/ml-model-tracker

A simple tracker for recording the changing state of an ML model in progress.

Primary LanguagePython

The idea here is that I want to be able to create a semi automated 'checkpoint' when working on training ML models.

This thing should:

  • Be imported into a notebook.
  • Automatically commit when you get to a certain point, saving the code at point of training.
  • Make the commit hash available so you can save data in S3 against it, meaning the data is retained.
  • Allow you to save whatever params are important in a dict.
  • Save the params, and the results of the model test into a csv so you have a log.