A template repository to make all machine learning projects.
Most people find it hard to package their python code and do not know how to set up the repo for it.
If the repository is setup in a wrong way, it would become hard to package and deploy the code later on as well.
This repo gives you the batteries required to package your code, CI checks, auto build and deploy docs, easy PyPi publishing support, and docker files.
This serves as a template to quickly have these things setup in your repo. Machine Learning Repos created from this template can easily be deployed and shipped. It becomes hassle free and easy to debug too.
You can add your code in template_python
folder. Since this is a package make sure that imports are
from the root. i.e. from template_python import stuff
- Choose use this repo as template option in GitHub.
- This should create a repo in your GitHub account.
- You would need to edit some files in order to rename this from
template_python
to your required repo name. - Just replace
pip install git+git://github.com/oke-aditya/template_python.git
in themk-docs-build.yml
andmk-docs-deploy.yml
workflows in .github folder with your package git url. This will set up docs. - Please edit ALL the
.md
files to include description that you need. - Edit the
requirements.txt
andrequirements-extra.txt
(optionally). - Edit the
.gitingore
and.dockerignore
files if anything extra is needed. I have included most stuff in them. - Edit the
settings.ini
andsetup.py
(optionally). You perhaps need your name and different requirements. Again most stuff is there, you need to add very small tweaks. - Edit the
LICENSE
you may need a different one. - Edit the docs folder. It is built using mkdocs.org. You can refer to mkdocs to know more about docs editing. These are just minimalistic docs that do the job.
- Optionally Add tests to
tests
folder usingpytest
.
Also please read the README
files that are present in the folders. They will help and guide you to setup stuff too.
- If you face any issues setting up project do raise an issue. I would be happy to help.
- If you have a few suggestions and additions you would like please raise a PR. I would be happy to merge.
Note: - These repos might have diverted a little bit from this template.
- PyTorch CNN trainer.
- Fashion Intel
- PyTorch Image Similarity
- quicktext: A Fast & Simple Text Detection Framework
Raise a PR if you have built your project with this template and I will add it here !!
This template was created using a lot of repositories, it includes these
- fastai nbdev
- Made With ML boilerplate.
- PyTorch Lightning
- IceVision
Huge credit to these repos, it would be hard to make this without them.
This template is different from above as this lays emphasis on bundling code in python packages and containers thus ensuring portability.