eval_results.json

{
    "epoch": 0.86,
    "eval_logits/chosen": -3.2169618606567383,
    "eval_logits/rejected": -3.168179512023926,
    "eval_logps/chosen": -767.6728515625,
    "eval_logps/rejected": -1184.0101318359375,
    "eval_loss": 0.6929298043251038,
    "eval_rewards/accuracies": 0.5393907427787781,
    "eval_rewards/chosen": 0.00030664310907013714,
    "eval_rewards/margins": 0.000654319126624614,
    "eval_rewards/rejected": -0.00034767595934681594,
    "eval_runtime": 800.2666,
    "eval_samples": 7588,
    "eval_samples_per_second": 9.482,
    "eval_steps_per_second": 0.297
}