mbart50_ibo_en_news / trainer_state.json
Davlan's picture
add MT model
2259ca6
raw
history blame
1.08 kB
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 3.0,
"global_step": 2100,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.71,
"learning_rate": 3.809523809523809e-05,
"loss": 5.9819,
"step": 500
},
{
"epoch": 1.43,
"learning_rate": 2.6190476190476192e-05,
"loss": 2.5892,
"step": 1000
},
{
"epoch": 2.14,
"learning_rate": 1.4285714285714285e-05,
"loss": 1.7507,
"step": 1500
},
{
"epoch": 2.86,
"learning_rate": 2.3809523809523808e-06,
"loss": 1.3433,
"step": 2000
},
{
"epoch": 3.0,
"step": 2100,
"total_flos": 7634802405113856.0,
"train_loss": 2.8371728079659597,
"train_runtime": 899.693,
"train_samples_per_second": 23.335,
"train_steps_per_second": 2.334
}
],
"max_steps": 2100,
"num_train_epochs": 3,
"total_flos": 7634802405113856.0,
"trial_name": null,
"trial_params": null
}