Generally, NEVER use pip install something !!! This may lead to fatal result on your environment.

You have to specify which verison of something you intend to install.

From 0 to 100

Here is a stable version of Graph Research Environment provided by Dr. Xiaohang Zhao. This by no means is the latest version.

conda create --name GNN python=3.8.10
conda activate GNN
conda install tqdm ipykernel=5.3.4 numpy=1.20.2 numba pandas=1.2.4 scikit-learn=0.24.2 scikit-learn-intelex
# there will be a message saying that to run scripts with accelerated scikit-learn, using
# python -m sklearnex my_application.py
python -m ipykernel install --user --name=GNN
conda install -c conda-forge ipywidgets
jupyter nbextension enable --py widgetsnbextension
# jupyter contrib nbextension install --user

# install pytorch the lastest version is 1.9
conda install pytorch=1.8.0 torchvision torchaudio cudatoolkit=10.2 -c pytorch


# install PyTorch geometric
pip install torch-scatter==2.0.7 -f https://pytorch-geometric.com/whl/torch-1.8.0+cu102.html
pip install torch-sparse==0.6.9 -f https://pytorch-geometric.com/whl/torch-1.8.0+cu102.html
pip install torch-cluster==1.5.9 -f https://pytorch-geometric.com/whl/torch-1.8.0+cu102.html
pip install torch-spline-conv==1.2.1 -f https://pytorch-geometric.com/whl/torch-1.8.0+cu102.html
pip install torch-geometric==1.7.0

# install DGL
conda install -c dglteam dgl-cuda10.2
python -m dgl.backend.set_default_backend pytorch

# install other useful packages
pip install torch-geometric-temporal==0.37
pip install deepsnap==0.2.0
pip install captum==0.3.1


# remove the environment
conda deactivate
conda remove --name GNN --all 
jupyter kernelspec uninstall gnn
# remove all the caches created by conda
conda clean --all

Here is another choice provided by Jiaxuan You:

# Python environment (Optional)
conda create -n envname python=3.7
source activate envname

# Pytorch
# CUDA versions: cpu, cu92, cu101, cu102, cu101, cu111
pip install torch==1.8.0+cu101 -f https://download.pytorch.org/whl/torch_stable.html

# Pytorch Geometric
# CUDA versions: cpu, cu92, cu101, cu102, cu101, cu111
# TORCH versions: 1.4.0, 1.5.0, 1.6.0, 1.7.0, 1.8.0
CUDA=cu101
TORCH=1.8.0
pip install torch-scatter -f https://pytorch-geometric.com/whl/torch-${TORCH}+${CUDA}.html
pip install torch-sparse -f https://pytorch-geometric.com/whl/torch-${TORCH}+${CUDA}.html
pip install torch-cluster -f https://pytorch-geometric.com/whl/torch-${TORCH}+${CUDA}.html
pip install torch-spline-conv -f https://pytorch-geometric.com/whl/torch-${TORCH}+${CUDA}.html
pip install torch-geometric

from 50 to 100

Basically, you have to pre-check the version of (if you already directly use pip install xxx to install the following module):

• Pytorch
• Nvidia CUDA Compiler (NVCC)
• Nvidia GPU Driver
• Pytorch Geometric

check torch version:

conda activate GNN
python3
import torch
print(torch.__version__)
exit()

check CUDA Compiler Version:

nvcc

check Nvidia GPU Driver Version:

nvidia-smi

For more detail, check out Pytorch Docs

Common Issues