| { | |
| "best_metric": 0.6314101444719962, | |
| "best_model_checkpoint": "test-klue/ynat/run-2/checkpoint-348", | |
| "epoch": 4.0, | |
| "eval_steps": 500, | |
| "global_step": 348, | |
| "is_hyper_param_search": true, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "eval_f1": 0.5269240698183681, | |
| "eval_loss": 0.9749890565872192, | |
| "eval_runtime": 9.3462, | |
| "eval_samples_per_second": 253.471, | |
| "eval_steps_per_second": 2.033, | |
| "step": 87 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_f1": 0.5891366613697401, | |
| "eval_loss": 0.8359125852584839, | |
| "eval_runtime": 9.3824, | |
| "eval_samples_per_second": 252.493, | |
| "eval_steps_per_second": 2.025, | |
| "step": 174 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_f1": 0.6193320854823038, | |
| "eval_loss": 0.9428530335426331, | |
| "eval_runtime": 9.3908, | |
| "eval_samples_per_second": 252.268, | |
| "eval_steps_per_second": 2.023, | |
| "step": 261 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_f1": 0.6314101444719962, | |
| "eval_loss": 1.1509172916412354, | |
| "eval_runtime": 9.4376, | |
| "eval_samples_per_second": 251.016, | |
| "eval_steps_per_second": 2.013, | |
| "step": 348 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 348, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 4, | |
| "save_steps": 500, | |
| "total_flos": 0, | |
| "train_batch_size": 64, | |
| "trial_name": null, | |
| "trial_params": { | |
| "learning_rate": 8.256323465193199e-05, | |
| "num_train_epochs": 4, | |
| "per_device_train_batch_size": 64, | |
| "seed": 15 | |
| } | |
| } | |