eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.228182792663574,
    "eval_logits/rejected": -2.2233598232269287,
    "eval_logps/chosen": -34.08854293823242,
    "eval_logps/rejected": -37.60850143432617,
    "eval_loss": 24.80753517150879,
    "eval_rewards/accuracies": 0.5132890343666077,
    "eval_rewards/chosen": -0.005399257410317659,
    "eval_rewards/margins": 0.0037888663355261087,
    "eval_rewards/rejected": -0.009188123047351837,
    "eval_runtime": 145.6001,
    "eval_samples": 343,
    "eval_samples_per_second": 2.356,
    "eval_steps_per_second": 0.295
}