| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 0.16789087093389296, | |
| "global_step": 1600, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.04, | |
| "eval_accuracy": 0.6216116513450103, | |
| "eval_f1": 0.5350163270245339, | |
| "eval_loss": 1.2418967485427856, | |
| "eval_runtime": 545.576, | |
| "eval_samples_per_second": 248.431, | |
| "eval_steps_per_second": 0.971, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 9.994805876180483e-06, | |
| "loss": 1.4969, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "eval_accuracy": 0.741024657291682, | |
| "eval_f1": 0.7346338559187492, | |
| "eval_loss": 1.0078158378601074, | |
| "eval_runtime": 547.2541, | |
| "eval_samples_per_second": 247.669, | |
| "eval_steps_per_second": 0.968, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.9895592864638e-06, | |
| "loss": 1.0807, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "eval_accuracy": 0.7675633401702844, | |
| "eval_f1": 0.7670752144773497, | |
| "eval_loss": 0.9420500993728638, | |
| "eval_runtime": 547.3649, | |
| "eval_samples_per_second": 247.619, | |
| "eval_steps_per_second": 0.968, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 9.984323189926549e-06, | |
| "loss": 0.9987, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "eval_accuracy": 0.7890333338252004, | |
| "eval_f1": 0.785083992813201, | |
| "eval_loss": 0.900003969669342, | |
| "eval_runtime": 547.6023, | |
| "eval_samples_per_second": 247.512, | |
| "eval_steps_per_second": 0.968, | |
| "step": 1600 | |
| } | |
| ], | |
| "max_steps": 953000, | |
| "num_train_epochs": 100, | |
| "total_flos": 6.580345049950003e+16, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |