I find the process described in the repo instructions fairly daunting, both in terms of time required as in steps required.
Do you see any problem in the following proposed alternative method?
1.- Download the respective package_xxx.whl from the project to the /submission_src folder
2.- From your python code, in python, just before the execution of anything your main.py code (in the first lines), run the installation of the .whl:
cmd = 'pip install package_xxx.whl'
I have made my experiments using the latest version of the pulled image with some very specific needed packages and it seems to run smoothly.
This way also, the size of the docker image doesn’t get so bloated for everyone, only at the expense of making submissions slightly larger.
The process you describe is fine, though we do prefer that you try to go through the official means if you can. Also note that with great power comes great responsibility - we won’t be able to fully support or debug if things go wrong in your custom environment. Please also document the libraries you are vendoring - what versions, hashes, etc.
If you would like, I can help guide you through the process of adding packages - it may look daunting but it can be done!
Thank you very much for you prompt reply @chrisk-dd
I’m happy to learn that I am not inventing some nonsense here
On the other hand, I think I must apologize for using the term daunting when referring to the updating process. I understand the reasons behind and it is well documented, so it is indeed doable
In order to be more precise, what I wanted to express is that, in the process of trial-and-error that (we) participants are involved, having flexibility to decide which packages to use and which ones to discard is of some practical importance.
Naturally, the shorter the cycles, the better quality results one may obtain.
Once those extra packages are consolidated, of course it is wiser to have them “officialized” through the updating procedure