pytorch1.4.0+torch_scatter+torch_sparse+torch_cluster+torch_spline_conv安装

环境：
CUDA10.0.130
CUDNN7.6.5
具体参见Ubuntu显卡驱动安装、CUDA+CUDNN安装文章，本文不赘述。

需要配置环境变量：

vim .bashrc

export PATH=/usr/local/cuda-10.0/bin:$PATH
export LD_LIBRARY_PATH=/usr/local/cuda-10.0/lib64:$LD_LIBRARY_PATH
export CUDA_HOME=/usr/local/cuda-10.0:$CUDA_HOME
export CPATH="/usr/local/cuda-10.0/include:$CPATH"

source .bashrc

可通过以下命令查看：

echo $PATH
echo $CPATH

---

以pytorch1.4.0+cu100为例：

pip --default-timeout=100 install -i https://pypi.tuna.tsinghua.edu.cn/simple torch==1.4.0+cu100 torchvision==0.5.0+cu100 -f https://download.pytorch.org/whl/torch_stable.html

为避免torch-sparse等安装报错，先行下载：
https://pytorch-geometric.com/whl/torch-1.4.0.html
具体根据torch版本改变网址即可。
下载完毕后，安装如下：

cd 下载
pip install torch_scatter-2.0.3+cu100-cp36-cp36m-linux_x86_64.whl
pip install torch_sparse-0.6.0+cu100-cp36-cp36m-linux_x86_64.whl
pip install torch_cluster-1.5.4+cu100-cp36-cp36m-linux_x86_64.whl
pip install torch_spline_conv-latest+cu100-cp36-cp36m-linux_x86_64.whl
pip --default-timeout=100 install -i https://pypi.tuna.tsinghua.edu.cn/simple torch-geometric==1.4.1

以上，安装了torch_scatter,torch_sparse,torch_cluster,torch_spline_conv,torch_geometric包及相关依赖。