Examples

This section contains examples of Stained Glass Core in action.

• Cancer image classification

  ---

  
  The purpose of this notebook is to demonstrate how to use Stained Glass Core to create a Stained Glass for a pretrained base model using a pure distillation loss.

  View Notebook

• Computing the Voronoi l2 Distance of a LLM's Embedding Table

  ---

  The inverse power law geometric catalyst loss uses a parameter inverse_l2_power_law_characteristic_length which is recommended to be set as the 5th percentile of the Voronoi l2 distances between embeddings.

  View Notebook

• Creating Stained Glass for LLMs Recipe (with logging)

  ---

  This notebook provides a full recipe for creating a Stained Glass Transform for LLMs with additional logging to track the evolution of metrics, parameters, generated text, and obfuscation scores during training.

  View Notebook

• Inference with Mistral 7B

  ---

  This notebook demonstrates using a pre-trained Stained Glass Transform to transform text inputs for a Mistral 7B model.

  View Notebook

• TensorFlow Adapter

  ---

  The notebook demonstrates how to use Stained Glass Core to create a Stained Glass Transform for a TensorFlow model using TensorFlowAdapter, a PyTorch-based TensorFlow model wrapper.

  View Notebook