{ "best_metric": null, "best_model_checkpoint": null, "epoch": 10.0, "global_step": 540, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.02, "learning_rate": 1.5151515151515152e-06, "loss": 1.7524, "step": 1 }, { "epoch": 0.09, "learning_rate": 7.5757575757575764e-06, "loss": 1.7511, "step": 5 }, { "epoch": 0.19, "learning_rate": 1.5151515151515153e-05, "loss": 1.5404, "step": 10 }, { "epoch": 0.28, "learning_rate": 2.272727272727273e-05, "loss": 1.118, "step": 15 }, { "epoch": 0.37, "learning_rate": 3.0303030303030306e-05, "loss": 0.6504, "step": 20 }, { "epoch": 0.46, "learning_rate": 3.787878787878788e-05, "loss": 0.8655, "step": 25 }, { "epoch": 0.56, "learning_rate": 4.545454545454546e-05, "loss": 0.4171, "step": 30 }, { "epoch": 0.65, "learning_rate": 4.999808023410233e-05, "loss": 0.3723, "step": 35 }, { "epoch": 0.74, "learning_rate": 4.997648625366471e-05, "loss": 0.3472, "step": 40 }, { "epoch": 0.83, "learning_rate": 4.993091938082206e-05, "loss": 0.1754, "step": 45 }, { "epoch": 0.93, "learning_rate": 4.986142335149978e-05, "loss": 0.2462, "step": 50 }, { "epoch": 1.0, "eval_PREDICATE_f1": 0.9416058394160584, "eval_PREDICATE_number": 129, "eval_PREDICATE_precision": 0.8896551724137931, "eval_PREDICATE_recall": 1.0, "eval_loss": 0.15542413294315338, "eval_overall_accuracy": 0.9486208761492698, "eval_overall_f1": 0.8, "eval_overall_precision": 0.8135593220338984, "eval_overall_recall": 0.7868852459016393, "eval_runtime": 0.5007, "eval_samples_per_second": 191.714, "eval_steps_per_second": 23.964, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_f1": 0.0, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_number": 1, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_precision": 0.0, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_recall": 0.0, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.6115702479338844, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_number": 73, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.7708333333333334, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.5068493150684932, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.6190476190476191, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 41, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.6046511627906976, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.6341463414634146, "step": 54 }, { "epoch": 1.02, "learning_rate": 4.976806486926926e-05, "loss": 0.216, "step": 55 }, { "epoch": 1.11, "learning_rate": 4.9650933541324506e-05, "loss": 0.1322, "step": 60 }, { "epoch": 1.2, "learning_rate": 4.951014179247555e-05, "loss": 0.1814, "step": 65 }, { "epoch": 1.3, "learning_rate": 4.934582475724101e-05, "loss": 0.1226, "step": 70 }, { "epoch": 1.39, "learning_rate": 4.9158140150143484e-05, "loss": 0.1595, "step": 75 }, { "epoch": 1.48, "learning_rate": 4.8947268114332274e-05, "loss": 0.172, "step": 80 }, { "epoch": 1.57, "learning_rate": 4.8713411048678635e-05, "loss": 0.1202, "step": 85 }, { "epoch": 1.67, "learning_rate": 4.8456793413509634e-05, "loss": 0.1816, "step": 90 }, { "epoch": 1.76, "learning_rate": 4.817766151516693e-05, "loss": 0.164, "step": 95 }, { "epoch": 1.85, "learning_rate": 4.787628326959747e-05, "loss": 0.1345, "step": 100 }, { "epoch": 1.94, "learning_rate": 4.755294794520276e-05, "loss": 0.1863, "step": 105 }, { "epoch": 2.0, "eval_PREDICATE_f1": 0.9446494464944649, "eval_PREDICATE_number": 129, "eval_PREDICATE_precision": 0.9014084507042254, "eval_PREDICATE_recall": 0.9922480620155039, "eval_loss": 0.1267632097005844, "eval_overall_accuracy": 0.9556517036235803, "eval_overall_f1": 0.8220338983050847, "eval_overall_precision": 0.8508771929824561, "eval_overall_recall": 0.7950819672131147, "eval_runtime": 0.5235, "eval_samples_per_second": 183.39, "eval_steps_per_second": 22.924, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_f1": 0.0, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_number": 1, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_precision": 0.0, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_recall": 0.0, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.6440677966101694, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_number": 73, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.8444444444444444, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.5205479452054794, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.6829268292682927, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 41, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.6829268292682927, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.6829268292682927, "step": 