| { | |
| "best_metric": 0.11326823383569717, | |
| "best_model_checkpoint": "vit-base-patch16-224-in21k-crack-detectorVITmain50epochs\\checkpoint-6688", | |
| "epoch": 6.999775734469612, | |
| "eval_steps": 500, | |
| "global_step": 7803, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 8.976660682226213e-08, | |
| "loss": 1.3961, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.7953321364452426e-07, | |
| "loss": 1.395, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.692998204667864e-07, | |
| "loss": 1.3978, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 3.590664272890485e-07, | |
| "loss": 1.3963, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 4.4883303411131064e-07, | |
| "loss": 1.3907, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 5.385996409335728e-07, | |
| "loss": 1.396, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 6.283662477558349e-07, | |
| "loss": 1.3885, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 7.18132854578097e-07, | |
| "loss": 1.387, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 8.078994614003591e-07, | |
| "loss": 1.3894, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 8.976660682226213e-07, | |
| "loss": 1.3806, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.874326750448833e-07, | |
| "loss": 1.383, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.0771992818671456e-06, | |
| "loss": 1.3764, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.1669658886894075e-06, | |
| "loss": 1.3688, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.2567324955116697e-06, | |
| "loss": 1.3708, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.3464991023339318e-06, | |
| "loss": 1.366, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.436265709156194e-06, | |
| "loss": 1.371, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.5260323159784561e-06, | |
| "loss": 1.3558, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.6157989228007182e-06, | |
| "loss": 1.3584, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.7055655296229805e-06, | |
| "loss": 1.3409, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.7953321364452425e-06, | |
| "loss": 1.346, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8850987432675046e-06, | |
| "loss": 1.3355, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.9748653500897667e-06, | |
| "loss": 1.3323, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 2.064631956912029e-06, | |
| "loss": 1.3231, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 2.1543985637342912e-06, | |
| "loss": 1.3129, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 2.244165170556553e-06, | |
| "loss": 1.3178, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 2.333931777378815e-06, | |
| "loss": 1.2953, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 2.423698384201077e-06, | |
| "loss": 1.2938, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 2.5134649910233395e-06, | |
| "loss": 1.2849, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 2.6032315978456015e-06, | |
| "loss": 1.2813, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 2.6929982046678636e-06, | |
| "loss": 1.2811, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 2.7827648114901257e-06, | |
| "loss": 1.2526, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 2.872531418312388e-06, | |
| "loss": 1.2438, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 2.9622980251346502e-06, | |
| "loss": 1.2145, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 3.0520646319569123e-06, | |
| "loss": 1.2146, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 3.1418312387791743e-06, | |
| "loss": 1.1883, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 3.2315978456014364e-06, | |
| "loss": 1.1891, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 3.321364452423698e-06, | |
| "loss": 1.1723, | |
| "step": 370 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 3.411131059245961e-06, | |
| "loss": 1.1417, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 3.500897666068223e-06, | |
| "loss": 1.1338, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 3.590664272890485e-06, | |
| "loss": 1.1004, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 3.680430879712747e-06, | |
| "loss": 1.0873, | |
| "step": 410 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 3.770197486535009e-06, | |
| "loss": 1.0358, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 3.859964093357271e-06, | |
| "loss": 1.0517, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 3.949730700179533e-06, | |
| "loss": 1.0139, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 4.039497307001795e-06, | |
| "loss": 1.0294, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 4.129263913824058e-06, | |
| "loss": 1.0171, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 4.21903052064632e-06, | |
| "loss": 0.9827, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 4.3087971274685824e-06, | |
| "loss": 0.9382, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 4.3985637342908445e-06, | |
| "loss": 0.9145, | |
| "step": 490 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 4.488330341113106e-06, | |
| "loss": 0.9276, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 4.578096947935368e-06, | |
| "loss": 0.8801, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 4.66786355475763e-06, | |
| "loss": 0.8906, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 4.757630161579892e-06, | |
| "loss": 0.9025, | |
| "step": 530 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 4.847396768402154e-06, | |
| "loss": 0.8621, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 4.937163375224417e-06, | |
| "loss": 0.8653, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 5.026929982046679e-06, | |
| "loss": 0.8292, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 5.116696588868941e-06, | |
| "loss": 0.8195, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 5.206463195691203e-06, | |
| "loss": 0.8026, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 5.296229802513465e-06, | |
| "loss": 0.841, | |
| "step": 590 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 5.385996409335727e-06, | |
| "loss": 0.778, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 5.475763016157989e-06, | |
| "loss": 0.7706, | |
| "step": 610 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 5.565529622980251e-06, | |
| "loss": 0.733, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 5.655296229802514e-06, | |
| "loss": 0.7437, | |
| "step": 630 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 5.745062836624776e-06, | |
| "loss": 0.7644, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 5.834829443447038e-06, | |
| "loss": 0.7125, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 5.9245960502693004e-06, | |
| "loss": 0.7146, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 6.0143626570915625e-06, | |
| "loss": 0.7201, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 6.1041292639138246e-06, | |
| "loss": 0.6851, | |
| "step": 680 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.193895870736087e-06, | |
| "loss": 0.6417, | |
| "step": 690 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.283662477558349e-06, | |
| "loss": 0.6656, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.373429084380611e-06, | |
| "loss": 0.6122, | |
| "step": 710 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 6.463195691202873e-06, | |
| "loss": 0.6536, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 6.552962298025135e-06, | |
| "loss": 0.6315, | |
| "step": 730 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 6.642728904847396e-06, | |
| "loss": 0.6618, | |
| "step": 740 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 6.732495511669659e-06, | |
| "loss": 0.6346, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 6.822262118491922e-06, | |
| "loss": 0.6866, | |
| "step": 760 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 6.912028725314183e-06, | |
| "loss": 0.6093, | |
| "step": 770 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 7.001795332136446e-06, | |
| "loss": 0.6116, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 7.091561938958707e-06, | |
| "loss": 0.602, | |
| "step": 790 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 7.18132854578097e-06, | |
| "loss": 0.6266, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.271095152603231e-06, | |
| "loss": 0.6082, | |
| "step": 810 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 7.360861759425494e-06, | |
| "loss": 0.6194, | |
| "step": 820 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 7.4506283662477555e-06, | |
| "loss": 0.6277, | |
| "step": 830 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 7.540394973070018e-06, | |
| "loss": 0.609, | |
| "step": 840 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 7.630161579892281e-06, | |
| "loss": 0.6252, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 7.719928186714543e-06, | |
| "loss": 0.5813, | |
| "step": 860 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 7.809694793536805e-06, | |
| "loss": 0.5799, | |
| "step": 870 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 7.899461400359067e-06, | |
| "loss": 0.6398, | |
| "step": 880 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 7.98922800718133e-06, | |
| "loss": 0.5549, | |
| "step": 890 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 8.07899461400359e-06, | |
| "loss": 0.6023, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 8.168761220825854e-06, | |
| "loss": 0.582, | |
| "step": 910 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 8.258527827648117e-06, | |
| "loss": 0.544, | |
| "step": 920 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 8.348294434470378e-06, | |
| "loss": 0.5912, | |
| "step": 930 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 8.43806104129264e-06, | |
| "loss": 0.5461, | |
| "step": 940 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 8.527827648114902e-06, | |
| "loss": 0.5238, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 8.617594254937165e-06, | |
| "loss": 0.5558, | |
| "step": 960 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 8.707360861759426e-06, | |
| "loss": 0.5371, | |
| "step": 970 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 8.797127468581689e-06, | |
| "loss": 0.5529, | |
| "step": 980 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 8.88689407540395e-06, | |
| "loss": 0.5691, | |
| "step": 990 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 8.976660682226211e-06, | |
| "loss": 0.5567, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 9.066427289048474e-06, | |
| "loss": 0.5135, | |
| "step": 1010 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 9.156193895870736e-06, | |
| "loss": 0.5099, | |
| "step": 1020 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 9.245960502692998e-06, | |
| "loss": 0.5422, | |
| "step": 1030 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 9.33572710951526e-06, | |
| "loss": 0.5106, | |
| "step": 1040 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 9.425493716337523e-06, | |
| "loss": 0.4892, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 9.515260323159784e-06, | |
| "loss": 0.4541, | |
| "step": 1060 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 9.605026929982047e-06, | |
| "loss": 0.4675, | |
| "step": 1070 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 9.694793536804308e-06, | |
| "loss": 0.5123, | |
| "step": 1080 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 9.784560143626571e-06, | |
| "loss": 0.5089, | |
| "step": 1090 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 9.874326750448834e-06, | |
| "loss": 0.5233, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 9.964093357271095e-06, | |
| "loss": 0.5622, | |
| "step": 1110 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": { | |
| "accuracy": 0.8736263736263736 | |
| }, | |
| "eval_f1": { | |
| "f1": 0.8653360894930927 | |
| }, | |
| "eval_loss": 0.4108576774597168, | |
| "eval_precision": { | |
| "precision": 0.8669392826942903 | |
| }, | |
| "eval_recall": { | |
| "recall": 0.8641606609285752 | |
| }, | |
