NNDA_ZH / checkpoint-700 /trainer_state.json
xiulinyang's picture
Adding model checkpoints and config files
cb5bad4
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 19.013333333333332,
"global_step": 700,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.0,
"learning_rate": 4.9999999999999996e-06,
"loss": 10.1072,
"step": 1
},
{
"epoch": 1.01,
"learning_rate": 0.00025,
"loss": 6.9379,
"step": 50
},
{
"epoch": 2.02,
"learning_rate": 0.0005,
"loss": 4.9342,
"step": 100
},
{
"epoch": 4.0,
"learning_rate": 0.0005833333333333333,
"loss": 4.3732,
"step": 150
},
{
"epoch": 5.02,
"learning_rate": 0.0005555555555555556,
"loss": 4.0714,
"step": 200
},
{
"epoch": 6.03,
"learning_rate": 0.0005277777777777777,
"loss": 3.9498,
"step": 250
},
{
"epoch": 8.01,
"learning_rate": 0.0005,
"loss": 3.8436,
"step": 300
},
{
"epoch": 9.02,
"learning_rate": 0.00047222222222222224,
"loss": 3.625,
"step": 350
},
{
"epoch": 11.0,
"learning_rate": 0.00044444444444444436,
"loss": 3.5012,
"step": 400
},
{
"epoch": 12.02,
"learning_rate": 0.00041666666666666664,
"loss": 3.3399,
"step": 450
},
{
"epoch": 13.03,
"learning_rate": 0.00038888888888888887,
"loss": 3.2438,
"step": 500
},
{
"epoch": 15.01,
"learning_rate": 0.0003611111111111111,
"loss": 3.1962,
"step": 550
},
{
"epoch": 16.02,
"learning_rate": 0.0003333333333333333,
"loss": 3.0903,
"step": 600
},
{
"epoch": 18.0,
"learning_rate": 0.00030555555555555555,
"loss": 3.0601,
"step": 650
},
{
"epoch": 19.01,
"learning_rate": 0.0002777777777777778,
"loss": 2.9726,
"step": 700
}
],
"max_steps": 1200,
"num_train_epochs": 9223372036854775807,
"total_flos": 1.90034559369216e+17,
"trial_name": null,
"trial_params": null
}