eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -1.1469733715057373,
    "eval_logits/rejected": -0.4162125587463379,
    "eval_logps/chosen": -372.8635559082031,
    "eval_logps/rejected": -469.4237060546875,
    "eval_loss": 0.5017749071121216,
    "eval_rewards/accuracies": 0.765625,
    "eval_rewards/chosen": -1.1026902198791504,
    "eval_rewards/margins": 0.9646469354629517,
    "eval_rewards/rejected": -2.0673370361328125,
    "eval_runtime": 91.4963,
    "eval_samples": 2000,
    "eval_samples_per_second": 21.859,
    "eval_steps_per_second": 0.35
}