all_results.json

{
    "epoch": 4.0,
    "eval_logits/chosen": -2.1421945095062256,
    "eval_logits/rejected": -2.1374316215515137,
    "eval_logps/chosen": -34.35047149658203,
    "eval_logps/rejected": -38.03379440307617,
    "eval_loss": 0.6879659295082092,
    "eval_rewards/accuracies": 0.5834717750549316,
    "eval_rewards/chosen": -0.2843308448791504,
    "eval_rewards/margins": 0.1811266541481018,
    "eval_rewards/rejected": -0.4654575288295746,
    "eval_runtime": 145.1525,
    "eval_samples": 343,
    "eval_samples_per_second": 2.363,
    "eval_steps_per_second": 0.296,
    "train_loss": 0.19153660760297403,
    "train_runtime": 10767.2812,
    "train_samples": 3079,
    "train_samples_per_second": 1.144,
    "train_steps_per_second": 0.143
}