LibMTL: A Python Library for Deep Multi-Task Learning
Authors: Baijiong Lin, Yu Zhang; Published in Journal of Machine Learning Research, 24(209):1−7, 2023.
Abstract
This article introduces LibMTL, a Python library based on PyTorch that offers a comprehensive, reproducible, and extensible implementation framework for Multi-Task Learning (MTL). LibMTL supports various settings and approaches in MTL, providing state-of-the-art methods including 13 optimization strategies and 8 architectures. The modular design of LibMTL ensures ease of use and extensibility, allowing users to quickly develop new MTL methods, fairly compare them with existing ones, and apply MTL algorithms to real-world applications. The source code and detailed documentation for LibMTL can be found at https://github.com/median-research-group/LibMTL and https://libmtl.readthedocs.io, respectively.
[abs]