[Bug] Optional nodes `NunchakuPulidApply` and `NunchakuPulidLoader` import failed

[yhf1407]

Checklist

• 1. I have searched for related issues and FAQs (https://github.com/mit-han-lab/nunchaku/blob/main/docs/faq.md) but was unable to find a solution.
• 2. The issue persists in the latest version.
• 3. Please note that without environment information and a minimal reproducible example, it will be difficult for us to reproduce and address the issue, which may delay our response.
• 4. If your report is a question rather than a bug, please submit it as a discussion at https://github.com/mit-han-lab/ComfyUI-nunchaku/discussions/new/choose. Otherwise, this issue will be closed.
• 5. I will do my best to describe the issue in English.

Describe the Bug

Nunchaku version: 0.3.1
ComfyUI-nunchaku version: 0.3.1
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/alter-list.json
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/model-list.json
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/github-stats.json
Optional nodes NunchakuPulidApply and NunchakuPulidLoader import failed:
Traceback (most recent call last):
File "D:\anaconda3\envs\ceshi\Lib\site-packages\insightface_init_.py", line 8, in
import onnxruntime
File "D:\anaconda3\envs\ceshi\Lib\site-packages\onnxruntime_init_.py", line 61, in
raise import_capi_exception
File "D:\anaconda3\envs\ceshi\Lib\site-packages\onnxruntime_init_.py", line 24, in
from onnxruntime.capi._pybind_state import (
File "D:\anaconda3\envs\ceshi\Lib\site-packages\onnxruntime\capi_pybind_state.py", line 32, in
from .onnxruntime_pybind11_state import * # noqa
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ImportError: DLL load failed while importing onnxruntime_pybind11_state: 动态链接库(DLL)初始化例程失败。

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File "D:\cf\ComfyUI\custom_nodes\ComfyUI-nunchaku_init_.py", line 52, in
from .nodes.models.pulid import NunchakuPulidApply, NunchakuPulidLoader
File "D:\cf\ComfyUI\custom_nodes\ComfyUI-nunchaku\nodes\models\pulid.py", line 8, in
from nunchaku.pipeline.pipeline_flux_pulid import PuLIDPipeline
File "D:\anaconda3\envs\ceshi\Lib\site-packages\nunchaku\pipeline_init_.py", line 1, in
from .pipeline_flux_pulid import PuLIDFluxPipeline
File "D:\anaconda3\envs\ceshi\Lib\site-packages\nunchaku\pipeline\pipeline_flux_pulid.py", line 9, in
import insightface
File "D:\anaconda3\envs\ceshi\Lib\site-packages\insightface_init_.py", line 10, in
raise ImportError(
ImportError: Unable to import dependency onnxruntime.

Environment

pytorch version: 2.7.0+cu128
xformers version: 0.0.30
Enabled fp16 accumulation.
Set vram state to: NORMAL_VRAM
Device: cuda:0 NVIDIA GeForce RTX 4070 : cudaMallocAsync
Using xformers attention
Python version: 3.12.4 | packaged by Anaconda, Inc. | (main, Jun 18 2024, 15:03:56) [MSC v.1929 64 bit (AMD64)]
ComfyUI version: 0.3.44
ComfyUI frontend version: 1.23.4

Reproduction Steps

uu.json

[BobMind758]

刚刚到处找方法终于解决了，现在忘记了，主要问题就是依赖编译失败，手动下载轮子安装https://www.piwheels.org/simple/filterpy/filterpy-1.4.5-py3-none-any.whl#sha256=227c2659b318061bb7ad5ec00f1b29ca3a8264f42b18c7be0d14487d521484d8

[nyukers]

Nunchaku version: 0.3.2.dev20250708
ComfyUI-nunchaku version: 0.3.3
ComfyUI-nunchaku 0.3.3 is not compatible with nunchaku 0.3.2.dev20250708. Please update nunchaku to a supported version in ['v0.3.1'].
Downgrade it?

[NBSTpeterhill]

刚刚到处找方法终于解决了，现在忘记了，主要问题就是依赖编译失败，手动下载轮子安装https://www.piwheels.org/simple/filterpy/filterpy-1.4.5-py3-none-any.whl#sha256=227c2659b318061bb7ad5ec00f1b29ca3a8264f42b18c7be0d14487d521484d8

Seems not work and the install of this wheel broken the [impact-comfyUI] and [impact-subpack] and [easy-use] nodes imported, Then LORA loader loss its function--Lora has no effect at all in 0.3.3 version, which works well in 0.3.2 version.

Also

Nunchaku version: 0.3.2.dev20250708
ComfyUI-nunchaku version: 0.3.3
ComfyUI-nunchaku 0.3.3 is not compatible with nunchaku 0.3.2.dev20250708. Please update nunchaku to a supported version in ['v0.3.1'].
Downgrade it?

If we were successed to solve this problem, then the other nodes lose there functions.
because [pandas] and [opencv-pytorch] need [numpy<2], but most of other nodes such as [impact---adetailers] and [easy-use] need numpy > 2.2.6

Waiting for the solution.
The LORA-Kontext-FLUX still not work in 0.3.3 version.

[nyukers]

Ok, in my case Nunchaki import failed, PuLID dependencies crashed. The same behavior with 0.3.2 and 0.3.3 versions.(
Interestingly, the manager Comfyui does not allow you to choose a version other than nigthly and latest.

[lmxyy]

For this PuLID-related issue, please refer to nunchaku-ai/nunchaku#392 to installinsightface.
You need to get insightface and onnxruntime correctly installed. The vanilla PuLID also requires them. You can also refer to their issues.

[NBSTpeterhill]

Ok, in my case Nunchaki import failed, PuLID dependencies crashed. The same behavior with 0.3.2 and 0.3.3 versions.(好的，在我的例子中，Nunnchaki 导入失败，PuLID 依赖项崩溃了。0.3.2 和 0.3.3 版本也出现了同样的情况。( Interestingly, the manager Comfyui does not allow you to choose a version other than nigthly and **latest.**有趣的是，管理器 Comfyui 不允许您选择除 nightthly 和 latest 之外的版本。

you are right, 0.3.2 version release needed.

[NBSTpeterhill]

For this PuLID-related issue, please refer to mit-han-lab/nunchaku#392 to installinsightface.对于与 PuLID 相关的问题，请参阅 mit-han-lab/nunchaku#392 来安装 insightface 。 You need to get insightface and onnxruntime correctly installed. The vanilla PuLID also requires them. You can also refer to their issues.您需要正确安装 insightface 和 onnxruntime 。原生 PuLID 也需要它们。您也可以参考它们的问题 。

I saw the #392 question, he said there was a CV2 problem, that is PuLID not successfully install problem. While, once you install all the wheel successfully (insightface and onnx). Then here will be CV8 problem and opencvtorch bugs.

ReActor, PuLID, these two [Face change] nodes are too old to match the AI and ComfyUI recently.

Otherwise, Kontext-FLUX itself can do the same function and better than that two old nodes.
So I do not know why does nunchaku want to hard-link the underlying code and configuration environment with PuLID? Why not us a soft link or optional, it is much better than the current embarrassing situation.