all_results.json

{
    "epoch": 2.999297541394882,
    "eval_logits/chosen": 0.11780353635549545,
    "eval_logits/rejected": 0.21422991156578064,
    "eval_logps/chosen": -1.6016647815704346,
    "eval_logps/rejected": -2.2330212593078613,
    "eval_loss": 0.6684653759002686,
    "eval_rewards/accuracies": 0.6505934596061707,
    "eval_rewards/chosen": -1.6016647815704346,
    "eval_rewards/margins": 0.6313564777374268,
    "eval_rewards/rejected": -2.2330212593078613,
    "eval_runtime": 42.5859,
    "eval_samples": 1345,
    "eval_samples_per_second": 31.583,
    "eval_sft_loss": 1.5896871089935303,
    "eval_steps_per_second": 7.913,
    "total_flos": 0.0,
    "train_loss": 0.6685084665272595,
    "train_runtime": 33774.3849,
    "train_samples": 59790,
    "train_samples_per_second": 5.311,
    "train_steps_per_second": 0.166
}