all_results.json

{
    "epoch": 0.9982631930527722,
    "eval_logits/chosen": -1.4595378637313843,
    "eval_logits/rejected": -1.4681200981140137,
    "eval_logps/chosen": -1.496623158454895,
    "eval_logps/rejected": -2.0228073596954346,
    "eval_loss": 1.3633201122283936,
    "eval_rewards/accuracies": 0.8373983502388,
    "eval_rewards/chosen": -14.966232299804688,
    "eval_rewards/margins": 5.261841297149658,
    "eval_rewards/rejected": -20.22807502746582,
    "eval_runtime": 87.8052,
    "eval_samples": 1961,
    "eval_samples_per_second": 22.334,
    "eval_steps_per_second": 1.401,
    "total_flos": 0.0,
    "train_loss": 1.9910774169693157,
    "train_runtime": 11451.8288,
    "train_samples": 59876,
    "train_samples_per_second": 5.229,
    "train_steps_per_second": 0.041
}