TQL / last-checkpoint /trainer_state.json
naman1011's picture
Training in progress, step 6000, checkpoint
8e6ae5c
raw
history blame
642 Bytes
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 2.3455824863174355,
"eval_steps": 500,
"global_step": 6000,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 1.17,
"learning_rate": 0.0001,
"loss": 0.204,
"step": 3000
},
{
"epoch": 2.35,
"learning_rate": 0.0001,
"loss": 0.1084,
"step": 6000
}
],
"logging_steps": 3000,
"max_steps": 12790,
"num_train_epochs": 5,
"save_steps": 3000,
"total_flos": 78856039956480.0,
"trial_name": null,
"trial_params": null
}