108 }, { "epoch": 2.04, "learning_rate": 4.72079658851938e-05, "loss": 0.0882, "step": 110 }, { "epoch": 2.13, "learning_rate": 4.684166820971779e-05, "loss": 0.0917, "step": 115 }, { "epoch": 2.22, "learning_rate": 4.6454406498043105e-05, "loss": 0.1179, "step": 120 }, { "epoch": 2.31, "learning_rate": 4.604655245110684e-05, "loss": 0.0599, "step": 125 }, { "epoch": 2.41, "learning_rate": 4.561849753474951e-05, "loss": 0.0923, "step": 130 }, { "epoch": 2.5, "learning_rate": 4.517065260397887e-05, "loss": 0.0903, "step": 135 }, { "epoch": 2.59, "learning_rate": 4.470344750862368e-05, "loss": 0.0941, "step": 140 }, { "epoch": 2.69, "learning_rate": 4.421733068075595e-05, "loss": 0.0881, "step": 145 }, { "epoch": 2.78, "learning_rate": 4.371276870427753e-05, "loss": 0.0653, "step": 150 }, { "epoch": 2.87, "learning_rate": 4.319024586708428e-05, "loss": 0.1036, "step": 155 }, { "epoch": 2.96, "learning_rate": 4.265026369623761e-05, "loss": 0.0668, "step": 160 }, { "epoch": 3.0, "eval_PREDICATE_f1": 0.9584905660377359, "eval_PREDICATE_number": 129, "eval_PREDICATE_precision": 0.9338235294117647, "eval_PREDICATE_recall": 0.9844961240310077, "eval_loss": 0.1287979632616043, "eval_overall_accuracy": 0.9599783666846944, "eval_overall_f1": 0.8458333333333333, "eval_overall_precision": 0.8601694915254238, "eval_overall_recall": 0.8319672131147541, "eval_runtime": 0.4971, "eval_samples_per_second": 193.12, "eval_steps_per_second": 24.14, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_f1": 0.0, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_number": 1, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_precision": 0.0, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_recall": 0.0, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.6929133858267718, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_number": 73, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.8148148148148148, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.6027397260273972, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.7356321839080461, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 41, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.6956521739130435, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.7804878048780488, "step": 162 }, { "epoch": 3.06, "learning_rate": 4.209334047658956e-05, "loss": 0.0458, "step": 165 }, { "epoch": 3.15, "learning_rate": 4.152001075332341e-05, "loss": 0.044, "step": 170 }, { "epoch": 3.24, "learning_rate": 4.09308248188874e-05, "loss": 0.0489, "step": 175 }, { "epoch": 3.33, "learning_rate": 4.0326348184813826e-05, "loss": 0.056, "step": 180 }, { "epoch": 3.43, "learning_rate": 3.970716103893065e-05, "loss": 0.05, "step": 185 }, { "epoch": 3.52, "learning_rate": 3.907385768848656e-05, "loss": 0.0403, "step": 190 }, { "epoch": 3.61, "learning_rate": 3.842704598972384e-05, "loss": 0.0573, "step": 195 }, { "epoch": 3.7, "learning_rate": 3.776734676444678e-05, "loss": 0.0624, "step": 200 }, { "epoch": 3.8, "learning_rate": 3.709539320414544e-05, "loss": 0.0465, "step": 205 }, { "epoch": 3.89, "learning_rate": 3.641183026224675e-05, "loss": 0.0677, "step": 210 }, { "epoch": 3.98, "learning_rate": 3.5717314035076355e-05, "loss": 0.039, "step": 215 }, { "epoch": 4.0, "eval_PREDICATE_f1": 0.9407407407407407, "eval_PREDICATE_number": 129, "eval_PREDICATE_precision": 0.900709219858156, "eval_PREDICATE_recall": 0.9844961240310077, "eval_loss": 0.16953788697719574, "eval_overall_accuracy": 0.9556517036235803, "eval_overall_f1": 0.8309572301425663, "eval_overall_precision": 0.8259109311740891, "eval_overall_recall": 0.8360655737704918, "eval_runtime": 0.5165, "eval_samples_per_second": 185.878, "eval_steps_per_second": 23.235, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_f1": 0.0, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_number": 1, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_precision": 0.0, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_recall": 0.0, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.6499999999999999, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_number": 73, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.8297872340425532, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.5342465753424658, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.76, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 41, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.6440677966101694, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.926829268292683, "step": 216 }, { "epoch": 4.07, "learning_rate": 