eval_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -1.1438143253326416,
    "eval_logits/rejected": -1.0389952659606934,
    "eval_logps/chosen": -116.79230499267578,
    "eval_logps/rejected": -110.2491226196289,
    "eval_loss": 0.5636938810348511,
    "eval_rewards/accuracies": 0.7720000147819519,
    "eval_rewards/chosen": 0.22718606889247894,
    "eval_rewards/margins": 0.688110888004303,
    "eval_rewards/rejected": -0.46092477440834045,
    "eval_runtime": 65.8827,
    "eval_samples_per_second": 15.178,
    "eval_steps_per_second": 1.897
}