CU130
[INFO] Executing: C:\Cfui\scripts\venv\Scripts\python.exe -m pip install torch==2.9.1+cu130 torchvision torchaudio xformers --index-url https://download.pytorch.org/whl/cu130
the automatic script is doing this so it doesn't boot, it works a few days ago i had to for uninstall a bunch and reinstall 129 and something or other (i struggle to explain but i think you understand)
Hi,
Now the script is updated to CUDA 13.0, i'm not sure to understand your problem.
it doesn't work on my 4090, unless i have to install cuda 13.0 for it to work? my downloas folder for a fresh windows install has always use 12.9 when i ran comfyui after your auto install, it doesn't boot it does this then i press a key and it exits
i think the installer is just installing CU130 even though it recognised i didn't have cuda 13 to begin with i installed all the prerequisites being visual studio , git, 7 zip, python, cuda 12.9 then it auto installed 13 making it unable to boot. i'll run it again later in a seporate folder and install cuda 13 see if that fixes everything.
Checking installation type...
[INFO] Activating venv environment...
Starting ComfyUI (Performance Mode)...
[START] Security scan
[DONE] Security scan
ComfyUI-Manager: installing dependencies done.
** ComfyUI startup time: 2025-12-12 12:52:18.546
** Platform: Windows
** Python version: 3.12.6 (tags/v3.12.6:a4a2d2b, Sep 6 2024, 20:11:23) [MSC v.1940 64 bit (AMD64)]
** Python executable: C:\Comfyui\scripts\venv\Scripts\python.exe
** ComfyUI Path: C:\Comfyui\ComfyUI
** ComfyUI Base Folder Path: C:\Comfyui\ComfyUI
** User directory: C:\Comfyui\ComfyUI\user
** ComfyUI-Manager config path: C:\Comfyui\ComfyUI\user_manager\config.ini
** Log path: C:\Comfyui\ComfyUI\user\comfyui.log
Prestartup times for custom nodes:
0.0 seconds: C:\Comfyui\ComfyUI\custom_nodes\rgthree-comfy
0.0 seconds: C:\Comfyui\ComfyUI\custom_nodes\ComfyUI-Easy-Use
2.4 seconds: C:\Comfyui\ComfyUI\custom_nodes\ComfyUI-Manager
C:\Comfyui\scripts\venv\Lib\site-packages\torch\cuda_init.py:63: FutureWarning: The pynvml package is deprecated. Please install nvidia-ml-py instead. If you did not install pynvml directly, please report this to the maintainers of the package that installed pynvml for you.
import pynvml # type: ignore[import]
Checkpoint files will always be loaded safely.
Press any key to continue . .
yeah cuda 13 sorted it. Grok sent me round all sorts making me uninstall things, then pytorch wasn't working then sage wasn't working cant believe it didn't jsut say update to cuda 13 , i've been using this for months on 12.9 so it didn't occur to me, thanks for your efforts :)