English

SHARCS: Efficient Transformers through Routing with Dynamic Width Sub-networks

Machine Learning 2023-10-19 v1 Artificial Intelligence Computation and Language

Abstract

We introduce SHARCS for adaptive inference that takes into account the hardness of input samples. SHARCS can train a router on any transformer network, enabling the model to direct different samples to sub-networks with varying widths. Our experiments demonstrate that: (1) SHARCS outperforms or complements existing per-sample adaptive inference methods across various classification tasks in terms of accuracy vs. FLOPs; (2) SHARCS generalizes across different architectures and can be even applied to compressed and efficient transformer encoders to further improve their efficiency; (3) SHARCS can provide a 2 times inference speed up at an insignificant drop in accuracy.

Keywords

Cite

@article{arxiv.2310.12126,
  title  = {SHARCS: Efficient Transformers through Routing with Dynamic Width Sub-networks},
  author = {Mohammadreza Salehi and Sachin Mehta and Aditya Kusupati and Ali Farhadi and Hannaneh Hajishirzi},
  journal= {arXiv preprint arXiv:2310.12126},
  year   = {2023}
}
R2 v1 2026-06-28T12:54:38.778Z