m2m100_418M_bam_fr_news / trainer_state.json
Davlan's picture
add MT model
28666b7
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 10.0,
"global_step": 3020,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 1.66,
"learning_rate": 4.1721854304635764e-05,
"loss": 2.378,
"step": 500
},
{
"epoch": 3.31,
"learning_rate": 3.3443708609271526e-05,
"loss": 1.4014,
"step": 1000
},
{
"epoch": 4.97,
"learning_rate": 2.5165562913907287e-05,
"loss": 0.8795,
"step": 1500
},
{
"epoch": 6.62,
"learning_rate": 1.688741721854305e-05,
"loss": 0.5271,
"step": 2000
},
{
"epoch": 8.28,
"learning_rate": 8.609271523178809e-06,
"loss": 0.3473,
"step": 2500
},
{
"epoch": 9.93,
"learning_rate": 3.3112582781456954e-07,
"loss": 0.231,
"step": 3000
},
{
"epoch": 10.0,
"step": 3020,
"total_flos": 7709575043579904.0,
"train_loss": 0.9559793366501663,
"train_runtime": 1464.4836,
"train_samples_per_second": 20.574,
"train_steps_per_second": 2.062
}
],
"max_steps": 3020,
"num_train_epochs": 10,
"total_flos": 7709575043579904.0,
"trial_name": null,
"trial_params": null
}