如何使用 pyinstaller 使用 torch 模块制作 exe 文件？

Question

[Environment]

• vscode
• Anaconda
• Window

I successfully ran main_for_exe.py in vscode.

When I make exe file using pyinstaller and then ran exe file, however, I have the below error message.

enter image description here

There were some solutions: "add module to hiddenimports"

enter image description here

but, it does not work.

Basic modules of python such as "os" and "sys" are successfully loaded, but "No module named numpy" error message appeared. So, I think that modules installed by 'pip' and 'conda' aren't loaded.

How do I solve this?

Answer 1

No matter base or other virtual conda environments, we need to first activate it then with the command pip install executed in integrated Terminal, the module can get its right location, or it will throw No module error.

Now open your project in VS Code, after activating anaconda environment, run pip show numpy to check if its location is anaconda\lib\site-packages . If not, please reinstall it, then regenerate.exe, the error should go away.

Reference: Select and activate Environment .

Answer 2

I solved this problem!

I just upgraded the pyinstaller package.

It seems that the pyinstaller package, before upgrade, cannot find the site-package path where the modules installed by anaconda are located. (Generally site-packe path: "~/anaconda3/lib/site-packages")

In this case, there are two solutions.

1. First, you just upgrade pyinstaller.
2. Second, you add site-package path to hookspath and then run "pyinstaller main.spec".