6 papers
SPRI: SVD-Partitioned Residual Initialization for Data-Constrained MoE Upcycling
Weiqiao Shan, Ruixiang Mao, Yuang Li +10
Mixture-of-Experts (MoE) models enable efficient scaling, but training them from scratch remains prohibitively expensive. MoE upcycling mitigates this cost by converting pretrained…
Enhancing Speech Large Language Models with Prompt-Aware Mixture of Audio Encoders
Weiqiao Shan, Yuang Li, Yuhao Zhang +9
Connecting audio encoders with large language models (LLMs) allows the LLM to perform various audio understanding tasks, such as automatic speech recognition (ASR) and audio captio…
Leveraging Unit Language Guidance to Advance Speech Modeling in Textless Speech-to-Speech Translation
Yuhao Zhang, Xiangnan Ma, Kaiqi Kou +7
The success of building textless speech-to-speech translation (S2ST) models has attracted much attention. However, S2ST still faces two main challenges: 1) extracting linguistic fe…
Optimizing Speech Multi-View Feature Fusion through Conditional Computation
Weiqiao Shan, Yuhao Zhang, Yuchen Han +7
Recent advancements have highlighted the efficacy of self-supervised learning (SSL) features in various speech-related tasks, providing lightweight and versatile multi-view speech…
Early Exit Is a Natural Capability in Transformer-based Models: An Empirical Study on Early Exit without Joint Optimization
Weiqiao Shan, Long Meng, Tong Zheng +5
Large language models (LLMs) exhibit exceptional performance across various downstream tasks. However, they encounter limitations due to slow inference speeds stemming from their e…
PartialFormer: Modeling Part Instead of Whole for Machine Translation
Tong Zheng, Bei Li, Huiwen Bao +4
The design choices in Transformer feed-forward neural networks have resulted in significant computational and parameter overhead. In this work, we emphasize the importance of hidde…