zephyr-dpo-qlora-uf-ours-5e-7 / eval_results.json
just1nseo's picture
End of training
de916d5 verified
{
"epoch": 3.0,
"eval_logits/chosen": -2.668572187423706,
"eval_logits/rejected": -2.635917901992798,
"eval_logps/chosen": -297.7778625488281,
"eval_logps/rejected": -285.26239013671875,
"eval_loss": 0.65846848487854,
"eval_rewards/accuracies": 0.6359999775886536,
"eval_rewards/chosen": -0.13184425234794617,
"eval_rewards/margins": 0.13499023020267487,
"eval_rewards/margins_max": 0.7123479843139648,
"eval_rewards/margins_min": -0.4480258822441101,
"eval_rewards/margins_std": 0.4010625183582306,
"eval_rewards/rejected": -0.26683446764945984,
"eval_runtime": 427.8892,
"eval_samples": 2000,
"eval_samples_per_second": 4.674,
"eval_steps_per_second": 0.292
}