roberta-base-mrpc / trainer_state.json
Jeremiah Zhou
End of training
d1d1516
raw
history blame contribute delete
839 Bytes
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 5.0,
"global_step": 1150,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 2.17,
"learning_rate": 1.2025901942645698e-05,
"loss": 0.4285,
"step": 500
},
{
"epoch": 4.35,
"learning_rate": 2.7752081406105456e-06,
"loss": 0.1748,
"step": 1000
},
{
"epoch": 5.0,
"step": 1150,
"total_flos": 1206364188825600.0,
"train_loss": 0.2764597121528957,
"train_runtime": 134.1282,
"train_samples_per_second": 136.735,
"train_steps_per_second": 8.574
}
],
"max_steps": 1150,
"num_train_epochs": 5,
"total_flos": 1206364188825600.0,
"trial_name": null,
"trial_params": null
}