| "eval_runtime": 236.9507, | |
| "eval_samples_per_second": 75.273, | |
| "eval_steps_per_second": 4.706, | |
| "step": 1114 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.0053859964093358e-05, | |
| "loss": 0.4957, | |
| "step": 1120 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 1.0143626570915619e-05, | |
| "loss": 0.5729, | |
| "step": 1130 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 1.0233393177737882e-05, | |
| "loss": 0.4884, | |
| "step": 1140 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 1.0323159784560143e-05, | |
| "loss": 0.4744, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 1.0412926391382406e-05, | |
| "loss": 0.4683, | |
| "step": 1160 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 1.0502692998204669e-05, | |
| "loss": 0.5, | |
| "step": 1170 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 1.059245960502693e-05, | |
| "loss": 0.4649, | |
| "step": 1180 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 1.0682226211849193e-05, | |
| "loss": 0.508, | |
| "step": 1190 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 1.0771992818671454e-05, | |
| "loss": 0.4898, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 1.0861759425493717e-05, | |
| "loss": 0.4722, | |
| "step": 1210 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 1.0951526032315979e-05, | |
| "loss": 0.4814, | |
| "step": 1220 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 1.1041292639138241e-05, | |
| "loss": 0.4505, | |
| "step": 1230 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 1.1131059245960503e-05, | |
| "loss": 0.4505, | |
| "step": 1240 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 1.1220825852782766e-05, | |
| "loss": 0.4187, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 1.1310592459605028e-05, | |
| "loss": 0.4603, | |
| "step": 1260 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 1.140035906642729e-05, | |
| "loss": 0.4721, | |
| "step": 1270 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 1.1490125673249553e-05, | |
| "loss": 0.4368, | |
| "step": 1280 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 1.1579892280071814e-05, | |
| "loss": 0.4722, | |
| "step": 1290 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 1.1669658886894077e-05, | |
| "loss": 0.4718, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 1.1759425493716338e-05, | |
| "loss": 0.4899, | |
| "step": 1310 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 1.1849192100538601e-05, | |
| "loss": 0.4758, | |
| "step": 1320 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 1.1938958707360862e-05, | |
| "loss": 0.4523, | |
| "step": 1330 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 1.2028725314183125e-05, | |
| "loss": 0.485, | |
| "step": 1340 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 1.2118491921005388e-05, | |
| "loss": 0.4977, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 1.2208258527827649e-05, | |
| "loss": 0.4305, | |
| "step": 1360 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 1.2298025134649912e-05, | |
| "loss": 0.4221, | |
| "step": 1370 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 1.2387791741472173e-05, | |
| "loss": 0.4975, | |
| "step": 1380 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 1.2477558348294434e-05, | |
| "loss": 0.4506, | |
| "step": 1390 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 1.2567324955116697e-05, | |
| "loss": 0.4326, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 1.2657091561938959e-05, | |
| "loss": 0.4479, | |
| "step": 1410 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 1.2746858168761221e-05, | |
| "loss": 0.4468, | |
| "step": 1420 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 1.2836624775583483e-05, | |
| "loss": 0.402, | |
| "step": 1430 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 1.2926391382405746e-05, | |
| "loss": 0.4711, | |
| "step": 1440 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 1.3016157989228009e-05, | |
| "loss": 0.4686, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 1.310592459605027e-05, | |
| "loss": 0.4639, | |
| "step": 1460 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 1.3195691202872531e-05, | |
| "loss": 0.435, | |
| "step": 1470 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 1.3285457809694792e-05, | |
| "loss": 0.3868, | |
| "step": 1480 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 1.3375224416517057e-05, | |
| "loss": 0.4303, | |
| "step": 1490 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 1.3464991023339318e-05, | |
| "loss": 0.432, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 1.355475763016158e-05, | |
| "loss": 0.4206, | |
| "step": 1510 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 1.3644524236983844e-05, | |
| "loss": 0.4282, | |
| "step": 1520 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 1.3734290843806105e-05, | |
| "loss": 0.4284, | |
| "step": 1530 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 1.3824057450628366e-05, | |
| "loss": 0.4096, | |
| "step": 1540 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 1.3913824057450627e-05, | |
| "loss": 0.4259, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 1.4003590664272892e-05, | |
| "loss": 0.4436, | |
| "step": 1560 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 1.4093357271095153e-05, | |
| "loss": 0.4453, | |
| "step": 1570 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 1.4183123877917415e-05, | |
| "loss": 0.4065, | |
| "step": 1580 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 1.4272890484739679e-05, | |
| "loss": 0.4011, | |
| "step": 1590 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 1.436265709156194e-05, | |
| "loss": 0.4488, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 1.4452423698384202e-05, | |
| "loss": 0.4122, | |
| "step": 1610 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 1.4542190305206463e-05, | |
| "loss": 0.429, | |
| "step": 1620 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 1.4631956912028727e-05, | |
| "loss": 0.4377, | |
| "step": 1630 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 1.4721723518850989e-05, | |
| "loss": 0.4186, | |
| "step": 1640 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 1.481149012567325e-05, | |
| "loss": 0.4063, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 1.4901256732495511e-05, | |
| "loss": 0.4055, | |
| "step": 1660 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 1.4991023339317776e-05, | |
| "loss": 0.4173, | |
| "step": 1670 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 1.5080789946140037e-05, | |
| "loss": 0.4361, | |
| "step": 1680 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 1.5170556552962298e-05, | |
| "loss": 0.4528, | |
| "step": 1690 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 1.5260323159784563e-05, | |
| "loss": 0.3665, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 1.5350089766606824e-05, | |
| "loss": 0.3913, | |
| "step": 1710 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 1.5439856373429085e-05, | |
| "loss": 0.4002, | |
| "step": 1720 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 1.5529622980251346e-05, | |
| "loss": 0.4004, | |
| "step": 1730 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 1.561938958707361e-05, | |
| "loss": 0.4432, | |
| "step": 1740 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 1.5709156193895872e-05, | |
| "loss": 0.3743, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 1.5798922800718133e-05, | |
| "loss": 0.3972, | |
| "step": 1760 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 1.5888689407540398e-05, | |
| "loss": 0.368, | |
| "step": 1770 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 1.597845601436266e-05, | |
| "loss": 0.4525, | |
| "step": 1780 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 1.606822262118492e-05, | |
| "loss": 0.3962, | |
| "step": 1790 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 1.615798922800718e-05, | |
| "loss": 0.3888, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 1.6247755834829446e-05, | |
| "loss": 0.4111, | |
| "step": 1810 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 1.6337522441651707e-05, | |
| "loss": 0.4084, | |
| "step": 1820 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 1.642728904847397e-05, | |
| "loss": 0.4181, | |
| "step": 1830 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "learning_rate": 1.6517055655296233e-05, | |
| "loss": 0.4022, | |
| "step": 1840 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 1.6606822262118494e-05, | |
| "loss": 0.3886, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 1.67, | |
| "learning_rate": 1.6696588868940756e-05, | |
| "loss": 0.4092, | |
| "step": 1860 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 1.6786355475763017e-05, | |
| "loss": 0.3843, | |
| "step": 1870 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 1.687612208258528e-05, | |
| "loss": 0.4237, | |
| "step": 1880 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 1.6965888689407543e-05, | |
| "loss": 0.3801, | |
| "step": 1890 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 1.7055655296229804e-05, | |
| "loss": 0.385, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 1.7145421903052065e-05, | |
| "loss": 0.4452, | |
| "step": 1910 | |
| }, | |
| { | |
| "epoch": 1.72, | |
| "learning_rate": 1.723518850987433e-05, | |
| "loss": 0.5147, | |
| "step": 1920 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 1.732495511669659e-05, | |
| "loss": 0.353, | |
| "step": 1930 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 1.7414721723518852e-05, | |
| "loss": 0.3816, | |
| "step": 1940 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 1.7504488330341113e-05, | |
| "loss": 0.361, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 1.7594254937163378e-05, | |
| "loss": 0.3884, | |
| "step": 1960 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 1.768402154398564e-05, | |
| "loss": 0.3391, | |
| "step": 1970 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 1.77737881508079e-05, | |
| "loss": 0.4211, | |
| "step": 1980 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 1.786355475763016e-05, | |
| "loss": 0.3796, | |
| "step": 1990 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 1.7953321364452423e-05, | |
| "loss": 0.3881, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 1.8043087971274687e-05, | |
| "loss": 0.3904, | |
| "step": 2010 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 1.813285457809695e-05, | |
| "loss": 0.3845, | |
| "step": 2020 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 1.822262118491921e-05, | |
| "loss": 0.3721, | |
| "step": 2030 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 1.831238779174147e-05, | |
| "loss": 0.4141, | |
| "step": 2040 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 1.8402154398563732e-05, | |
| "loss": 0.3682, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 1.8491921005385997e-05, | |
| "loss": 0.3826, | |
| "step": 2060 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 1.8581687612208258e-05, | |
| "loss": 0.3588, | |
| "step": 2070 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 1.867145421903052e-05, | |
| "loss": 0.3663, | |
| "step": 2080 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 1.8761220825852784e-05, | |
| "loss": 0.4186, | |
| "step": 2090 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 1.8850987432675045e-05, | |
| "loss": 0.3756, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 1.8940754039497306e-05, | |
| "loss": 0.3655, | |
| "step": 2110 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 1.9030520646319568e-05, | |
| "loss": 0.4158, | |
| "step": 2120 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 1.9120287253141832e-05, | |
| "loss": 0.3985, | |
| "step": 2130 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 1.9210053859964093e-05, | |
| "loss": 0.3955, | |
| "step": 2140 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 1.9299820466786355e-05, | |
| "loss": 0.3757, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 1.9389587073608616e-05, | |
| "loss": 0.3712, | |
