microsoft / BackwardCompatibilityML

Project for open sourcing research efforts on Backward Compatibility in Machine Learning
https://backwardcompatibilityml.readthedocs.io/en/latest/
MIT License
71 stars 14 forks source link

Fix examples to use realistic training scenarios for h2, and fix widget re-instantiation failure #124

Closed ilmarinen closed 3 years ago

ilmarinen commented 3 years ago
  1. In the compatibility analysis examples, make sure that h1 is trained on a fraction of the dataset, while h2 is trained on all of the dataset.
  2. Fix a bug where the model comparison widget would fail on re-instantiation and needed the notebook / kernel to be re-started.