| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 10.0, | |
| "global_step": 7880, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 4.7429305912596405e-05, | |
| "loss": 4.0354, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 4.42159383033419e-05, | |
| "loss": 3.1227, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 4.1002570694087404e-05, | |
| "loss": 2.8006, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 2.54, | |
| "learning_rate": 3.7789203084832907e-05, | |
| "loss": 2.5518, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 3.17, | |
| "learning_rate": 3.45758354755784e-05, | |
| "loss": 2.4603, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 3.81, | |
| "learning_rate": 3.1362467866323906e-05, | |
| "loss": 2.3477, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 4.44, | |
| "learning_rate": 2.8149100257069412e-05, | |
| "loss": 2.2727, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 5.08, | |
| "learning_rate": 2.493573264781491e-05, | |
| "loss": 2.2234, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 5.71, | |
| "learning_rate": 2.1722365038560414e-05, | |
| "loss": 2.1827, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 6.35, | |
| "learning_rate": 1.8508997429305914e-05, | |
| "loss": 2.1235, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 6.98, | |
| "learning_rate": 1.5295629820051416e-05, | |
| "loss": 2.1156, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 7.61, | |
| "learning_rate": 1.2082262210796916e-05, | |
| "loss": 2.0648, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 8.25, | |
| "learning_rate": 8.868894601542417e-06, | |
| "loss": 2.0744, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 8.88, | |
| "learning_rate": 5.6555269922879175e-06, | |
| "loss": 2.0276, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 9.52, | |
| "learning_rate": 2.442159383033419e-06, | |
| "loss": 2.0262, | |
| "step": 7500 | |
| } | |
| ], | |
| "max_steps": 7880, | |
| "num_train_epochs": 10, | |
| "total_flos": 502540376306640.0, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |