Project required dependency

[airyyan]

Does the project have a corresponding docker image or can the project file generate a requirement.txt file? Once the code is downloaded, installing the corresponding dependencies is a headache. thank you！

[frankligy]

Hi @airyyan,

Sorry for the inconvenience, I can help you build the dockers once I get a bit free time. But in the meantime, I was wondering what are the problem that you were encountering with? What if just start a fresh conda environment with python version set as 3.6, and start with first installing the tensorflow (if using deepimmuno-cnn) or pytorch (if using gan) with the correct version number, and then downgrade numpy and pandas if needed. I'll also test it and see if there's any mysterious conflict going on.

Frank

[airyyan]

Now the project is running, but from the running results, can only GPU be used? Or is there something wrong with the way I use it?
So far I've just downloaded the code and installed the dependencies. Run: python3 deepimmuno-cnn.py --mode "single" --epitope "HPPLMNVER" -- HLA "HLA-A*0201"
If I am deploying it to a GPU server, do I also need to pass the "python3 deepimmuno-cnn.py" instruction every time I use it?

[frankligy]

Hi @airyyan,

For the CNN model, it's been a while but I assume you asked "can only CPU be used", right? I never tried or tested CNN model on GPU because it's so lightweight. For the GAN model, it will automatically utilize GPU if you have or bounce back to CPU.

Regarding how to run the code, yes I believe how to run a python code won't change no matter we use GPU or CPU. Also I just want to say, since the code is so lightweight, you can easily modify this API in the way you like, you can remove the command line argument and just put all functions into your existing code, or adding a shebang line to make the python script an executable, etc.

Best,
Frank