End-to-End-ML-Recipe (employee burnout prediction)

Setup conda env

Setup conda virtual environment from the environment.yml file of this repo.

conda env create -f environment.yml
conda activate emp_burnout

Setup package env

Install the codebase as a package (setup.py)

pip install .

Usage

• For running the servers (MLFlow server, Minio, NGINX), start docker-compose as:

  docker-compose up 

  • Make sure to configure the volumes (check host paths) in the docker-compose.yml file as needed.
  • The server is only needed for tracking purposes. If you don't need mlflow tracking, you can opt to ignore this step.

• For the client REST apis, start the app with uvicorn using:

  uvicorn emp_burnout.app:app

• Head over to localhost:8000/docs to view the swagger ui for the exposed REST apis.

• For the job config files, refer to configs/train.yml and configs/predict.yml.

  • The train.yml also contains the hyperparameters for training. One can modify them too.

TODOS:

• Groundwork
• Environment setup
• Ingestion to DB
• Docker setup
• MLFlow Server setup
• Training Job
• Batch prediction Job
• Single input prediction Job
• Cleanup
• Pydantic config parsing
• REST APIs
• Update README