Demo how to
deploy a Mlflow app to Heroku cloud
. Main point : Show how to train/manage ML tasks as experiments with various parameter collections, save the model config, train results and visualize them via Mlflow framework. For more data infra/ETL development ideas, please visit :
- Daas (Data as a service) repo : Data infra build -> ETL build -> DS application demo
- Airflow Heroku demo : airflow-heroku-dev
Quick-Start-docker
$ cd ~ && git clone https://github.com/yennanliu/mlflow-heroku-dev.git
$ cd ~ && cd mlflow-heroku-dev
$ docker build -t mlflow . && docker run -p 5000:5000 mlflow
# visit the MLflow UI via http://0.0.0.0:5000
Quick-Start-deploy-to-heroku
$ cd ~ && git clone https://github.com/yennanliu/mlflow-heroku-dev.git
$ cd ~ && cd mlflow-heroku-dev
$ heroku create mlflow-heroku
$ git add . && git commit -m 'update for heroku deploy' && git push origin
# Set the stack of your app to container
$ heroku stack:set container
# login to heroku container
$ heroku login
$ heroku container:login
# push dockrized app to heroku
#$ git push heroku master
# push to heroku container
$ heroku container:push web -a mlflow-heroku
- Intro to Mlflow
- Deploy app built from Docker Images to Heroku