| { | |
| "best_metric": 0.8256599140577041, | |
| "best_model_checkpoint": "BERTForDetectingDepression-Twitter2015/run-2/checkpoint-1101", | |
| "epoch": 3.0, | |
| "eval_steps": 500, | |
| "global_step": 1101, | |
| "is_hyper_param_search": true, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.7808471454880295, | |
| "eval_loss": 0.4515644609928131, | |
| "eval_runtime": 1.6125, | |
| "eval_samples_per_second": 1010.242, | |
| "eval_steps_per_second": 63.256, | |
| "step": 367 | |
| }, | |
| { | |
| "epoch": 1.3623978201634879, | |
| "grad_norm": 3.5979933738708496, | |
| "learning_rate": 2.9911866451614284e-06, | |
| "loss": 0.5208, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.8066298342541437, | |
| "eval_loss": 0.448224276304245, | |
| "eval_runtime": 1.6212, | |
| "eval_samples_per_second": 1004.815, | |
| "eval_steps_per_second": 62.917, | |
| "step": 734 | |
| }, | |
| { | |
| "epoch": 2.7247956403269757, | |
| "grad_norm": 4.29625940322876, | |
| "learning_rate": 5.026786209006726e-07, | |
| "loss": 0.325, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.8256599140577041, | |
| "eval_loss": 0.4320577085018158, | |
| "eval_runtime": 1.678, | |
| "eval_samples_per_second": 970.776, | |
| "eval_steps_per_second": 60.785, | |
| "step": 1101 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 1101, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 3, | |
| "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": 90217286439240.0, | |
| "train_batch_size": 4, | |
| "trial_name": null, | |
| "trial_params": { | |
| "learning_rate": 5.479694669422184e-06, | |
| "num_train_epochs": 3, | |
| "per_device_train_batch_size": 4, | |
| "seed": 23 | |
| } | |
| } | |