summ / trainer_state.json
aashour's picture
push test model
38bf2a4
raw
history blame
2.12 kB
{
"best_metric": 0.061439353972673416,
"best_model_checkpoint": "/tmp/debug_covidfact_gatortron/checkpoint-1248",
"epoch": 6.0,
"global_step": 1248,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 1.0,
"eval_accuracy": 0.9041591320072333,
"eval_loss": 0.2822079658508301,
"eval_runtime": 10.5311,
"eval_samples_per_second": 52.511,
"eval_steps_per_second": 6.647,
"step": 208
},
{
"epoch": 2.0,
"eval_accuracy": 0.9710669077757685,
"eval_loss": 0.10182330012321472,
"eval_runtime": 10.4772,
"eval_samples_per_second": 52.781,
"eval_steps_per_second": 6.681,
"step": 416
},
{
"epoch": 2.4,
"learning_rate": 3.798076923076923e-05,
"loss": 0.193,
"step": 500
},
{
"epoch": 3.0,
"eval_accuracy": 0.9674502712477396,
"eval_loss": 0.12143626064062119,
"eval_runtime": 10.3087,
"eval_samples_per_second": 53.644,
"eval_steps_per_second": 6.79,
"step": 624
},
{
"epoch": 4.0,
"eval_accuracy": 0.9801084990958409,
"eval_loss": 0.07419443130493164,
"eval_runtime": 10.2802,
"eval_samples_per_second": 53.793,
"eval_steps_per_second": 6.809,
"step": 832
},
{
"epoch": 4.81,
"learning_rate": 2.5961538461538464e-05,
"loss": 0.025,
"step": 1000
},
{
"epoch": 5.0,
"eval_accuracy": 0.976491862567812,
"eval_loss": 0.07501477003097534,
"eval_runtime": 10.2703,
"eval_samples_per_second": 53.845,
"eval_steps_per_second": 6.816,
"step": 1040
},
{
"epoch": 6.0,
"eval_accuracy": 0.976491862567812,
"eval_loss": 0.061439353972673416,
"eval_runtime": 10.3008,
"eval_samples_per_second": 53.685,
"eval_steps_per_second": 6.796,
"step": 1248
}
],
"max_steps": 2080,
"num_train_epochs": 10,
"total_flos": 1.3910473494715392e+16,
"trial_name": null,
"trial_params": null
}