| { | |
| "best_metric": 2.216588020324707, | |
| "best_model_checkpoint": "cmarkea/distilcamembert-base-emoji-pred/checkpoint-2804", | |
| "epoch": 2.0, | |
| "global_step": 2804, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.8811222063718498e-05, | |
| "loss": 2.5341, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 1.7622444127436998e-05, | |
| "loss": 2.329, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.3152823623873673, | |
| "eval_loss": 2.2571091651916504, | |
| "eval_runtime": 107.5303, | |
| "eval_samples_per_second": 104.24, | |
| "eval_steps_per_second": 3.264, | |
| "step": 1402 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 1.6433666191155494e-05, | |
| "loss": 2.2509, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 1.524488825487399e-05, | |
| "loss": 2.1789, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 1.4056110318592488e-05, | |
| "loss": 2.1714, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.32375769470960836, | |
| "eval_loss": 2.216588020324707, | |
| "eval_runtime": 107.7873, | |
| "eval_samples_per_second": 103.992, | |
| "eval_steps_per_second": 3.256, | |
| "step": 2804 | |
| } | |
| ], | |
| "max_steps": 8412, | |
| "num_train_epochs": 6, | |
| "total_flos": 1.188348885233664e+16, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |