paper-with-me

Papers

WBMM: Windowed Batch Matrix Multiplication for Efficient Large Receptive Field Convolution

2026-07-02 · Wan Song, Wei Zhou, Rui Wang, Jun Yu, Toru Kurihara, Jiajia Xu, Shu Zhan arxiv

Large kernel depthwise convolutions achieve strong performance but suffer from significant degradation as kernel size grows due to irregular memory access from gather-based computation; while Large Kernel Acceleration (LKA) helps on small feature maps, it becomes counterproductive on large feature maps, even slower than non-accelerated implementations. We propose Windowed Batch Matrix Multiplication (WBMM), which partitions input into contiguous windows and indexes a compact relative position bias table to construct weight matrices, enabling regular memory access via batched matrix multiplication. This yields a unique property: WBMM's throughput improves with larger windows, opposite to depthwise convolutions that degrade with larger kernels. Operator-level benchmarks show WBMM with 14x14 windows outperforms 5x5 depthwise convolution baselines in speed while providing a 7.8x larger per-layer receptive field. Combined with inter-block cross-window communication and hierarchical window reparameterization, WBMM achieves comparable or higher accuracy on ImageNet-1K, COCO, and ADE20K with 1.31-1.88x training speedup, and demonstrates consistent advantages across GPU, CPU, and edge devices without requiring specialized acceleration kernels. Our code is available at http://github.com/wansong-s/WBMM

📄 PDF Abstract BibTeX arXiv:2607.02097

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

Memory-Efficient Backpropagation through Large Linear Layers

2022-01-31 · Daniel Bershatsky, Aleksandr Mikhalev, Alexandr Katrutsa, Julia Gusak 외

In modern neural networks like Transformers, linear layers require significant memory to store activations during backward pass. This study proposes a memory reduction approach to perform backpropagation through linear l…

Model Compression

Efficient batchwise dropout training using submatrices

2015-02-09 · Ben Graham, Jeremy Reizenstein, Leigh Robinson

Dropout is a popular technique for regularizing artificial neural networks. Dropout networks are generally trained by minibatch gradient descent with a dropout mask turning off some of the units---a different pattern of …

GFormer: Accelerating Large Language Models with Optimized Transformers on Gaudi Processors

2024-12-19 · Chengming Zhang, Xinheng Ding, Baixi Sun, Xiaodong Yu 외

Heterogeneous hardware like Gaudi processor has been developed to enhance computations, especially matrix operations for Transformer-based large language models (LLMs) for generative AI tasks. However, our analysis indic…

MME

Trace norm regularization and faster inference for embedded speech recognition RNNs

2017-10-25 · ICLR 2018 1 · Markus Kliegl, Siddharth Goyal, Kexin Zhao, Kavya Srinet 외

We propose and evaluate new techniques for compressing and speeding up dense matrix multiplications as found in the fully connected and recurrent layers of neural networks for embedded large vocabulary continuous speech …

speech-recognitionSpeech Recognition

A Unified Coded Deep Neural Network Training Strategy Based on Generalized PolyDot Codes for Matrix Multiplication

2018-11-27 · Sanghamitra Dutta, Ziqian Bai, Haewon Jeong, Tze Meng Low 외

This paper has two contributions. First, we propose a novel coded matrix multiplication technique called Generalized PolyDot codes that advances on existing methods for coded matrix multiplication under storage and commu…