Metadata-Version: 2.1
Name: lerg
Version: 0.0.1
Summary: A unified approach to explain the output of any conditional text generation model.
Home-page: https://github.com/Pascalson/LERG
Author: Yi-Lin Tuan
Author-email: ytuan@cs.ucsb.edu
License: MIT
Project-URL: Bug Tracker, https://github.com/Pascalson/LERG/issues
Platform: UNKNOWN
Classifier: Programming Language :: Python :: 3
Classifier: License :: OSI Approved :: MIT License
Classifier: Operating System :: OS Independent
Requires-Python: >=3.6
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: matplotlib
Requires-Dist: numpy
Requires-Dist: torch
Requires-Dist: tqdm
Requires-Dist: transformers

LERG (Local Explanation of Response Generation) is a unified approach to explain the output of any conditional text generation model. LERG extends and unifies game theory based approach (Shapley value) and local explanations to consider the relationship among output time steps, thus being suitable for text generation.

