| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 0.12367030270238935, | |
| "global_step": 800, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 9.999226963512678e-06, | |
| "loss": 1.1361, | |
| "step": 5 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 9.998453927025357e-06, | |
| "loss": 0.7353, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 9.997680890538034e-06, | |
| "loss": 0.6909, | |
| "step": 15 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 9.996907854050712e-06, | |
| "loss": 0.6498, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 9.99613481756339e-06, | |
| "loss": 0.6414, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 9.995361781076068e-06, | |
| "loss": 0.6415, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.994588744588745e-06, | |
| "loss": 0.6317, | |
| "step": 35 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.993815708101423e-06, | |
| "loss": 0.6378, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.9930426716141e-06, | |
| "loss": 0.6347, | |
| "step": 45 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.992269635126779e-06, | |
| "loss": 0.5924, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.991496598639456e-06, | |
| "loss": 0.6046, | |
| "step": 55 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.990723562152135e-06, | |
| "loss": 0.6045, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.989950525664813e-06, | |
| "loss": 0.6, | |
| "step": 65 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.98917748917749e-06, | |
| "loss": 0.5504, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.988404452690169e-06, | |
| "loss": 0.5747, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.987631416202846e-06, | |
| "loss": 0.5526, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.986858379715523e-06, | |
| "loss": 0.5958, | |
| "step": 85 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.9860853432282e-06, | |
| "loss": 0.608, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.985312306740878e-06, | |
| "loss": 0.5988, | |
| "step": 95 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.984539270253557e-06, | |
| "loss": 0.5861, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.983766233766234e-06, | |
| "loss": 0.5749, | |
| "step": 105 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.982993197278913e-06, | |
| "loss": 0.5498, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.98222016079159e-06, | |
| "loss": 0.5841, | |
| "step": 115 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.981447124304268e-06, | |
| "loss": 0.5973, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.980674087816947e-06, | |
| "loss": 0.5954, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.979901051329624e-06, | |
| "loss": 0.527, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.979128014842301e-06, | |
| "loss": 0.5321, | |
| "step": 135 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.978354978354979e-06, | |
| "loss": 0.5781, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.977581941867656e-06, | |
| "loss": 0.5119, | |
| "step": 145 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.976808905380335e-06, | |
| "loss": 0.5271, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.976035868893012e-06, | |
| "loss": 0.5814, | |
| "step": 155 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 9.97526283240569e-06, | |
| "loss": 0.518, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.974489795918369e-06, | |
| "loss": 0.5335, | |
| "step": 165 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.973716759431046e-06, | |
| "loss": 0.5062, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.972943722943725e-06, | |
| "loss": 0.5253, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.972170686456402e-06, | |
| "loss": 0.5856, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.97139764996908e-06, | |
| "loss": 0.5196, | |
| "step": 185 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.970624613481757e-06, | |
| "loss": 0.4764, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.969851576994434e-06, | |
| "loss": 0.5254, | |
| "step": 195 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.969078540507111e-06, | |
| "loss": 0.5442, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "eval_accuracy": 0.5804400673190799, | |
| "eval_accuracy_sklearn": 0.5804400673190799, | |
| "eval_f1": 0.5294915349019279, | |
| "eval_loss": 0.7918509840965271, | |
| "eval_precision": 0.6370946036872561, | |
| "eval_recall": 0.45298409281186464, | |
| "eval_runtime": 4914.2737, | |
| "eval_samples_per_second": 16.323, | |
| "eval_steps_per_second": 2.04, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.96830550401979e-06, | |
| "loss": 0.5163, | |
| "step": 205 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.967532467532468e-06, | |
| "loss": 0.5044, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.966759431045147e-06, | |
| "loss": 0.5078, | |
| "step": 215 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.965986394557824e-06, | |
| "loss": 0.4623, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.965213358070501e-06, | |
| "loss": 0.5359, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.96444032158318e-06, | |
| "loss": 0.5068, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.963667285095858e-06, | |
| "loss": 0.5029, | |
| "step": 235 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.962894248608535e-06, | |
| "loss": 0.5084, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.962121212121212e-06, | |
| "loss": 0.4783, | |
| "step": 245 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.96134817563389e-06, | |
| "loss": 0.5216, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.960575139146569e-06, | |
| "loss": 0.54, | |
| "step": 255 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.959802102659246e-06, | |
| "loss": 0.5494, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.959029066171925e-06, | |
| "loss": 0.5401, | |
| "step": 265 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.958256029684602e-06, | |
| "loss": 0.5073, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.95748299319728e-06, | |
| "loss": 0.4598, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.956709956709958e-06, | |
