all_results.json

{
    "epoch": 4.0,
    "eval_logits/chosen": -1.5872344970703125,
    "eval_logits/rejected": -1.58331298828125,
    "eval_logps/chosen": -39.727821350097656,
    "eval_logps/rejected": -44.08587646484375,
    "eval_loss": 0.9912604689598083,
    "eval_rewards/accuracies": 0.5510797500610352,
    "eval_rewards/chosen": -0.056932687759399414,
    "eval_rewards/margins": 0.008759919553995132,
    "eval_rewards/rejected": -0.06569261103868484,
    "eval_runtime": 145.7755,
    "eval_samples": 343,
    "eval_samples_per_second": 2.353,
    "eval_steps_per_second": 0.295,
    "train_loss": 0.7654441864459546,
    "train_runtime": 10797.8934,
    "train_samples": 3079,
    "train_samples_per_second": 1.141,
    "train_steps_per_second": 0.143
}