3.501251113212521e-05, "loss": 0.0332, "step": 220 }, { "epoch": 4.17, "learning_rate": 3.429809803622551e-05, "loss": 0.0549, "step": 225 }, { "epoch": 4.26, "learning_rate": 3.357476045424997e-05, "loss": 0.0214, "step": 230 }, { "epoch": 4.35, "learning_rate": 3.2843192658957774e-05, "loss": 0.0352, "step": 235 }, { "epoch": 4.44, "learning_rate": 3.210409682261866e-05, "loss": 0.0521, "step": 240 }, { "epoch": 4.54, "learning_rate": 3.135818234305511e-05, "loss": 0.0267, "step": 245 }, { "epoch": 4.63, "learning_rate": 3.060616516274921e-05, "loss": 0.0454, "step": 250 }, { "epoch": 4.72, "learning_rate": 2.984876708166782e-05, "loss": 0.0234, "step": 255 }, { "epoch": 4.81, "learning_rate": 2.908671506446566e-05, "loss": 0.034, "step": 260 }, { "epoch": 4.91, "learning_rate": 2.8320740542731212e-05, "loss": 0.0198, "step": 265 }, { "epoch": 5.0, "learning_rate": 2.7551578712945208e-05, "loss": 0.0187, "step": 270 }, { "epoch": 5.0, "eval_PREDICATE_f1": 0.9465648854961832, "eval_PREDICATE_number": 129, "eval_PREDICATE_precision": 0.9323308270676691, "eval_PREDICATE_recall": 0.9612403100775194, "eval_loss": 0.1955076903104782, "eval_overall_accuracy": 0.955110870740941, "eval_overall_f1": 0.8195329087048832, "eval_overall_precision": 0.8502202643171806, "eval_overall_recall": 0.7909836065573771, "eval_runtime": 0.5123, "eval_samples_per_second": 187.387, "eval_steps_per_second": 23.423, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_f1": 0.0, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_number": 1, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_precision": 0.0, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_recall": 0.0, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.6511627906976744, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_number": 73, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.75, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.5753424657534246, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.6835443037974684, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 41, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.7105263157894737, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.6585365853658537, "step": 270 }, { "epoch": 5.09, "learning_rate": 2.6779967830825454e-05, "loss": 0.0178, "step": 275 }, { "epoch": 5.19, "learning_rate": 2.600664850273538e-05, "loss": 0.0231, "step": 280 }, { "epoch": 5.28, "learning_rate": 2.5232362974836394e-05, "loss": 0.0149, "step": 285 }, { "epoch": 5.37, "learning_rate": 2.4457854420666278e-05, "loss": 0.0237, "step": 290 }, { "epoch": 5.46, "learning_rate": 2.3683866227827603e-05, "loss": 0.0134, "step": 295 }, { "epoch": 5.56, "learning_rate": 2.2911141284470466e-05, "loss": 0.0154, "step": 300 }, { "epoch": 5.65, "learning_rate": 2.214042126625486e-05, "loss": 0.0361, "step": 305 }, { "epoch": 5.74, "learning_rate": 2.137244592447658e-05, "loss": 0.0088, "step": 310 }, { "epoch": 5.83, "learning_rate": 2.0607952376040353e-05, "loss": 0.042, "step": 315 }, { "epoch": 5.93, "learning_rate": 1.9847674395961406e-05, "loss": 0.0216, "step": 320 }, { "epoch": 6.0, "eval_PREDICATE_f1": 0.9501915708812261, "eval_PREDICATE_number": 129, "eval_PREDICATE_precision": 0.9393939393939394, "eval_PREDICATE_recall": 0.9612403100775194, "eval_loss": 0.2082541584968567, "eval_overall_accuracy": 0.9561925365062196, "eval_overall_f1": 0.8252631578947369, "eval_overall_precision": 0.8484848484848485, "eval_overall_recall": 0.8032786885245902, "eval_runtime": 0.4958, "eval_samples_per_second": 193.643, "eval_steps_per_second": 24.205, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_f1": 0.0, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_number": 1, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_precision": 0.0, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_recall": 0.0, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.6717557251908397, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_number": 73, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.7586206896551724, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.6027397260273972, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.6829268292682927, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 41, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.6829268292682927, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.6829268292682927, "step": 324 }, { "epoch": 6.02, "learning_rate": 1.909234171307466e-05, "loss": 0.0091, "step": 325 }, { "epoch": 6.11, "learning_rate": 1.8342679309627543e-05, "loss": 0.0164, "step": 330 }, { "epoch": 