Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2601.15473 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866908780715835392 |
|---|---|
| author | Seddik, Fahd Elbedewy, Abdulrahman Sami, Gaser Abdelmoniem, Mohamed Zakaria, Yahia |
| author_facet | Seddik, Fahd Elbedewy, Abdulrahman Sami, Gaser Abdelmoniem, Mohamed Zakaria, Yahia |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_15473 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Panther: Faster and Cheaper Computations with Randomized Numerical Linear Algebra Seddik, Fahd Elbedewy, Abdulrahman Sami, Gaser Abdelmoniem, Mohamed Zakaria, Yahia Machine Learning Artificial Intelligence 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. |
| title | Panther: Faster and Cheaper Computations with Randomized Numerical Linear Algebra |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2601.15473 |