Enhancing Computation Efficiency in Large Language Models through Weight and Activation Quantization
It addresses the deployment challenges of LLMs for practitioners by enhancing computational efficiency, though it is incremental as it builds on existing quantization methods.
This paper tackles the computational inefficiency of deploying large language models by proposing post-training quantization techniques, including W4A8 quantization with methods like AQAS, SLAC, and dINT, achieving task accuracies comparable to full-precision models and a 2× hardware efficiency improvement.
Large Language Models (LLMs) are proficient in natural language processing tasks, but their deployment is often restricted by extensive parameter sizes and computational demands. This paper focuses on post-training quantization (PTQ) in LLMs, specifically 4-bit weight and 8-bit activation (W4A8) quantization, to enhance computational efficiency -- a topic less explored compared to weight-only quantization. We present two innovative techniques: activation-quantization-aware scaling (AQAS) and sequence-length-aware calibration (SLAC) to enhance PTQ by considering the combined effects on weights and activations and aligning calibration sequence lengths to target tasks. Moreover, we introduce dINT, a hybrid data format combining integer and denormal representations, to address the underflow issue in W4A8 quantization, where small values are rounded to zero. Through rigorous evaluations of LLMs, including OPT and LLaMA, we demonstrate that our techniques significantly boost task accuracies to levels comparable with full-precision models. By developing arithmetic units compatible with dINT, we further confirm that our methods yield a 2$\times$ hardware efficiency improvement compared to 8-bit integer MAC unit.