oh-dcft-v3.1-llama-3.1-8b-qwen / all_results.json
sedrickkeh's picture
End of training
b2df411 verified
raw
history blame
360 Bytes
{
"epoch": 2.9991537376586743,
"eval_loss": 0.440873384475708,
"eval_runtime": 447.5541,
"eval_samples_per_second": 26.674,
"eval_steps_per_second": 0.418,
"total_flos": 2786674505416704.0,
"train_loss": 0.4018145801655057,
"train_runtime": 71328.9114,
"train_samples_per_second": 9.54,
"train_steps_per_second": 0.075
}