all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": 14.445918083190918,
    "eval_logits/rejected": 14.679619789123535,
    "eval_logps/chosen": -4.605433940887451,
    "eval_logps/rejected": -4.770407676696777,
    "eval_loss": 3.9434103965759277,
    "eval_rewards/accuracies": 0.6472868323326111,
    "eval_rewards/chosen": -46.05434036254883,
    "eval_rewards/margins": 1.6497403383255005,
    "eval_rewards/rejected": -47.704078674316406,
    "eval_runtime": 439.853,
    "eval_samples_per_second": 112.481,
    "eval_steps_per_second": 1.76,
    "total_flos": 4181538261958656.0,
    "train_loss": 4.104885466542899,
    "train_runtime": 31282.7939,
    "train_samples_per_second": 30.049,
    "train_steps_per_second": 0.235
}