mlflow-elasticsearchstore

Mlflow plugin to use ElasticSearch as backend for MLflow tracking service. To use this plugin you need a running instance of Elasticsearch 6.X.

Run 'pip install mlflow-elasticsearchstore' to register the plugin as an entrypoint with Elasticsearch backend.

$ pip install mlflow-elasticsearchstore

Development

In a python environment (you can use the one where mlflow is already installed):

$ git clone git clone https://github.com/criteo/mlflow-elasticsearchstore.git
$ cd mlflow-elasticsearch
$ pip install .

How To

mlflow-elasticsearchstore can now be used with the "elasticsearch" scheme, in the same python environment :

$ mlflow server --host $MLFLOW_HOST --backend-store-uri elasticsearch://$USER:$PASSWORD@$ELASTICSEARCH_HOST:$ELASTICSEARCH_PORT --port $MLFLOW_PORT --default-artifact-root $ARTIFACT_LOCATION