This repository provides necessary artefacts to quicky and easily deploy an MLflow Tracking server on Azure.
MLflow is an open source platform for managing the end-to-end machine learning lifecycle.
- Ensure you are logged-in in the azure cli.
- Run
az login
to login. - Run
az account set -s <SUBSCRIPTION_ID>
to set target azure subscription.
- Run
- Ensure you are in the
deploy-aci
folder. - Open
deploy-aci.sh
and inspect/change top parameters, if necessary. - Run
./deploy-aci.sh
- Validate deployment by navigating to the ACI IP:port (default: 5000). NOTE, that it takes a few moments for the server to startup.
- You can retrieve IP and port of the deployed Tracking Server on ACI by running:
az container show --name <ACI_NAME> --resource-group <ACI_RESOURCE_GROUP> --output table
Documentaion for mlflow repository and scripts
- Mlflow server deployed on a container instance with:
- storage container - with blob and file store
- docker image for mlflow server
- Version 1.8.0 of mlflow
- Models saved to blob storage and meta data saved to SQLlite db inside the container
Build image
docker build --tag jaredmagrath/mlflowserver-azure:1.8.2 .\mlflow-tracking-docker\
Push to dockerhub
docker push jaredmagrath/mlflowserver-azure:1.8.2