eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.531444787979126,
    "eval_logits/rejected": -2.5150935649871826,
    "eval_logps/chosen": -73.5765609741211,
    "eval_logps/rejected": -81.2323989868164,
    "eval_loss": 0.6325609087944031,
    "eval_rewards/accuracies": 0.3492063581943512,
    "eval_rewards/chosen": 0.3567197322845459,
    "eval_rewards/margins": 0.39729830622673035,
    "eval_rewards/rejected": -0.04057859256863594,
    "eval_runtime": 114.1594,
    "eval_samples": 2000,
    "eval_samples_per_second": 17.519,
    "eval_steps_per_second": 0.552
}