all_results.json

{
    "epoch": 2.9158878504672896,
    "eval_logits/chosen": -2.440101385116577,
    "eval_logits/rejected": -2.4343435764312744,
    "eval_logps/chosen": -248.78541564941406,
    "eval_logps/rejected": -190.92059326171875,
    "eval_loss": 0.5820934176445007,
    "eval_rewards/accuracies": 0.8125,
    "eval_rewards/chosen": 2.0596745014190674,
    "eval_rewards/margins": 1.809003233909607,
    "eval_rewards/rejected": 0.2506714165210724,
    "eval_runtime": 50.0587,
    "eval_samples_per_second": 15.182,
    "eval_steps_per_second": 0.24,
    "total_flos": 919378820333568.0,
    "train_loss": 0.36702350164071107,
    "train_runtime": 2733.1708,
    "train_samples_per_second": 7.506,
    "train_steps_per_second": 0.029
}