How to use the PeRFlow accelerated version mentioned in the paper?

Hi, the paper mentioned the accelerated version trained with PeRFlow (which can use step=8). How can we use this accelerated version? Is it the current open-source model? Thanks.

论文里提到了使用 PePFlow 训练的加速版本（可以用step=8）。请问这个加速版本我们如何使用呢，是否就是当前开源的这个模型？

The issue does not name a file, test, or entry point. Read the paper's PeRFlow reference and compare it with the current open-source model's documented usage; done means documenting whether the accelerated model is available and how to run it with step=8.

Written by the indexing model from the issue text.