all_results.json

{
    "epoch": 2.999297541394882,
    "eval_logits/chosen": 0.015653692185878754,
    "eval_logits/rejected": 0.0926935076713562,
    "eval_logps/chosen": -4.668216705322266,
    "eval_logps/rejected": -5.721502304077148,
    "eval_loss": 0.5805643200874329,
    "eval_rewards/accuracies": 0.7292284965515137,
    "eval_rewards/chosen": -4.668216705322266,
    "eval_rewards/margins": 1.0532848834991455,
    "eval_rewards/rejected": -5.721502304077148,
    "eval_runtime": 42.6158,
    "eval_samples": 1345,
    "eval_samples_per_second": 31.561,
    "eval_sft_loss": 4.639245510101318,
    "eval_steps_per_second": 7.908,
    "total_flos": 0.0,
    "train_loss": 0.5697717436673725,
    "train_runtime": 31532.5259,
    "train_samples": 59790,
    "train_samples_per_second": 5.688,
    "train_steps_per_second": 0.178
}