all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -1.0169163942337036,
    "eval_logits/rejected": -0.8867645263671875,
    "eval_logps/chosen": -412.1428527832031,
    "eval_logps/rejected": -479.4647521972656,
    "eval_loss": 0.02242046222090721,
    "eval_rewards/accuracies": 0.6754999756813049,
    "eval_rewards/chosen": -0.18013788759708405,
    "eval_rewards/margins": 0.08771497756242752,
    "eval_rewards/rejected": -0.26785287261009216,
    "eval_runtime": 713.9076,
    "eval_samples": 2000,
    "eval_samples_per_second": 2.801,
    "eval_steps_per_second": 1.401,
    "train_loss": 0.028396672180453324,
    "train_runtime": 172620.3313,
    "train_samples": 61135,
    "train_samples_per_second": 0.354,
    "train_steps_per_second": 0.089
}