | |
| "epoch": 2.77, | |
| "learning_rate": 6.140724946695097e-05, | |
| "loss": 0.0696, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 2.77, | |
| "eval_accuracy": 0.9184, | |
| "eval_loss": 0.3137047588825226, | |
| "eval_runtime": 902.4586, | |
| "eval_samples_per_second": 2.77, | |
| "eval_steps_per_second": 0.347, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 2.79, | |
| "learning_rate": 6.0341151385927506e-05, | |
| "loss": 0.0716, | |
| "step": 1310 | |
| }, | |
| { | |
| "epoch": 2.81, | |
| "learning_rate": 5.927505330490405e-05, | |
| "loss": 0.0409, | |
| "step": 1320 | |
| }, | |
| { | |
| "epoch": 2.84, | |
| "learning_rate": 5.82089552238806e-05, | |
| "loss": 0.0595, | |
| "step": 1330 | |
| }, | |
| { | |
| "epoch": 2.86, | |
| "learning_rate": 5.714285714285714e-05, | |
| "loss": 0.1415, | |
| "step": 1340 | |
| }, | |
| { | |
| "epoch": 2.88, | |
| "learning_rate": 5.6076759061833696e-05, | |
| "loss": 0.1064, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 2.9, | |
| "learning_rate": 5.5010660980810236e-05, | |
| "loss": 0.0783, | |
| "step": 1360 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "learning_rate": 5.394456289978679e-05, | |
| "loss": 0.0889, | |
| "step": 1370 | |
| }, | |
| { | |
| "epoch": 2.94, | |
| "learning_rate": 5.287846481876333e-05, | |
| "loss": 0.1406, | |
| "step": 1380 | |
| }, | |
| { | |
| "epoch": 2.96, | |
| "learning_rate": 5.181236673773987e-05, | |
| "loss": 0.0857, | |
| "step": 1390 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "learning_rate": 5.074626865671642e-05, | |
| "loss": 0.0684, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "eval_accuracy": 0.9208, | |
| "eval_loss": 0.28956088423728943, | |
| "eval_runtime": 904.414, | |
| "eval_samples_per_second": 2.764, | |
| "eval_steps_per_second": 0.346, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 4.9680170575692967e-05, | |
| "loss": 0.038, | |
| "step": 1410 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 4.861407249466951e-05, | |
| "loss": 0.0321, | |
| "step": 1420 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 4.7547974413646055e-05, | |
| "loss": 0.0125, | |
| "step": 1430 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "learning_rate": 4.64818763326226e-05, | |
| "loss": 0.0197, | |
| "step": 1440 | |
| }, | |
| { | |
| "epoch": 3.09, | |
| "learning_rate": 4.541577825159915e-05, | |
| "loss": 0.0289, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "learning_rate": 4.43496801705757e-05, | |
| "loss": 0.0273, | |
| "step": 1460 | |
| }, | |
| { | |
| "epoch": 3.13, | |
| "learning_rate": 4.328358208955224e-05, | |
| "loss": 0.0118, | |
| "step": 1470 | |
| }, | |
| { | |
| "epoch": 3.16, | |
| "learning_rate": 4.2217484008528785e-05, | |
| "loss": 0.0105, | |
| "step": 1480 | |
| }, | |
| { | |
| "epoch": 3.18, | |
| "learning_rate": 4.115138592750533e-05, | |
| "loss": 0.008, | |
| "step": 1490 | |
| }, | |
| { | |
| "epoch": 3.2, | |
| "learning_rate": 4.008528784648188e-05, | |
| "loss": 0.0084, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 3.2, | |
| "eval_accuracy": 0.9308, | |
| "eval_loss": 0.26776641607284546, | |
| "eval_runtime": 908.0277, | |
| "eval_samples_per_second": 2.753, | |
| "eval_steps_per_second": 0.345, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 3.22, | |
| "learning_rate": 3.901918976545843e-05, | |
| "loss": 0.0076, | |
| "step": 1510 | |
| }, | |
| { | |
| "epoch": 3.24, | |
| "learning_rate": 3.7953091684434974e-05, | |
| "loss": 0.0087, | |
| "step": 1520 | |
| }, | |
| { | |
| "epoch": 3.26, | |
| "learning_rate": 3.6886993603411515e-05, | |
| "loss": 0.019, | |
| "step": 1530 | |
| }, | |
| { | |
| "epoch": 3.28, | |
| "learning_rate": 3.582089552238806e-05, | |
| "loss": 0.0182, | |
| "step": 1540 | |
| }, | |
| { | |
| "epoch": 3.3, | |
| "learning_rate": 3.47547974413646e-05, | |
| "loss": 0.0401, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 3.33, | |
| "learning_rate": 3.368869936034115e-05, | |
| "loss": 0.0461, | |
| "step": 1560 | |
| }, | |
| { | |
| "epoch": 3.35, | |
| "learning_rate": 3.26226012793177e-05, | |
| "loss": 0.0272, | |
| "step": 1570 | |
| }, | |
| { | |
| "epoch": 3.37, | |
| "learning_rate": 3.1556503198294245e-05, | |
| "loss": 0.0333, | |
| "step": 1580 | |
| }, | |
| { | |
| "epoch": 3.39, | |
| "learning_rate": 3.0490405117270792e-05, | |
| "loss": 0.0699, | |
| "step": 1590 | |
| }, | |
| { | |
| "epoch": 3.41, | |
| "learning_rate": 2.9424307036247333e-05, | |
