all_results.json

{
    "epoch": 2.9724770642201834,
    "eval_logits/chosen": -2.20833683013916,
    "eval_logits/rejected": -2.1962075233459473,
    "eval_logps/chosen": -315.7890319824219,
    "eval_logps/rejected": -252.42320251464844,
    "eval_loss": 0.4124302268028259,
    "eval_rewards/accuracies": 0.8557692170143127,
    "eval_rewards/chosen": 3.4337260723114014,
    "eval_rewards/margins": 3.355928659439087,
    "eval_rewards/rejected": 0.07779762148857117,
    "eval_runtime": 53.6255,
    "eval_samples_per_second": 14.34,
    "eval_steps_per_second": 0.242,
    "total_flos": 954757539692544.0,
    "train_loss": 0.21779433532077588,
    "train_runtime": 2944.3137,
    "train_samples_per_second": 7.049,
    "train_steps_per_second": 0.028
}