在 AWS 上部署 django 时出现 Pytorch 导入错误

Question

My django app works locally. When i try to deploy it and then open it on AWS, i get the following error:

I included torch==1.5.1 in my requirements.txt

Requirements.txt:

When I comment out the parts of the application that require torch and remove it from requirements.txt, it works fine. What can I do to ensure that torch is installed like the other modules in requirements.txt? Here is the link to requirements.txt

Answer 1

By default torch will require CUDA and GPU . Depending on your ec2 instance this may not be viable option, thus leading to the errors you are observing.

You can install non-cuda version of torch as follows for tests and see if this will have any benefits:

pip3 install torch==1.5.1+cpu torchvision==0.6.1+cpu -f https://download.pytorch.org/whl/torch_stable.html