eval_results.json

{
    "epoch": 1.0,
    "eval_dpo_losses": 0.6380388140678406,
    "eval_logits/chosen": -2.6435492038726807,
    "eval_logits/rejected": -2.6096348762512207,
    "eval_logps/chosen": -261.558349609375,
    "eval_logps/rejected": -255.8979949951172,
    "eval_loss": 0.6949998140335083,
    "eval_positive_losses": 0.5819965600967407,
    "eval_rewards/accuracies": 0.7059999704360962,
    "eval_rewards/chosen": 0.22896993160247803,
    "eval_rewards/margins": 0.12941709160804749,
    "eval_rewards/margins_max": 0.5133692622184753,
    "eval_rewards/margins_min": -0.18140342831611633,
    "eval_rewards/margins_std": 0.2328445017337799,
    "eval_rewards/rejected": 0.09955283999443054,
    "eval_runtime": 858.1484,
    "eval_samples": 4000,
    "eval_samples_per_second": 4.661,
    "eval_steps_per_second": 0.291
}