| "loss": 0.4913, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.955936920222636e-06, | |
| "loss": 0.4947, | |
| "step": 285 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.955163883735313e-06, | |
| "loss": 0.4806, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.95439084724799e-06, | |
| "loss": 0.4659, | |
| "step": 295 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.953617810760668e-06, | |
| "loss": 0.4555, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.952844774273347e-06, | |
| "loss": 0.4606, | |
| "step": 305 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.952071737786024e-06, | |
| "loss": 0.4905, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.951298701298701e-06, | |
| "loss": 0.4423, | |
| "step": 315 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.95052566481138e-06, | |
| "loss": 0.4855, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.949752628324058e-06, | |
| "loss": 0.486, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.948979591836737e-06, | |
| "loss": 0.4774, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.948206555349414e-06, | |
| "loss": 0.4909, | |
| "step": 335 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.947433518862091e-06, | |
| "loss": 0.47, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.946660482374768e-06, | |
| "loss": 0.4496, | |
| "step": 345 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.945887445887446e-06, | |
| "loss": 0.5146, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.945114409400125e-06, | |
| "loss": 0.4876, | |
| "step": 355 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.944341372912802e-06, | |
| "loss": 0.4747, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.94356833642548e-06, | |
| "loss": 0.4614, | |
| "step": 365 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.942795299938158e-06, | |
| "loss": 0.4755, | |
| "step": 370 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.942022263450836e-06, | |
| "loss": 0.4785, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.941249226963513e-06, | |
| "loss": 0.4581, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.940476190476192e-06, | |
| "loss": 0.4671, | |
| "step": 385 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.93970315398887e-06, | |
| "loss": 0.4327, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.938930117501547e-06, | |
| "loss": 0.4906, | |
| "step": 395 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.938157081014226e-06, | |
| "loss": 0.5006, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "eval_accuracy": 0.5519790562862308, | |
| "eval_accuracy_sklearn": 0.5519790562862308, | |
| "eval_f1": 0.39240549130993435, | |
| "eval_loss": 0.969095766544342, | |
| "eval_precision": 0.6691460531626593, | |
| "eval_recall": 0.27759837340031096, | |
| "eval_runtime": 4903.2377, | |
| "eval_samples_per_second": 16.36, | |
| "eval_steps_per_second": 2.045, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.937384044526903e-06, | |
| "loss": 0.4531, | |
| "step": 405 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.93661100803958e-06, | |
| "loss": 0.4467, | |
| "step": 410 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.935837971552257e-06, | |
| "loss": 0.4753, | |
| "step": 415 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.935064935064936e-06, | |
| "loss": 0.428, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.934291898577614e-06, | |
| "loss": 0.4418, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.933518862090291e-06, | |
| "loss": 0.5087, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.93274582560297e-06, | |
| "loss": 0.4775, | |
| "step": 435 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.931972789115647e-06, | |
| "loss": 0.4923, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.931199752628325e-06, | |
| "loss": 0.4721, | |
| "step": 445 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.930426716141004e-06, | |
| "loss": 0.4718, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.929653679653681e-06, | |
| "loss": 0.4392, | |
| "step": 455 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.928880643166358e-06, | |
| "loss": 0.4315, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.928107606679036e-06, | |
| "loss": 0.4641, | |
| "step": 465 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.927334570191713e-06, | |
| "loss": 0.4417, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.926561533704392e-06, | |
| "loss": 0.454, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.925788497217069e-06, | |
| "loss": 0.4414, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 9.925015460729748e-06, | |
| "loss": 0.4802, | |
| "step": 485 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.924242424242425e-06, | |
| "loss": 0.4262, | |
| "step": 490 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.923469387755103e-06, | |
| "loss": 0.4543, | |
| "step": 495 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.922696351267782e-06, | |
| "loss": 0.407, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.921923314780459e-06, | |
| "loss": 0.4328, | |
| "step": 505 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.921150278293136e-06, | |
| "loss": 0.4589, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.920377241805814e-06, | |
| "loss": 0.4588, | |
| "step": 515 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.919604205318491e-06, | |
| "loss": 0.4273, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.91883116883117e-06, | |
| "loss": 0.4688, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.918058132343847e-06, | |
| "loss": 0.4324, | |
| "step": 530 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.917285095856525e-06, | |
| "loss": 0.4846, | |
| "step": 535 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.916512059369204e-06, | |
| "loss": 0.4366, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 9.91573902288188e-06, | |
| "loss": 0.442, | |
| "step": 545 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.91496598639456e-06, | |
| "loss": 0.4241, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.914192949907237e-06, | |
| "loss": 0.4574, | |