| "step": 2160 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 1.947935368043088e-05, | |
| "loss": 0.3723, | |
| "step": 2170 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 1.9569120287253142e-05, | |
| "loss": 0.367, | |
| "step": 2180 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 1.9658886894075403e-05, | |
| "loss": 0.3659, | |
| "step": 2190 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 1.9748653500897668e-05, | |
| "loss": 0.3752, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 1.983842010771993e-05, | |
| "loss": 0.318, | |
| "step": 2210 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 1.992818671454219e-05, | |
| "loss": 0.3696, | |
| "step": 2220 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": { | |
| "accuracy": 0.9249271137026239 | |
| }, | |
| "eval_f1": { | |
| "f1": 0.9207602119769538 | |
| }, | |
| "eval_loss": 0.2492757886648178, | |
| "eval_precision": { | |
| "precision": 0.9199949527238288 | |
| }, | |
| "eval_recall": { | |
| "recall": 0.9219224853720074 | |
| }, | |
| "eval_runtime": 166.1808, | |
| "eval_samples_per_second": 107.329, | |
| "eval_steps_per_second": 6.71, | |
| "step": 2229 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 2.001795332136445e-05, | |
| "loss": 0.3681, | |
| "step": 2230 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 2.0107719928186716e-05, | |
| "loss": 0.321, | |
| "step": 2240 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 2.0197486535008977e-05, | |
| "loss": 0.332, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 2.0287253141831238e-05, | |
| "loss": 0.3343, | |
| "step": 2260 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 2.0377019748653503e-05, | |
| "loss": 0.3341, | |
| "step": 2270 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 2.0466786355475764e-05, | |
| "loss": 0.3473, | |
| "step": 2280 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 2.0556552962298025e-05, | |
| "loss": 0.3479, | |
| "step": 2290 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 2.0646319569120286e-05, | |
| "loss": 0.3235, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 2.073608617594255e-05, | |
| "loss": 0.3455, | |
| "step": 2310 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 2.0825852782764812e-05, | |
| "loss": 0.3675, | |
| "step": 2320 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 2.0915619389587073e-05, | |
| "loss": 0.3267, | |
| "step": 2330 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 2.1005385996409338e-05, | |
| "loss": 0.3588, | |
| "step": 2340 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 2.10951526032316e-05, | |
| "loss": 0.3762, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 2.118491921005386e-05, | |
| "loss": 0.3474, | |
| "step": 2360 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 2.1274685816876122e-05, | |
| "loss": 0.3614, | |
| "step": 2370 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 2.1364452423698386e-05, | |
| "loss": 0.3611, | |
| "step": 2380 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 2.1454219030520648e-05, | |
| "loss": 0.4051, | |
| "step": 2390 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 2.154398563734291e-05, | |
| "loss": 0.3314, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 2.163375224416517e-05, | |
| "loss": 0.3427, | |
| "step": 2410 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 2.1723518850987435e-05, | |
| "loss": 0.2886, | |
| "step": 2420 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 2.1813285457809696e-05, | |
| "loss": 0.3262, | |
| "step": 2430 | |
| }, | |
| { | |
| "epoch": 2.19, | |
| "learning_rate": 2.1903052064631957e-05, | |
| "loss": 0.3516, | |
| "step": 2440 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 2.199281867145422e-05, | |
| "loss": 0.3744, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 2.21, | |
| "learning_rate": 2.2082585278276483e-05, | |
| "loss": 0.3545, | |
| "step": 2460 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 2.2172351885098744e-05, | |
| "loss": 0.2993, | |
| "step": 2470 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 2.2262118491921005e-05, | |
| "loss": 0.3412, | |
| "step": 2480 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 2.235188509874327e-05, | |
| "loss": 0.3823, | |
| "step": 2490 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 2.244165170556553e-05, | |
| "loss": 0.3206, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 2.2531418312387792e-05, | |
| "loss": 0.2969, | |
| "step": 2510 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 2.2621184919210057e-05, | |
| "loss": 0.3306, | |
| "step": 2520 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 2.2710951526032318e-05, | |
| "loss": 0.3641, | |
| "step": 2530 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 2.280071813285458e-05, | |
| "loss": 0.3588, | |
| "step": 2540 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 2.289048473967684e-05, | |
| "loss": 0.3632, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 2.2980251346499105e-05, | |
| "loss": 0.3072, | |
| "step": 2560 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 2.3070017953321366e-05, | |
| "loss": 0.3606, | |
| "step": 2570 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 2.3159784560143628e-05, | |
| "loss": 0.4114, | |
| "step": 2580 | |
| }, | |
| { | |
| "epoch": 2.32, | |
| "learning_rate": 2.3249551166965892e-05, | |
| "loss": 0.282, | |
| "step": 2590 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 2.3339317773788153e-05, | |
| "loss": 0.3336, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 2.3429084380610415e-05, | |
| "loss": 0.3406, | |
| "step": 2610 | |
| }, | |
| { | |
| "epoch": 2.35, | |
| "learning_rate": 2.3518850987432676e-05, | |
| "loss": 0.3433, | |
| "step": 2620 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 2.360861759425494e-05, | |
| "loss": 0.3169, | |
| "step": 2630 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "learning_rate": 2.3698384201077202e-05, | |
| "loss": 0.3855, | |
| "step": 2640 | |
| }, | |
| { | |
| "epoch": 2.38, | |
| "learning_rate": 2.3788150807899463e-05, | |
| "loss": 0.3036, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 2.3877917414721724e-05, | |
| "loss": 0.3076, | |
| "step": 2660 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 2.396768402154399e-05, | |
| "loss": 0.3052, | |
| "step": 2670 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 2.405745062836625e-05, | |
| "loss": 0.3253, | |
| "step": 2680 | |
| }, | |
| { | |
| "epoch": 2.41, | |
| "learning_rate": 2.414721723518851e-05, | |
| "loss": 0.3235, | |
| "step": 2690 | |
| }, | |
| { | |
| "epoch": 2.42, | |
| "learning_rate": 2.4236983842010776e-05, | |
| "loss": 0.2835, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 2.4326750448833037e-05, | |
| "loss": 0.2824, | |
| "step": 2710 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 2.4416517055655298e-05, | |
| "loss": 0.3183, | |
| "step": 2720 | |
| }, | |
| { | |
| "epoch": 2.45, | |
| "learning_rate": 2.450628366247756e-05, | |
| "loss": 0.3281, | |
| "step": 2730 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 2.4596050269299824e-05, | |
| "loss": 0.2939, | |
| "step": 2740 | |
| }, | |
| { | |
| "epoch": 2.47, | |
| "learning_rate": 2.4685816876122085e-05, | |
| "loss": 0.3679, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 2.48, | |
| "learning_rate": 2.4775583482944346e-05, | |
| "loss": 0.3165, | |
| "step": 2760 | |
| }, | |
| { | |
| "epoch": 2.48, | |
| "learning_rate": 2.4865350089766608e-05, | |
| "loss": 0.3146, | |
| "step": 2770 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "learning_rate": 2.495511669658887e-05, | |
| "loss": 0.2981, | |
| "step": 2780 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 2.5044883303411134e-05, | |
| "loss": 0.3099, | |
| "step": 2790 | |
| }, | |
| { | |
| "epoch": 2.51, | |
| "learning_rate": 2.5134649910233395e-05, | |
| "loss": 0.267, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "learning_rate": 2.5224416517055656e-05, | |
| "loss": 0.3317, | |
| "step": 2810 | |
| }, | |
| { | |
| "epoch": 2.53, | |
| "learning_rate": 2.5314183123877917e-05, | |
| "loss": 0.3241, | |
| "step": 2820 | |
| }, | |
| { | |
| "epoch": 2.54, | |
| "learning_rate": 2.5403949730700182e-05, | |
| "loss": 0.371, | |
| "step": 2830 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 2.5493716337522443e-05, | |
| "loss": 0.3431, | |
| "step": 2840 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "learning_rate": 2.5583482944344704e-05, | |
| "loss": 0.3218, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 2.57, | |
| "learning_rate": 2.5673249551166965e-05, | |
| "loss": 0.3222, | |
| "step": 2860 | |
| }, | |
| { | |
| "epoch": 2.57, | |
| "learning_rate": 2.5763016157989227e-05, | |
| "loss": 0.3176, | |
| "step": 2870 | |
| }, | |
| { | |
| "epoch": 2.58, | |
| "learning_rate": 2.585278276481149e-05, | |
| "loss": 0.3256, | |
| "step": 2880 | |
| }, | |
| { | |
| "epoch": 2.59, | |
| "learning_rate": 2.5942549371633752e-05, | |
| "loss": 0.2971, | |
| "step": 2890 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "learning_rate": 2.6032315978456017e-05, | |
| "loss": 0.253, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 2.61, | |
| "learning_rate": 2.6122082585278278e-05, | |
| "loss": 0.3261, | |
| "step": 2910 | |
| }, | |
| { | |
| "epoch": 2.62, | |
| "learning_rate": 2.621184919210054e-05, | |
| "loss": 0.3325, | |
| "step": 2920 | |
| }, | |
| { | |
| "epoch": 2.63, | |
| "learning_rate": 2.63016157989228e-05, | |
| "loss": 0.3005, | |
| "step": 2930 | |
| }, | |
| { | |
| "epoch": 2.64, | |
| "learning_rate": 2.6391382405745062e-05, | |
| "loss": 0.2897, | |
| "step": 2940 | |
| }, | |
| { | |
| "epoch": 2.65, | |
| "learning_rate": 2.6481149012567323e-05, | |
| "loss": 0.3249, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 2.66, | |
| "learning_rate": 2.6570915619389584e-05, | |
| "loss": 0.3647, | |
| "step": 2960 | |
| }, | |
| { | |
| "epoch": 2.66, | |
| "learning_rate": 2.6660682226211852e-05, | |
| "loss": 0.3034, | |
| "step": 2970 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 2.6750448833034114e-05, | |
| "loss": 0.2735, | |
| "step": 2980 | |
| }, | |
| { | |
| "epoch": 2.68, | |
| "learning_rate": 2.6840215439856375e-05, | |
| "loss": 0.3122, | |
| "step": 2990 | |
| }, | |
| { | |
| "epoch": 2.69, | |
| "learning_rate": 2.6929982046678636e-05, | |
| "loss": 0.2622, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 2.7, | |
| "learning_rate": 2.7019748653500897e-05, | |
| "loss": 0.322, | |
| "step": 3010 | |
| }, | |
| { | |
| "epoch": 2.71, | |
| "learning_rate": 2.710951526032316e-05, | |
| "loss": 0.2671, | |
| "step": 3020 | |
| }, | |
| { | |
| "epoch": 2.72, | |
| "learning_rate": 2.719928186714542e-05, | |
| "loss": 0.3392, | |
| "step": 3030 | |
| }, | |
| { | |
| "epoch": 2.73, | |
| "learning_rate": 2.7289048473967688e-05, | |
| "loss": 0.2507, | |
| "step": 3040 | |
| }, | |
| { | |
| "epoch": 2.74, | |
| "learning_rate": 2.737881508078995e-05, | |
| "loss": 0.294, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 2.746858168761221e-05, | |
| "loss": 0.3119, | |
| "step": 3060 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 2.755834829443447e-05, | |
| "loss": 0.2956, | |
| "step": 3070 | |
| }, | |
| { | |
| "epoch": 2.76, | |
| "learning_rate": 2.7648114901256732e-05, | |
| "loss": 0.2744, | |
| "step": 3080 | |
| }, | |
| { | |
| "epoch": 2.77, | |
| "learning_rate": 2.7737881508078994e-05, | |
| "loss": 0.279, | |
| "step": 3090 | |
| }, | |
| { | |
| "epoch": 2.78, | |
| "learning_rate": 2.7827648114901255e-05, | |
| "loss": 0.379, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 2.79, | |
| "learning_rate": 2.7917414721723523e-05, | |
| "loss": 0.2667, | |
| "step": 3110 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "learning_rate": 2.8007181328545784e-05, | |
| "loss": 0.2732, | |
| "step": 3120 | |
| }, | |
| { | |
| "epoch": 2.81, | |
| "learning_rate": 2.8096947935368045e-05, | |
| "loss": 0.311, | |
| "step": 3130 | |
| }, | |
| { | |
| "epoch": 2.82, | |
| "learning_rate": 2.8186714542190307e-05, | |
