all_results.json

{
    "epoch": 3.97,
    "eval_logits/chosen": -0.056332726031541824,
    "eval_logits/rejected": 0.04286394640803337,
    "eval_logps/chosen": -307.5767822265625,
    "eval_logps/rejected": -279.65240478515625,
    "eval_loss": 0.001864879042841494,
    "eval_rewards/accuracies": 0.47099998593330383,
    "eval_rewards/chosen": -0.007448107469826937,
    "eval_rewards/margins": -0.001194409909658134,
    "eval_rewards/rejected": -0.006253697443753481,
    "eval_runtime": 412.367,
    "eval_samples": 2000,
    "eval_samples_per_second": 4.85,
    "eval_steps_per_second": 1.213,
    "train_loss": 0.0009980735903222775,
    "train_runtime": 2694.3955,
    "train_samples": 61135,
    "train_samples_per_second": 1.485,
    "train_steps_per_second": 0.092
}