Hymba: A Hybrid-head Architecture for Small Language Models
원문 초록
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.
함께 살펴볼 논문
공통 태그를 근거로 연결했습니다. 실제 인용 관계나 논문의 우수성을 뜻하지 않습니다.
Optimal Sensor Scheduling and Selection for Continuous-Discrete Kalman Filtering with Auxiliary Dynamics
Precision Spatio-Temporal Feature Fusion for Robust Remote Sensing Change Detection
Bridging Robustness and Generalization Against Word Substitution Attacks in NLP via the Growth Bound Matrix Approach
MambaFusion: Height-Fidelity Dense Global Fusion for Multi-modal 3D Object Detection
Spatio-Temporal State Space Model For Efficient Event-Based Optical Flow
Structured State Space Model Dynamics and Parametrization for Spiking Neural Networks
토론
이 논문과 연결된 포럼 글이 아직 없습니다.