eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.22452712059021,
    "eval_logits/rejected": -2.219710350036621,
    "eval_logps/chosen": -34.03827667236328,
    "eval_logps/rejected": -37.573402404785156,
    "eval_loss": 0.9946814179420471,
    "eval_rewards/accuracies": 0.5710132718086243,
    "eval_rewards/chosen": -0.00037224299740046263,
    "eval_rewards/margins": 0.005306063685566187,
    "eval_rewards/rejected": -0.005678307265043259,
    "eval_runtime": 145.3784,
    "eval_samples": 343,
    "eval_samples_per_second": 2.359,
    "eval_steps_per_second": 0.296
}