English

Masked Gated Linear Unit

Machine Learning 2025-07-01 v1 Computation and Language

Abstract

Gated Linear Units (GLUs) have become essential components in the feed-forward networks of state-of-the-art Large Language Models (LLMs). However, they require twice as many memory reads compared to feed-forward layers without gating, due to the use of separate weight matrices for the gate and value streams. To address this bottleneck, we introduce Masked Gated Linear Units (MGLUs), a novel family of GLUs with an efficient kernel implementation. The core contribution of MGLUs include: (1) the Mixture of Element-wise Gating (MoEG) architecture that learns multiple binary masks, each determining gate or value assignments at the element level on a single shared weight matrix resulting in reduced memory transfer, and (2) FlashMGLU, a hardware-friendly kernel that yields up to a 19.7 ×\times inference-time speed-up over a naive PyTorch MGLU and is 47% more memory-efficient and 34% faster than standard GLUs despite added architectural complexity on an RTX5090 GPU. In LLM experiments, the Swish-activated variant SwiMGLU preserves its memory advantages while matching - or even surpassing - the downstream accuracy of the SwiGLU baseline.

Keywords

Cite

@article{arxiv.2506.23225,
  title  = {Masked Gated Linear Unit},
  author = {Yukito Tajima and Nakamasa Inoue and Yusuke Sekikawa and Ikuro Sato and Rio Yokota},
  journal= {arXiv preprint arXiv:2506.23225},
  year   = {2025}
}
R2 v1 2026-07-01T03:38:27.806Z