all_results.json

{
    "epoch": 3.0,
    "eval_logits/chosen": -1.5187865495681763,
    "eval_logits/rejected": -1.4330620765686035,
    "eval_logps/chosen": -1948.4508056640625,
    "eval_logps/rejected": -2216.220703125,
    "eval_loss": 1.8388876914978027,
    "eval_rewards/accuracies": 0.6904761791229248,
    "eval_rewards/chosen": -16.632295608520508,
    "eval_rewards/margins": 2.9380877017974854,
    "eval_rewards/margins_max": 11.99801254272461,
    "eval_rewards/margins_min": -5.294867038726807,
    "eval_rewards/margins_std": 7.78206729888916,
    "eval_rewards/rejected": -19.570384979248047,
    "eval_runtime": 281.8693,
    "eval_samples": 2000,
    "eval_samples_per_second": 7.095,
    "eval_steps_per_second": 0.224,
    "train_loss": 0.1103198329137612,
    "train_runtime": 9245.0119,
    "train_samples": 5678,
    "train_samples_per_second": 1.843,
    "train_steps_per_second": 0.115
}