machine_learning_docker

In this repository I showcase examples of using docker to make handling ML research and production easier. As of writing there are three projects

Flask

I serve a pytorch based image classifier using flask. This is essentially the same code as in the official pytorch tutorial

However, I also add a jupyter server running a notebook that access the above service. Furthermore, both of these are in docker containers so running them is quite easy.

Multiple Jupyter Kernels in Docker

There are two projects for this one. The first one called multiple_conda_environments is simple and is meant as a tutorial. The second jupter_ml_docker is actually useful Docker image that contains tensorflow and pytorch kernels. The docker-compose file manages two containers from this image - one running the jupyter notebook server with these two kernels and the other running tensorboard.