facebookresearch / facebookresearch/meshflow
Installation fails on Windows 11 due to torch/torchvision version conflict
- Dominant language
- Python
- Stars
- 512
- Forks
- 27
- PR merge metrics
- No merged PRs in 30d
Description
I am trying to install MeshFlow on Windows 11 with Python 3.10 and RTX 5090 GPU. I followed the official instructions exactly:
git clone https://github.com/facebookresearch/meshflow.git
cd meshflow
pip install -r requirements.txt
However, I get a dependency conflict immediately:
ERROR: Cannot install -r requirements.txt (line 5) and torch==2.8.0 because these package versions have conflicting dependencies.
The conflict is caused by:
The user requested torch==2.8.0
torchvision 0.22.1+cu128 depends on torch==2.7.1+cu128
torchvision 0.22.1 depends on torch==2.7.1
According to the official PyTorch compatibility table, torch==2.8.0 requires torchvision==0.23.x, while torchvision==0.22.1 requires torch==2.7.x. So requirements.txt contains an incompatible pair: torch==2.8.0 + torchvision==0.22.1.
I tried manually installing torch==2.7.1+cu128 and torchvision==0.22.1+cu128, but then I got another error when importing torch:
The procedure entry point ?toScalarType@TypeMeta@caffe2@@QEAQA?AW4ScalarType@c10@@XZ could not be located in the dynamic link library torchvision/_C.pyd
This indicates a mismatch between torch and torchvision even after forcing the versions.
Full error log from pip install:
[paste the full output of pip install here]
Full error log from import test:
[paste the full output of python -c "import torch; import torchvision" here]
Questions:
Is there a specific combination of torch/torchvision that works on Windows 11?
Are you using custom builds of PyTorch that differ from the official ones?
Could you please update requirements.txt with a working set of versions for Windows?
If not, could you provide a Dockerfile or recommend using conda instead?
Thank you for your help.
Contributor guide
Research direction
Start with requirements.txt, especially line 5, and reproduce the documented pip install on Windows 11 with Python 3.10. Then run the issue's import test for torch and torchvision and compare the declared versions with the official compatibility table. Done means a documented, compatible requirements set installs cleanly and both packages import successfully, or the issue is resolved with a supported installation path.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python, pytorch
- Domain
- machine-learning
- Issue type
- Bug
- Difficulty
- 3/5
- Estimated time
- 1-2 days
- Activity status
- Quiet
- Clarity
- Mostly clear
- Newbie friendliness
- 42/100