eval_results.json

{
    "epoch": 1.971563981042654,
    "eval_logits/chosen": 94.27800750732422,
    "eval_logits/rejected": 88.3936767578125,
    "eval_logps/chosen": -441.594482421875,
    "eval_logps/rejected": -473.087646484375,
    "eval_loss": 0.4694259464740753,
    "eval_rewards/accuracies": 0.7291666865348816,
    "eval_rewards/chosen": -3.94303035736084,
    "eval_rewards/margins": 1.613731026649475,
    "eval_rewards/rejected": -5.556761264801025,
    "eval_runtime": 20.9081,
    "eval_samples": 750,
    "eval_samples_per_second": 35.871,
    "eval_steps_per_second": 1.148
}