6.2, "learning_rate": 1.7599406725428617e-05, "loss": 0.01, "step": 335 }, { "epoch": 6.3, "learning_rate": 1.6863237367220058e-05, "loss": 0.0103, "step": 340 }, { "epoch": 6.39, "learning_rate": 1.613487782393661e-05, "loss": 0.0151, "step": 345 }, { "epoch": 6.48, "learning_rate": 1.5415027188508573e-05, "loss": 0.0039, "step": 350 }, { "epoch": 6.57, "learning_rate": 1.4704376386859448e-05, "loss": 0.0211, "step": 355 }, { "epoch": 6.67, "learning_rate": 1.4003607514742529e-05, "loss": 0.0257, "step": 360 }, { "epoch": 6.76, "learning_rate": 1.3313393183052747e-05, "loss": 0.0072, "step": 365 }, { "epoch": 6.85, "learning_rate": 1.2634395872242433e-05, "loss": 0.0171, "step": 370 }, { "epoch": 6.94, "learning_rate": 1.1967267296460208e-05, "loss": 0.0176, "step": 375 }, { "epoch": 7.0, "eval_PREDICATE_f1": 0.9465648854961832, "eval_PREDICATE_number": 129, "eval_PREDICATE_precision": 0.9323308270676691, "eval_PREDICATE_recall": 0.9612403100775194, "eval_loss": 0.220278799533844, "eval_overall_accuracy": 0.9578150351541374, "eval_overall_f1": 0.8356997971602433, "eval_overall_precision": 0.8273092369477911, "eval_overall_recall": 0.8442622950819673, "eval_runtime": 0.4918, "eval_samples_per_second": 195.192, "eval_steps_per_second": 24.399, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_f1": 0.0, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_number": 1, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_precision": 0.0, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_recall": 0.0, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.6906474820143884, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_number": 73, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.7272727272727273, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.6575342465753424, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.7472527472527474, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 41, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.68, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.8292682926829268, "step": 378 }, { "epoch": 7.04, "learning_rate": 1.131264777802387e-05, "loss": 0.008, "step": 380 }, { "epoch": 7.13, "learning_rate": 1.0671165632827097e-05, "loss": 0.0118, "step": 385 }, { "epoch": 7.22, "learning_rate": 1.0043436567270312e-05, "loss": 0.0111, "step": 390 }, { "epoch": 7.31, "learning_rate": 9.43006308729432e-06, "loss": 0.0085, "step": 395 }, { "epoch": 7.41, "learning_rate": 8.831633920083968e-06, "loss": 0.0163, "step": 400 }, { "epoch": 7.5, "learning_rate": 8.248723448996942e-06, "loss": 0.0045, "step": 405 }, { "epoch": 7.59, "learning_rate": 7.681891162260015e-06, "loss": 0.0077, "step": 410 }, { "epoch": 7.69, "learning_rate": 7.1316811159619305e-06, "loss": 0.0078, "step": 415 }, { "epoch": 7.78, "learning_rate": 6.598621411858319e-06, "loss": 0.0147, "step": 420 }, { "epoch": 7.87, "learning_rate": 6.083223690489901e-06, "loss": 0.0145, "step": 425 }, { "epoch": 7.96, "learning_rate": 5.585982640100415e-06, "loss": 0.0037, "step": 430 }, { "epoch": 8.0, "eval_PREDICATE_f1": 0.9465648854961832, "eval_PREDICATE_number": 129, "eval_PREDICATE_precision": 0.9323308270676691, "eval_PREDICATE_recall": 0.9612403100775194, "eval_loss": 0.2312764674425125, "eval_overall_accuracy": 0.9567333693888589, "eval_overall_f1": 0.8329896907216495, "eval_overall_precision": 0.8381742738589212, "eval_overall_recall": 0.8278688524590164, "eval_runtime": 0.5097, "eval_samples_per_second": 188.335, "eval_steps_per_second": 23.542, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_f1": 0.0, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_number": 1, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_precision": 0.0, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_recall": 0.0, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.6865671641791045, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_number": 73, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.7540983606557377, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.6301369863013698, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.7272727272727273, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 41, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.6808510638297872, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.7804878048780488, "step": 432 }, { "epoch": 8.06, "learning_rate": 5.107375521825791e-06, "loss": 0.0066, "step": 435 }, { "epoch": 8.15, "learning_rate": 4.647861711610024e-06, "loss": 0.0083, "step": 440 }, { "epoch": 8.24, "learning_rate": 4.207882259287707e-06, "loss": 0.0091, "step": 