Design a dockerized custom ML model

Logisitic Regression created from scratch is used as an example

Features include:

1. Dockerization compatibility
2. Proper testing
3. Modularization
4. Offline training support

Project Structure

ml_project/

    │── data/                   Data folder (bind-mounted in Docker)
        
        │── processed/
        
            │── testing_data/
            
            │── training_data/

    │── models/                 Trained models saved here
    
        │── offline_model.pkl
        
        │── saved_model.pkl

    │── tests/                  Unit tests

       │──  test_classifier.py       

       │──  test_ml_datapipeline.py 

       │──  test_optimizerandloss.py 

       │──  test_pipeline.py 

       │──  test_preprocessing.py 

       │──  test_utils.py 

    │── tools/
    
        │── install.sh
        
        │── test_build.sh

    │── utils/                  

       ├── utils.py            General utility functions

       ├── preprocessing.py    Preprocessing functions
    
    │── docker-compose.yml

    │── config.yml

    │── ml_model.py             Main model training script

    │── ml_datapipeline.py      Make dataset according to config yml.

    │── Dockerfile              Docker setup

    │── requirements.txt        Dependencies

    │── README.md               Documentation