DNAN_EN / checkpoint-700 /trainer_state.json
xiulinyang's picture
Adding model checkpoints and config files
26a4811
{
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 25.020833333333332,
"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": 11.0218,
"step": 1
},
{
"epoch": 1.02,
"learning_rate": 0.00025,
"loss": 7.9982,
"step": 50
},
{
"epoch": 3.02,
"learning_rate": 0.0005,
"loss": 5.8003,
"step": 100
},
{
"epoch": 5.01,
"learning_rate": 0.0005833333333333333,
"loss": 5.17,
"step": 150
},
{
"epoch": 7.01,
"learning_rate": 0.0005555555555555556,
"loss": 4.9158,
"step": 200
},
{
"epoch": 9.01,
"learning_rate": 0.0005277777777777777,
"loss": 4.7809,
"step": 250
},
{
"epoch": 11.0,
"learning_rate": 0.0005,
"loss": 4.6208,
"step": 300
},
{
"epoch": 12.02,
"learning_rate": 0.00047222222222222224,
"loss": 4.3481,
"step": 350
},
{
"epoch": 14.02,
"learning_rate": 0.00044444444444444436,
"loss": 4.2378,
"step": 400
},
{
"epoch": 16.02,
"learning_rate": 0.00041666666666666664,
"loss": 4.0729,
"step": 450
},
{
"epoch": 18.01,
"learning_rate": 0.00038888888888888887,
"loss": 3.941,
"step": 500
},
{
"epoch": 20.01,
"learning_rate": 0.0003611111111111111,
"loss": 3.826,
"step": 550
},
{
"epoch": 22.0,
"learning_rate": 0.0003333333333333333,
"loss": 3.7276,
"step": 600
},
{
"epoch": 24.0,
"learning_rate": 0.00030555555555555555,
"loss": 3.6415,
"step": 650
},
{
"epoch": 25.02,
"learning_rate": 0.0002777777777777778,
"loss": 3.4952,
"step": 700
}
],
"max_steps": 1200,
"num_train_epochs": 9223372036854775807,
"total_flos": 1.936696541184e+17,
"trial_name": null,
"trial_params": null
}