all_results.json

{
    "epoch": 2.97196261682243,
    "eval_logits/chosen": -2.3927104473114014,
    "eval_logits/rejected": -2.3628900051116943,
    "eval_logps/chosen": -255.4082489013672,
    "eval_logps/rejected": -247.4369659423828,
    "eval_loss": 0.5902177691459656,
    "eval_rewards/accuracies": 0.7552083134651184,
    "eval_rewards/chosen": 1.1360557079315186,
    "eval_rewards/margins": 1.9774025678634644,
    "eval_rewards/rejected": -0.8413467407226562,
    "eval_runtime": 100.4193,
    "eval_samples_per_second": 15.137,
    "eval_steps_per_second": 0.239,
    "total_flos": 1874604243025920.0,
    "train_loss": 0.33037466410570926,
    "train_runtime": 5730.1243,
    "train_samples_per_second": 7.16,
    "train_steps_per_second": 0.028
}