all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -1.532152771949768,
    "eval_logits/rejected": -1.4972456693649292,
    "eval_logps/chosen": -261.3055114746094,
    "eval_logps/rejected": -286.0805358886719,
    "eval_loss": 0.6392669677734375,
    "eval_rewards/accuracies": 0.6036585569381714,
    "eval_rewards/chosen": -1.0972425937652588,
    "eval_rewards/margins": 0.2594923973083496,
    "eval_rewards/rejected": -1.356735110282898,
    "eval_runtime": 169.8112,
    "eval_samples": 1962,
    "eval_samples_per_second": 11.554,
    "eval_steps_per_second": 1.449,
    "total_flos": 0.0,
    "train_loss": 0.6256769998495513,
    "train_runtime": 22377.6313,
    "train_samples": 60029,
    "train_samples_per_second": 2.683,
    "train_steps_per_second": 0.042
}