| "loss": 0.3404, | |
| "step": 3140 | |
| }, | |
| { | |
| "epoch": 2.83, | |
| "learning_rate": 2.8276481149012568e-05, | |
| "loss": 0.2479, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 2.83, | |
| "learning_rate": 2.836624775583483e-05, | |
| "loss": 0.3057, | |
| "step": 3160 | |
| }, | |
| { | |
| "epoch": 2.84, | |
| "learning_rate": 2.845601436265709e-05, | |
| "loss": 0.3075, | |
| "step": 3170 | |
| }, | |
| { | |
| "epoch": 2.85, | |
| "learning_rate": 2.8545780969479358e-05, | |
| "loss": 0.3008, | |
| "step": 3180 | |
| }, | |
| { | |
| "epoch": 2.86, | |
| "learning_rate": 2.863554757630162e-05, | |
| "loss": 0.2372, | |
| "step": 3190 | |
| }, | |
| { | |
| "epoch": 2.87, | |
| "learning_rate": 2.872531418312388e-05, | |
| "loss": 0.2617, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 2.88, | |
| "learning_rate": 2.8815080789946142e-05, | |
| "loss": 0.3, | |
| "step": 3210 | |
| }, | |
| { | |
| "epoch": 2.89, | |
| "learning_rate": 2.8904847396768403e-05, | |
| "loss": 0.2749, | |
| "step": 3220 | |
| }, | |
| { | |
| "epoch": 2.9, | |
| "learning_rate": 2.8994614003590664e-05, | |
| "loss": 0.2742, | |
| "step": 3230 | |
| }, | |
| { | |
| "epoch": 2.91, | |
| "learning_rate": 2.9084380610412926e-05, | |
| "loss": 0.2844, | |
| "step": 3240 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "learning_rate": 2.9174147217235194e-05, | |
| "loss": 0.2909, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "learning_rate": 2.9263913824057455e-05, | |
| "loss": 0.3045, | |
| "step": 3260 | |
| }, | |
| { | |
| "epoch": 2.93, | |
| "learning_rate": 2.9353680430879716e-05, | |
| "loss": 0.2875, | |
| "step": 3270 | |
| }, | |
| { | |
| "epoch": 2.94, | |
| "learning_rate": 2.9443447037701977e-05, | |
| "loss": 0.2963, | |
| "step": 3280 | |
| }, | |
| { | |
| "epoch": 2.95, | |
| "learning_rate": 2.953321364452424e-05, | |
| "loss": 0.314, | |
| "step": 3290 | |
| }, | |
| { | |
| "epoch": 2.96, | |
| "learning_rate": 2.96229802513465e-05, | |
| "loss": 0.3005, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 2.97, | |
| "learning_rate": 2.971274685816876e-05, | |
| "loss": 0.3127, | |
| "step": 3310 | |
| }, | |
| { | |
| "epoch": 2.98, | |
| "learning_rate": 2.9802513464991022e-05, | |
| "loss": 0.3113, | |
| "step": 3320 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "learning_rate": 2.989228007181329e-05, | |
| "loss": 0.2886, | |
| "step": 3330 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 2.998204667863555e-05, | |
| "loss": 0.321, | |
| "step": 3340 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": { | |
| "accuracy": 0.9410742318905584 | |
| }, | |
| "eval_f1": { | |
| "f1": 0.9378351696354423 | |
| }, | |
| "eval_loss": 0.18490658700466156, | |
| "eval_precision": { | |
| "precision": 0.9376663800198108 | |
| }, | |
| "eval_recall": { | |
| "recall": 0.9381803360170522 | |
| }, | |
| "eval_runtime": 165.2584, | |
| "eval_samples_per_second": 107.928, | |
| "eval_steps_per_second": 6.747, | |
| "step": 3344 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 3.0071813285457812e-05, | |
| "loss": 0.3252, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 3.0161579892280074e-05, | |
| "loss": 0.2945, | |
| "step": 3360 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 3.0251346499102335e-05, | |
| "loss": 0.2584, | |
| "step": 3370 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 3.0341113105924596e-05, | |
| "loss": 0.2878, | |
| "step": 3380 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 3.0430879712746857e-05, | |
| "loss": 0.2618, | |
| "step": 3390 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 3.0520646319569125e-05, | |
| "loss": 0.3356, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 3.06, | |
| "learning_rate": 3.061041292639138e-05, | |
| "loss": 0.2706, | |
| "step": 3410 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "learning_rate": 3.070017953321365e-05, | |
| "loss": 0.2821, | |
| "step": 3420 | |
| }, | |
| { | |
| "epoch": 3.08, | |
| "learning_rate": 3.0789946140035906e-05, | |
| "loss": 0.2634, | |
| "step": 3430 | |
| }, | |
| { | |
| "epoch": 3.09, | |
| "learning_rate": 3.087971274685817e-05, | |
| "loss": 0.2927, | |
| "step": 3440 | |
| }, | |
| { | |
| "epoch": 3.09, | |
| "learning_rate": 3.096947935368043e-05, | |
| "loss": 0.2406, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 3.1, | |
| "learning_rate": 3.105924596050269e-05, | |
| "loss": 0.2871, | |
| "step": 3460 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "learning_rate": 3.114901256732496e-05, | |
| "loss": 0.2389, | |
| "step": 3470 | |
| }, | |
| { | |
| "epoch": 3.12, | |
| "learning_rate": 3.123877917414722e-05, | |
| "loss": 0.2782, | |
| "step": 3480 | |
| }, | |
| { | |
| "epoch": 3.13, | |
| "learning_rate": 3.132854578096948e-05, | |
| "loss": 0.2864, | |
| "step": 3490 | |
| }, | |
| { | |
| "epoch": 3.14, | |
| "learning_rate": 3.1418312387791744e-05, | |
| "loss": 0.29, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 3.15, | |
| "learning_rate": 3.1508078994614e-05, | |
| "loss": 0.2089, | |
| "step": 3510 | |
| }, | |
| { | |
| "epoch": 3.16, | |
| "learning_rate": 3.159784560143627e-05, | |
| "loss": 0.2716, | |
| "step": 3520 | |
| }, | |
| { | |
| "epoch": 3.17, | |
| "learning_rate": 3.1687612208258525e-05, | |
| "loss": 0.2981, | |
| "step": 3530 | |
| }, | |
| { | |
| "epoch": 3.18, | |
| "learning_rate": 3.1777378815080796e-05, | |
| "loss": 0.2265, | |
| "step": 3540 | |
| }, | |
| { | |
| "epoch": 3.18, | |
| "learning_rate": 3.1867145421903054e-05, | |
| "loss": 0.1997, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 3.19, | |
| "learning_rate": 3.195691202872532e-05, | |
| "loss": 0.2707, | |
| "step": 3560 | |
| }, | |
| { | |
| "epoch": 3.2, | |
| "learning_rate": 3.2046678635547576e-05, | |
| "loss": 0.2917, | |
| "step": 3570 | |
| }, | |
| { | |
| "epoch": 3.21, | |
| "learning_rate": 3.213644524236984e-05, | |
| "loss": 0.2966, | |
| "step": 3580 | |
| }, | |
| { | |
| "epoch": 3.22, | |
| "learning_rate": 3.22262118491921e-05, | |
| "loss": 0.265, | |
| "step": 3590 | |
| }, | |
| { | |
| "epoch": 3.23, | |
| "learning_rate": 3.231597845601436e-05, | |
| "loss": 0.2386, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 3.24, | |
| "learning_rate": 3.240574506283663e-05, | |
| "loss": 0.2613, | |
| "step": 3610 | |
| }, | |
| { | |
| "epoch": 3.25, | |
| "learning_rate": 3.249551166965889e-05, | |
| "loss": 0.2737, | |
| "step": 3620 | |
| }, | |
| { | |
| "epoch": 3.26, | |
| "learning_rate": 3.258527827648115e-05, | |
| "loss": 0.2965, | |
| "step": 3630 | |
| }, | |
| { | |
| "epoch": 3.27, | |
| "learning_rate": 3.2675044883303415e-05, | |
| "loss": 0.2683, | |
| "step": 3640 | |
| }, | |
| { | |
| "epoch": 3.27, | |
| "learning_rate": 3.276481149012567e-05, | |
| "loss": 0.2379, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 3.28, | |
| "learning_rate": 3.285457809694794e-05, | |
| "loss": 0.3282, | |
| "step": 3660 | |
| }, | |
| { | |
| "epoch": 3.29, | |
| "learning_rate": 3.2944344703770195e-05, | |
| "loss": 0.2589, | |
| "step": 3670 | |
| }, | |
| { | |
| "epoch": 3.3, | |
| "learning_rate": 3.3034111310592466e-05, | |
| "loss": 0.2955, | |
| "step": 3680 | |
| }, | |
| { | |
| "epoch": 3.31, | |
| "learning_rate": 3.3123877917414724e-05, | |
| "loss": 0.2584, | |
| "step": 3690 | |
| }, | |
| { | |
| "epoch": 3.32, | |
| "learning_rate": 3.321364452423699e-05, | |
| "loss": 0.2643, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 3.33, | |
| "learning_rate": 3.330341113105925e-05, | |
| "loss": 0.2987, | |
| "step": 3710 | |
| }, | |
| { | |
| "epoch": 3.34, | |
| "learning_rate": 3.339317773788151e-05, | |
| "loss": 0.3, | |
| "step": 3720 | |
| }, | |
| { | |
| "epoch": 3.35, | |
| "learning_rate": 3.348294434470377e-05, | |
| "loss": 0.2488, | |
| "step": 3730 | |
| }, | |
| { | |
| "epoch": 3.36, | |
| "learning_rate": 3.3572710951526034e-05, | |
| "loss": 0.252, | |
| "step": 3740 | |
| }, | |
| { | |
| "epoch": 3.36, | |
| "learning_rate": 3.36624775583483e-05, | |
| "loss": 0.286, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 3.37, | |
| "learning_rate": 3.375224416517056e-05, | |
| "loss": 0.2801, | |
| "step": 3760 | |
| }, | |
| { | |
| "epoch": 3.38, | |
| "learning_rate": 3.384201077199282e-05, | |
| "loss": 0.2463, | |
| "step": 3770 | |
| }, | |
| { | |
| "epoch": 3.39, | |
| "learning_rate": 3.3931777378815085e-05, | |
| "loss": 0.2824, | |
| "step": 3780 | |
| }, | |
| { | |
| "epoch": 3.4, | |
| "learning_rate": 3.402154398563734e-05, | |
| "loss": 0.2847, | |
| "step": 3790 | |
| }, | |
| { | |
| "epoch": 3.41, | |
| "learning_rate": 3.411131059245961e-05, | |
| "loss": 0.3061, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 3.42, | |
| "learning_rate": 3.4201077199281866e-05, | |
| "loss": 0.2799, | |
| "step": 3810 | |
| }, | |
| { | |
| "epoch": 3.43, | |
| "learning_rate": 3.429084380610413e-05, | |
| "loss": 0.2532, | |
| "step": 3820 | |
| }, | |
| { | |
| "epoch": 3.44, | |
| "learning_rate": 3.4380610412926395e-05, | |
| "loss": 0.2761, | |
| "step": 3830 | |
| }, | |
| { | |
| "epoch": 3.44, | |
| "learning_rate": 3.447037701974866e-05, | |
| "loss": 0.3087, | |
| "step": 3840 | |
| }, | |
| { | |
| "epoch": 3.45, | |
| "learning_rate": 3.456014362657092e-05, | |
| "loss": 0.2841, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 3.46, | |
| "learning_rate": 3.464991023339318e-05, | |
| "loss": 0.2191, | |
| "step": 3860 | |
| }, | |
| { | |
| "epoch": 3.47, | |
| "learning_rate": 3.473967684021544e-05, | |
| "loss": 0.328, | |
| "step": 3870 | |
| }, | |
| { | |
| "epoch": 3.48, | |
| "learning_rate": 3.4829443447037704e-05, | |
| "loss": 0.2377, | |
| "step": 3880 | |
| }, | |
| { | |
| "epoch": 3.49, | |
| "learning_rate": 3.491921005385996e-05, | |
| "loss": 0.2844, | |
| "step": 3890 | |
| }, | |
| { | |
| "epoch": 3.5, | |
| "learning_rate": 3.500897666068223e-05, | |
| "loss": 0.2944, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 3.51, | |
| "learning_rate": 3.509874326750449e-05, | |
| "loss": 0.2371, | |
| "step": 3910 | |
| }, | |
| { | |
| "epoch": 3.52, | |
| "learning_rate": 3.5188509874326756e-05, | |
| "loss": 0.2552, | |
| "step": 3920 | |
| }, | |
| { | |
| "epoch": 3.53, | |
| "learning_rate": 3.5278276481149014e-05, | |
| "loss": 0.2492, | |
| "step": 3930 | |
| }, | |
| { | |
| "epoch": 3.53, | |
| "learning_rate": 3.536804308797128e-05, | |
| "loss": 0.253, | |
| "step": 3940 | |
| }, | |
| { | |
| "epoch": 3.54, | |
| "learning_rate": 3.5457809694793536e-05, | |
| "loss": 0.2794, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 3.55, | |
| "learning_rate": 3.55475763016158e-05, | |
| "loss": 0.2253, | |
| "step": 3960 | |
| }, | |
| { | |
| "epoch": 3.56, | |
| "learning_rate": 3.5637342908438065e-05, | |
| "loss": 0.2894, | |
| "step": 3970 | |
| }, | |
| { | |
| "epoch": 3.57, | |
| "learning_rate": 3.572710951526032e-05, | |
| "loss": 0.2076, | |
| "step": 3980 | |
| }, | |
| { | |
| "epoch": 3.58, | |
| "learning_rate": 3.581687612208259e-05, | |
| "loss": 0.2241, | |
| "step": 3990 | |
| }, | |
| { | |
| "epoch": 3.59, | |
| "learning_rate": 3.5906642728904846e-05, | |
| "loss": 0.2547, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 3.6, | |
| "learning_rate": 3.599640933572711e-05, | |
| "loss": 0.2471, | |
| "step": 4010 | |
| }, | |
| { | |
| "epoch": 3.61, | |
| "learning_rate": 3.6086175942549375e-05, | |
| "loss": 0.2585, | |
| "step": 4020 | |
| }, | |
| { | |
| "epoch": 3.62, | |
| "learning_rate": 3.617594254937163e-05, | |
| "loss": 0.2617, | |
| "step": 4030 | |
| }, | |
| { | |
| "epoch": 3.62, | |
| "learning_rate": 3.62657091561939e-05, | |
| "loss": 0.2724, | |
| "step": 4040 | |
| }, | |
| { | |
| "epoch": 3.63, | |
| "learning_rate": 3.635547576301616e-05, | |
| "loss": 0.2353, | |
| "step": 4050 | |
| }, | |
| { | |
| "epoch": 3.64, | |
| "learning_rate": 3.644524236983842e-05, | |
| "loss": 0.2288, | |
| "step": 4060 | |
| }, | |
| { | |
| "epoch": 3.65, | |
| "learning_rate": 3.6535008976660684e-05, | |
| "loss": 0.2431, | |
| "step": 4070 | |
| }, | |
| { | |
| "epoch": 3.66, | |
