GQSA: Group Quantization and Sparsity for Accelerating Large Language Model Inference
This work addresses the need for efficient model compression in LLMs, offering a novel co-design approach that is incremental but provides specific gains over existing methods.
The paper tackles the problem of accelerating large language model inference by proposing GQSA, a compression technique that integrates quantization and sparsification, resulting in improved accuracy and speed, such as outperforming W2 quantization by 1.26× and 2:4 pruning by 2.35× in inference speed.
Model compression has emerged as a mainstream solution to reduce memory usage and computational overhead. This paper presents Group Quantization and Sparse Acceleration (GQSA), a novel compression technique tailored for LLMs. Traditional methods typically focus exclusively on either quantization or sparsification, but relying on a single strategy often results in significant performance loss at high compression rates. In contrast, GQSA integrates quantization and sparsification in a tightly coupled manner, leveraging GPU-friendly structured group sparsity and quantization for efficient acceleration. Building upon system-algorithm co-design principles, we propose a two-stage sparse optimization strategy that ensures the performance superiority of the compressed model. On the engine side, we introduce a "task-centric" parallel strategy, which, to the best of our knowledge, is the first application in the domain of sparse computing. Compared to the traditional 2:4 sparse method, the GQSA offers a more flexible and adjustable sparsity rate, as well as a higher weight compression rate, and is efficiently compatible with weight-only quantization methods. Experimental results demonstrate that, under the GQSA W4S50% compression setting, the model's accuracy surpasses that of both 2:4 pruning and W2 quantization. Furthermore, at the inference level, GQSA outperforms W2 by 1.26$\times$ and 2:4 pruning by 2.35$\times$ in terms of speed.