MLflow Integration in Gathr
MLflow is an open-source, unified platform to manage end-to-end ML and GenAI workflows. It streamlines the machine learning process, from experimentation to production deployment, providing tools for tracking experiments, sharing models, and managing model deployment.
Integrate your MLflow instance with Gathr by connecting to the MLflow Tracking Server, allowing access to MLflow models directly within Gathr’s interface.
Utilize MLflow models in ETL applications with the MLflow Inference processor to generate inferences on your prediction data.
MLflow Integration Prerequisites
MLflow Tracking Server Version 2.9.2 is supported.
MLflow model should be hosted in user’s cluster.
Obtain the hosted model’s endpoint URL.
For LLM Models: AI Gateway URL & Model Route Name.
For ML Models: Model Serving URL.
Ensure the model services are up and running before using them to make inferences within Gathr.
Set Up MLflow Connection
Navigate to Models > MLflow Tab > Create Connection to create an MLflow connection.
You can also create a connection from the Connections page and then navigate to the Models > MLflow tab.
If creating from Connections page, select MLflow as the ‘Component Type’ to create the connection.
Create MLflow Connection
Enter the necessary configuration parameters to create the connection.
Connection Name
Provide a unique name for your MLflow connection. The saved connection can be identified in Gathr with the name you provide.
Scope
Define the scope of connection to customize its accessibility.
Organization: Accessible to organization users across all Gathr projects for usage in applications.
Project: Accessible to organization users limited to projects that are specified by the connection owner for usage in applications.
Private: Accessible only to the connection owner for usage in applications.
MLflow Tracking URL
URL where the MLflow Tracking server is hosted.
If your MLflow server is hosted locally, the URL might look like http://localhost:5000
. If it’s hosted on a server accessible over the internet, it could look something like http://mlflow.example.com
.
Example: http://mlflow-server:5000
In this example:
http
indicates the protocol.mlflow-server
is the hostname or IP address where your MLflow Tracking server is hosted.5000
is the port number where the MLflow Tracking server is listening.
Replace mlflow-server and 5000 with the actual hostname or IP address and port number of your MLflow Tracking server.
MLflow Username
Username of MLflow to create connection.
MLflow Password
Password of MLflow to create connection.
Check with your MLflow server administrator for the authentication method in place and obtain the necessary credentials.
Test Connection
After entering all the configuration details, verify that the connection to the specified resources is correctly configured.
Click on the Test button.
If the provided details are correct and all required services are operational, you will receive a confirmation message indicating that the connection to the specified resources is successfully established.
If any of the provided details are incorrect, or if any required services are not running or accessible, you will receive an appropriate message indicating an issue with the connection setup. You will need to review and correct the configuration details accordingly.
Click on the Create button once the configuration details are provided.
Models will be listed according to the MLflow Tracking URL provided in the MLflow connection.
Refer to the MLflow Models Listing Page for details on listed models.
See the MLflow Inference Processor for configuration details for use in ETL applications.
If you have any feedback on Gathr documentation, please email us!