eval_results.json

{
    "epoch": 1.971563981042654,
    "eval_logits/chosen": -16.5866641998291,
    "eval_logits/rejected": -15.948689460754395,
    "eval_logps/chosen": -363.826171875,
    "eval_logps/rejected": -363.9148864746094,
    "eval_loss": 0.5950456857681274,
    "eval_rewards/accuracies": 0.71875,
    "eval_rewards/chosen": 0.22188298404216766,
    "eval_rewards/margins": 0.26559606194496155,
    "eval_rewards/rejected": -0.043713077902793884,
    "eval_runtime": 6.8121,
    "eval_samples": 750,
    "eval_samples_per_second": 110.098,
    "eval_steps_per_second": 3.523
}