Edge AI

Nabla: Adaptive Multi-Domain DVFS for Energy-Efficient LLM Inference on Edge SoCs
Nabla is a residual multi-domain DVFS framework for energy-efficient edge LLM inference. It uses an interference-free execution …
Yaotian Hu, Chuang Hu, Zheng Gong, Qianlong Sang, Zheng Zhang, Xiaobo Zhou, Dazhao Cheng
QDiTx: Arbitrary Bit-Width Quantization Framework for Diffusion Transformer
QDiTx is an arbitrary-bit-width weight-quantization framework for Diffusion Transformers. It addresses accumulated quantization error …
Qianlong Sang, Puyi He, Tianyu Tu, Huanghuang Liang, Yaqi Xia, Yili Gong, Chuang Hu, Xiaobo Zhou, Dazhao Cheng