ComfyUI crashes during generation

[IckyTheBiggy]

Expected Behavior

For the UI to not crash during generation.

Actual Behavior

The model crashes after about 10–15 seconds after queuing.

Steps to Reproduce

Download the flux model any 16bit or 8bit happens on both, then get any of the flux dev workflows from the examples page it happens both on the regular and checkpoint version workflow. Then try queueing the image generation and wait for 10–15 seconds, you will get a reconnecting error message popup and the Comfy UI process will be terminated.

Debug Logs

Starting server

To see the GUI go to: http://127.0.0.1:8188
got prompt
Using pytorch attention in VAE
Using pytorch attention in VAE
VAE load device: cuda:0, offload device: cpu, dtype: torch.bfloat16
model weight dtype torch.float8_e4m3fn, manual cast: torch.bfloat16
model_type FLUX
CLIP/text encoder model load device: cuda:0, offload device: cpu, current: cpu, dtype: torch.float16
Requested to load FluxClipModel_
loaded partially 4245.425 4244.8984375 0
fish: Job 1, 'python main.py' terminated by signal SIGILL (Illegal instruction)

Other

PC Specs:
I9 11900k
Rtx 3070ti (8GB Vram)
32GB Ram

I'm using a Linux OS to run this.
I have also tried using the lighter flux model, but the same thing happens.

[ltdrdata]

It seems that your python environment is damaged. Try reinstall venv from scratch.

[IckyTheBiggy]

The issue still seems to persist, I cleared the pip cache to make sure but no success. Also, I tested it on Windows which works perfectly fine, so it seems to be a Linux specific problem. It always crashes at the same point when looking at the graph, it never seems to go into the sampler stage it stays and never proceeds past the prompt stage.