9 hours ago, derp456789 said:

Has any attempted to run this on older hardware and ran into the following error ?
Fatal Python error: Illegal instruction

As far as i can tell it looks like something to do with Kornia and maybe the avx support ?
Unraid 7.3.1
dual xeon E5-2680 v2
RTX 3090
Repo: mmartial/comfyui-nvidia-docker:ubuntu24_cuda13.2-latest

Snippet of the error
Fatal Python error: Illegal instruction

Stack (most recent call first):

File "<frozen importlib._bootstrap>", line 488 in callwith_frames_removed

File "<frozen importlib._bootstrap_external>", line 1297 in exec_module

File "<frozen importlib._bootstrap>", line 935 in loadunlocked

File "<frozen importlib._bootstrap>", line 1331 in findand_load_unlocked

File "<frozen importlib._bootstrap>", line 1360 in findand_load

File "/comfy/mnt/venv/lib/python3.12/site-packages/kornia_rs/__init__.py", line 1 in <module>

File "<frozen importlib._bootstrap>", line 488 in callwith_frames_removed

File "<frozen importlib._bootstrap_external>", line 995 in exec_module

File "<frozen importlib._bootstrap>", line 935 in loadunlocked

File "<frozen importlib._bootstrap>", line 1331 in findand_load_unlocked

File "<frozen importlib._bootstrap>", line 1360 in findand_load

File "/comfy/mnt/venv/lib/python3.12/site-packages/kornia/io/io.py", line 24 in <module>

File "<frozen importlib._bootstrap>", line 488 in callwith_frames_removed

File "<frozen importlib._bootstrap_external>", line 995 in exec_module

File "<frozen importlib._bootstrap>", line 935 in loadunlocked

File "<frozen importlib._bootstrap>", line 1331 in findand_load_unlocked

File "<frozen importlib._bootstrap>", line 1360 in findand_load
/comfyui-nvidia_init.bash: line 953: 2609 Illegal instruction ${COMFY_CMDLINE_BASE} ${COMFY_CMDLINE_EXTRA}

!! ERROR: ComfyUI failed or exited with an error

!! Exiting script (ID: 97)

!! ERROR: subscript failed

!! Exiting script (ID: 1)

The crash comes from ComfyUI itself so I would not be able to comment too much on what the issue is.

When I see this kind of behavior, I rename the run folder to run.old and re-run the container which will re-create a new installation.

In general this fixed things for me, knowing that the custom nodes and images are in basedir, everything in "run" should be able to be deleted.

3 hours ago, martial said:

The crash comes from ComfyUI itself so I would not be able to comment too much on what the issue is.

When I see this kind of behavior, I rename the run folder to run.old and re-run the container which will re-create a new installation.

In general this fixed things for me, knowing that the custom nodes and images are in basedir, everything in "run" should be able to be deleted.

Got it thanks for the quick reply
I've attempted a fresh install ( removing the appdata files ) plus a lower version and no luck
open to any suggestions

23 hours ago, derp456789 said:

Got it thanks for the quick reply
I've attempted a fresh install ( removing the appdata files ) plus a lower version and no luck
open to any suggestions

When you deleted the run folder you downloaded a fresh install of ComfyUI during the next startup, so we can expect this to be something else.

Doing a quick gpt search seems to say that it might be a CPU issue and you might have to compile kornia from source.

You can try to use the userscript (check the README for instructions) to attempt to run the build manually: /comfy/mnt/venv/bin/pip install kornia-rs --force-reinstall --no-cache-dir --no-binary :all: (also copied from the GPT answer)

The Fatal Python error: Illegal instruction coming directly from kornia_rs means that the pre-compiled Rust binaries in that library are attempting to use CPU instruction sets (specifically AVX, AVX2, or FMA3) that your host machine's CPU does not support, or that your Docker/VM hypervisor configuration is masking from the container.

Not sure exactly what is happening, google is telling me that the wrong version of pytorch, not sure the google solutions will work as a i don't know that i can console in and correct this as it's happening during the container startup.

