all_results.json

{
    "epoch": 3.0,
    "eval_logits/chosen": -2.9668068885803223,
    "eval_logits/rejected": -2.877352237701416,
    "eval_logps/chosen": -266.8033752441406,
    "eval_logps/rejected": -229.52769470214844,
    "eval_loss": 0.5254282355308533,
    "eval_rewards/accuracies": 0.7400000095367432,
    "eval_rewards/chosen": -0.12839017808437347,
    "eval_rewards/margins": 0.7881025671958923,
    "eval_rewards/rejected": -0.9164927005767822,
    "eval_runtime": 537.7252,
    "eval_samples": 2000,
    "eval_samples_per_second": 3.719,
    "eval_steps_per_second": 1.86,
    "train_loss": 0.5633651232629111,
    "train_runtime": 75701.2005,
    "train_samples": 61966,
    "train_samples_per_second": 2.456,
    "train_steps_per_second": 0.038
}