all_results.json

{
    "epoch": 2.999297541394882,
    "eval_logits/chosen": -0.36871230602264404,
    "eval_logits/rejected": -0.29547765851020813,
    "eval_logps/chosen": -4.4231061935424805,
    "eval_logps/rejected": -5.1417670249938965,
    "eval_loss": 0.05035458132624626,
    "eval_rewards/accuracies": 0.6862017512321472,
    "eval_rewards/chosen": -4.4231061935424805,
    "eval_rewards/margins": 0.7186610102653503,
    "eval_rewards/rejected": -5.1417670249938965,
    "eval_runtime": 44.074,
    "eval_samples": 1345,
    "eval_samples_per_second": 30.517,
    "eval_sft_loss": 4.028082370758057,
    "eval_steps_per_second": 7.646,
    "total_flos": 0.0,
    "train_loss": 0.0619825580417259,
    "train_runtime": 31817.9407,
    "train_samples": 59790,
    "train_samples_per_second": 5.637,
    "train_steps_per_second": 0.176
}