qwen2.5-0.5b-expo-L2EXPO-ES-10 / eval_results.json
hZzy's picture
End of training
8ac4c6a verified
raw
history blame contribute delete
595 Bytes
{
"epoch": 1.7005196032120926,
"eval_dpo_loss": 18.911605834960938,
"eval_logits": -0.8236067295074463,
"eval_logps": -75.27147674560547,
"eval_loss": 38.52634048461914,
"eval_objective": 37.43656921386719,
"eval_ranking_idealized": 0.5212215185165405,
"eval_ranking_idealized_expo": 0.5212215185165405,
"eval_ranking_simple": 0.5295031070709229,
"eval_regularize": 37.43656921386719,
"eval_runtime": 308.2182,
"eval_samples": 5790,
"eval_samples_per_second": 18.785,
"eval_steps_per_second": 1.567,
"eval_wo_beta": 14.68159294128418
}