| { | |
| "best_metric": 82.23530648613082, | |
| "best_model_checkpoint": "/home/jcanete/ft-data/all_results/sqac/roberta_large_bne/epochs_2_bs_32_lr_3e-5/checkpoint-1000", | |
| "epoch": 2.0, | |
| "global_step": 1034, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.39, | |
| "eval_exact_match": 55.847639484978544, | |
| "eval_f1": 74.93127077149742, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "eval_exact_match": 60.032188841201716, | |
| "eval_f1": 79.17500100284947, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.5522243713733078e-05, | |
| "loss": 1.3484, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "eval_exact_match": 63.09012875536481, | |
| "eval_f1": 80.62559091106165, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "eval_exact_match": 63.09012875536481, | |
| "eval_f1": 80.4832094493146, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 1.0154738878143134e-06, | |
| "loss": 0.3737, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "eval_exact_match": 64.91416309012875, | |
| "eval_f1": 82.23530648613082, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "step": 1034, | |
| "total_flos": 3.0615870790865664e+16, | |
| "train_loss": 0.8424169625258308, | |
| "train_runtime": 640.0293, | |
| "train_samples_per_second": 51.632, | |
| "train_steps_per_second": 1.616 | |
| } | |
| ], | |
| "max_steps": 1034, | |
| "num_train_epochs": 2, | |
| "total_flos": 3.0615870790865664e+16, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |