all_results.json

{
    "epoch": 2.981366459627329,
    "eval_logits/chosen": -2.2292356491088867,
    "eval_logits/rejected": -2.2232847213745117,
    "eval_logps/chosen": -249.73751831054688,
    "eval_logps/rejected": -233.29318237304688,
    "eval_loss": 0.5883417725563049,
    "eval_rewards/accuracies": 0.8402777910232544,
    "eval_rewards/chosen": 1.9964463710784912,
    "eval_rewards/margins": 2.374969959259033,
    "eval_rewards/rejected": -0.3785232603549957,
    "eval_runtime": 74.456,
    "eval_samples_per_second": 15.311,
    "eval_steps_per_second": 0.242,
    "total_flos": 1414680891359232.0,
    "train_loss": 0.34148518641789755,
    "train_runtime": 4236.5791,
    "train_samples_per_second": 7.263,
    "train_steps_per_second": 0.028
}