all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -0.9851527214050293,
    "eval_logits/rejected": -0.9165840148925781,
    "eval_logps/chosen": -336.4683837890625,
    "eval_logps/rejected": -384.2898864746094,
    "eval_loss": 0.5675918459892273,
    "eval_rewards/accuracies": 0.73046875,
    "eval_rewards/chosen": -0.5684676766395569,
    "eval_rewards/margins": 0.5406599044799805,
    "eval_rewards/rejected": -1.1091275215148926,
    "eval_runtime": 143.0981,
    "eval_samples": 2000,
    "eval_samples_per_second": 13.976,
    "eval_steps_per_second": 0.224,
    "train_loss": 0.6029723678173879,
    "train_runtime": 7448.3301,
    "train_samples": 61135,
    "train_samples_per_second": 8.208,
    "train_steps_per_second": 0.064
}