all_results.json

{
    "epoch": 3.0,
    "eval_logits/chosen": -2.085988998413086,
    "eval_logits/rejected": -1.9190013408660889,
    "eval_logps/chosen": -248.65191650390625,
    "eval_logps/rejected": -198.58494567871094,
    "eval_loss": 0.5269633531570435,
    "eval_rewards/accuracies": 0.78125,
    "eval_rewards/chosen": -0.12098389863967896,
    "eval_rewards/margins": 0.8767741918563843,
    "eval_rewards/rejected": -0.9977580308914185,
    "eval_runtime": 49.9631,
    "eval_samples": 2000,
    "eval_samples_per_second": 40.03,
    "eval_steps_per_second": 0.32,
    "train_loss": 0.5643668570057567,
    "train_runtime": 8096.9375,
    "train_samples": 61966,
    "train_samples_per_second": 22.959,
    "train_steps_per_second": 0.359
}