| "learning_rate": 3.662477558348294e-05, | |
| "loss": 0.2693, | |
| "step": 4080 | |
| }, | |
| { | |
| "epoch": 3.67, | |
| "learning_rate": 3.671454219030521e-05, | |
| "loss": 0.2903, | |
| "step": 4090 | |
| }, | |
| { | |
| "epoch": 3.68, | |
| "learning_rate": 3.6804308797127465e-05, | |
| "loss": 0.2658, | |
| "step": 4100 | |
| }, | |
| { | |
| "epoch": 3.69, | |
| "learning_rate": 3.6894075403949736e-05, | |
| "loss": 0.2175, | |
| "step": 4110 | |
| }, | |
| { | |
| "epoch": 3.7, | |
| "learning_rate": 3.6983842010771994e-05, | |
| "loss": 0.3072, | |
| "step": 4120 | |
| }, | |
| { | |
| "epoch": 3.7, | |
| "learning_rate": 3.707360861759426e-05, | |
| "loss": 0.305, | |
| "step": 4130 | |
| }, | |
| { | |
| "epoch": 3.71, | |
| "learning_rate": 3.7163375224416516e-05, | |
| "loss": 0.2475, | |
| "step": 4140 | |
| }, | |
| { | |
| "epoch": 3.72, | |
| "learning_rate": 3.725314183123878e-05, | |
| "loss": 0.2682, | |
| "step": 4150 | |
| }, | |
| { | |
| "epoch": 3.73, | |
| "learning_rate": 3.734290843806104e-05, | |
| "loss": 0.2593, | |
| "step": 4160 | |
| }, | |
| { | |
| "epoch": 3.74, | |
| "learning_rate": 3.74326750448833e-05, | |
| "loss": 0.3293, | |
| "step": 4170 | |
| }, | |
| { | |
| "epoch": 3.75, | |
| "learning_rate": 3.752244165170557e-05, | |
| "loss": 0.2422, | |
| "step": 4180 | |
| }, | |
| { | |
| "epoch": 3.76, | |
| "learning_rate": 3.761220825852783e-05, | |
| "loss": 0.2018, | |
| "step": 4190 | |
| }, | |
| { | |
| "epoch": 3.77, | |
| "learning_rate": 3.770197486535009e-05, | |
| "loss": 0.2802, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 3.78, | |
| "learning_rate": 3.7791741472172355e-05, | |
| "loss": 0.2298, | |
| "step": 4210 | |
| }, | |
| { | |
| "epoch": 3.79, | |
| "learning_rate": 3.788150807899461e-05, | |
| "loss": 0.2512, | |
| "step": 4220 | |
| }, | |
| { | |
| "epoch": 3.79, | |
| "learning_rate": 3.797127468581688e-05, | |
| "loss": 0.2503, | |
| "step": 4230 | |
| }, | |
| { | |
| "epoch": 3.8, | |
| "learning_rate": 3.8061041292639135e-05, | |
| "loss": 0.2244, | |
| "step": 4240 | |
| }, | |
| { | |
| "epoch": 3.81, | |
| "learning_rate": 3.815080789946141e-05, | |
| "loss": 0.2456, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 3.82, | |
| "learning_rate": 3.8240574506283664e-05, | |
| "loss": 0.2085, | |
| "step": 4260 | |
| }, | |
| { | |
| "epoch": 3.83, | |
| "learning_rate": 3.833034111310593e-05, | |
| "loss": 0.2367, | |
| "step": 4270 | |
| }, | |
| { | |
| "epoch": 3.84, | |
| "learning_rate": 3.842010771992819e-05, | |
| "loss": 0.2174, | |
| "step": 4280 | |
| }, | |
| { | |
| "epoch": 3.85, | |
| "learning_rate": 3.850987432675045e-05, | |
| "loss": 0.2256, | |
| "step": 4290 | |
| }, | |
| { | |
| "epoch": 3.86, | |
| "learning_rate": 3.859964093357271e-05, | |
| "loss": 0.2344, | |
| "step": 4300 | |
| }, | |
| { | |
| "epoch": 3.87, | |
| "learning_rate": 3.8689407540394974e-05, | |
| "loss": 0.2107, | |
| "step": 4310 | |
| }, | |
| { | |
| "epoch": 3.88, | |
| "learning_rate": 3.877917414721723e-05, | |
| "loss": 0.2797, | |
| "step": 4320 | |
| }, | |
| { | |
| "epoch": 3.88, | |
| "learning_rate": 3.88689407540395e-05, | |
| "loss": 0.2568, | |
| "step": 4330 | |
| }, | |
| { | |
| "epoch": 3.89, | |
| "learning_rate": 3.895870736086176e-05, | |
| "loss": 0.2312, | |
| "step": 4340 | |
| }, | |
| { | |
| "epoch": 3.9, | |
| "learning_rate": 3.9048473967684026e-05, | |
| "loss": 0.2269, | |
| "step": 4350 | |
| }, | |
| { | |
| "epoch": 3.91, | |
| "learning_rate": 3.9138240574506283e-05, | |
| "loss": 0.2412, | |
| "step": 4360 | |
| }, | |
| { | |
| "epoch": 3.92, | |
| "learning_rate": 3.922800718132855e-05, | |
| "loss": 0.2626, | |
| "step": 4370 | |
| }, | |
| { | |
| "epoch": 3.93, | |
| "learning_rate": 3.9317773788150806e-05, | |
| "loss": 0.2995, | |
| "step": 4380 | |
| }, | |
| { | |
| "epoch": 3.94, | |
| "learning_rate": 3.940754039497307e-05, | |
| "loss": 0.2435, | |
| "step": 4390 | |
| }, | |
| { | |
| "epoch": 3.95, | |
| "learning_rate": 3.9497307001795335e-05, | |
| "loss": 0.2317, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 3.96, | |
| "learning_rate": 3.95870736086176e-05, | |
| "loss": 0.1932, | |
| "step": 4410 | |
| }, | |
| { | |
| "epoch": 3.97, | |
| "learning_rate": 3.967684021543986e-05, | |
| "loss": 0.2498, | |
| "step": 4420 | |
| }, | |
| { | |
| "epoch": 3.97, | |
| "learning_rate": 3.976660682226212e-05, | |
| "loss": 0.2792, | |
| "step": 4430 | |
| }, | |
| { | |
| "epoch": 3.98, | |
| "learning_rate": 3.985637342908438e-05, | |
| "loss": 0.2554, | |
| "step": 4440 | |
| }, | |
| { | |
| "epoch": 3.99, | |
| "learning_rate": 3.9946140035906645e-05, | |
| "loss": 0.3403, | |
| "step": 4450 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": { | |
| "accuracy": 0.9382148463781117 | |
| }, | |
| "eval_f1": { | |
| "f1": 0.9358183038090213 | |
| }, | |
| "eval_loss": 0.18279214203357697, | |
| "eval_precision": { | |
| "precision": 0.9351989305945833 | |
| }, | |
| "eval_recall": { | |
| "recall": 0.9374254756058774 | |
| }, | |
| "eval_runtime": 165.3018, | |
| "eval_samples_per_second": 107.9, | |
| "eval_steps_per_second": 6.745, | |
| "step": 4459 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 4.00359066427289e-05, | |
| "loss": 0.2936, | |
| "step": 4460 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "learning_rate": 4.0125673249551174e-05, | |
| "loss": 0.2723, | |
| "step": 4470 | |
| }, | |
| { | |
| "epoch": 4.02, | |
| "learning_rate": 4.021543985637343e-05, | |
| "loss": 0.2307, | |
| "step": 4480 | |
| }, | |
| { | |
| "epoch": 4.03, | |
| "learning_rate": 4.0305206463195696e-05, | |
| "loss": 0.2567, | |
| "step": 4490 | |
| }, | |
| { | |
| "epoch": 4.04, | |
| "learning_rate": 4.0394973070017954e-05, | |
| "loss": 0.2273, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 4.05, | |
| "learning_rate": 4.048473967684022e-05, | |
| "loss": 0.2638, | |
| "step": 4510 | |
| }, | |
| { | |
| "epoch": 4.05, | |
| "learning_rate": 4.0574506283662476e-05, | |
| "loss": 0.2625, | |
| "step": 4520 | |
| }, | |
| { | |
| "epoch": 4.06, | |
| "learning_rate": 4.066427289048474e-05, | |
| "loss": 0.2432, | |
| "step": 4530 | |
| }, | |
| { | |
| "epoch": 4.07, | |
| "learning_rate": 4.0754039497307006e-05, | |
| "loss": 0.2345, | |
| "step": 4540 | |
| }, | |
| { | |
| "epoch": 4.08, | |
| "learning_rate": 4.084380610412927e-05, | |
| "loss": 0.2272, | |
| "step": 4550 | |
| }, | |
| { | |
| "epoch": 4.09, | |
| "learning_rate": 4.093357271095153e-05, | |
| "loss": 0.2217, | |
| "step": 4560 | |
| }, | |
| { | |
| "epoch": 4.1, | |
| "learning_rate": 4.102333931777379e-05, | |
| "loss": 0.2195, | |
| "step": 4570 | |
| }, | |
| { | |
| "epoch": 4.11, | |
| "learning_rate": 4.111310592459605e-05, | |
| "loss": 0.2637, | |
| "step": 4580 | |
| }, | |
| { | |
| "epoch": 4.12, | |
| "learning_rate": 4.1202872531418315e-05, | |
| "loss": 0.2668, | |
| "step": 4590 | |
| }, | |
| { | |
| "epoch": 4.13, | |
| "learning_rate": 4.129263913824057e-05, | |
| "loss": 0.2742, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 4.14, | |
| "learning_rate": 4.1382405745062844e-05, | |
| "loss": 0.2128, | |
| "step": 4610 | |
| }, | |
| { | |
| "epoch": 4.14, | |
| "learning_rate": 4.14721723518851e-05, | |
| "loss": 0.241, | |
| "step": 4620 | |
| }, | |
| { | |
| "epoch": 4.15, | |
| "learning_rate": 4.156193895870737e-05, | |
| "loss": 0.2332, | |
| "step": 4630 | |
| }, | |
| { | |
| "epoch": 4.16, | |
| "learning_rate": 4.1651705565529625e-05, | |
| "loss": 0.2359, | |
| "step": 4640 | |
| }, | |
| { | |
| "epoch": 4.17, | |
| "learning_rate": 4.174147217235189e-05, | |
| "loss": 0.2219, | |
| "step": 4650 | |
| }, | |
| { | |
| "epoch": 4.18, | |
| "learning_rate": 4.183123877917415e-05, | |
| "loss": 0.2325, | |
| "step": 4660 | |
| }, | |
| { | |
| "epoch": 4.19, | |
| "learning_rate": 4.192100538599641e-05, | |
| "loss": 0.2627, | |
| "step": 4670 | |
| }, | |
| { | |
| "epoch": 4.2, | |
| "learning_rate": 4.2010771992818676e-05, | |
| "loss": 0.196, | |
| "step": 4680 | |
| }, | |
| { | |
| "epoch": 4.21, | |
| "learning_rate": 4.2100538599640934e-05, | |
| "loss": 0.234, | |
| "step": 4690 | |
| }, | |
| { | |
| "epoch": 4.22, | |
| "learning_rate": 4.21903052064632e-05, | |
| "loss": 0.1998, | |
| "step": 4700 | |
| }, | |
| { | |
| "epoch": 4.23, | |
| "learning_rate": 4.228007181328546e-05, | |
| "loss": 0.2579, | |
| "step": 4710 | |
| }, | |
| { | |
| "epoch": 4.23, | |
| "learning_rate": 4.236983842010772e-05, | |
| "loss": 0.2678, | |
| "step": 4720 | |
| }, | |
| { | |
| "epoch": 4.24, | |
| "learning_rate": 4.2459605026929986e-05, | |
| "loss": 0.2431, | |
| "step": 4730 | |
| }, | |
| { | |
| "epoch": 4.25, | |
| "learning_rate": 4.2549371633752244e-05, | |
| "loss": 0.2274, | |
| "step": 4740 | |
| }, | |
| { | |
| "epoch": 4.26, | |
| "learning_rate": 4.263913824057451e-05, | |
| "loss": 0.2325, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 4.27, | |
| "learning_rate": 4.272890484739677e-05, | |
| "loss": 0.2121, | |
| "step": 4760 | |
| }, | |
| { | |
| "epoch": 4.28, | |
| "learning_rate": 4.281867145421903e-05, | |
| "loss": 0.2466, | |
| "step": 4770 | |
| }, | |
| { | |
| "epoch": 4.29, | |
| "learning_rate": 4.2908438061041295e-05, | |
| "loss": 0.2331, | |
| "step": 4780 | |
| }, | |
| { | |
| "epoch": 4.3, | |
| "learning_rate": 4.299820466786356e-05, | |
| "loss": 0.221, | |
| "step": 4790 | |
| }, | |
| { | |
| "epoch": 4.31, | |
| "learning_rate": 4.308797127468582e-05, | |
| "loss": 0.2563, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 4.31, | |
| "learning_rate": 4.317773788150808e-05, | |
| "loss": 0.2564, | |
| "step": 4810 | |
| }, | |
| { | |
| "epoch": 4.32, | |
| "learning_rate": 4.326750448833034e-05, | |
| "loss": 0.2369, | |
| "step": 4820 | |
| }, | |
| { | |
| "epoch": 4.33, | |
| "learning_rate": 4.3357271095152605e-05, | |
| "loss": 0.1929, | |
| "step": 4830 | |
| }, | |
| { | |
| "epoch": 4.34, | |
| "learning_rate": 4.344703770197487e-05, | |
| "loss": 0.2639, | |
| "step": 4840 | |
| }, | |
| { | |
| "epoch": 4.35, | |
| "learning_rate": 4.353680430879713e-05, | |
| "loss": 0.232, | |
| "step": 4850 | |
| }, | |
| { | |
| "epoch": 4.36, | |
| "learning_rate": 4.362657091561939e-05, | |
| "loss": 0.2453, | |
| "step": 4860 | |
| }, | |
| { | |
| "epoch": 4.37, | |
| "learning_rate": 4.371633752244165e-05, | |
| "loss": 0.2145, | |
| "step": 4870 | |
| }, | |
| { | |
| "epoch": 4.38, | |
| "learning_rate": 4.3806104129263914e-05, | |
| "loss": 0.2249, | |
| "step": 4880 | |
| }, | |
| { | |
| "epoch": 4.39, | |
| "learning_rate": 4.389587073608618e-05, | |
| "loss": 0.2861, | |
| "step": 4890 | |
| }, | |
| { | |
| "epoch": 4.4, | |
| "learning_rate": 4.398563734290844e-05, | |
| "loss": 0.2452, | |
| "step": 4900 | |
| }, | |
| { | |
| "epoch": 4.4, | |
| "learning_rate": 4.40754039497307e-05, | |
| "loss": 0.2512, | |
| "step": 4910 | |
| }, | |
| { | |
| "epoch": 4.41, | |
| "learning_rate": 4.4165170556552966e-05, | |
| "loss": 0.2373, | |
| "step": 4920 | |
| }, | |
| { | |
| "epoch": 4.42, | |
| "learning_rate": 4.4254937163375224e-05, | |
| "loss": 0.2262, | |
| "step": 4930 | |
| }, | |
| { | |
| "epoch": 4.43, | |
| "learning_rate": 4.434470377019749e-05, | |
| "loss": 0.2243, | |
| "step": 4940 | |
| }, | |
| { | |
| "epoch": 4.44, | |
| "learning_rate": 4.4434470377019746e-05, | |
| "loss": 0.2228, | |
| "step": 4950 | |
| }, | |
| { | |
| "epoch": 4.45, | |
| "learning_rate": 4.452423698384201e-05, | |
| "loss": 0.2331, | |
| "step": 4960 | |
| }, | |
| { | |
| "epoch": 4.46, | |
| "learning_rate": 4.4614003590664275e-05, | |
| "loss": 0.2706, | |
| "step": 4970 | |
| }, | |
| { | |
| "epoch": 4.47, | |
| "learning_rate": 4.470377019748654e-05, | |
| "loss": 0.2326, | |
| "step": 4980 | |
| }, | |
| { | |
| "epoch": 4.48, | |
| "learning_rate": 4.47935368043088e-05, | |
| "loss": 0.3095, | |
| "step": 4990 | |
| }, | |
| { | |
| "epoch": 4.49, | |
| "learning_rate": 4.488330341113106e-05, | |
| "loss": 0.2106, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 4.49, | |
| "learning_rate": 4.497307001795332e-05, | |
| "loss": 0.2494, | |
| "step": 5010 | |
| }, | |
