Paper

Hymba: A Hybrid-head Architecture for Small Language Models

Matthijs Van Keirsbilck, Ameya Sunil Mahabaleshwarkar, Byeon, Wonmin, Min-Hung, Chen, Kautz, Jan, Molchanov, Pavlo, Diao, Shizhe, Lin, Yingyan, Shih-Yang, Liu, Chen, Zijia, Fu, Yonggan, Dong, Xin, Suhara, Yoshi

2024

Overview

We propose Hymba, a family of small language models featuring a hybrid-head parallel architecture that integrates transformer attention mechanisms with state space models (SSMs) for enhanced efficiency. Attention heads provide high-resolution recall, while SSM heads enable efficient context summarization. Additionally, we introduce learnable meta tokens that are prepended to prompts, storing critical information and alleviating the \"forced-to-attend\" burden associated with attention mechanisms. This model is further optimized by incorporating cross-layer key-value (KV) sharing and partial sliding window attention, resulting in a compact cache size. During development, we conducted a controlled study comparing various architectures under identical settings and observed significant advantages of our proposed architecture. Notably, Hymba achieves state-of-the-art results for small LMs: Our Hymba-1.5B-Base model surpasses all sub-2B public models in performance and even outperforms Llama-3.2-3B with 1.32% higher average accuracy, an 11.67x cache size reduction, and 3.49x throughput.

• Publisher: Cornell University Library, arXiv.org
• Subject: State space models