RedPajama-INCITE-Base-3B-v1-wikipedia-8bit
/
flyte_4dtz63w
/local_flytekit
/5e5d536b4c662b7a5699a1dfbf8a9c95
/trainer_state.json
{ | |
"best_metric": null, | |
"best_model_checkpoint": null, | |
"epoch": 6.666666666666667, | |
"global_step": 10, | |
"is_hyper_param_search": false, | |
"is_local_process_zero": true, | |
"is_world_process_zero": true, | |
"log_history": [ | |
{ | |
"epoch": 6.67, | |
"learning_rate": 2e-05, | |
"loss": 2.0296, | |
"step": 10 | |
}, | |
{ | |
"epoch": 6.67, | |
"step": 10, | |
"total_flos": 4181427486720.0, | |
"train_loss": 2.029572677612305, | |
"train_runtime": 431.4795, | |
"train_samples_per_second": 2.967, | |
"train_steps_per_second": 0.023 | |
} | |
], | |
"max_steps": 10, | |
"num_train_epochs": 10, | |
"total_flos": 4181427486720.0, | |
"trial_name": null, | |
"trial_params": null | |
} | |