More steps for Wan2.2

[Cuogeihong]

Hello, how to set "denoising_step_list" and "boundary_step_index" for more inference steps on Wan2.2? We want
to achieve higher video quality with 8steps or 16steps infer, are there several practical presets?

请问wan2.2使用8步或16步推理，应该如何设置t值以及在哪一步切换high noise与low noise模型？

[helloyongyang]

LightX2V/lightx2v/models/runners/wan/wan_runner.py

Line 553 in 84e6109

if self.scheduler.timesteps[self.scheduler.step_index] >= self.boundary_timestep:

You can hardcode here.

[Cuogeihong]

my bad, I mean the optimal parameter presets, for 4 steps: [1000, 750, 500, 250] and 2, and what for 8 and 16steps?