ubuntu16.04 借助anaconda安装多版本 pytorch

参考博客1
Anaconda下pip离线安装Pytorch(亲测可用）
https://blog.csdn.net/caracol/article/details/94832202
参考博客2
Ubuntu16.04安装pytorch经验整理
https://blog.csdn.net/qq_42165939/article/details/86181585?utm_medium=distribute.pc_relevant.none-task-blog-BlogCommendFromMachineLearnPai2-2.edu_weight&depth_1-utm_source=distribute.pc_relevant.none-task-blog-BlogCommendFromMachineLearnPai2-2.edu_weight

下载清华镜像 pytorch linux
https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/pytorch/linux-64/
torchvision-0.4.0-py36_cu100.tar.bz2
pytorch-1.4.0-py3.6_cuda10.0.130_cudnn7.6.3_0.tar.bz2
安装
conda activate pytorch
cd downloads
conda install torchvision-0.4.0-py36_cu100.tar.bz2
conda install pytorch-1.4.0-py3.6_cuda10.0.130_cudnn7.6.3_0.tar.bz2
成功界面
Downloading and Extracting Packages
######################################################################## | 100%
Preparing transaction: done
Verifying transaction: done
Executing transaction: done
进入python
import torch
import torchvision
看是否正确安装
1.非root用戶import torch遇到ImportError: libmkl_intel_lp64.so: cannot open shared object file: No such file
https://blog.csdn.net/Jean001100/article/details/106613553
将路径添加到环境变量
2.import torchvision 失败
torchvison 版本换个
3.importError: No module named six，已解决
https://blog.csdn.net/Jason43/article/details/84834990

根据博客2已经把包安装的源更改了

---

在安装一个pytorch 0.3
用 anaconda 导航创建pytorch2 python3.6
conda activate pytorch2 激活虚拟环境
conda install pytorch=0.3.1(在线安装)或者
pip install pytorch-0.3.1-py36_cuda9.1.85_cudnn7.0.5_2.tar.bz2 (从这里下的镜像离线安装 https://pytorch.org/get-started/previous-versions/)
pip install torchvision0.2.1 安装torchvision
检验是否成功
python
进入python
import torch 不报错
import torchvision 不报错
torch.cuda.is_available() 检查cuda 是否可用 TRUE
exit()离开python
nvcc -V 检查cuda 版本
nvidia-smi 检查显卡驱动