all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": 1.6140714883804321,
    "eval_logits/rejected": 2.3821487426757812,
    "eval_logps/chosen": -368.41351318359375,
    "eval_logps/rejected": -468.14495849609375,
    "eval_loss": 0.5042223334312439,
    "eval_rewards/accuracies": 0.75390625,
    "eval_rewards/chosen": -1.049981713294983,
    "eval_rewards/margins": 0.9979785680770874,
    "eval_rewards/rejected": -2.0479602813720703,
    "eval_runtime": 91.1991,
    "eval_samples": 2000,
    "eval_samples_per_second": 21.93,
    "eval_steps_per_second": 0.351,
    "train_loss": 0.5379065808890754,
    "train_runtime": 5396.8094,
    "train_samples": 61135,
    "train_samples_per_second": 11.328,
    "train_steps_per_second": 0.089
}