torch-scatter failing to install
See original GitHub issue📚 Installation
Environment
- OS: Ubuntu 18.04
- Python version: 3.6
- PyTorch version: 1.7.1
- CUDA/cuDNN version: 10.1
- GCC version: 7.5.0
- How did you try to install PyTorch Geometric and its extensions (wheel, source): With pip3
- Any other relevant information:
Checklist
- [yes] I followed the installation guide.
- [yes] I cannot find my error message in the FAQ.
- [yes] I set up CUDA correctly and can compile CUDA code via
nvcc
. - [yes] I do have multiple CUDA versions on my machine.
Additional context
I installed torch 1.7.1 with
pip3 install torch==1.7.1+cu101 -f https://download.pytorch.org/whl/torch_stable.html
successfully
Then torch-scatter with
pip3 install --no-index torch-scatter -f https://pytorch-geometric.com/whl/torch-1.7.0+cu101.html
also successfully
but then, for torch-sparse,
pip3 install --no-index torch-sparse -f https://pytorch-geometric.com/whl/torch-1.7.0+cu101.html
I get this error message
Could not find a version that satisfies the requirement scipy (from torch-sparse) (from versions: ) No matching distribution found for scipy (from torch-sparse)
I tried to repeat the last pip3 after running a pip3 install of scipy, but it did not work. I also tried repeating the sequence with versions 1.4.0, 1.5.0, and 1.6.0 of torch, but it did not work either.
Issue Analytics
- State:
- Created 3 years ago
- Comments:6 (3 by maintainers)
Top GitHub Comments
It worked! After installing torch-sparse, the cluster and spline-conv worked smoothly. Then I had problems with the torch-geometric package that seemed to be related to some confusing issues with llvmlite, but I fixed this by upgrading pip3 to version 21.0.1. Then, the last problem seemed to be having different cuda versions (10.2 for torch vs. 10.1 for torch-sparse but I could fix this by reinstalling torch with version 1.7.0 and cuda 10.1). Seems to work well now when running the examples of your tutorials. Thanks for your support and for the amazing tool you have developed for ML with graphs.
Can you try to run without the
--no-index
option?