-
Notifications
You must be signed in to change notification settings - Fork 2k
Open
Description
While setting up Wan2.1 on Windows 10/11, I encountered several installation issues:
- PyTorch → CPU-only build installs by default, CUDA not detected.
flash_attn
→ fails to build on Windows.- No clear guide for CUDA/PyTorch installation on Windows.
Steps to Reproduce
-
Clone the repo on Windows.
-
Run:
pip install -r requirements.txt
-
Installation fails (see screenshot below).
Expected Behavior
-
Clear installation instructions for Windows users.
-
A dedicated
requirements-win.txt
with correct dependencies. -
Correct PyTorch + CUDA install command, e.g.:
pip3 install torch torchvision --index-url https://download.pytorch.org/whl/cu128
Proposed Solution
- Add a Windows Installation Guide section to the README.
- Provide a
requirements-win.txt
with Windows-compatible dependencies. - Mention known incompatibilities (e.g.,
flash_attn
not supported on Windows). - I have created a solution and pushed it to a new branch:
update/installation_guide
.
System Info
- OS: Windows 11 Pro 64-bit
- GPU: NVIDIA RTX (CUDA 12.8 supported)
- Python: 3.10 / 3.11
Additional Context
Windows users currently need to manually adjust dependencies. Adding official support will improve adoption.
Screenshot of installation error:

Metadata
Metadata
Assignees
Labels
No labels