ニュース&特集
Advances to low-bit quantization enable LLMs on edge devices
| Shijie Cao, Lingxiao Ma, と Ting Cao
Advances in low-bit quantization techniques enable efficient operation of LLMs on resource-constrained edge devices. Discover how innovations like T-MAC, Ladder, and LUT Tensor Core improve computational efficiency and enhance hardware compatibility.
Unified Database: Laying the foundation for large language model vertical applications
Unified databases offer better knowledge transfer between multimodal data types. They provide substantial corpus support for large language models and are poised to drive innovation in underlying hardware, laying the foundation for data-enhanced AI.
Research Focus: Week of November 22, 2023
A new deep-learning compiler for dynamic sparsity; Tongue Tap could make tongue gestures viable for VR/AR headsets; Ranking LLM-Generated Loop Invariants for Program Verification; Assessing the limits of zero-shot foundation models in single-cell biology.
Efficient and hardware-friendly neural architecture search with SpaceEvo
| Li Lyna Zhang, Jiahang Xu, Quanlu Zhang, Yuqing Yang, Ting Cao, と Mao Yang
A persistent challenge in deep learning is optimizing neural network models for diverse hardware configurations, balancing performance and low latency. Learn how SpaceEvo automates hardware-aware neural architecture search to fine-tune DNN models for swift execution on diverse devices.
Building a “heavy metal quartet” of AI compilers
A new quartet of AI compilers: Rammer, Roller, Welder, and Grinder, tackle a range of compiler optimization challenges based on the same tile abstraction, providing a comprehensive solution to connect AI models with hardware accelerators.