all_results.json

{
    "epoch": 1.0,
    "eval_logits/chosen": -2.2347500324249268,
    "eval_logits/rejected": -2.242765188217163,
    "eval_logps/chosen": -481.0046691894531,
    "eval_logps/rejected": -493.4847717285156,
    "eval_loss": 0.5999384522438049,
    "eval_rewards/accuracies": 0.6735000014305115,
    "eval_rewards/chosen": -1.482552409172058,
    "eval_rewards/margins": 0.4411955773830414,
    "eval_rewards/rejected": -1.9237478971481323,
    "eval_runtime": 197.0628,
    "eval_samples": 2000,
    "eval_samples_per_second": 10.149,
    "eval_steps_per_second": 5.075,
    "train_loss": 0.6145847465156994,
    "train_runtime": 171708.6447,
    "train_samples": 61135,
    "train_samples_per_second": 0.356,
    "train_steps_per_second": 0.044
}