深度学习下载失败总结ing

在查找资料过程中出现的报错

跳过前面的……

pip install --upgrade --ignore-installed tensorflow

安装tensorflow
失败

https好像过期了，换http

要先移除所有再添加

更新conda 
conda update -n base -c defaults conda

conda config --add channels http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/pytorch/
conda config --add channels http://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/
conda config --add channels http://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/
conda config --set show_channel_urls yes

移除所有网站channels
conda config --remove-key channels

查看网站channels
conda config --show-sources

版本不匹配出现的问题：gpu=1.15.0的版本

(pytorch) C:\Users\86185>conda install tensorflow-gpu=1.15.0
Channels:
 - http://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free
 - http://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main
 - http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/pytorch
 - defaults
 - https://mirrors.ustc.edu.cn/anaconda/pkgs/main
Platform: win-64
Collecting package metadata (repodata.json): done
Solving environment: / warning  libmamba Added empty dependency for problem type SOLVER_RULE_UPDATE
failed

LibMambaUnsatisfiableError: Encountered problems while solving:
  - package tensorflow-gpu-1.15.0-h0d30ee6_0 requires tensorflow 1.15.0.*, but none of the providers can be installed

Could not solve for environment specs
The following packages are incompatible
├─ pin-1 is installable and it requires
│  └─ python 3.11.* , which can be installed;
└─ tensorflow-gpu 1.15.0**  is not installable because it requires
   └─ tensorflow 1.15.0.*  but there are no viable options
      ├─ tensorflow 1.15.0 would require
      │  └─ python 3.6.* , which conflicts with any installable versions previously reported;
      └─ tensorflow 1.15.0 would require
         └─ python 3.7.* , which conflicts with any installable versions previously reported.

说明我的3.11的python不匹配tensorflow的版本

找匹配python与tensorflow的网站：

从源代码构建  |  TensorFlow

pytorch：
在命令提示符中输入nvidia-smi可得到

 NVIDIA-SMI 512.36       Driver Version: 512.36       CUDA Version: 11.6

也就是说当前电脑最多只能安装到CUDA11.6版本

需要用conda，用pip的话比较麻烦

 在官网找到相对应的，例如为为nvidia芯片，则
conda install pytorch torchvision torchaudio pytorch-cuda=11.6 -c pytorch -c nvidia

经查询，tensorflow最新可匹配的为python3.9的版本

结果是卡了半天进度条都不动，因为这些包都是国外网站的，卡是正常的，一般是最后403或者其他的下载失败

tensorflow与Pytorch最终下载鉴定都为失败，可能要换个网络试试不知道行不行，懂得都懂；

或者去官网看看吧

PyTorch

在 Windows 环境中从源代码构建  |  TensorFlow