all_results.json

{
    "epoch": 1.971563981042654,
    "eval_logits/chosen": 101.30475616455078,
    "eval_logits/rejected": 95.51566314697266,
    "eval_logps/chosen": -407.296630859375,
    "eval_logps/rejected": -427.6639099121094,
    "eval_loss": 0.4977857172489166,
    "eval_rewards/accuracies": 0.71875,
    "eval_rewards/chosen": -2.1814193725585938,
    "eval_rewards/margins": 1.098227858543396,
    "eval_rewards/rejected": -3.2796471118927,
    "eval_runtime": 119.8728,
    "eval_samples": 750,
    "eval_samples_per_second": 6.257,
    "eval_steps_per_second": 0.2,
    "total_flos": 0.0,
    "train_loss": 0.5053261174605443,
    "train_runtime": 2173.3813,
    "train_samples": 6750,
    "train_samples_per_second": 6.212,
    "train_steps_per_second": 0.048
}