qwen2.5-0.5b-expo-L1EXPO / eval_results.json
hZzy's picture
End of training
672ff36 verified
raw
history blame contribute delete
565 Bytes
{
"epoch": 1.9992126604204392,
"eval_dpo_loss": 0.6928920149803162,
"eval_logits": -1.9894039630889893,
"eval_logps": -98.45726776123047,
"eval_loss": 0.008759393356740475,
"eval_objective": 0.008833886124193668,
"eval_ranking_idealized": 0.6022099256515503,
"eval_ranking_idealized_expo": 0.5207182168960571,
"eval_ranking_simple": 0.5179557800292969,
"eval_regularize": 0.008831382729113102,
"eval_runtime": 450.9955,
"eval_samples": 5790,
"eval_samples_per_second": 12.838,
"eval_steps_per_second": 3.211
}