all_results.json

{
    "epoch": 1.986206896551724,
    "eval_kl": 0.0009412201470695436,
    "eval_logps/chosen": -11.28572518484933,
    "eval_logps/rejected": -10.654654366629464,
    "eval_loss": 0.5029926300048828,
    "eval_rewards/chosen": 0.000681033224931785,
    "eval_rewards/margins": 0.0012661761471203395,
    "eval_rewards/rejected": -0.0005851429221885544,
    "eval_runtime": 41.2133,
    "eval_samples_per_second": 10.822,
    "eval_steps_per_second": 0.679,
    "total_flos": 2.0165964255461376e+18,
    "train_loss": 0.5037083285942412,
    "train_runtime": 19753.9777,
    "train_samples_per_second": 4.463,
    "train_steps_per_second": 0.006
}