COMPACT: Common-token Optimized Model Pruning Across Channels and Tokens

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Kwek, Eugene, Yin, Wenpeng
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911202295152640
author Kwek, Eugene
Yin, Wenpeng
author_facet Kwek, Eugene
Yin, Wenpeng
contents Making large language models (LLMs) more efficient in memory, latency, and serving cost is crucial for edge deployment, interactive applications, and sustainable inference at scale. Pruning is a promising technique, but existing pruning methods are limited: width pruning often breaks the standard transformer layout, requiring custom inference code, while depth pruning can cause abrupt accuracy drops. Also, while many pruning approaches are effective against LLMs, they struggle to maintain performance on small language models (SLMs). In this work, we propose COMPACT, which jointly (i) prunes rare vocabulary to shrink embedding/LM head layers and (ii) prunes FFN intermediate channels using common-token-weighted activations, aligning importance with the post-pruning token distribution. COMPACT inherits strengths of both depth and width pruning, such as: deployment-friendliness (keeps a standard transformer architecture), scale-adaptivity (trade off vocab. vs. FFN pruning), competitive pruning times, and strong memory savings alongside throughput gains. Experiments across Qwen, LLaMA, and Gemma families (0.5B-70B) show state-of-the-art downstream performance, with substantial reductions in parameters, GPU memory, and latency.
format Preprint
id arxiv_https___arxiv_org_abs_2509_06836
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle COMPACT: Common-token Optimized Model Pruning Across Channels and Tokens
Kwek, Eugene
Yin, Wenpeng
Computation and Language
Artificial Intelligence
Machine Learning
Making large language models (LLMs) more efficient in memory, latency, and serving cost is crucial for edge deployment, interactive applications, and sustainable inference at scale. Pruning is a promising technique, but existing pruning methods are limited: width pruning often breaks the standard transformer layout, requiring custom inference code, while depth pruning can cause abrupt accuracy drops. Also, while many pruning approaches are effective against LLMs, they struggle to maintain performance on small language models (SLMs). In this work, we propose COMPACT, which jointly (i) prunes rare vocabulary to shrink embedding/LM head layers and (ii) prunes FFN intermediate channels using common-token-weighted activations, aligning importance with the post-pruning token distribution. COMPACT inherits strengths of both depth and width pruning, such as: deployment-friendliness (keeps a standard transformer architecture), scale-adaptivity (trade off vocab. vs. FFN pruning), competitive pruning times, and strong memory savings alongside throughput gains. Experiments across Qwen, LLaMA, and Gemma families (0.5B-70B) show state-of-the-art downstream performance, with substantial reductions in parameters, GPU memory, and latency.
title COMPACT: Common-token Optimized Model Pruning Across Channels and Tokens
topic Computation and Language
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2509.06836