Showing cs.LGShow all
2 papers · 1 filter
cs.LG2025
PrefixQuant: Eliminating Outliers by Prefixed Tokens for Large Language Models Quantization
Mengzhao Chen, Yi Liu, Jiahao Wang +3
Existing weight-activation quantization methods for Large Language Models (LLMs) primarily address channel-wise outliers but often neglect token-wise outliers, which limits the acc…
cs.LG2024
DCP: Learning Accelerator Dataflow for Neural Network via Propagation
Peng Xu, Wenqi Shao, Mingyu Ding +1
Deep neural network (DNN) hardware (HW) accelerators have achieved great success in improving DNNs' performance and efficiency. One key reason is dataflow in executing a DNN layer,…