/mldeploy

An attempt to deploy machine learning models via docker

Primary LanguageDockerfileApache License 2.0Apache-2.0

mldeploy

An attempt to deploy machine learning models via docker. In this docker image most common build tools and common machine learning frameworks have been included

Docker image details

Includes following high level packages for CPU only

  1. miniconda with python 3.8.8
  2. Pytorch
  3. Tensorflow
  4. keras
  5. sklearn
  6. spaCy
  7. spaCy en_core_web_trf package for english
  8. nltk
  9. Huggingface transformers
  10. sentencepiece
  11. sentence_transformers from SBERT
  12. charting packages (plotly, matplotlib, bokeh, seaborn)

How to use

docker run -it -v your_persitent_location_on_disk:/workspace/app llearnell/ubuntu-ml Once the container is running, you will be provided with a /bin/bash prompt within /workspace/app directory All python environment already set.

Where are the downloaded transformer models stored

They are stored in /workspace/app/data/ folder. This is one level inside the persistent_location_on_disk you provided while running the docker container.

Why is the image size so large ?

I wanted a system where i have to do minimal steps and not wait for dependency resolution by conda everytime i replicate the environment.

Docker container can be found here: https://hub.docker.com/repository/docker/llearnell/ubuntu-ml