all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.6727941036224365,
    "eval_logits/rejected": -2.6590042114257812,
    "eval_logps/chosen": -70.43208312988281,
    "eval_logps/rejected": -92.74579620361328,
    "eval_loss": 0.24523001909255981,
    "eval_rewards/accuracies": 0.913241982460022,
    "eval_rewards/chosen": -0.7311916947364807,
    "eval_rewards/margins": 2.0473380088806152,
    "eval_rewards/rejected": -2.778529405593872,
    "eval_runtime": 288.4496,
    "eval_samples": 876,
    "eval_samples_per_second": 3.037,
    "eval_steps_per_second": 3.037,
    "train_loss": 0.3616804119643815,
    "train_runtime": 2568.179,
    "train_samples": 3505,
    "train_samples_per_second": 1.365,
    "train_steps_per_second": 0.042
}