| "loss": 0.0072, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 3.41, | |
| "eval_accuracy": 0.9292, | |
| "eval_loss": 0.2749463617801666, | |
| "eval_runtime": 908.1664, | |
| "eval_samples_per_second": 2.753, | |
| "eval_steps_per_second": 0.345, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 3.43, | |
| "learning_rate": 2.835820895522388e-05, | |
| "loss": 0.0238, | |
| "step": 1610 | |
| }, | |
| { | |
| "epoch": 3.45, | |
| "learning_rate": 2.7292110874200428e-05, | |
| "loss": 0.0098, | |
| "step": 1620 | |
| }, | |
| { | |
| "epoch": 3.48, | |
| "learning_rate": 2.6226012793176975e-05, | |
| "loss": 0.0162, | |
| "step": 1630 | |
| }, | |
| { | |
| "epoch": 3.5, | |
| "learning_rate": 2.515991471215352e-05, | |
| "loss": 0.0075, | |
| "step": 1640 | |
| }, | |
| { | |
| "epoch": 3.52, | |
| "learning_rate": 2.4093816631130063e-05, | |
| "loss": 0.0326, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 3.54, | |
| "learning_rate": 2.302771855010661e-05, | |
| "loss": 0.0141, | |
| "step": 1660 | |
| }, | |
| { | |
| "epoch": 3.56, | |
| "learning_rate": 2.1961620469083158e-05, | |
| "loss": 0.0083, | |
| "step": 1670 | |
| }, | |
| { | |
| "epoch": 3.58, | |
| "learning_rate": 2.0895522388059702e-05, | |
| "loss": 0.011, | |
| "step": 1680 | |
| }, | |
| { | |
| "epoch": 3.6, | |
| "learning_rate": 1.982942430703625e-05, | |
| "loss": 0.0067, | |
| "step": 1690 | |
| }, | |
| { | |
| "epoch": 3.62, | |
| "learning_rate": 1.8763326226012797e-05, | |
| "loss": 0.0585, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 3.62, | |
| "eval_accuracy": 0.9316, | |
| "eval_loss": 0.2734415531158447, | |
| "eval_runtime": 910.1726, | |
| "eval_samples_per_second": 2.747, | |
| "eval_steps_per_second": 0.344, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 3.65, | |
| "learning_rate": 1.769722814498934e-05, | |
| "loss": 0.007, | |
| "step": 1710 | |
| }, | |
| { | |
| "epoch": 3.67, | |
| "learning_rate": 1.6631130063965885e-05, | |
| "loss": 0.0061, | |
| "step": 1720 | |
| }, | |
| { | |
| "epoch": 3.69, | |
| "learning_rate": 1.5565031982942432e-05, | |
| "loss": 0.0061, | |
| "step": 1730 | |
| }, | |
| { | |
| "epoch": 3.71, | |
| "learning_rate": 1.4498933901918976e-05, | |
| "loss": 0.0159, | |
| "step": 1740 | |
| }, | |
| { | |
| "epoch": 3.73, | |
| "learning_rate": 1.3432835820895523e-05, | |
| "loss": 0.0154, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 3.75, | |
| "learning_rate": 1.2366737739872069e-05, | |
| "loss": 0.0071, | |
| "step": 1760 | |
| }, | |
| { | |
| "epoch": 3.77, | |
| "learning_rate": 1.1300639658848615e-05, | |
| "loss": 0.0067, | |
| "step": 1770 | |
| }, | |
| { | |
| "epoch": 3.8, | |
| "learning_rate": 1.023454157782516e-05, | |
| "loss": 0.0387, | |
| "step": 1780 | |
| }, | |
| { | |
| "epoch": 3.82, | |
| "learning_rate": 9.168443496801706e-06, | |
| "loss": 0.0492, | |
| "step": 1790 | |
| }, | |
| { | |
| "epoch": 3.84, | |
| "learning_rate": 8.102345415778252e-06, | |
| "loss": 0.0152, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 3.84, | |
| "eval_accuracy": 0.9332, | |
| "eval_loss": 0.26903361082077026, | |
| "eval_runtime": 906.1705, | |
| "eval_samples_per_second": 2.759, | |
| "eval_steps_per_second": 0.345, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 3.86, | |
| "learning_rate": 7.0362473347547975e-06, | |
| "loss": 0.0054, | |
| "step": 1810 | |
| }, | |
| { | |
| "epoch": 3.88, | |
| "learning_rate": 5.970149253731343e-06, | |
| "loss": 0.006, | |
| "step": 1820 | |
| }, | |
| { | |
| "epoch": 3.9, | |
| "learning_rate": 4.90405117270789e-06, | |
| "loss": 0.0257, | |
| "step": 1830 | |
| }, | |
| { | |
| "epoch": 3.92, | |
| "learning_rate": 3.837953091684435e-06, | |
| "loss": 0.0076, | |
| "step": 1840 | |
| }, | |
| { | |
| "epoch": 3.94, | |
| "learning_rate": 2.771855010660981e-06, | |
| "loss": 0.0055, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 3.97, | |
| "learning_rate": 1.7057569296375267e-06, | |
| "loss": 0.0064, | |
| "step": 1860 | |
| }, | |
| { | |
| "epoch": 3.99, | |
| "learning_rate": 6.396588486140725e-07, | |
| "loss": 0.0054, | |
| "step": 1870 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "step": 1876, | |
| "total_flos": 0.0, | |
| "train_loss": 0.2672756195687917, | |
| "train_runtime": 47803.3846, | |
| "train_samples_per_second": 0.628, | |
| "train_steps_per_second": 0.039 | |
| } | |
| ], | |
| "max_steps": 1876, | |
| "num_train_epochs": 4, | |
| "total_flos": 0.0, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |