all_results.json

{
    "epoch": 3.0,
    "eval_logits/chosen": -1.588410496711731,
    "eval_logits/rejected": -1.3595237731933594,
    "eval_logps/chosen": -383.10223388671875,
    "eval_logps/rejected": -666.2252807617188,
    "eval_loss": 0.005952088162302971,
    "eval_rewards/accuracies": 0.996632993221283,
    "eval_rewards/chosen": 3.6820499897003174,
    "eval_rewards/margins": 14.352936744689941,
    "eval_rewards/rejected": -10.670886993408203,
    "eval_runtime": 577.443,
    "eval_samples": 9500,
    "eval_samples_per_second": 16.452,
    "eval_steps_per_second": 0.514,
    "train_loss": 0.053706495013128505,
    "train_runtime": 110737.0914,
    "train_samples": 188284,
    "train_samples_per_second": 5.101,
    "train_steps_per_second": 0.08
}