all_results.json

{
    "epoch": 2.0,
    "eval_logits/chosen": -2.6612839698791504,
    "eval_logits/rejected": -2.6444251537323,
    "eval_logps/chosen": -71.76615142822266,
    "eval_logps/rejected": -80.43990325927734,
    "eval_loss": 1.1726570129394531,
    "eval_rewards/accuracies": 0.3472222089767456,
    "eval_rewards/chosen": 2.6978304386138916,
    "eval_rewards/margins": 2.016267776489258,
    "eval_rewards/rejected": 0.6815626621246338,
    "eval_runtime": 113.5073,
    "eval_samples": 2000,
    "eval_samples_per_second": 17.62,
    "eval_steps_per_second": 0.555,
    "total_flos": 0.0,
    "train_loss": 0.6661292189920406,
    "train_runtime": 6507.4651,
    "train_samples": 18340,
    "train_samples_per_second": 5.637,
    "train_steps_per_second": 0.088
}