| { | |
| "best_metric": 0.9214503654296459, | |
| "best_model_checkpoint": "final_models/glue_models/structroberta_s2_50ep//finetune/control_raising_control/checkpoint-600", | |
| "epoch": 10.0, | |
| "global_step": 1030, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.94, | |
| "eval_accuracy": 0.9008369445800781, | |
| "eval_f1": 0.9063976863934542, | |
| "eval_loss": 0.4983138144016266, | |
| "eval_mcc": 0.8099430728731271, | |
| "eval_runtime": 28.5479, | |
| "eval_samples_per_second": 468.756, | |
| "eval_steps_per_second": 58.603, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 3.88, | |
| "eval_accuracy": 0.8905245661735535, | |
| "eval_f1": 0.8989446092294957, | |
| "eval_loss": 0.6522421836853027, | |
| "eval_mcc": 0.7956233558781975, | |
| "eval_runtime": 28.4619, | |
| "eval_samples_per_second": 470.173, | |
| "eval_steps_per_second": 58.78, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 4.85, | |
| "learning_rate": 2.5728155339805826e-05, | |
| "loss": 0.0708, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 5.83, | |
| "eval_accuracy": 0.917276918888092, | |
| "eval_f1": 0.9214503654296459, | |
| "eval_loss": 0.5907565951347351, | |
| "eval_mcc": 0.8416442761632831, | |
| "eval_runtime": 28.5402, | |
| "eval_samples_per_second": 468.882, | |
| "eval_steps_per_second": 58.619, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 7.77, | |
| "eval_accuracy": 0.8982962369918823, | |
| "eval_f1": 0.9057544491378713, | |
| "eval_loss": 0.7629005312919617, | |
| "eval_mcc": 0.810114823371703, | |
| "eval_runtime": 28.587, | |
| "eval_samples_per_second": 468.115, | |
| "eval_steps_per_second": 58.523, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 9.71, | |
| "learning_rate": 1.4563106796116506e-06, | |
| "loss": 0.0001, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 9.71, | |
| "eval_accuracy": 0.8985951542854309, | |
| "eval_f1": 0.9059923796328369, | |
| "eval_loss": 0.7728782892227173, | |
| "eval_mcc": 0.8105854793640443, | |
| "eval_runtime": 28.6363, | |
| "eval_samples_per_second": 467.31, | |
| "eval_steps_per_second": 58.422, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "step": 1030, | |
| "total_flos": 6018384229632000.0, | |
| "train_loss": 0.034445077063316856, | |
| "train_runtime": 499.6416, | |
| "train_samples_per_second": 131.494, | |
| "train_steps_per_second": 2.061 | |
| } | |
| ], | |
| "max_steps": 1030, | |
| "num_train_epochs": 10, | |
| "total_flos": 6018384229632000.0, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |