machine-learning-apps / mlops-dashboard

Prototype MLOps Dashboard Using GitHub Pages

Purpose

Give developers an incentive to provide a standardized ML specification for machine learning projects, to enable the following:

• An entrypoint for users to run ML workloads on the compute of their choice ( or on GitHub directly? )
• Provide mechanisms to make ML workflows extensible and facilitate greater collaboration / experimentation.
• Transperency by organizing information into a common format
• Reproduceability

One way of providing this incentive is to render the information associated with the standardized specification into a dashboard which facilitates the consumption of this information. Furthermore, this dashboard should facilitate interacting with the ML workflow, including running and deploying models.

The Prototype

This initial prototype uses Jekyll and GitHub Actions to render assets (metadata in the form of YAML & JSON, notebooks, etc) into a MLOps Dashboard. The metadata is located in a repo in a pre-determined directory structure. When metadata is created or changed in a repo, a GitHub Action refreshes the GitHub page.