This directory contains jupyter notebooks for tutorials of InterpretDL.
Usages examples are moved to examples.
Usage examples do not discuss the details of these algorithms. Therefore, for both practical and academic purposes, we are preparing a series of tutorials to introduce the designs and motivations of interpreters and trustworthiness evaluators.
The available (and planning) tutorials are listed below:
-
Getting Started. This tutorial includes the installation and basic ideas of InterpretDL.
-
NLP Explanations. There are four tutorials for NLP tutorials, using Ernie2.0 in English (on NBViewer), Bert in English (on NBViewer), BiLSTM in Chinese (on NBViewer) and Ernie1.0 in Chinese (on NBViewer) as examples. For text visualizations, NBViewer gives better and colorful rendering results.
-
Input Gradient Interpreters. This tutorial introduces the input gradient based interpretation algorithms.
-
LIME and Its Variants Part1 (LIME) | Part2 (GLIME). This tutorial introduces the LIME algorithms and many advanced improvements based on LIME.
-
GradCAM on Object Detection Models. This tutorial shows how to use GradCAM to explain object detection models. Mask-RCNN and PPYOLOE are used as models.
-
Transformers (to appear).
-
Trustworthiness Evaluation Tutorials (to appear).
-
Dataset-Level Tutorials (to appear).
