mlflow-mysql-minio

Set up a MLFlow instance with MySQL and MinIO backends using Docker compose for development and testing purposes.

References

1. https://github.com/clemens33/mlflow

How to use

1. Clone the repo
2. Create and update the paths as mentioned with the #TODO in the docker-compose file
3. Navigate inside mlflow_mysql_minio folder and type docker-compose up --build --remove-orphans -d. Assuming Docker and Docker compose are installed and running.
4. To stop the the containers type docker-compose down

Access the different services

1. MLFlow

   • Purpose: To track metrics and artifacts generated from ML experiments
   • Access URL: http://127.0.0.1:5000

2. MYSQL Adminer

   • Purpose: To view contents of the MySQL database which contains the entity related data for MLFlow e.g. metrics
   • Access URL: http://127.0.0.1:8080
   • Access details:
     • System: MySQL
     • Server: mlflow-db
     • Username: mlflow
     • Password: mlflow123
     • Database: mlflow

3. MinIO web console

   • Purpose: To view artifacts that are stored in MLFlow e.g. models, images, etc.
   • Access URL: http://127.0.0.1:9001
   • Access details (root):
     • Username: mlflow_artifacts
     • Password: mlflow_artifacts_secret
     • Bucket name: ml/
     • MLFlow artifacts path: ml/mlflow/