CHESS: Optimizing LLM Inference via Channel-Wise Thresholding and Selective Sparsification

TL;DR

CHESS introduces a novel channel-wise thresholding and selective sparsification method to reduce activation during LLM inference, significantly accelerating performance with minimal accuracy loss on multiple tasks.

Contribution

This work presents a new activation sparsification framework that models performance impact and applies channel-wise thresholds and selective layer sparsification for efficient LLM inference.

Findings

Achieves up to 1.27x speedup in LLM inference.

Reduces performance degradation across eight downstream tasks.

Activates fewer parameters than existing sparsification methods.

Abstract

Deploying large language models (LLMs) on edge devices presents significant challenges due to the substantial computational overhead and memory requirements. Activation sparsification can mitigate these resource challenges by reducing the number of activated neurons during inference. Existing methods typically employ thresholding-based sparsification based on the statistics of activation tensors. However, they do not model the impact of activation sparsification on performance, resulting in suboptimal performance degradation. To address the limitations, this paper reformulates the activation sparsification problem to explicitly capture the relationship between activation sparsity and model performance. Then, this paper proposes CHESS, a general activation sparsification approach via CHannel-wise thrEsholding and Selective Sparsification. First, channel-wise thresholding assigns a unique threshold to each activation channel in the feed-forward network (FFN) layers. Then, selective sparsification involves applying thresholding-based activation sparsification to specific layers within the attention modules. Finally, we detail the implementation of sparse kernels to accelerate LLM inference. Experimental results demonstrate that the proposed CHESS achieves lower performance degradation over eight downstream tasks while activating fewer parameters than existing methods, thus speeding up the LLM inference by up to 1.27x.