Running MLFlow locally
This repo contains code to run MLFlow locally in two configurations.
Docker
This approach to running MLFlow locally builds the MLFlow server with a MySQL DB and S3 Storage with a single docker-compose file. The S3 storage and MySQL DB are persistent. Currently, it runs version 1.30.0 of MLFlow.
The Docker folder contains a docker-compose.yml file. To build and run the containers just run:
docker-compose up -d --build
This will start three Docker containers
- MLFlow Server
- S3 storage -> MinIO Instance
- MySQL Database
- mc container which will run the setup for the S3 storage
Python
This approach is based on the newly released version 2.0.0 of MLFlow, which allows running MLFlow directly on the user machine without docker. All the files required for the MLFlow ui will be stored in the file system (mlruns)