|
{ |
|
"best_metric": null, |
|
"best_model_checkpoint": null, |
|
"epoch": 5.0, |
|
"global_step": 4585, |
|
"is_hyper_param_search": false, |
|
"is_local_process_zero": true, |
|
"is_world_process_zero": true, |
|
"log_history": [ |
|
{ |
|
"epoch": 1.0, |
|
"learning_rate": 1.6000000000000003e-05, |
|
"loss": 0.6441, |
|
"step": 917 |
|
}, |
|
{ |
|
"epoch": 1.0, |
|
"eval_accuracy": 0.6838235294117647, |
|
"eval_combined_score": 0.7480253018237863, |
|
"eval_f1": 0.8122270742358079, |
|
"eval_loss": 0.6369990706443787, |
|
"eval_runtime": 13.1449, |
|
"eval_samples_per_second": 31.039, |
|
"eval_steps_per_second": 3.88, |
|
"step": 917 |
|
}, |
|
{ |
|
"epoch": 2.0, |
|
"learning_rate": 1.2e-05, |
|
"loss": 0.6451, |
|
"step": 1834 |
|
}, |
|
{ |
|
"epoch": 2.0, |
|
"eval_accuracy": 0.6838235294117647, |
|
"eval_combined_score": 0.7480253018237863, |
|
"eval_f1": 0.8122270742358079, |
|
"eval_loss": 0.6552556157112122, |
|
"eval_runtime": 13.1957, |
|
"eval_samples_per_second": 30.919, |
|
"eval_steps_per_second": 3.865, |
|
"step": 1834 |
|
}, |
|
{ |
|
"epoch": 3.0, |
|
"learning_rate": 8.000000000000001e-06, |
|
"loss": 0.6428, |
|
"step": 2751 |
|
}, |
|
{ |
|
"epoch": 3.0, |
|
"eval_accuracy": 0.6838235294117647, |
|
"eval_combined_score": 0.7480253018237863, |
|
"eval_f1": 0.8122270742358079, |
|
"eval_loss": 0.6332471966743469, |
|
"eval_runtime": 13.1447, |
|
"eval_samples_per_second": 31.039, |
|
"eval_steps_per_second": 3.88, |
|
"step": 2751 |
|
}, |
|
{ |
|
"epoch": 4.0, |
|
"learning_rate": 4.000000000000001e-06, |
|
"loss": 0.6476, |
|
"step": 3668 |
|
}, |
|
{ |
|
"epoch": 4.0, |
|
"eval_accuracy": 0.6838235294117647, |
|
"eval_combined_score": 0.7480253018237863, |
|
"eval_f1": 0.8122270742358079, |
|
"eval_loss": 0.6247974634170532, |
|
"eval_runtime": 13.1288, |
|
"eval_samples_per_second": 31.077, |
|
"eval_steps_per_second": 3.885, |
|
"step": 3668 |
|
}, |
|
{ |
|
"epoch": 5.0, |
|
"learning_rate": 0.0, |
|
"loss": 0.6499, |
|
"step": 4585 |
|
}, |
|
{ |
|
"epoch": 5.0, |
|
"eval_accuracy": 0.6838235294117647, |
|
"eval_combined_score": 0.7480253018237863, |
|
"eval_f1": 0.8122270742358079, |
|
"eval_loss": 0.6273975968360901, |
|
"eval_runtime": 13.1363, |
|
"eval_samples_per_second": 31.059, |
|
"eval_steps_per_second": 3.882, |
|
"step": 4585 |
|
}, |
|
{ |
|
"epoch": 5.0, |
|
"step": 4585, |
|
"total_flos": 1.709162120343552e+16, |
|
"train_loss": 0.6458905701548868, |
|
"train_runtime": 2431.5157, |
|
"train_samples_per_second": 7.543, |
|
"train_steps_per_second": 1.886 |
|
} |
|
], |
|
"max_steps": 4585, |
|
"num_train_epochs": 5, |
|
"total_flos": 1.709162120343552e+16, |
|
"trial_name": null, |
|
"trial_params": null |
|
} |
|
|