eval_results.json

{
    "epoch": 0.9985553308292401,
    "eval_logits/chosen": -0.3882608115673065,
    "eval_logits/rejected": -0.3577967882156372,
    "eval_logps/chosen": -161.60928344726562,
    "eval_logps/rejected": -169.34393310546875,
    "eval_loss": 1.4395084381103516,
    "eval_nll_loss": 0.28411635756492615,
    "eval_rewards/accuracies": 0.6326087117195129,
    "eval_rewards/chosen": -16.160930633544922,
    "eval_rewards/margins": 0.773465096950531,
    "eval_rewards/rejected": -16.93439483642578,
    "eval_runtime": 76.7453,
    "eval_samples": 1826,
    "eval_samples_per_second": 23.793,
    "eval_steps_per_second": 1.498
}