| { | |
| "best_metric": 0.980825229383001, | |
| "best_model_checkpoint": "omarmomen/sf_babylm_1/finetune/relative_position_control/checkpoint-400", | |
| "epoch": 10.0, | |
| "global_step": 1420, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.41, | |
| "eval_accuracy": 0.9216228127479553, | |
| "eval_f1": 0.9270501080358061, | |
| "eval_loss": 0.4737815260887146, | |
| "eval_mcc": 0.8527382673557574, | |
| "eval_runtime": 30.4369, | |
| "eval_samples_per_second": 594.409, | |
| "eval_steps_per_second": 74.318, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 2.82, | |
| "eval_accuracy": 0.9807097315788269, | |
| "eval_f1": 0.980825229383001, | |
| "eval_loss": 0.08114158362150192, | |
| "eval_mcc": 0.9614892142400866, | |
| "eval_runtime": 30.4073, | |
| "eval_samples_per_second": 594.988, | |
| "eval_steps_per_second": 74.39, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 3.52, | |
| "learning_rate": 3.23943661971831e-05, | |
| "loss": 0.0425, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 4.23, | |
| "eval_accuracy": 0.9766194820404053, | |
| "eval_f1": 0.9769545083083628, | |
| "eval_loss": 0.11881698668003082, | |
| "eval_mcc": 0.9536421310802985, | |
| "eval_runtime": 30.3659, | |
| "eval_samples_per_second": 595.799, | |
| "eval_steps_per_second": 74.491, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 5.63, | |
| "eval_accuracy": 0.9430134892463684, | |
| "eval_f1": 0.9460689438719464, | |
| "eval_loss": 0.407075434923172, | |
| "eval_mcc": 0.8917702307190046, | |
| "eval_runtime": 30.3806, | |
| "eval_samples_per_second": 595.512, | |
| "eval_steps_per_second": 74.455, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 7.04, | |
| "learning_rate": 1.4788732394366198e-05, | |
| "loss": 0.0001, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 7.04, | |
| "eval_accuracy": 0.9644041657447815, | |
| "eval_f1": 0.9655172413793104, | |
| "eval_loss": 0.2293713390827179, | |
| "eval_mcc": 0.930749953390411, | |
| "eval_runtime": 30.5501, | |
| "eval_samples_per_second": 592.207, | |
| "eval_steps_per_second": 74.042, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 8.45, | |
| "eval_accuracy": 0.9323458075523376, | |
| "eval_f1": 0.9365540120257101, | |
| "eval_loss": 0.5338804721832275, | |
| "eval_mcc": 0.8724016613441335, | |
| "eval_runtime": 30.3498, | |
| "eval_samples_per_second": 596.116, | |
| "eval_steps_per_second": 74.531, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 9.86, | |
| "eval_accuracy": 0.951028048992157, | |
| "eval_f1": 0.9530521407376007, | |
| "eval_loss": 0.35129231214523315, | |
| "eval_mcc": 0.9054283286330445, | |
| "eval_runtime": 30.3142, | |
| "eval_samples_per_second": 596.816, | |
| "eval_steps_per_second": 74.618, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "step": 1420, | |
| "total_flos": 2323809310863360.0, | |
| "train_loss": 0.015042712267907992, | |
| "train_runtime": 512.0019, | |
| "train_samples_per_second": 177.109, | |
| "train_steps_per_second": 2.773 | |
| } | |
| ], | |
| "max_steps": 1420, | |
| "num_train_epochs": 10, | |
| "total_flos": 2323809310863360.0, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |