qwen2.5-0.5b-expo-DPO-ES-100 / eval_results.json
hZzy's picture
End of training
57a1b6c verified
{
"epoch": 2.69248937175248,
"eval_dpo_loss": 213.3031463623047,
"eval_logits": -0.6268911957740784,
"eval_logps": -80.2667236328125,
"eval_loss": 226.3468017578125,
"eval_objective": 213.3031463623047,
"eval_ranking_idealized": 0.5212215185165405,
"eval_ranking_idealized_expo": 0.5212215185165405,
"eval_ranking_simple": 0.5398550629615784,
"eval_regularize": 213.3031463623047,
"eval_runtime": 308.4568,
"eval_samples": 5790,
"eval_samples_per_second": 18.771,
"eval_steps_per_second": 1.566,
"eval_wo_beta": 6.621464252471924
}