qwen2.5-0.5b-expo-L1EXPO-ES-1 / eval_results.json
hZzy's picture
End of training
020800f verified
raw
history blame contribute delete
594 Bytes
{
"epoch": 1.9839395370807746,
"eval_dpo_loss": 2.573472738265991,
"eval_logits": -0.6935593485832214,
"eval_logps": -80.17532348632812,
"eval_loss": 4.835423946380615,
"eval_objective": 4.811424732208252,
"eval_ranking_idealized": 0.5295031070709229,
"eval_ranking_idealized_expo": 0.5212215185165405,
"eval_ranking_simple": 0.5248447060585022,
"eval_regularize": 4.811424732208252,
"eval_runtime": 308.0567,
"eval_samples": 5790,
"eval_samples_per_second": 18.795,
"eval_steps_per_second": 1.568,
"eval_wo_beta": 13.93560791015625
}