all_results.json

{
    "epoch": 2.993762993762994,
    "eval_logits/chosen": -2.64923095703125,
    "eval_logits/rejected": -2.6339383125305176,
    "eval_logps/chosen": -243.1698455810547,
    "eval_logps/rejected": -225.00645446777344,
    "eval_loss": 0.47353264689445496,
    "eval_rewards/accuracies": 0.8217592835426331,
    "eval_rewards/chosen": 0.5419993996620178,
    "eval_rewards/margins": 2.8439457416534424,
    "eval_rewards/rejected": -2.3019466400146484,
    "eval_runtime": 228.1965,
    "eval_samples_per_second": 14.987,
    "eval_steps_per_second": 0.237,
    "total_flos": 4244978440077312.0,
    "train_loss": 0.33558651275104945,
    "train_runtime": 13885.8564,
    "train_samples_per_second": 6.648,
    "train_steps_per_second": 0.026
}