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
Includes following high level packages for CPU only
- miniconda with python 3.8.8
- Pytorch
- Tensorflow
- keras
- sklearn
- spaCy
- spaCy en_core_web_trf package for english
- nltk
- Huggingface transformers
- sentencepiece
- sentence_transformers from SBERT
- charting packages (plotly, matplotlib, bokeh, seaborn)
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.
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.
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