Anaconda environment installation problem

Recently, I am learning reinforcement learning , Of course, it uses python 了 , Don't bother the first few guys who just started running didn't use tensorflow Your code is ok , There is nothing wrong with reporting , The environment installation is also normal .

But learn DQN It's different from the beginning , install tensorflow There are always problems , Here are my problems and solutions , Draw lessons from .

1 conda and pip

conda and pip Can be used to download third-party libraries , such as pip install tensorflow and conda install tensorflow, But if you use Anaconda Manage resources , Then I recommend you use conda install tensorflow, And use it first activate xxx Enter your environment , And then use conda install tensorflow Install the third-party package .

Anaconda Have strong environmental management ability , It can be configured with different python route , And manage third-party packages for each path , Like the path 1 Of python The version is 3.5, It installs numpy、pandas Two bags ; route 2 Of python The version is also 3.5, It only installs tensorflow package , All this comes from Anaconda Management of .

that pip and conda What's the difference ？ This is my understanding ：

Under normal circumstances （ For example, I ） One will be installed first python Version or multiple reinstallations Anaconda（ There is no need to install python, because Anaconda You can also install different versions python）. So using pip Install third-party packages such as pip install tensorflow, Its installation position is python Under the path instead of Anaconda The next path , If you use conda install tensorflow, Then the installation path is Anaconda Under the path , This makes it easy to manage , It is also convenient to run different code in different packages .

Put a picture to show ：

Remember this path , One version is python3.5.x, One is python3.6.13, Then, for example, you need to python3.6.13 Install packages in, such as pytz. Enter the first python3-6-13, Install again , Specific commands ：

activate python3-6-13
conda install pytz

Then here is python The path of , use pip If installed, the high probability may be here ;

I will find that I have many bags here , But it's all because I didn't know at the beginning pip and conda The difference leads to both pip Installed , And then in pycharm When running in, it uses Anaconda The environment , You will find that Mingming I pip install tensorflow 了 , Why can't I find it tensorflow The fault of .

2 Anaconda Environment transplantation

But I don't know why I recently used Anaconda Install third-party libraries, such as conda install xxx It didn't work , The specific reason has not been found , That's how I solved it , My old computer installed Anaconda There is no problem installing any libraries , So I put the whole old computer Anaconda3 The folder is copied to the new computer , And then in pycharm Use in Anconda I used the old computer when I was Anaconda Environmental Science , It's also possible .

In this picture Anaconda3 It was installed on the new computer Anaconda Environmental Science , and Anaconda The folder is copied from the old computer , It is feasible. , I tested. , You just need to pycharm When configuring the compiler in Anaconda The file corresponding to the folder is ok .

3 Qt Report errors

Old computer copy Anaconda There is one after the file that needs to be modified , Namely Anaconda In the catalog qt.conf file .

The content is like this , To change the starting path , Otherwise, when you use python Make drawings, such as matplotlib mapping , Will report a mistake ;

It's because the path hasn't been changed , Change to corresponding Anaconda Just go down the path .