conda安装pytorch1.0.0 GPU+CPU版本

1、

conda create -n torch100 gdal python=3.6

2、

# CUDA 10.0
conda install pytorch==1.0.0 torchvision==0.2.1 cudatoolkit=10.0 -c pytorch

# CUDA 9.0
conda install pytorch==1.0.0 torchvision==0.2.1 cudatoolkit=0.0 -c pytorch

# CUDA 8.0
conda install pytorch==1.0.0 torchvision==0.2.1 cudatoolkit=8.0 -c pytorch

# CPU Only
conda install pytorch-cpu==1.0.0 torchvision-cpu==0.2.1 cpuonly -c pytorch

安装完如import报错：

Traceback (most recent call last):
  File "", line 1, in
  File "/home/qqlee/anaconda3/envs/torch100/lib/python3.6/site-packages/torch/__init__.py", line 84, in
    from torch._C import *
ImportError: /home/qqlee/anaconda3/envs/torch100/lib/python3.6/site-packages/torch/lib/libmkldnn.so.0: undefined symbol: cblas_sgemm_alloc

则：

conda install -c anaconda mkl

3、测试

python
import torch
torch.__version__          #输出Pytorch的版本
torch.cuda.is_available()   #看看能否正常调用CUDA
# 输出True代表可以正常调用

4、版本对应关系

pytorch 不同版本对应的cuda_一个小小记事簿-CSDN博客