{ "best_metric": null, "best_model_checkpoint": null, "epoch": 18.06451612903226, "global_step": 140, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.9, "eval_accuracy": 0.7493506493506493, "eval_f1": 0.0, "eval_loss": 2.0015974044799805, "eval_precision": 0.0, "eval_recall": 0.0, "eval_runtime": 0.0742, "eval_samples_per_second": 107.752, "eval_steps_per_second": 13.469, "step": 7 }, { "epoch": 1.94, "eval_accuracy": 0.7506493506493507, "eval_f1": 0.0, "eval_loss": 1.2346057891845703, "eval_precision": 0.0, "eval_recall": 0.0, "eval_runtime": 0.0731, "eval_samples_per_second": 109.424, "eval_steps_per_second": 13.678, "step": 15 }, { "epoch": 2.97, "eval_accuracy": 0.7506493506493507, "eval_f1": 0.0, "eval_loss": 1.0348587036132812, "eval_precision": 0.0, "eval_recall": 0.0, "eval_runtime": 0.0828, "eval_samples_per_second": 96.596, "eval_steps_per_second": 12.074, "step": 23 }, { "epoch": 4.0, "eval_accuracy": 0.7532467532467533, "eval_f1": 0.0, "eval_loss": 0.8629260659217834, "eval_precision": 0.0, "eval_recall": 0.0, "eval_runtime": 0.0723, "eval_samples_per_second": 110.649, "eval_steps_per_second": 13.831, "step": 31 }, { "epoch": 4.9, "eval_accuracy": 0.8064935064935065, "eval_f1": 0.18965517241379312, "eval_loss": 0.7592337727546692, "eval_precision": 0.25, "eval_recall": 0.1527777777777778, "eval_runtime": 0.0625, "eval_samples_per_second": 128.063, "eval_steps_per_second": 16.008, "step": 38 }, { "epoch": 5.94, "eval_accuracy": 0.8402597402597403, "eval_f1": 0.3356643356643357, "eval_loss": 0.6848217248916626, "eval_precision": 0.3380281690140845, "eval_recall": 0.3333333333333333, "eval_runtime": 0.0632, "eval_samples_per_second": 126.518, "eval_steps_per_second": 15.815, "step": 46 }, { "epoch": 6.97, "eval_accuracy": 0.8545454545454545, "eval_f1": 0.4081632653061225, "eval_loss": 0.6335278153419495, "eval_precision": 0.4, "eval_recall": 0.4166666666666667, "eval_runtime": 0.058, "eval_samples_per_second": 137.846, "eval_steps_per_second": 17.231, "step": 54 }, { "epoch": 8.0, "eval_accuracy": 0.861038961038961, "eval_f1": 0.4697986577181208, "eval_loss": 0.5917361974716187, "eval_precision": 0.45454545454545453, "eval_recall": 0.4861111111111111, "eval_runtime": 0.0572, "eval_samples_per_second": 139.965, "eval_steps_per_second": 17.496, "step": 62 }, { "epoch": 8.9, "eval_accuracy": 0.8727272727272727, "eval_f1": 0.4966442953020133, "eval_loss": 0.559086799621582, "eval_precision": 0.4805194805194805, "eval_recall": 0.5138888888888888, "eval_runtime": 0.0653, "eval_samples_per_second": 122.43, "eval_steps_per_second": 15.304, "step": 69 }, { "epoch": 9.94, "eval_accuracy": 0.8831168831168831, "eval_f1": 0.5490196078431373, "eval_loss": 0.5236128568649292, "eval_precision": 0.5185185185185185, "eval_recall": 0.5833333333333334, "eval_runtime": 0.0798, "eval_samples_per_second": 100.3, "eval_steps_per_second": 12.537, "step": 77 }, { "epoch": 10.97, "eval_accuracy": 0.8935064935064935, "eval_f1": 0.5866666666666667, "eval_loss": 0.49883803725242615, "eval_precision": 0.5641025641025641, "eval_recall": 0.6111111111111112, "eval_runtime": 0.082, "eval_samples_per_second": 97.527, "eval_steps_per_second": 12.191, "step": 85 }, { "epoch": 12.0, "eval_accuracy": 0.8935064935064935, "eval_f1": 0.5605095541401275, "eval_loss": 0.4820162355899811, "eval_precision": 0.5176470588235295, "eval_recall": 0.6111111111111112, "eval_runtime": 0.0889, "eval_samples_per_second": 90.026, "eval_steps_per_second": 11.253, "step": 93 }, { "epoch": 12.9, "eval_accuracy": 0.8935064935064935, "eval_f1": 0.576923076923077, "eval_loss": 0.4691239893436432, "eval_precision": 0.5357142857142857, "eval_recall": 0.625, "eval_runtime": 0.1004, "eval_samples_per_second": 79.701, "eval_steps_per_second": 9.963, "step": 100 }, { "epoch": 13.94, "eval_accuracy": 0.8961038961038961, "eval_f1": 0.5859872611464968, "eval_loss": 0.4579147398471832, "eval_precision": 0.5411764705882353, "eval_recall": 0.6388888888888888, "eval_runtime": 0.0904, "eval_samples_per_second": 88.539, "eval_steps_per_second": 11.067, "step": 108 }, { "epoch": 14.97, "eval_accuracy": 0.9012987012987013, "eval_f1": 0.5874999999999999, "eval_loss": 0.44733989238739014, "eval_precision": 0.5340909090909091, "eval_recall": 0.6527777777777778, "eval_runtime": 0.0796, "eval_samples_per_second": 100.53, "eval_steps_per_second": 12.566, "step": 116 }, { "epoch": 16.0, "eval_accuracy": 0.9, "eval_f1": 0.5822784810126581, "eval_loss": 0.44305217266082764, "eval_precision": 0.5348837209302325, "eval_recall": 0.6388888888888888, "eval_runtime": 0.0612, "eval_samples_per_second": 130.641, "eval_steps_per_second": 16.33, "step": 124 }, { "epoch": 16.9, "eval_accuracy": 0.9, "eval_f1": 0.5822784810126581, "eval_loss": 0.442893922328949, "eval_precision": 0.5348837209302325, "eval_recall": 0.6388888888888888, "eval_runtime": 0.0713, "eval_samples_per_second": 112.198, "eval_steps_per_second": 14.025, "step": 131 }, { "epoch": 17.94, "eval_accuracy": 0.8987012987012987, "eval_f1": 0.5732484076433121, "eval_loss": 0.4412069320678711, "eval_precision": 0.5294117647058824, "eval_recall": 0.625, "eval_runtime": 0.0835, "eval_samples_per_second": 95.82, "eval_steps_per_second": 11.977, "step": 139 }, { "epoch": 18.06, "eval_accuracy": 0.8987012987012987, "eval_f1": 0.5732484076433121, "eval_loss": 0.4411640167236328, "eval_precision": 0.5294117647058824, "eval_recall": 0.625, "eval_runtime": 0.0624, "eval_samples_per_second": 128.169, "eval_steps_per_second": 16.021, "step": 140 }, { "epoch": 18.06, "step": 140, "total_flos": 142786456035984.0, "train_loss": 0.6083726610456194, "train_runtime": 458.469, "train_samples_per_second": 5.278, "train_steps_per_second": 0.305 } ], "max_steps": 140, "num_train_epochs": 20, "total_flos": 142786456035984.0, "trial_name": null, "trial_params": null }