| { | |
| "best_metric": 1.169450044631958, | |
| "best_model_checkpoint": "./results/checkpoint-50", | |
| "epoch": 5.0, | |
| "eval_steps": 500, | |
| "global_step": 50, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "grad_norm": 2.3909294605255127, | |
| "learning_rate": 4.000000000000001e-06, | |
| "loss": 1.2093, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_loss": 1.2005947828292847, | |
| "eval_runtime": 1.0773, | |
| "eval_samples_per_second": 9.283, | |
| "eval_steps_per_second": 1.857, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 2.311216354370117, | |
| "learning_rate": 3e-06, | |
| "loss": 1.1937, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_loss": 1.1875404119491577, | |
| "eval_runtime": 1.0727, | |
| "eval_samples_per_second": 9.322, | |
| "eval_steps_per_second": 1.864, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 2.0154950618743896, | |
| "learning_rate": 2.0000000000000003e-06, | |
| "loss": 1.182, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_loss": 1.1778172254562378, | |
| "eval_runtime": 1.0761, | |
| "eval_samples_per_second": 9.293, | |
| "eval_steps_per_second": 1.859, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "grad_norm": 2.287374973297119, | |
| "learning_rate": 1.0000000000000002e-06, | |
| "loss": 1.1739, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_loss": 1.1716408729553223, | |
| "eval_runtime": 1.0824, | |
| "eval_samples_per_second": 9.239, | |
| "eval_steps_per_second": 1.848, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "grad_norm": 2.37646746635437, | |
| "learning_rate": 0.0, | |
| "loss": 1.1691, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_loss": 1.169450044631958, | |
| "eval_runtime": 1.0858, | |
| "eval_samples_per_second": 9.21, | |
| "eval_steps_per_second": 1.842, | |
| "step": 50 | |
| } | |
| ], | |
| "logging_steps": 10, | |
| "max_steps": 50, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 5, | |
| "save_steps": 500, | |
| "stateful_callbacks": { | |
| "TrainerControl": { | |
| "args": { | |
| "should_epoch_stop": false, | |
| "should_evaluate": false, | |
| "should_log": false, | |
| "should_save": true, | |
| "should_training_stop": true | |
| }, | |
| "attributes": {} | |
| } | |
| }, | |
| "total_flos": 2185441679769600.0, | |
| "train_batch_size": 1, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |