/backstage-mnist

Example ML training scripts to populate mlflow/backstage plugin.

Primary LanguagePython

MLFlow examples.

This will hold a few examples for training models and submitting them to the mlflow tracker, to be surfaced via the MLFlow Backstage plugin. The hard-coded paths should indicate the state of this repo.

  • component-info.yaml contains the required configuration for a Backstage component. You can "register an existing component" in backstage and give it the path to this file.
  • src/mnist_trainer.py - trains an image classifier on the mnist dataset.
  • src/backstage_utils.py - some helpers for reading component-info.yaml and configuring the mlflow sdk appropriately.

With Backstage and the MLFlow server running, the mnist_trainer script will produce views like this:

run view preview 1

run view preview 2