| { | |
| "epoch": 4.5, | |
| "learning_rate": 4.5062836624775585e-05, | |
| "loss": 0.234, | |
| "step": 5020 | |
| }, | |
| { | |
| "epoch": 4.51, | |
| "learning_rate": 4.515260323159784e-05, | |
| "loss": 0.283, | |
| "step": 5030 | |
| }, | |
| { | |
| "epoch": 4.52, | |
| "learning_rate": 4.5242369838420114e-05, | |
| "loss": 0.2626, | |
| "step": 5040 | |
| }, | |
| { | |
| "epoch": 4.53, | |
| "learning_rate": 4.533213644524237e-05, | |
| "loss": 0.2352, | |
| "step": 5050 | |
| }, | |
| { | |
| "epoch": 4.54, | |
| "learning_rate": 4.5421903052064636e-05, | |
| "loss": 0.2748, | |
| "step": 5060 | |
| }, | |
| { | |
| "epoch": 4.55, | |
| "learning_rate": 4.5511669658886894e-05, | |
| "loss": 0.2849, | |
| "step": 5070 | |
| }, | |
| { | |
| "epoch": 4.56, | |
| "learning_rate": 4.560143626570916e-05, | |
| "loss": 0.2296, | |
| "step": 5080 | |
| }, | |
| { | |
| "epoch": 4.57, | |
| "learning_rate": 4.5691202872531417e-05, | |
| "loss": 0.2231, | |
| "step": 5090 | |
| }, | |
| { | |
| "epoch": 4.58, | |
| "learning_rate": 4.578096947935368e-05, | |
| "loss": 0.2526, | |
| "step": 5100 | |
| }, | |
| { | |
| "epoch": 4.58, | |
| "learning_rate": 4.5870736086175946e-05, | |
| "loss": 0.2441, | |
| "step": 5110 | |
| }, | |
| { | |
| "epoch": 4.59, | |
| "learning_rate": 4.596050269299821e-05, | |
| "loss": 0.236, | |
| "step": 5120 | |
| }, | |
| { | |
| "epoch": 4.6, | |
| "learning_rate": 4.605026929982047e-05, | |
| "loss": 0.2485, | |
| "step": 5130 | |
| }, | |
| { | |
| "epoch": 4.61, | |
| "learning_rate": 4.614003590664273e-05, | |
| "loss": 0.1969, | |
| "step": 5140 | |
| }, | |
| { | |
| "epoch": 4.62, | |
| "learning_rate": 4.622980251346499e-05, | |
| "loss": 0.2454, | |
| "step": 5150 | |
| }, | |
| { | |
| "epoch": 4.63, | |
| "learning_rate": 4.6319569120287255e-05, | |
| "loss": 0.2486, | |
| "step": 5160 | |
| }, | |
| { | |
| "epoch": 4.64, | |
| "learning_rate": 4.640933572710951e-05, | |
| "loss": 0.2304, | |
| "step": 5170 | |
| }, | |
| { | |
| "epoch": 4.65, | |
| "learning_rate": 4.6499102333931784e-05, | |
| "loss": 0.2321, | |
| "step": 5180 | |
| }, | |
| { | |
| "epoch": 4.66, | |
| "learning_rate": 4.658886894075404e-05, | |
| "loss": 0.2205, | |
| "step": 5190 | |
| }, | |
| { | |
| "epoch": 4.66, | |
| "learning_rate": 4.667863554757631e-05, | |
| "loss": 0.2107, | |
| "step": 5200 | |
| }, | |
| { | |
| "epoch": 4.67, | |
| "learning_rate": 4.6768402154398565e-05, | |
| "loss": 0.2215, | |
| "step": 5210 | |
| }, | |
| { | |
| "epoch": 4.68, | |
| "learning_rate": 4.685816876122083e-05, | |
| "loss": 0.222, | |
| "step": 5220 | |
| }, | |
| { | |
| "epoch": 4.69, | |
| "learning_rate": 4.694793536804309e-05, | |
| "loss": 0.2271, | |
| "step": 5230 | |
| }, | |
| { | |
| "epoch": 4.7, | |
| "learning_rate": 4.703770197486535e-05, | |
| "loss": 0.2342, | |
| "step": 5240 | |
| }, | |
| { | |
| "epoch": 4.71, | |
| "learning_rate": 4.7127468581687616e-05, | |
| "loss": 0.2478, | |
| "step": 5250 | |
| }, | |
| { | |
| "epoch": 4.72, | |
| "learning_rate": 4.721723518850988e-05, | |
| "loss": 0.2691, | |
| "step": 5260 | |
| }, | |
| { | |
| "epoch": 4.73, | |
| "learning_rate": 4.730700179533214e-05, | |
| "loss": 0.2059, | |
| "step": 5270 | |
| }, | |
| { | |
| "epoch": 4.74, | |
| "learning_rate": 4.7396768402154403e-05, | |
| "loss": 0.2085, | |
| "step": 5280 | |
| }, | |
| { | |
| "epoch": 4.75, | |
| "learning_rate": 4.748653500897666e-05, | |
| "loss": 0.1888, | |
| "step": 5290 | |
| }, | |
| { | |
| "epoch": 4.75, | |
| "learning_rate": 4.7576301615798926e-05, | |
| "loss": 0.2933, | |
| "step": 5300 | |
| }, | |
| { | |
| "epoch": 4.76, | |
| "learning_rate": 4.7666068222621184e-05, | |
| "loss": 0.2519, | |
| "step": 5310 | |
| }, | |
| { | |
| "epoch": 4.77, | |
| "learning_rate": 4.775583482944345e-05, | |
| "loss": 0.2142, | |
| "step": 5320 | |
| }, | |
| { | |
| "epoch": 4.78, | |
| "learning_rate": 4.784560143626571e-05, | |
| "loss": 0.1972, | |
| "step": 5330 | |
| }, | |
| { | |
| "epoch": 4.79, | |
| "learning_rate": 4.793536804308798e-05, | |
| "loss": 0.2134, | |
| "step": 5340 | |
| }, | |
| { | |
| "epoch": 4.8, | |
| "learning_rate": 4.8025134649910235e-05, | |
| "loss": 0.2924, | |
| "step": 5350 | |
| }, | |
| { | |
| "epoch": 4.81, | |
| "learning_rate": 4.81149012567325e-05, | |
| "loss": 0.2207, | |
| "step": 5360 | |
| }, | |
| { | |
| "epoch": 4.82, | |
| "learning_rate": 4.820466786355476e-05, | |
| "loss": 0.2369, | |
| "step": 5370 | |
| }, | |
| { | |
| "epoch": 4.83, | |
| "learning_rate": 4.829443447037702e-05, | |
| "loss": 0.242, | |
| "step": 5380 | |
| }, | |
| { | |
| "epoch": 4.84, | |
| "learning_rate": 4.838420107719928e-05, | |
| "loss": 0.2023, | |
| "step": 5390 | |
| }, | |
| { | |
| "epoch": 4.84, | |
| "learning_rate": 4.847396768402155e-05, | |
| "loss": 0.2413, | |
| "step": 5400 | |
| }, | |
| { | |
| "epoch": 4.85, | |
| "learning_rate": 4.856373429084381e-05, | |
| "loss": 0.2757, | |
| "step": 5410 | |
| }, | |
| { | |
| "epoch": 4.86, | |
| "learning_rate": 4.8653500897666074e-05, | |
| "loss": 0.2481, | |
| "step": 5420 | |
| }, | |
| { | |
| "epoch": 4.87, | |
| "learning_rate": 4.874326750448833e-05, | |
| "loss": 0.2487, | |
| "step": 5430 | |
| }, | |
| { | |
| "epoch": 4.88, | |
| "learning_rate": 4.8833034111310596e-05, | |
| "loss": 0.1747, | |
| "step": 5440 | |
| }, | |
| { | |
| "epoch": 4.89, | |
| "learning_rate": 4.8922800718132854e-05, | |
| "loss": 0.2537, | |
| "step": 5450 | |
| }, | |
| { | |
| "epoch": 4.9, | |
| "learning_rate": 4.901256732495512e-05, | |
| "loss": 0.2722, | |
| "step": 5460 | |
| }, | |
| { | |
| "epoch": 4.91, | |
| "learning_rate": 4.9102333931777383e-05, | |
| "loss": 0.2225, | |
| "step": 5470 | |
| }, | |
| { | |
| "epoch": 4.92, | |
| "learning_rate": 4.919210053859965e-05, | |
| "loss": 0.1949, | |
| "step": 5480 | |
| }, | |
| { | |
| "epoch": 4.92, | |
| "learning_rate": 4.9281867145421906e-05, | |
| "loss": 0.2416, | |
| "step": 5490 | |
| }, | |
| { | |
| "epoch": 4.93, | |
| "learning_rate": 4.937163375224417e-05, | |
| "loss": 0.2197, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 4.94, | |
| "learning_rate": 4.946140035906643e-05, | |
| "loss": 0.2939, | |
| "step": 5510 | |
| }, | |
| { | |
| "epoch": 4.95, | |
| "learning_rate": 4.955116696588869e-05, | |
| "loss": 0.2786, | |
| "step": 5520 | |
| }, | |
| { | |
| "epoch": 4.96, | |
| "learning_rate": 4.964093357271095e-05, | |
| "loss": 0.2187, | |
| "step": 5530 | |
| }, | |
| { | |
| "epoch": 4.97, | |
| "learning_rate": 4.9730700179533215e-05, | |
| "loss": 0.2518, | |
| "step": 5540 | |
| }, | |
| { | |
| "epoch": 4.98, | |
| "learning_rate": 4.982046678635548e-05, | |
| "loss": 0.205, | |
| "step": 5550 | |
| }, | |
| { | |
| "epoch": 4.99, | |
| "learning_rate": 4.991023339317774e-05, | |
| "loss": 0.2277, | |
| "step": 5560 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "learning_rate": 5e-05, | |
| "loss": 0.2399, | |
| "step": 5570 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": { | |
| "accuracy": 0.9539694998878673 | |
| }, | |
| "eval_f1": { | |
| "f1": 0.9516195088904471 | |
| }, | |
| "eval_loss": 0.1283472627401352, | |
| "eval_precision": { | |
| "precision": 0.9525864748624857 | |
| }, | |
| "eval_recall": { | |
| "recall": 0.9506985041313745 | |
| }, | |
| "eval_runtime": 164.7789, | |
| "eval_samples_per_second": 108.242, | |
| "eval_steps_per_second": 6.767, | |
| "step": 5573 | |
| }, | |
| { | |
| "epoch": 5.01, | |
| "learning_rate": 4.999002593257531e-05, | |
| "loss": 0.2288, | |
| "step": 5580 | |
| }, | |
| { | |
| "epoch": 5.01, | |
| "learning_rate": 4.9980051865150615e-05, | |
| "loss": 0.2216, | |
| "step": 5590 | |
| }, | |
| { | |
| "epoch": 5.02, | |
| "learning_rate": 4.9970077797725914e-05, | |
| "loss": 0.3012, | |
| "step": 5600 | |
| }, | |
| { | |
| "epoch": 5.03, | |
| "learning_rate": 4.9960103730301214e-05, | |
| "loss": 0.1937, | |
| "step": 5610 | |
| }, | |
| { | |
| "epoch": 5.04, | |
| "learning_rate": 4.995012966287652e-05, | |
| "loss": 0.1837, | |
| "step": 5620 | |
| }, | |
| { | |
| "epoch": 5.05, | |
| "learning_rate": 4.9940155595451826e-05, | |
| "loss": 0.2226, | |
| "step": 5630 | |
| }, | |
| { | |
| "epoch": 5.06, | |
| "learning_rate": 4.993018152802713e-05, | |
| "loss": 0.2332, | |
| "step": 5640 | |
| }, | |
| { | |
| "epoch": 5.07, | |
| "learning_rate": 4.992020746060244e-05, | |
| "loss": 0.2046, | |
| "step": 5650 | |
| }, | |
| { | |
| "epoch": 5.08, | |
| "learning_rate": 4.991023339317774e-05, | |
| "loss": 0.219, | |
| "step": 5660 | |
| }, | |
| { | |
| "epoch": 5.09, | |
| "learning_rate": 4.9900259325753044e-05, | |
| "loss": 0.2049, | |
| "step": 5670 | |
| }, | |
| { | |
| "epoch": 5.1, | |
| "learning_rate": 4.989028525832835e-05, | |
| "loss": 0.2299, | |
| "step": 5680 | |
| }, | |
| { | |
| "epoch": 5.1, | |
| "learning_rate": 4.988031119090365e-05, | |
| "loss": 0.2195, | |
| "step": 5690 | |
| }, | |
| { | |
| "epoch": 5.11, | |
| "learning_rate": 4.9870337123478956e-05, | |
| "loss": 0.221, | |
| "step": 5700 | |
| }, | |
| { | |
| "epoch": 5.12, | |
| "learning_rate": 4.986036305605426e-05, | |
| "loss": 0.2159, | |
| "step": 5710 | |
| }, | |
| { | |
| "epoch": 5.13, | |
| "learning_rate": 4.985038898862957e-05, | |
| "loss": 0.2517, | |
| "step": 5720 | |
| }, | |
| { | |
| "epoch": 5.14, | |
| "learning_rate": 4.984041492120487e-05, | |
| "loss": 0.1899, | |
| "step": 5730 | |
| }, | |
| { | |
| "epoch": 5.15, | |
| "learning_rate": 4.9830440853780174e-05, | |
| "loss": 0.188, | |
| "step": 5740 | |
| }, | |
| { | |
| "epoch": 5.16, | |
| "learning_rate": 4.982046678635548e-05, | |
| "loss": 0.2629, | |
| "step": 5750 | |
| }, | |
| { | |
| "epoch": 5.17, | |
| "learning_rate": 4.9810492718930786e-05, | |
| "loss": 0.1955, | |
| "step": 5760 | |
| }, | |
| { | |
| "epoch": 5.18, | |
| "learning_rate": 4.9800518651506086e-05, | |
| "loss": 0.211, | |
| "step": 5770 | |
| }, | |
| { | |
| "epoch": 5.19, | |
| "learning_rate": 4.979054458408139e-05, | |
| "loss": 0.2094, | |
| "step": 5780 | |
| }, | |
| { | |
| "epoch": 5.19, | |
| "learning_rate": 4.978057051665669e-05, | |
| "loss": 0.1961, | |
| "step": 5790 | |
| }, | |
| { | |
| "epoch": 5.2, | |
| "learning_rate": 4.9770596449232e-05, | |
| "loss": 0.2094, | |
| "step": 5800 | |
| }, | |
| { | |
| "epoch": 5.21, | |
| "learning_rate": 4.9760622381807304e-05, | |
| "loss": 0.1873, | |
| "step": 5810 | |
| }, | |
| { | |
| "epoch": 5.22, | |
| "learning_rate": 4.975064831438261e-05, | |
| "loss": 0.2139, | |
| "step": 5820 | |
| }, | |
| { | |
| "epoch": 5.23, | |
| "learning_rate": 4.9740674246957916e-05, | |
| "loss": 0.2314, | |
| "step": 5830 | |
| }, | |
| { | |
| "epoch": 5.24, | |
| "learning_rate": 4.9730700179533215e-05, | |
| "loss": 0.2177, | |
| "step": 5840 | |
| }, | |
| { | |
| "epoch": 5.25, | |
| "learning_rate": 4.9720726112108515e-05, | |
| "loss": 0.2068, | |
| "step": 5850 | |
| }, | |
| { | |
| "epoch": 5.26, | |
| "learning_rate": 4.971075204468382e-05, | |
| "loss": 0.1825, | |
| "step": 5860 | |
| }, | |
| { | |
| "epoch": 5.27, | |
| "learning_rate": 4.970077797725913e-05, | |
| "loss": 0.2621, | |
| "step": 5870 | |
| }, | |
| { | |
| "epoch": 5.27, | |
| "learning_rate": 4.969080390983443e-05, | |
| "loss": 0.1762, | |
| "step": 5880 | |
| }, | |
| { | |
| "epoch": 5.28, | |
| "learning_rate": 4.968082984240974e-05, | |
| "loss": 0.2026, | |
| "step": 5890 | |
| }, | |
| { | |
| "epoch": 5.29, | |
| "learning_rate": 4.967085577498504e-05, | |
| "loss": 0.1896, | |
| "step": 5900 | |
| }, | |
| { | |
| "epoch": 5.3, | |
| "learning_rate": 4.9660881707560345e-05, | |
| "loss": 0.2083, | |
| "step": 5910 | |
| }, | |
| { | |
| "epoch": 5.31, | |
| "learning_rate": 4.965090764013565e-05, | |
| "loss": 0.2006, | |
| "step": 5920 | |
| }, | |
| { | |
| "epoch": 5.32, | |
| "learning_rate": 4.964093357271095e-05, | |
| "loss": 0.2536, | |
| "step": 5930 | |
| }, | |
| { | |
| "epoch": 5.33, | |
| "learning_rate": 4.963095950528626e-05, | |
| "loss": 0.233, | |
| "step": 5940 | |
| }, | |
| { | |
| "epoch": 5.34, | |
| "learning_rate": 4.962098543786156e-05, | |
| "loss": 0.1991, | |
