all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -1.563957691192627,
    "eval_logits/rejected": -1.5160561800003052,
    "eval_logps/chosen": -331.9000549316406,
    "eval_logps/rejected": -348.5905456542969,
    "eval_loss": 0.5600629448890686,
    "eval_rewards/accuracies": 0.699999988079071,
    "eval_rewards/chosen": -0.40962255001068115,
    "eval_rewards/margins": 0.5067135691642761,
    "eval_rewards/rejected": -0.9163362383842468,
    "eval_runtime": 469.9532,
    "eval_samples": 2000,
    "eval_samples_per_second": 4.256,
    "eval_steps_per_second": 1.064,
    "train_loss": 0.589801928247341,
    "train_runtime": 31982.1679,
    "train_samples": 61135,
    "train_samples_per_second": 1.912,
    "train_steps_per_second": 0.239
}