|
{ |
|
"best_metric": 18.8732, |
|
"best_model_checkpoint": "/local1/hfs/gs_stuff/ft-wmt14/0-0002/checkpoint-5000", |
|
"epoch": 0.01774311168046785, |
|
"eval_steps": 5000, |
|
"global_step": 5000, |
|
"is_hyper_param_search": false, |
|
"is_local_process_zero": true, |
|
"is_world_process_zero": true, |
|
"log_history": [ |
|
{ |
|
"epoch": 0.01774311168046785, |
|
"grad_norm": 1.9640109539031982, |
|
"learning_rate": 0.0, |
|
"loss": 2.4878, |
|
"step": 5000 |
|
}, |
|
{ |
|
"epoch": 0.01774311168046785, |
|
"eval_bleu": 18.8732, |
|
"eval_gen_len": 28.2253, |
|
"eval_loss": 1.947550892829895, |
|
"eval_runtime": 259.3729, |
|
"eval_samples_per_second": 11.566, |
|
"eval_steps_per_second": 1.446, |
|
"step": 5000 |
|
}, |
|
{ |
|
"epoch": 0.01774311168046785, |
|
"step": 5000, |
|
"total_flos": 1786854363365376.0, |
|
"train_loss": 2.4878458984375, |
|
"train_runtime": 952.7348, |
|
"train_samples_per_second": 83.969, |
|
"train_steps_per_second": 5.248 |
|
} |
|
], |
|
"logging_steps": 5000, |
|
"max_steps": 5000, |
|
"num_input_tokens_seen": 0, |
|
"num_train_epochs": 1, |
|
"save_steps": 5000, |
|
"total_flos": 1786854363365376.0, |
|
"train_batch_size": 8, |
|
"trial_name": null, |
|
"trial_params": null |
|
} |
|
|