SDQ: Sparse Decomposed Quantization for LLM Inference
This addresses the problem of enabling wider adaptation of LLMs by reducing their resource demands, representing an incremental improvement over existing compression methods.
The paper tackles the high compute and memory requirements of large language models (LLMs) by proposing SDQ, a method combining structured sparsity and quantization, achieving 4x effective compute throughput with less than 1% quality drop.
Recently, large language models (LLMs) have shown surprising performance in task-specific workloads as well as general tasks with the given prompts. However, to achieve unprecedented performance, recent LLMs use billions to trillions of parameters, which hinder the wide adaptation of those models due to their extremely large compute and memory requirements. To resolve the issue, various model compression methods are being actively investigated. In this work, we propose SDQ (Sparse Decomposed Quantization) to exploit both structured sparsity and quantization to achieve both high compute and memory efficiency. From our evaluations, we observe that SDQ can achieve 4x effective compute throughput with <1% quality drop.