all_results.json

{
    "epoch": 3.0,
    "eval_logits/chosen": -2.2545087337493896,
    "eval_logits/rejected": -2.2175943851470947,
    "eval_logps/chosen": -183.74139404296875,
    "eval_logps/rejected": -203.22280883789062,
    "eval_loss": 0.552519679069519,
    "eval_rewards/accuracies": 0.734375,
    "eval_rewards/chosen": -2.5933074951171875,
    "eval_rewards/margins": 1.6603199243545532,
    "eval_rewards/rejected": -4.253627300262451,
    "eval_runtime": 127.5607,
    "eval_samples_per_second": 16.016,
    "eval_steps_per_second": 0.251,
    "total_flos": 2546799910846464.0,
    "train_loss": 0.29666884402158084,
    "train_runtime": 7681.0779,
    "train_samples_per_second": 7.181,
    "train_steps_per_second": 0.028
}