mirror of
https://github.com/Comfy-Org/desktop.git
synced 2026-11-05 07:57:46 +00:00
[GH-ISSUE #1647] [Bug]: AssertionError: Torch not compiled with CUDA enabled #487
Labels
No labels
area: desktop-app
Bounty
bug
bug-cop:answered
bug-cop:response-received
component: installer
component: preload
custom node
dependencies
Design
documentation
duplicate
enhancement
expected
frontend
good first issue
installation
Low-Priority
performance
Planned
platform: cross-platform
platform: windows
Potential Bug
Potential Bug
pull-request
question
refactor
Release
stale
User Support
V1
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set.
Reference
Comfy-Org/desktop#487
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Originally created by @1LucasGraham on GitHub (Mar 11, 2026).
Original GitHub issue: https://github.com/Comfy-Org/desktop/issues/1647
App Version
v0.8.18
Expected Behavior
How to solve the problem of installing and starting ComfyUI desktop version for the first time?
Actual Behavior
How to solve the problem of installing and starting ComfyUI desktop version for the first time?
Steps to Reproduce
How to solve the problem of installing and starting ComfyUI desktop version for the first time?
Debug Logs
Browser Logs
No response
Settings JSON
No response
Other
┆Issue is synchronized with this Notion page by Unito
@dosubot[bot] commented on GitHub (Mar 11, 2026):
Hi @1LucasGraham! I'm Dosu and I’m helping the desktop team.
This error occurs when PyTorch was installed without CUDA support, even though your RTX 4070 and drivers are working correctly.
Quick fix: Open the ComfyUI virtual environment and reinstall PyTorch with CUDA support:
You can verify the fix worked by running:
This should return
True.If this doesn't work, a VPN can sometimes block PyTorch downloads during installation. If you were using a VPN when you first installed ComfyUI Desktop, try:
%APPDATA%\ComfyUI)This is a known issue that happens when the CUDA-enabled PyTorch fails to download properly during initial setup.
To reply, just mention @dosu.
Share context across your team and agents. Try Dosu.
@1LucasGraham commented on GitHub (Mar 11, 2026):
Thanks, trying to reinstall it.
@1LucasGraham commented on GitHub (Mar 11, 2026):
Run
python -c "import torch; print(torch.cuda.is_available())"Result
False@dosu
@dosubot[bot] commented on GitHub (Mar 11, 2026):
Let's verify the installation actually has CUDA support. Can you run these commands and share the output?
The
torch.__version__should show something like2.x.x+cu128- the+cu128suffix indicates CUDA support. If it shows just2.x.xwithout the+cusuffix, it installed the CPU-only version.Also, I noticed you used
pip uninstallinstead ofuv pip uninstallin your screenshot. Since ComfyUI Desktop usesuvfor package management, try usinguvconsistently:If it still doesn't work, your RTX 4070 with driver 595.71 should support CUDA 13.0. You could try the newer CUDA build:
To reply, just mention @dosu.
Docs are dead. Just use Dosu.
@1LucasGraham commented on GitHub (Mar 11, 2026):
Abandon the desktop version.
Manual installation successful.
Tip: Pytorch download always times out.
@dosu
@dosubot[bot] commented on GitHub (Mar 11, 2026):
Glad you got ComfyUI working with the manual installation!
The PyTorch download timeout issue is a known problem, often caused by network conditions or VPN interference during the desktop app's setup. The PyTorch packages are quite large (~2GB), so slower or unstable connections can fail before completing.
Thanks for reporting this - it helps the team understand the installation pain points. If you ever want to try the desktop version again, disabling VPN and ensuring a stable network connection during initial setup usually helps.
To reply, just mention @dosu.
Share context across your team and agents. Try Dosu.