| "step": 5950 | |
| }, | |
| { | |
| "epoch": 5.35, | |
| "learning_rate": 4.961101137043687e-05, | |
| "loss": 0.1913, | |
| "step": 5960 | |
| }, | |
| { | |
| "epoch": 5.36, | |
| "learning_rate": 4.960103730301217e-05, | |
| "loss": 0.2553, | |
| "step": 5970 | |
| }, | |
| { | |
| "epoch": 5.36, | |
| "learning_rate": 4.9591063235587475e-05, | |
| "loss": 0.22, | |
| "step": 5980 | |
| }, | |
| { | |
| "epoch": 5.37, | |
| "learning_rate": 4.958108916816278e-05, | |
| "loss": 0.1851, | |
| "step": 5990 | |
| }, | |
| { | |
| "epoch": 5.38, | |
| "learning_rate": 4.957111510073809e-05, | |
| "loss": 0.2366, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 5.39, | |
| "learning_rate": 4.956114103331339e-05, | |
| "loss": 0.1834, | |
| "step": 6010 | |
| }, | |
| { | |
| "epoch": 5.4, | |
| "learning_rate": 4.955116696588869e-05, | |
| "loss": 0.1908, | |
| "step": 6020 | |
| }, | |
| { | |
| "epoch": 5.41, | |
| "learning_rate": 4.954119289846399e-05, | |
| "loss": 0.1859, | |
| "step": 6030 | |
| }, | |
| { | |
| "epoch": 5.42, | |
| "learning_rate": 4.95312188310393e-05, | |
| "loss": 0.2187, | |
| "step": 6040 | |
| }, | |
| { | |
| "epoch": 5.43, | |
| "learning_rate": 4.9521244763614605e-05, | |
| "loss": 0.1675, | |
| "step": 6050 | |
| }, | |
| { | |
| "epoch": 5.44, | |
| "learning_rate": 4.951127069618991e-05, | |
| "loss": 0.227, | |
| "step": 6060 | |
| }, | |
| { | |
| "epoch": 5.45, | |
| "learning_rate": 4.950129662876522e-05, | |
| "loss": 0.2286, | |
| "step": 6070 | |
| }, | |
| { | |
| "epoch": 5.45, | |
| "learning_rate": 4.9491322561340517e-05, | |
| "loss": 0.2063, | |
| "step": 6080 | |
| }, | |
| { | |
| "epoch": 5.46, | |
| "learning_rate": 4.9481348493915816e-05, | |
| "loss": 0.193, | |
| "step": 6090 | |
| }, | |
| { | |
| "epoch": 5.47, | |
| "learning_rate": 4.947137442649112e-05, | |
| "loss": 0.1823, | |
| "step": 6100 | |
| }, | |
| { | |
| "epoch": 5.48, | |
| "learning_rate": 4.946140035906643e-05, | |
| "loss": 0.2303, | |
| "step": 6110 | |
| }, | |
| { | |
| "epoch": 5.49, | |
| "learning_rate": 4.9451426291641735e-05, | |
| "loss": 0.1933, | |
| "step": 6120 | |
| }, | |
| { | |
| "epoch": 5.5, | |
| "learning_rate": 4.944145222421704e-05, | |
| "loss": 0.2238, | |
| "step": 6130 | |
| }, | |
| { | |
| "epoch": 5.51, | |
| "learning_rate": 4.943147815679234e-05, | |
| "loss": 0.1762, | |
| "step": 6140 | |
| }, | |
| { | |
| "epoch": 5.52, | |
| "learning_rate": 4.9421504089367646e-05, | |
| "loss": 0.2441, | |
| "step": 6150 | |
| }, | |
| { | |
| "epoch": 5.53, | |
| "learning_rate": 4.941153002194295e-05, | |
| "loss": 0.1548, | |
| "step": 6160 | |
| }, | |
| { | |
| "epoch": 5.53, | |
| "learning_rate": 4.940155595451825e-05, | |
| "loss": 0.2052, | |
| "step": 6170 | |
| }, | |
| { | |
| "epoch": 5.54, | |
| "learning_rate": 4.939158188709356e-05, | |
| "loss": 0.1747, | |
| "step": 6180 | |
| }, | |
| { | |
| "epoch": 5.55, | |
| "learning_rate": 4.9381607819668864e-05, | |
| "loss": 0.1918, | |
| "step": 6190 | |
| }, | |
| { | |
| "epoch": 5.56, | |
| "learning_rate": 4.937163375224417e-05, | |
| "loss": 0.1776, | |
| "step": 6200 | |
| }, | |
| { | |
| "epoch": 5.57, | |
| "learning_rate": 4.936165968481947e-05, | |
| "loss": 0.2143, | |
| "step": 6210 | |
| }, | |
| { | |
| "epoch": 5.58, | |
| "learning_rate": 4.9351685617394776e-05, | |
| "loss": 0.2272, | |
| "step": 6220 | |
| }, | |
| { | |
| "epoch": 5.59, | |
| "learning_rate": 4.934171154997008e-05, | |
| "loss": 0.2238, | |
| "step": 6230 | |
| }, | |
| { | |
| "epoch": 5.6, | |
| "learning_rate": 4.933173748254539e-05, | |
| "loss": 0.215, | |
| "step": 6240 | |
| }, | |
| { | |
| "epoch": 5.61, | |
| "learning_rate": 4.932176341512069e-05, | |
| "loss": 0.2505, | |
| "step": 6250 | |
| }, | |
| { | |
| "epoch": 5.62, | |
| "learning_rate": 4.9311789347695994e-05, | |
| "loss": 0.193, | |
| "step": 6260 | |
| }, | |
| { | |
| "epoch": 5.62, | |
| "learning_rate": 4.9301815280271294e-05, | |
| "loss": 0.2112, | |
| "step": 6270 | |
| }, | |
| { | |
| "epoch": 5.63, | |
| "learning_rate": 4.92918412128466e-05, | |
| "loss": 0.1776, | |
| "step": 6280 | |
| }, | |
| { | |
| "epoch": 5.64, | |
| "learning_rate": 4.9281867145421906e-05, | |
| "loss": 0.1934, | |
| "step": 6290 | |
| }, | |
| { | |
| "epoch": 5.65, | |
| "learning_rate": 4.927189307799721e-05, | |
| "loss": 0.1629, | |
| "step": 6300 | |
| }, | |
| { | |
| "epoch": 5.66, | |
| "learning_rate": 4.926191901057252e-05, | |
| "loss": 0.2348, | |
| "step": 6310 | |
| }, | |
| { | |
| "epoch": 5.67, | |
| "learning_rate": 4.925194494314782e-05, | |
| "loss": 0.2408, | |
| "step": 6320 | |
| }, | |
| { | |
| "epoch": 5.68, | |
| "learning_rate": 4.924197087572312e-05, | |
| "loss": 0.1725, | |
| "step": 6330 | |
| }, | |
| { | |
| "epoch": 5.69, | |
| "learning_rate": 4.923199680829842e-05, | |
| "loss": 0.1752, | |
| "step": 6340 | |
| }, | |
| { | |
| "epoch": 5.7, | |
| "learning_rate": 4.922202274087373e-05, | |
| "loss": 0.185, | |
| "step": 6350 | |
| }, | |
| { | |
| "epoch": 5.71, | |
| "learning_rate": 4.9212048673449036e-05, | |
| "loss": 0.2098, | |
| "step": 6360 | |
| }, | |
| { | |
| "epoch": 5.71, | |
| "learning_rate": 4.920207460602434e-05, | |
| "loss": 0.2003, | |
| "step": 6370 | |
| }, | |
| { | |
| "epoch": 5.72, | |
| "learning_rate": 4.919210053859965e-05, | |
| "loss": 0.2272, | |
| "step": 6380 | |
| }, | |
| { | |
| "epoch": 5.73, | |
| "learning_rate": 4.918212647117495e-05, | |
| "loss": 0.2392, | |
| "step": 6390 | |
| }, | |
| { | |
| "epoch": 5.74, | |
| "learning_rate": 4.917215240375025e-05, | |
| "loss": 0.2252, | |
| "step": 6400 | |
| }, | |
| { | |
| "epoch": 5.75, | |
| "learning_rate": 4.916217833632555e-05, | |
| "loss": 0.1832, | |
| "step": 6410 | |
| }, | |
| { | |
| "epoch": 5.76, | |
| "learning_rate": 4.915220426890086e-05, | |
| "loss": 0.1667, | |
| "step": 6420 | |
| }, | |
| { | |
| "epoch": 5.77, | |
| "learning_rate": 4.9142230201476165e-05, | |
| "loss": 0.2217, | |
| "step": 6430 | |
| }, | |
| { | |
| "epoch": 5.78, | |
| "learning_rate": 4.913225613405147e-05, | |
| "loss": 0.1898, | |
| "step": 6440 | |
| }, | |
| { | |
| "epoch": 5.79, | |
| "learning_rate": 4.912228206662677e-05, | |
| "loss": 0.2141, | |
| "step": 6450 | |
| }, | |
| { | |
| "epoch": 5.8, | |
| "learning_rate": 4.911230799920208e-05, | |
| "loss": 0.1524, | |
| "step": 6460 | |
| }, | |
| { | |
| "epoch": 5.8, | |
| "learning_rate": 4.9102333931777383e-05, | |
| "loss": 0.1946, | |
| "step": 6470 | |
| }, | |
| { | |
| "epoch": 5.81, | |
| "learning_rate": 4.909235986435268e-05, | |
| "loss": 0.2343, | |
| "step": 6480 | |
| }, | |
| { | |
| "epoch": 5.82, | |
| "learning_rate": 4.908238579692799e-05, | |
| "loss": 0.1926, | |
| "step": 6490 | |
| }, | |
| { | |
| "epoch": 5.83, | |
| "learning_rate": 4.9072411729503295e-05, | |
| "loss": 0.1683, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 5.84, | |
| "learning_rate": 4.9062437662078595e-05, | |
| "loss": 0.188, | |
| "step": 6510 | |
| }, | |
| { | |
| "epoch": 5.85, | |
| "learning_rate": 4.90524635946539e-05, | |
| "loss": 0.2279, | |
| "step": 6520 | |
| }, | |
| { | |
| "epoch": 5.86, | |
| "learning_rate": 4.904248952722921e-05, | |
| "loss": 0.1946, | |
| "step": 6530 | |
| }, | |
| { | |
| "epoch": 5.87, | |
| "learning_rate": 4.903251545980451e-05, | |
| "loss": 0.1914, | |
| "step": 6540 | |
| }, | |
| { | |
| "epoch": 5.88, | |
| "learning_rate": 4.902254139237982e-05, | |
| "loss": 0.1413, | |
| "step": 6550 | |
| }, | |
| { | |
| "epoch": 5.88, | |
| "learning_rate": 4.901256732495512e-05, | |
| "loss": 0.2083, | |
| "step": 6560 | |
| }, | |
| { | |
| "epoch": 5.89, | |
| "learning_rate": 4.900259325753042e-05, | |
| "loss": 0.1975, | |
| "step": 6570 | |
| }, | |
| { | |
| "epoch": 5.9, | |
| "learning_rate": 4.8992619190105724e-05, | |
| "loss": 0.1888, | |
| "step": 6580 | |
| }, | |
| { | |
| "epoch": 5.91, | |
| "learning_rate": 4.898264512268103e-05, | |
| "loss": 0.1991, | |
| "step": 6590 | |
| }, | |
| { | |
| "epoch": 5.92, | |
| "learning_rate": 4.897267105525634e-05, | |
| "loss": 0.1945, | |
| "step": 6600 | |
| }, | |
| { | |
| "epoch": 5.93, | |
| "learning_rate": 4.896269698783164e-05, | |
| "loss": 0.1908, | |
| "step": 6610 | |
| }, | |
| { | |
| "epoch": 5.94, | |
| "learning_rate": 4.895272292040695e-05, | |
| "loss": 0.1826, | |
| "step": 6620 | |
| }, | |
| { | |
| "epoch": 5.95, | |
| "learning_rate": 4.894274885298225e-05, | |
| "loss": 0.1633, | |
| "step": 6630 | |
| }, | |
| { | |
| "epoch": 5.96, | |
| "learning_rate": 4.893277478555755e-05, | |
| "loss": 0.2038, | |
| "step": 6640 | |
| }, | |
| { | |
| "epoch": 5.97, | |
| "learning_rate": 4.8922800718132854e-05, | |
| "loss": 0.1872, | |
| "step": 6650 | |
| }, | |
| { | |
| "epoch": 5.97, | |
| "learning_rate": 4.891282665070816e-05, | |
| "loss": 0.1859, | |
| "step": 6660 | |
| }, | |
| { | |
| "epoch": 5.98, | |
| "learning_rate": 4.890285258328347e-05, | |
| "loss": 0.2032, | |
| "step": 6670 | |
| }, | |
| { | |
| "epoch": 5.99, | |
| "learning_rate": 4.889287851585877e-05, | |
| "loss": 0.2241, | |
| "step": 6680 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": { | |
| "accuracy": 0.9605292666517157 | |
| }, | |
| "eval_f1": { | |
| "f1": 0.9584812990028835 | |
| }, | |
| "eval_loss": 0.11326823383569717, | |
| "eval_precision": { | |
| "precision": 0.9604748305069212 | |
| }, | |
| "eval_recall": { | |
| "recall": 0.9567171466267748 | |
| }, | |
| "eval_runtime": 164.5128, | |
| "eval_samples_per_second": 108.417, | |
| "eval_steps_per_second": 6.778, | |
| "step": 6688 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "learning_rate": 4.888290444843407e-05, | |
| "loss": 0.2208, | |
| "step": 6690 | |
| }, | |
| { | |
| "epoch": 6.01, | |
| "learning_rate": 4.887293038100938e-05, | |
| "loss": 0.1563, | |
| "step": 6700 | |
| }, | |
| { | |
| "epoch": 6.02, | |
| "learning_rate": 4.8862956313584685e-05, | |
| "loss": 0.2045, | |
| "step": 6710 | |
| }, | |
| { | |
| "epoch": 6.03, | |
| "learning_rate": 4.8852982246159984e-05, | |
| "loss": 0.2119, | |
| "step": 6720 | |
| }, | |
| { | |
| "epoch": 6.04, | |
| "learning_rate": 4.884300817873529e-05, | |
| "loss": 0.1606, | |
| "step": 6730 | |
| }, | |
| { | |
| "epoch": 6.05, | |
| "learning_rate": 4.8833034111310596e-05, | |
| "loss": 0.1847, | |
| "step": 6740 | |
| }, | |
| { | |
| "epoch": 6.06, | |
| "learning_rate": 4.8823060043885896e-05, | |
| "loss": 0.1676, | |
| "step": 6750 | |
| }, | |
| { | |
| "epoch": 6.06, | |
| "learning_rate": 4.88130859764612e-05, | |
| "loss": 0.1854, | |
| "step": 6760 | |
| }, | |
| { | |
| "epoch": 6.07, | |
| "learning_rate": 4.880311190903651e-05, | |
| "loss": 0.2056, | |
| "step": 6770 | |
| }, | |
| { | |
| "epoch": 6.08, | |
| "learning_rate": 4.8793137841611814e-05, | |
| "loss": 0.1767, | |
| "step": 6780 | |
| }, | |
| { | |
| "epoch": 6.09, | |
| "learning_rate": 4.878316377418712e-05, | |
| "loss": 0.1851, | |
| "step": 6790 | |
| }, | |
| { | |
| "epoch": 6.1, | |
| "learning_rate": 4.877318970676242e-05, | |
| "loss": 0.183, | |
| "step": 6800 | |
| }, | |
| { | |
| "epoch": 6.11, | |
| "learning_rate": 4.876321563933772e-05, | |
| "loss": 0.1944, | |
| "step": 6810 | |
| }, | |
| { | |
| "epoch": 6.12, | |
| "learning_rate": 4.8753241571913026e-05, | |
| "loss": 0.1925, | |
| "step": 6820 | |
| }, | |
| { | |
| "epoch": 6.13, | |
| "learning_rate": 4.874326750448833e-05, | |
| "loss": 0.2301, | |
| "step": 6830 | |
| }, | |
| { | |
| "epoch": 6.14, | |
| "learning_rate": 4.873329343706364e-05, | |
| "loss": 0.1632, | |
| "step": 6840 | |
| }, | |
| { | |
| "epoch": 6.14, | |
| "learning_rate": 4.8723319369638944e-05, | |
| "loss": 0.1924, | |
| "step": 6850 | |
| }, | |
| { | |
| "epoch": 6.15, | |
| "learning_rate": 4.871334530221425e-05, | |
| "loss": 0.1527, | |
| "step": 6860 | |
| }, | |
| { | |
| "epoch": 6.16, | |
| "learning_rate": 4.870337123478955e-05, | |
| "loss": 0.2015, | |
| "step": 6870 | |
| }, | |
| { | |
| "epoch": 6.17, | |
| "learning_rate": 4.869339716736485e-05, | |
| "loss": 0.2006, | |
| "step": 6880 | |
| }, | |
| { | |
| "epoch": 6.18, | |
