all_results.json

{
    "epoch": 3.0,
    "eval_logits/chosen": -2.508760690689087,
    "eval_logits/rejected": -2.47902250289917,
    "eval_logps/chosen": -319.566650390625,
    "eval_logps/rejected": -311.4621276855469,
    "eval_loss": 0.7239906191825867,
    "eval_rewards/accuracies": 0.7639999985694885,
    "eval_rewards/chosen": -4.3843231201171875,
    "eval_rewards/margins": 3.5257680416107178,
    "eval_rewards/rejected": -7.910090446472168,
    "eval_runtime": 462.674,
    "eval_samples": 2000,
    "eval_samples_per_second": 4.323,
    "eval_steps_per_second": 0.27,
    "train_loss": 0.21807545795603994,
    "train_runtime": 74953.8161,
    "train_samples": 61966,
    "train_samples_per_second": 2.48,
    "train_steps_per_second": 0.078
}