Pytorch-Geometric环境配置

打开Anaconda Prompt

1、新建环境

conda create -n torch python=3.8

2、安装pytorch-gpu

conda activate torch

conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/pytorch/

conda install pytorch==1.10.1 torchvision==0.11.2 torchaudio==0.10.1 cudatoolkit=10.2

测试

python
import torch
torch.cuda.is_available()

若返回True则安装成功，退出python

exit()

3、安装工具包‍

点击这里下载轮子先 共四个轮子要安装

pip install 轮子们的绝对路径
：
pip install C:\...\torch_cluster-1.5.9-cp38-cp38-win_amd64.whl
pip install C:\...\torch_scatter-2.0.9-cp38-cp38-win_amd64.whl
pip install C:\...\torch_sparse-0.6.12-cp38-cp38-win_amd64.whl -i https://pypi.douban.com/simple
pip install C:\...\torch_spline_conv-1.2.1-cp38-cp38-win_amd64.whl

其中torch_sparse轮子用源是因为它要下载scipy

4、安装Geometric

pip install torch-geometric -i https://pypi.douban.com/simple

5、其他

若 import torch 报错 No module named 'numpy.core._multiarray_umath'为numpy版本过低所致，执行

pip install -i https://pypi.tuna.tsinghua.edu.cn/simple --upgrade numpy