== Pre-installing/Upgrading torch

== Installing torch using uv with backend: cu130

Installing: torch torchvision torchaudio

== Installing/Updating from ComfyUI's requirements

!! ERROR: ComfyUI requirements install/upgrade failed

!! Exiting script (ID: 100)

!! ERROR: subscript failed

!! Exiting script (ID: 1)

Edited July 18Jul 18 by WyoFarr

On 7/18/2026 at 5:49 AM, WyoFarr said:

Not sure exactly what is happening, google is telling me that the wrong version of pytorch, not sure the google solutions will work as a i don't know that i can console in and correct this as it's happening during the container startup.

== Pre-installing/Upgrading torch

== Installing torch using uv with backend: cu130

Installing: torch torchvision torchaudio

== Installing/Updating from ComfyUI's requirements

!! ERROR: ComfyUI requirements install/upgrade failed

!! Exiting script (ID: 100)

!! ERROR: subscript failed

!! Exiting script (ID: 1)

That log looks very bare, as if nothing happened.

The previous step completed (installing torch), then it goes into installing Comfy's requirements.txt which should run uv pip install -r and be a little bit more verbose, especially in case of error. Here it is like it was not able to start the command.

Was having problems finding this and found the issue listed under statistics.

ComfyUI-Nvidia-Docker is currently missing from Community Apps because the template is being rejected as invalid XML.

Apps → Statistics → Invalid Templates reports:

char '&' is not expected.

The issue appears to be the August 5 changelog line:

- Included full ffmpeg/sqlite3 for sub-scripts compilations & video support

The & needs to be escaped as & (or replaced with and).

11 hours ago, TwistUnraId said:

Was having problems finding this and found the issue listed under statistics.

The & needs to be escaped as & (or replaced with and).

Thanks for reporting it, it is fixed, it might take a couple hours to show up again.

Hey,

sorry to drop in but also got issues. No matter what Nvidia Driver is use or which repo im not able to get comy running again

mmartial/comfyui-nvidia-docker:ubuntu24_cuda12.9-latest

3090 FE with v580.173.02

RuntimeError: Detected that PyTorch and TorchAudio were compiled with different CUDA versions. PyTorch has CUDA version 13.0 whereas TorchAudio has CUDA version 12.9. Please install the TorchAudio version that matches your PyTorch version.

How to fix this ?

18 hours ago, schubdog said:

Hey,

sorry to drop in but also got issues. No matter what Nvidia Driver is use or which repo im not able to get comy running again

mmartial/comfyui-nvidia-docker:ubuntu24_cuda12.9-latest

3090 FE with v580.173.02

RuntimeError: Detected that PyTorch and TorchAudio were compiled with different CUDA versions. PyTorch has CUDA version 13.0 whereas TorchAudio has CUDA version 12.9. Please install the TorchAudio version that matches your PyTorch version.

How to fix this ?

This is a strange error, the "max" for now is cu130

I have it running with driver 610 with the 12.9 version and I do not see any error.

I would recommend attempting to add a new variable called TORCH_LOCK

torch==2.13.0+cu128 torchaudio==2.11.0+cu128 torchvision==0.28.0+cu128

(I am not sure those are the right values, I adapted them from my cu130 line below)

The next step would be to move the driver up (check with nvidia-smi in a terminal so you can get CUDA to at least 13.0) and try with

torch==2.13.0+cu130 torchaudio==2.11.0+cu130 torchvision==0.28.0+cu130

with the CUDA 13.0 (and up to CUDA 13.2) version

I'm trying to download models in the UI, but it downloads them into my browser itself. Is there any way to force it to download them into the appropriate dirs inside the comfy installation?

On 9/9/2026 at 5:24 PM, drbaltar said:

I'm trying to download models in the UI, but it downloads them into my browser itself. Is there any way to force it to download them into the appropriate dirs inside the comfy installation?

Not directly, the tool is just a frontend to ComfyUI, it does not know anything about your host or directory structure.

If you want to make your life easier, I recommend you look at custom nodes. I use "Lora Manager" https://github.com/willmiao/ComfyUI-Lora-Manager and I pay the author for its browser integration and CivitAI download capabilities. You do not have to pay if you are comfortable pasting the url in the tool's WebUI to do direct download..

There are likely other similar nodes I just started using this one and have not looked back.