Interpret

October 30, 2020 ยท View on GitHub

The PyTorch implementation the Smooth Grad and Integrated Gradients for NLP Models.

This repository based on the Allen NLP Interpret Module and describes how to integrate the Hugging Fase's PyTorch Model to it.

  • The explanation you can find in this article.
  • The example of usege in the notebook.

The example of the output:

Integrate your own task-specific model

To integrate your model simply override the forward_step and update_output methods of base class SaliencyInterpreter.