ZeroQuant-HERO: Hardware-Enhanced Robust Optimized Post-Training Quantization Framework for W8A8 Transformers

TL;DR

ZeroQuant-HERO introduces a hardware-aware post-training quantization framework for W8A8 transformers, optimizing memory and compute efficiency while allowing mode flexibility to improve accuracy.

Contribution

It is the first to integrate hardware considerations into robust post-training quantization for transformers, addressing memory and operator complexities.

Findings

Enhanced hardware performance in W8A8 quantized transformers

Flexible mode switching improves quantization accuracy

Addresses memory-bounded operators and per-token quantization challenges

Abstract

Quantization techniques are pivotal in reducing the memory and computational demands of deep neural network inference. Existing solutions, such as ZeroQuant, offer dynamic quantization for models like BERT and GPT but overlook crucial memory-bounded operators and the complexities of per-token quantization. Addressing these gaps, we present a novel, fully hardware-enhanced robust optimized post-training W8A8 quantization framework, ZeroQuant-HERO. This framework uniquely integrates both memory bandwidth and compute-intensive operators, aiming for optimal hardware performance. Additionally, it offers flexibility by allowing specific INT8 modules to switch to FP16/BF16 mode, enhancing accuracy.