chaObserv

I would convert this to a simplified version.

Extract sample functions and simplified the process, which makes it fit into k_diffusion. Instead of passing percent to sample function, SamplerSASolver node requires a model now to compute interval of...

Check the python in the ComfyUI_windows_portable\python_embeded. Windows portable is using that portable python to run comfy. Go to python_embeded then do .\python.exe -m pip list

It appears similar [here](https://github.com/comfyanonymous/ComfyUI/commit/916d1e14a93ef331adef7c0deff2fdcf443b05cf#commitcomment-151986557). In the CPU-only case, the noise from `default_noise_sampler` and the latent noise are both generated by the CPU and share the same seed. The noise sampler...

You can try the official comfy implementation: [here](https://github.com/ChenDarYen/ComfyUI-NAG).

This node is used for separating sigmas into two parts to use two (or more) different sampling settings on different intervals. - Step 1: 0.8901 to 0.7516 - Step 2:...

> What is the model? There is a embedded workflow in the image, so you can find it there.

Disabling this parameter is useful when chaining multiple samplers within a single denoising process. If a noise schedule is split into multiple segments to apply different settings, each segment, except...

I see. I think the main issue is that the sampler cannot know the current state of `latent_image`. In the chaining case where no new noise is added, `noise_scaling` is...