| "step": 555 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.913419913419914e-06, | |
| "loss": 0.4173, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.912646876932592e-06, | |
| "loss": 0.4202, | |
| "step": 565 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.911873840445269e-06, | |
| "loss": 0.4694, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.911100803957948e-06, | |
| "loss": 0.4595, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.910327767470625e-06, | |
| "loss": 0.464, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.909554730983303e-06, | |
| "loss": 0.4688, | |
| "step": 585 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.908781694495982e-06, | |
| "loss": 0.4932, | |
| "step": 590 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.908008658008659e-06, | |
| "loss": 0.4648, | |
| "step": 595 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.907235621521336e-06, | |
| "loss": 0.5136, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "eval_accuracy": 0.595349996883376, | |
| "eval_accuracy_sklearn": 0.595349996883376, | |
| "eval_f1": 0.6681389238209163, | |
| "eval_loss": 0.8079590201377869, | |
| "eval_precision": 0.5834404685379616, | |
| "eval_recall": 0.7816050711637363, | |
| "eval_runtime": 4915.3389, | |
| "eval_samples_per_second": 16.319, | |
| "eval_steps_per_second": 2.04, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.906462585034015e-06, | |
| "loss": 0.5246, | |
| "step": 605 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 9.905689548546693e-06, | |
| "loss": 0.4751, | |
| "step": 610 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.90491651205937e-06, | |
| "loss": 0.4224, | |
| "step": 615 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.904143475572047e-06, | |
| "loss": 0.3842, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.903370439084724e-06, | |
| "loss": 0.4524, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.902597402597403e-06, | |
| "loss": 0.4657, | |
| "step": 630 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.90182436611008e-06, | |
| "loss": 0.4596, | |
| "step": 635 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.90105132962276e-06, | |
| "loss": 0.4383, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.900278293135437e-06, | |
| "loss": 0.4351, | |
| "step": 645 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.899505256648114e-06, | |
| "loss": 0.4263, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.898732220160793e-06, | |
| "loss": 0.448, | |
| "step": 655 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.89795918367347e-06, | |
| "loss": 0.4372, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.897186147186148e-06, | |
| "loss": 0.4681, | |
| "step": 665 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.896413110698825e-06, | |
| "loss": 0.4783, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.895640074211502e-06, | |
| "loss": 0.4155, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.894867037724181e-06, | |
| "loss": 0.4398, | |
| "step": 680 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.894094001236859e-06, | |
| "loss": 0.4364, | |
| "step": 685 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.893320964749536e-06, | |
| "loss": 0.4175, | |
| "step": 690 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.892547928262215e-06, | |
| "loss": 0.4763, | |
| "step": 695 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.891774891774892e-06, | |
| "loss": 0.4629, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.891001855287571e-06, | |
| "loss": 0.5083, | |
| "step": 705 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.890228818800249e-06, | |
| "loss": 0.4867, | |
| "step": 710 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.889455782312926e-06, | |
| "loss": 0.4855, | |
| "step": 715 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.888682745825603e-06, | |
| "loss": 0.4219, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.88790970933828e-06, | |
| "loss": 0.4676, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.88713667285096e-06, | |
| "loss": 0.3963, | |
| "step": 730 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.886363636363637e-06, | |
| "loss": 0.4347, | |
| "step": 735 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 9.885590599876314e-06, | |
| "loss": 0.4721, | |
| "step": 740 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.884817563388993e-06, | |
| "loss": 0.3992, | |
| "step": 745 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.88404452690167e-06, | |
| "loss": 0.413, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.883271490414348e-06, | |
| "loss": 0.4881, | |
| "step": 755 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.882498453927027e-06, | |
| "loss": 0.4478, | |
| "step": 760 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.881725417439704e-06, | |
| "loss": 0.4453, | |
| "step": 765 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.880952380952381e-06, | |
| "loss": 0.4252, | |
| "step": 770 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.88017934446506e-06, | |
| "loss": 0.411, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.879406307977738e-06, | |
| "loss": 0.435, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.878633271490415e-06, | |
| "loss": 0.3948, | |
| "step": 785 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.877860235003092e-06, | |
| "loss": 0.4004, | |
| "step": 790 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.877087198515771e-06, | |
| "loss": 0.4515, | |
| "step": 795 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.876314162028449e-06, | |
| "loss": 0.4117, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "eval_accuracy": 0.5725113756778657, | |
| "eval_accuracy_sklearn": 0.5725113756778657, | |
| "eval_f1": 0.5080130274465917, | |
| "eval_loss": 0.9323258996009827, | |
| "eval_precision": 0.6346884634688463, | |
| "eval_recall": 0.423490013156321, | |
| "eval_runtime": 4928.5167, | |
| "eval_samples_per_second": 16.276, | |
| "eval_steps_per_second": 2.034, | |
| "step": 800 | |
| } | |
| ], | |
| "max_steps": 64680, | |
| "num_train_epochs": 10, | |
| "total_flos": 9.50142335188992e+16, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |