eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.2379801273345947,
    "eval_logits/rejected": -2.135157823562622,
    "eval_logps/chosen": -1.0100419521331787,
    "eval_logps/rejected": -1.2764651775360107,
    "eval_loss": 1.5542570352554321,
    "eval_rewards/accuracies": 0.6215000152587891,
    "eval_rewards/chosen": -2.0200839042663574,
    "eval_rewards/margins": 0.5328465104103088,
    "eval_rewards/rejected": -2.5529303550720215,
    "eval_runtime": 610.686,
    "eval_samples": 2000,
    "eval_samples_per_second": 3.275,
    "eval_steps_per_second": 0.819
}