all_results.json

{
    "epoch": 4.0,
    "eval_logits/chosen": -2.1216881275177,
    "eval_logits/rejected": -2.1169610023498535,
    "eval_logps/chosen": -34.41923522949219,
    "eval_logps/rejected": -38.13933563232422,
    "eval_loss": 0.6746898293495178,
    "eval_rewards/accuracies": 0.6009136438369751,
    "eval_rewards/chosen": -0.2692793905735016,
    "eval_rewards/margins": 0.16662099957466125,
    "eval_rewards/rejected": -0.43590039014816284,
    "eval_runtime": 145.1817,
    "eval_samples": 343,
    "eval_samples_per_second": 2.363,
    "eval_steps_per_second": 0.296,
    "train_loss": 0.2216110908365869,
    "train_runtime": 10768.4727,
    "train_samples": 3079,
    "train_samples_per_second": 1.144,
    "train_steps_per_second": 0.143
}