Songwen Pei

LOAD: Enabling Low-Bit Group Quantization for Vision Transformers via Adaptive Data Formats

Vision Transformers (ViTs) achieve state-of-the-art performance across various vision tasks but pose significant deployment challenges due to their high computational complexity and memory footprint. Post-training quantization (PTQ) offers an efficient, retraining-free compressio...

LSR: Eliminating Loop Control Overhead for Energy-Efficient Processors

Loops constitute a small fraction of static code but dominate dynamic execution time and control overhead, posing significant challenges for energy-efficient mobile and embedded processors. This paper introduces Loop Semantic Recognition (LSR) , a microarchitectural technique tha...