| "learning_rate": 4.8683423099940155e-05, | |
| "loss": 0.1515, | |
| "step": 6890 | |
| }, | |
| { | |
| "epoch": 6.19, | |
| "learning_rate": 4.867344903251546e-05, | |
| "loss": 0.1969, | |
| "step": 6900 | |
| }, | |
| { | |
| "epoch": 6.2, | |
| "learning_rate": 4.866347496509077e-05, | |
| "loss": 0.2036, | |
| "step": 6910 | |
| }, | |
| { | |
| "epoch": 6.21, | |
| "learning_rate": 4.8653500897666074e-05, | |
| "loss": 0.1788, | |
| "step": 6920 | |
| }, | |
| { | |
| "epoch": 6.22, | |
| "learning_rate": 4.8643526830241373e-05, | |
| "loss": 0.1581, | |
| "step": 6930 | |
| }, | |
| { | |
| "epoch": 6.23, | |
| "learning_rate": 4.863355276281668e-05, | |
| "loss": 0.1733, | |
| "step": 6940 | |
| }, | |
| { | |
| "epoch": 6.23, | |
| "learning_rate": 4.8623578695391986e-05, | |
| "loss": 0.1951, | |
| "step": 6950 | |
| }, | |
| { | |
| "epoch": 6.24, | |
| "learning_rate": 4.8613604627967285e-05, | |
| "loss": 0.1607, | |
| "step": 6960 | |
| }, | |
| { | |
| "epoch": 6.25, | |
| "learning_rate": 4.860363056054259e-05, | |
| "loss": 0.2152, | |
| "step": 6970 | |
| }, | |
| { | |
| "epoch": 6.26, | |
| "learning_rate": 4.85936564931179e-05, | |
| "loss": 0.1884, | |
| "step": 6980 | |
| }, | |
| { | |
| "epoch": 6.27, | |
| "learning_rate": 4.85836824256932e-05, | |
| "loss": 0.196, | |
| "step": 6990 | |
| }, | |
| { | |
| "epoch": 6.28, | |
| "learning_rate": 4.85737083582685e-05, | |
| "loss": 0.1845, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 6.29, | |
| "learning_rate": 4.856373429084381e-05, | |
| "loss": 0.1394, | |
| "step": 7010 | |
| }, | |
| { | |
| "epoch": 6.3, | |
| "learning_rate": 4.8553760223419116e-05, | |
| "loss": 0.2307, | |
| "step": 7020 | |
| }, | |
| { | |
| "epoch": 6.31, | |
| "learning_rate": 4.854378615599442e-05, | |
| "loss": 0.142, | |
| "step": 7030 | |
| }, | |
| { | |
| "epoch": 6.32, | |
| "learning_rate": 4.853381208856972e-05, | |
| "loss": 0.1859, | |
| "step": 7040 | |
| }, | |
| { | |
| "epoch": 6.32, | |
| "learning_rate": 4.852383802114502e-05, | |
| "loss": 0.2221, | |
| "step": 7050 | |
| }, | |
| { | |
| "epoch": 6.33, | |
| "learning_rate": 4.851386395372033e-05, | |
| "loss": 0.1906, | |
| "step": 7060 | |
| }, | |
| { | |
| "epoch": 6.34, | |
| "learning_rate": 4.850388988629563e-05, | |
| "loss": 0.1679, | |
| "step": 7070 | |
| }, | |
| { | |
| "epoch": 6.35, | |
| "learning_rate": 4.849391581887094e-05, | |
| "loss": 0.1754, | |
| "step": 7080 | |
| }, | |
| { | |
| "epoch": 6.36, | |
| "learning_rate": 4.8483941751446245e-05, | |
| "loss": 0.1901, | |
| "step": 7090 | |
| }, | |
| { | |
| "epoch": 6.37, | |
| "learning_rate": 4.847396768402155e-05, | |
| "loss": 0.1671, | |
| "step": 7100 | |
| }, | |
| { | |
| "epoch": 6.38, | |
| "learning_rate": 4.846399361659685e-05, | |
| "loss": 0.2069, | |
| "step": 7110 | |
| }, | |
| { | |
| "epoch": 6.39, | |
| "learning_rate": 4.845401954917215e-05, | |
| "loss": 0.1864, | |
| "step": 7120 | |
| }, | |
| { | |
| "epoch": 6.4, | |
| "learning_rate": 4.8444045481747457e-05, | |
| "loss": 0.1979, | |
| "step": 7130 | |
| }, | |
| { | |
| "epoch": 6.41, | |
| "learning_rate": 4.843407141432276e-05, | |
| "loss": 0.1953, | |
| "step": 7140 | |
| }, | |
| { | |
| "epoch": 6.41, | |
| "learning_rate": 4.842409734689807e-05, | |
| "loss": 0.2274, | |
| "step": 7150 | |
| }, | |
| { | |
| "epoch": 6.42, | |
| "learning_rate": 4.8414123279473375e-05, | |
| "loss": 0.1881, | |
| "step": 7160 | |
| }, | |
| { | |
| "epoch": 6.43, | |
| "learning_rate": 4.8404149212048675e-05, | |
| "loss": 0.213, | |
| "step": 7170 | |
| }, | |
| { | |
| "epoch": 6.44, | |
| "learning_rate": 4.839417514462398e-05, | |
| "loss": 0.1931, | |
| "step": 7180 | |
| }, | |
| { | |
| "epoch": 6.45, | |
| "learning_rate": 4.838420107719928e-05, | |
| "loss": 0.191, | |
| "step": 7190 | |
| }, | |
| { | |
| "epoch": 6.46, | |
| "learning_rate": 4.8374227009774586e-05, | |
| "loss": 0.1893, | |
| "step": 7200 | |
| }, | |
| { | |
| "epoch": 6.47, | |
| "learning_rate": 4.836425294234989e-05, | |
| "loss": 0.1639, | |
| "step": 7210 | |
| }, | |
| { | |
| "epoch": 6.48, | |
| "learning_rate": 4.83542788749252e-05, | |
| "loss": 0.1494, | |
| "step": 7220 | |
| }, | |
| { | |
| "epoch": 6.49, | |
| "learning_rate": 4.83443048075005e-05, | |
| "loss": 0.2012, | |
| "step": 7230 | |
| }, | |
| { | |
| "epoch": 6.49, | |
| "learning_rate": 4.8334330740075804e-05, | |
| "loss": 0.246, | |
| "step": 7240 | |
| }, | |
| { | |
| "epoch": 6.5, | |
| "learning_rate": 4.832435667265111e-05, | |
| "loss": 0.1742, | |
| "step": 7250 | |
| }, | |
| { | |
| "epoch": 6.51, | |
| "learning_rate": 4.831438260522642e-05, | |
| "loss": 0.1408, | |
| "step": 7260 | |
| }, | |
| { | |
| "epoch": 6.52, | |
| "learning_rate": 4.8304408537801716e-05, | |
| "loss": 0.1961, | |
| "step": 7270 | |
| }, | |
| { | |
| "epoch": 6.53, | |
| "learning_rate": 4.829443447037702e-05, | |
| "loss": 0.2584, | |
| "step": 7280 | |
| }, | |
| { | |
| "epoch": 6.54, | |
| "learning_rate": 4.828446040295232e-05, | |
| "loss": 0.1761, | |
| "step": 7290 | |
| }, | |
| { | |
| "epoch": 6.55, | |
| "learning_rate": 4.827448633552763e-05, | |
| "loss": 0.2035, | |
| "step": 7300 | |
| }, | |
| { | |
| "epoch": 6.56, | |
| "learning_rate": 4.8264512268102934e-05, | |
| "loss": 0.2017, | |
| "step": 7310 | |
| }, | |
| { | |
| "epoch": 6.57, | |
| "learning_rate": 4.825453820067824e-05, | |
| "loss": 0.1807, | |
| "step": 7320 | |
| }, | |
| { | |
| "epoch": 6.58, | |
| "learning_rate": 4.8244564133253547e-05, | |
| "loss": 0.1653, | |
| "step": 7330 | |
| }, | |
| { | |
| "epoch": 6.58, | |
| "learning_rate": 4.823459006582885e-05, | |
| "loss": 0.1892, | |
| "step": 7340 | |
| }, | |
| { | |
| "epoch": 6.59, | |
| "learning_rate": 4.822461599840415e-05, | |
| "loss": 0.1663, | |
| "step": 7350 | |
| }, | |
| { | |
| "epoch": 6.6, | |
| "learning_rate": 4.821464193097945e-05, | |
| "loss": 0.1638, | |
| "step": 7360 | |
| }, | |
| { | |
| "epoch": 6.61, | |
| "learning_rate": 4.820466786355476e-05, | |
| "loss": 0.2148, | |
| "step": 7370 | |
| }, | |
| { | |
| "epoch": 6.62, | |
| "learning_rate": 4.8194693796130064e-05, | |
| "loss": 0.1801, | |
| "step": 7380 | |
| }, | |
| { | |
| "epoch": 6.63, | |
| "learning_rate": 4.818471972870537e-05, | |
| "loss": 0.2135, | |
| "step": 7390 | |
| }, | |
| { | |
| "epoch": 6.64, | |
| "learning_rate": 4.8174745661280676e-05, | |
| "loss": 0.1737, | |
| "step": 7400 | |
| }, | |
| { | |
| "epoch": 6.65, | |
| "learning_rate": 4.8164771593855976e-05, | |
| "loss": 0.1583, | |
| "step": 7410 | |
| }, | |
| { | |
| "epoch": 6.66, | |
| "learning_rate": 4.815479752643128e-05, | |
| "loss": 0.1403, | |
| "step": 7420 | |
| }, | |
| { | |
| "epoch": 6.67, | |
| "learning_rate": 4.814482345900658e-05, | |
| "loss": 0.223, | |
| "step": 7430 | |
| }, | |
| { | |
| "epoch": 6.67, | |
| "learning_rate": 4.813484939158189e-05, | |
| "loss": 0.1794, | |
| "step": 7440 | |
| }, | |
| { | |
| "epoch": 6.68, | |
| "learning_rate": 4.8124875324157194e-05, | |
| "loss": 0.182, | |
| "step": 7450 | |
| }, | |
| { | |
| "epoch": 6.69, | |
| "learning_rate": 4.81149012567325e-05, | |
| "loss": 0.1429, | |
| "step": 7460 | |
| }, | |
| { | |
| "epoch": 6.7, | |
| "learning_rate": 4.81049271893078e-05, | |
| "loss": 0.1657, | |
| "step": 7470 | |
| }, | |
| { | |
| "epoch": 6.71, | |
| "learning_rate": 4.8094953121883106e-05, | |
| "loss": 0.1623, | |
| "step": 7480 | |
| }, | |
| { | |
| "epoch": 6.72, | |
| "learning_rate": 4.808497905445841e-05, | |
| "loss": 0.2058, | |
| "step": 7490 | |
| }, | |
| { | |
| "epoch": 6.73, | |
| "learning_rate": 4.807500498703372e-05, | |
| "loss": 0.1683, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 6.74, | |
| "learning_rate": 4.806503091960902e-05, | |
| "loss": 0.1531, | |
| "step": 7510 | |
| }, | |
| { | |
| "epoch": 6.75, | |
| "learning_rate": 4.8055056852184324e-05, | |
| "loss": 0.1793, | |
| "step": 7520 | |
| }, | |
| { | |
| "epoch": 6.75, | |
| "learning_rate": 4.804508278475963e-05, | |
| "loss": 0.1785, | |
| "step": 7530 | |
| }, | |
| { | |
| "epoch": 6.76, | |
| "learning_rate": 4.803510871733493e-05, | |
| "loss": 0.2247, | |
| "step": 7540 | |
| }, | |
| { | |
| "epoch": 6.77, | |
| "learning_rate": 4.8025134649910235e-05, | |
| "loss": 0.1673, | |
| "step": 7550 | |
| }, | |
| { | |
| "epoch": 6.78, | |
| "learning_rate": 4.801516058248554e-05, | |
| "loss": 0.2173, | |
| "step": 7560 | |
| }, | |
| { | |
| "epoch": 6.79, | |
| "learning_rate": 4.800518651506085e-05, | |
| "loss": 0.1675, | |
| "step": 7570 | |
| }, | |
| { | |
| "epoch": 6.8, | |
| "learning_rate": 4.7995212447636154e-05, | |
| "loss": 0.2222, | |
| "step": 7580 | |
| }, | |
| { | |
| "epoch": 6.81, | |
| "learning_rate": 4.798523838021145e-05, | |
| "loss": 0.2106, | |
| "step": 7590 | |
| }, | |
| { | |
| "epoch": 6.82, | |
| "learning_rate": 4.797526431278675e-05, | |
| "loss": 0.177, | |
| "step": 7600 | |
| }, | |
| { | |
| "epoch": 6.83, | |
| "learning_rate": 4.796529024536206e-05, | |
| "loss": 0.1841, | |
| "step": 7610 | |
| }, | |
| { | |
| "epoch": 6.84, | |
| "learning_rate": 4.7955316177937365e-05, | |
| "loss": 0.16, | |
| "step": 7620 | |
| }, | |
| { | |
| "epoch": 6.84, | |
| "learning_rate": 4.794534211051267e-05, | |
| "loss": 0.1899, | |
| "step": 7630 | |
| }, | |
| { | |
| "epoch": 6.85, | |
| "learning_rate": 4.793536804308798e-05, | |
| "loss": 0.1864, | |
| "step": 7640 | |
| }, | |
| { | |
| "epoch": 6.86, | |
| "learning_rate": 4.792539397566328e-05, | |
| "loss": 0.137, | |
| "step": 7650 | |
| }, | |
| { | |
| "epoch": 6.87, | |
| "learning_rate": 4.791541990823858e-05, | |
| "loss": 0.1933, | |
| "step": 7660 | |
| }, | |
| { | |
| "epoch": 6.88, | |
| "learning_rate": 4.790544584081388e-05, | |
| "loss": 0.1269, | |
| "step": 7670 | |
| }, | |
| { | |
| "epoch": 6.89, | |
| "learning_rate": 4.789547177338919e-05, | |
| "loss": 0.2185, | |
| "step": 7680 | |
| }, | |
| { | |
| "epoch": 6.9, | |
| "learning_rate": 4.7885497705964495e-05, | |
| "loss": 0.1833, | |
| "step": 7690 | |
| }, | |
| { | |
| "epoch": 6.91, | |
| "learning_rate": 4.78755236385398e-05, | |
| "loss": 0.1457, | |
| "step": 7700 | |
| }, | |
| { | |
| "epoch": 6.92, | |
| "learning_rate": 4.78655495711151e-05, | |
| "loss": 0.1443, | |
| "step": 7710 | |
| }, | |
| { | |
| "epoch": 6.93, | |
| "learning_rate": 4.785557550369041e-05, | |
| "loss": 0.1887, | |
| "step": 7720 | |
| }, | |
| { | |
| "epoch": 6.93, | |
| "learning_rate": 4.784560143626571e-05, | |
| "loss": 0.1302, | |
| "step": 7730 | |
| }, | |
| { | |
| "epoch": 6.94, | |
| "learning_rate": 4.783562736884102e-05, | |
| "loss": 0.1741, | |
| "step": 7740 | |
| }, | |
| { | |
| "epoch": 6.95, | |
| "learning_rate": 4.782565330141632e-05, | |
| "loss": 0.1816, | |
| "step": 7750 | |
| }, | |
| { | |
| "epoch": 6.96, | |
| "learning_rate": 4.7815679233991625e-05, | |
| "loss": 0.1846, | |
| "step": 7760 | |
| }, | |
| { | |
| "epoch": 6.97, | |
| "learning_rate": 4.780570516656693e-05, | |
| "loss": 0.1754, | |
| "step": 7770 | |
| }, | |
| { | |
| "epoch": 6.98, | |
| "learning_rate": 4.779573109914223e-05, | |
| "loss": 0.1595, | |
| "step": 7780 | |
| }, | |
| { | |
| "epoch": 6.99, | |
| "learning_rate": 4.7785757031717536e-05, | |
| "loss": 0.185, | |
| "step": 7790 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "learning_rate": 4.777578296429284e-05, | |
| "loss": 0.2375, | |
| "step": 7800 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": { | |
| "accuracy": 0.9638932496075353 | |
| }, | |
| "eval_f1": { | |
| "f1": 0.9615520166695122 | |
| }, | |
| "eval_loss": 0.11376180499792099, | |
| "eval_precision": { | |
| "precision": 0.9628392910079963 | |
| }, | |
| "eval_recall": { | |
| "recall": 0.9605232709871921 | |
| }, | |
| "eval_runtime": 164.872, | |
| "eval_samples_per_second": 108.181, | |
| "eval_steps_per_second": 6.763, | |
| "step": 7803 | |
| } | |
| ], | |
| "logging_steps": 10, | |
| "max_steps": 55700, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 50, | |
| "save_steps": 500, | |
| "total_flos": 3.8698643448702075e+19, | |
| "train_batch_size": 16, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |