all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.4359169006347656,
    "eval_logits/rejected": -2.4180893898010254,
    "eval_logps/chosen": -74.32483673095703,
    "eval_logps/rejected": -86.18550872802734,
    "eval_loss": 0.6700397729873657,
    "eval_rewards/accuracies": 0.3154761791229248,
    "eval_rewards/chosen": 0.0016639787936583161,
    "eval_rewards/margins": 0.05223553627729416,
    "eval_rewards/rejected": -0.05057155340909958,
    "eval_runtime": 114.1287,
    "eval_samples": 2000,
    "eval_samples_per_second": 17.524,
    "eval_steps_per_second": 0.552,
    "total_flos": 0.0,
    "train_loss": 0.0,
    "train_runtime": 0.0351,
    "train_samples": 6113,
    "train_samples_per_second": 173972.916,
    "train_steps_per_second": 2732.112
}