445 }, { "epoch": 8.33, "learning_rate": 3.7878594652562575e-06, "loss": 0.009, "step": 450 }, { "epoch": 8.43, "learning_rate": 3.388196475144198e-06, "loss": 0.0047, "step": 455 }, { "epoch": 8.52, "learning_rate": 3.0092768928645375e-06, "loss": 0.0048, "step": 460 }, { "epoch": 8.61, "learning_rate": 2.651464412424667e-06, "loss": 0.0114, "step": 465 }, { "epoch": 8.7, "learning_rate": 2.315102468846142e-06, "loss": 0.0058, "step": 470 }, { "epoch": 8.8, "learning_rate": 2.0005139085293945e-06, "loss": 0.0049, "step": 475 }, { "epoch": 8.89, "learning_rate": 1.7080006793798176e-06, "loss": 0.0125, "step": 480 }, { "epoch": 8.98, "learning_rate": 1.4378435409925867e-06, "loss": 0.0089, "step": 485 }, { "epoch": 9.0, "eval_PREDICATE_f1": 0.9465648854961832, "eval_PREDICATE_number": 129, "eval_PREDICATE_precision": 0.9323308270676691, "eval_PREDICATE_recall": 0.9612403100775194, "eval_loss": 0.24090717732906342, "eval_overall_accuracy": 0.9588967009194159, "eval_overall_f1": 0.8418891170431212, "eval_overall_precision": 0.8436213991769548, "eval_overall_recall": 0.8401639344262295, "eval_runtime": 0.502, "eval_samples_per_second": 191.245, "eval_steps_per_second": 23.906, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_f1": 0.0, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_number": 1, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_precision": 0.0, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_recall": 0.0, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.7014925373134328, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_number": 73, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.7704918032786885, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.6438356164383562, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.7555555555555555, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 41, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.6938775510204082, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.8292682926829268, "step": 486 }, { "epoch": 9.07, "learning_rate": 1.1903017951744144e-06, "loss": 0.0055, "step": 490 }, { "epoch": 9.17, "learning_rate": 9.656130370609056e-07, "loss": 0.0089, "step": 495 }, { "epoch": 9.26, "learning_rate": 7.639929270683438e-07, "loss": 0.0064, "step": 500 }, { "epoch": 9.35, "learning_rate": 5.856349838988612e-07, "loss": 0.0057, "step": 505 }, { "epoch": 9.44, "learning_rate": 4.307103987976041e-07, "loss": 0.002, "step": 510 }, { "epoch": 9.54, "learning_rate": 2.9936787124022206e-07, "loss": 0.006, "step": 515 }, { "epoch": 9.63, "learning_rate": 1.917334662083714e-07, "loss": 0.0094, "step": 520 }, { "epoch": 9.72, "learning_rate": 1.0791049319021085e-07, "loss": 0.0044, "step": 525 }, { "epoch": 9.81, "learning_rate": 4.797940702205572e-08, "loss": 0.0092, "step": 530 }, { "epoch": 9.91, "learning_rate": 1.1997730666338247e-08, "loss": 0.009, "step": 535 }, { "epoch": 10.0, "learning_rate": 0.0, "loss": 0.0043, "step": 540 }, { "epoch": 10.0, "eval_PREDICATE_f1": 0.9465648854961832, "eval_PREDICATE_number": 129, "eval_PREDICATE_precision": 0.9323308270676691, "eval_PREDICATE_recall": 0.9612403100775194, "eval_loss": 0.2416924387216568, "eval_overall_accuracy": 0.9583558680367766, "eval_overall_f1": 0.8395061728395061, "eval_overall_precision": 0.8429752066115702, "eval_overall_recall": 0.8360655737704918, "eval_runtime": 0.5002, "eval_samples_per_second": 191.922, "eval_steps_per_second": 23.99, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_f1": 0.0, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_number": 1, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_precision": 0.0, "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_recall": 0.0, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.6917293233082706, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_number": 73, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.7666666666666667, "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.6301369863013698, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.7555555555555555, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 41, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.6938775510204082, "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.8292682926829268, "step": 540 }, { "epoch": 10.0, "step": 540, "total_flos": 127804763383596.0, "train_loss": 0.10899043403289936, "train_runtime": 173.8981, "train_samples_per_second": 49.397, "train_steps_per_second": 3.105 } ], "max_steps": 540, "num_train_epochs": 10, "total_flos": 127804763383596.0, "trial_name": null, "trial_params": null }