all_results.json

{
    "epoch": 1.971563981042654,
    "eval_logits/chosen": 92.7257308959961,
    "eval_logits/rejected": 86.90755462646484,
    "eval_logps/chosen": -470.2870178222656,
    "eval_logps/rejected": -508.6237487792969,
    "eval_loss": 0.2685789465904236,
    "eval_rewards/accuracies": 0.7291666865348816,
    "eval_rewards/chosen": -5.330941677093506,
    "eval_rewards/margins": 1.9966983795166016,
    "eval_rewards/rejected": -7.327640533447266,
    "eval_runtime": 118.8627,
    "eval_samples": 750,
    "eval_samples_per_second": 6.31,
    "eval_steps_per_second": 0.202,
    "total_flos": 0.0,
    "train_loss": 0.2568448650149199,
    "train_runtime": 2175.3939,
    "train_samples": 6750,
    "train_samples_per_second": 6.206,
    "train_steps_per_second": 0.048
}