| { | |
| "best_metric": 0.8564009070396423, | |
| "best_model_checkpoint": "./results/run-eb32b854/checkpoint-882", | |
| "epoch": 4.0, | |
| "global_step": 3528, | |
| "is_hyper_param_search": true, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 3.1798763444888135e-05, | |
| "loss": 2.0373, | |
| "step": 1 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_f1": 0.4929364503071642, | |
| "eval_loss": 0.8564009070396423, | |
| "eval_runtime": 19.2818, | |
| "eval_samples_per_second": 121.98, | |
| "eval_steps_per_second": 15.248, | |
| "step": 882 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 2.2791968808811233e-05, | |
| "loss": 0.9246, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_f1": 0.6146216192253161, | |
| "eval_loss": 0.8962076306343079, | |
| "eval_runtime": 19.3209, | |
| "eval_samples_per_second": 121.734, | |
| "eval_steps_per_second": 15.217, | |
| "step": 1764 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 1.3776158362287801e-05, | |
| "loss": 0.4402, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_f1": 0.6171442898225464, | |
| "eval_loss": 1.2047011852264404, | |
| "eval_runtime": 19.3331, | |
| "eval_samples_per_second": 121.656, | |
| "eval_steps_per_second": 15.207, | |
| "step": 2646 | |
| }, | |
| { | |
| "epoch": 3.4, | |
| "learning_rate": 4.760347915764371e-06, | |
| "loss": 0.169, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_f1": 0.6443113896695691, | |
| "eval_loss": 1.3850061893463135, | |
| "eval_runtime": 19.3579, | |
| "eval_samples_per_second": 121.501, | |
| "eval_steps_per_second": 15.188, | |
| "step": 3528 | |
| } | |
| ], | |
| "max_steps": 3528, | |
| "num_train_epochs": 4, | |
| "total_flos": 5472186620003712.0, | |
| "trial_name": null, | |
| "trial_params": { | |
| "learning_rate": 3.180777925533466e-05, | |
| "num_train_epochs": 4, | |
| "per_device_train_batch_size": 8, | |
| "warmup_steps": 0, | |
| "weight_decay": 0.120996580489383 | |
| } | |
| } | |