LAVS-TM / MLOPS

Complete Machine Learning in production using fastAPI, MLflow, Streamlit and Docker.

MLOPS

This project describe a complete machine learning project in production, with the training and the deployement phase.

AUTHORS

Alexandre Lemonnier & Victor Simonin

---

Usage

Docker

You can build the docker images with the following command:

docker-compose up -d --build

Once the docker images are built and started, you can access to the following services:

• the streamlit front-end at the following address http://localhost:8501.
• the web interface of the mlflow server at the following address http://localhost:5000.
• the uvicorn api at the following address http://localhost:8000.

To shutdown the docker containers, run the following command:

docker-compose down -v

Manual

The training of the model is in the src/train_model.py file. It is possible to train the model with the following command:

cd src
python train_model.py

A mlflow workflow is available to track the training. After a few training, to start the mlflow server, run the following command:

mlflow ui

To start the api, run the following command:

cd src
uvicorn api:app --reload

The streamlit front-end is available with the following command:

cd src
streamlit run model_app.py