DQ-Whisper：面向高效多语言语音识别的联合蒸馏与量化方法
DQ-Whisper: Joint Distillation and Quantization for Efficient Multilingual Speech Recognition
AI总结：
本文提出 DQ-Whisper 框架，通过动态匹配蒸馏与量化感知蒸馏联合压缩 Whisper，在几乎不损失性能的前提下将模型缩小最多 5.18 倍，有效提升小型模型的多语言语音识别能力。
摘要：
作为一种流行的多语言多任务预训练语音模型，Whisper 存在“多语言诅咒”问题。为增强小型 Whisper 模型的多语言能力，我们提出 DQ-Whisper，一种新颖的联合蒸馏与量化框架，用于压缩 Whisper 以实现高效推理。首先，我们提出一种新颖的动态匹配蒸馏策略。随后，引入量化感知蒸馏框架，将量化与蒸馏相结合。在多个多语言数据集上的实验结果表明，我们提出的蒸馏方法能够在不增加计算成本的情况下有效增强小型 Whisper 模型的多语言能力。模型规模最多可减少 5.18 倍，且性能下降极小。此外，量化与蒸馏兼容，可实现更高的压缩率。
英文摘要：
As a popular multilingual and multitask pre-trained speech model, Whisper has the problem of curse of multilinguality. To enhance multilingual capabilities in small Whisper models, we propose DQ-Whisper, a novel joint distillation and quantization framework to compress Whisper for efficient inference. Firstly, we propose a novel dynamic matching distillation strategy. Then, a quantization-aware distillation framework is introduced to integrate quantization with distillation. Experimental results on various multilingual datasets show that our suggested distillation approach can effectively enhance the multilingual capabilities of small Whisper models without increasing computational costs. Up to 5.18x reduction in model size is achieved with marginal performance degradation. In addition, quantization is compatible with distillation, which can result in a higher compression rate.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
