Ready to use configuration mlflow tracking server + local S3 storage based on Docker and Minio

Prerequisites:

For mlflow tracking server machine:

For client machine:

just install mlflow and boto3 libraries to your environment

It's possible to use the same machine as the server and the client

Configurations

Client and server are the same machine
Server on remote machine

Configuration Client and Server are the same machine

Go to the directory, where your server will run and clone project

git clone https://github.com/leoromanovich/mlflow_tracking_server && cd mlflow_tracking_server

Go to mlflow_server directory

cd mlflow_server

Run server

sudo docker-compose up --build

Server is ready, so add few lines to client code
Check localhost:5555 for MLflow server and localhost:9001 for Minio
Enjoy!

Configuration Server is remote machine

Connect to your remote machine
Go to the directory, where your server will run and clone project

git clone https://github.com/leoromanovich/mlflow_tracking_server && cd mlflow_tracking_server

Go to mlflow_server directory

cd mlflow_server

Run server

sudo docker-compose up --build

Server is ready, so add few lines to client code
Check your_server_ip:5555 for MLflow server and your_server_ip:9001
Enjoy!

Client configuration

Add this lines to your experiment

import os
import mlflow

# Use IP of your remote machine here
# Don't change if client and server is the same machine
server_ip = '0.0.0.0'

os.environ['AWS_ACCESS_KEY_ID'] = 'minio'
os.environ['AWS_SECRET_ACCESS_KEY'] = 'minio123'
os.environ['MLFLOW_S3_ENDPOINT_URL'] = f'http://{server_ip}:9001'

mlflow.set_tracking_uri(f"http://{server_ip}:5555")
mlflow.set_experiment("awesome-experiment")

What if I want to change

Aritfact-bucket and save artifacts there?

Open docker-compose.yml file and change all "mlflow-artifacts" entries to "new_bucket_name"

Acknowledgments

Mikhail Knyazev
Dmitry Kozlov
OCRV