eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.221278667449951,
    "eval_logits/rejected": -2.2164483070373535,
    "eval_logps/chosen": -34.091766357421875,
    "eval_logps/rejected": -37.60596466064453,
    "eval_loss": 0.9996774792671204,
    "eval_rewards/accuracies": 0.5477575063705444,
    "eval_rewards/chosen": -0.000572161574382335,
    "eval_rewards/margins": 0.00032129042665474117,
    "eval_rewards/rejected": -0.0008934520301409066,
    "eval_runtime": 145.6061,
    "eval_samples": 343,
    "eval_samples_per_second": 2.356,
    "eval_steps_per_second": 0.295
}