Panther: Faster and Cheaper Computations with Randomized Numerical Linear Algebra
Training modern deep learning models is increasingly constrained by GPU memory and compute limits. While Randomized Numerical Linear Algebra (RandNLA) offers proven techniques to compress these models, the lack of a unified, production-grade library prevents widely adopting these methods. We present Panther, a PyTorch-compatible library that consolidates established RandNLA algorithms into a single high-performance framework. Panther engineers efficient, drop-in replacements for standard components including sketched linear layers, 2D convolution, multi-head attention, and randomized matrix decompositions (such as pivoted CholeskyQR). By implementing a custom C++/CUDA backend (pawX), Panther provides an optimized implementation that can run on both CPUs and GPUs. We demonstrate the effectiveness of RandNLA techniques and Panther’s ease of adoption. By replacing standard PyTorch linear layers with Panther layers (requiring only a few lines of code) we achieve significant memory savings (up to 75%) on BERT while maintaining comparable loss. Source code is available (MIT License) at https://github.com/FahdSeddik/panther, along with demonstration video at https://youtu.be/7M3RQb4KWxs.
Thu 9 JulDisplayed time zone: Eastern Time (US & Canada) change
14:00 - 15:30 | SE and AI 3Tool Demonstrations / Ideas, Visions and Reflections / Industry Papers / Journal-First Paper / Research Papers at MB 3.210 Chair(s): Earl T. Barr University College London | ||
14:00 20mTalk | Model Context Protocol (MCP): Landscape, Security Threats, and Future Research Directions Journal-First Paper Xinyi Hou Huazhong University of Science and Technology, Yanjie Zhao Huazhong University of Science and Technology, Shenao Wang Huazhong University of Science and Technology, Haoyu Wang Huazhong University of Science and Technology | ||
14:20 20mTalk | Human-aligned AI Model Cards with Weighted Hierarchy Architecture Industry Papers Pengyue Yang The University of Sydney, Haolin Jin The University of Sydney, Qingwen Zeng The University of Sydney, Jiawen Wen The University of Sydney, Harry Rao Bytedance, Huaming Chen The University of Sydney | ||
14:40 10mTalk | Machine Learning in the Wild: Early Evidence of Non-Compliant ML-Automation in Open-Source Software Ideas, Visions and Reflections Zohaib Arshid University of Sannio, Italy, Daniele Bifolco University of Sannio, Fiorella Zampetti University of Sannio, Italy, Massimiliano Di Penta University of Sannio, Italy | ||
14:50 20mTalk | SMARLA: A Safety Monitoring Approach for Deep Reinforcement Learning Agents Journal-First Paper Amirhossein Zolfagharian University of Ottawa - School of Electrical Engineering & Computer Science (EECS), Manel Abdellatif École de Technologie Supérieure, Lionel Briand University of Ottawa, Canada; Lero centre, University of Limerick, Ireland, Ramesh S | ||
15:10 10mTalk | DevGen: Automated Generation of Virtual Device Models for Kernel Drivers via Large Language Models Ideas, Visions and Reflections Mingyu Wang Xidian University, Bin Yu Xidian University, Wenjian Lu Xidian University, kefeng gao Xidian University, Zhi Wang Xidian University, Cheng Wen Xidian University, Xu Lu Xidian University, Cong Tian Xidian University | ||
15:20 10mTalk | Panther: Faster and Cheaper Computations with Randomized Numerical Linear Algebra Tool Demonstrations Fahd Seddik University of British Columbia, Abdulrahman Elbedewy University of Texas at Austin, Gaser Elmasry Cairo University, Mohamed Abdelmoniem Noon, Yahia Zakaria Cairo University Pre-print | ||