Výsledky vyhledávání - "CAI, Changpeng"

Report

Faster Multi-GPU Training with PPLL: A Pipeline Parallelism Framework Leveraging Local Learning

Autor: Guo, Xiuyuan, Xu, Chengqi, Guo, Guinan, Zhu, Feiyu, Cai, Changpeng, Wang, Peizhe, Wei, Xiaoming, Su, Junhao, Gao, Jialin

Currently, training large-scale deep learning models is typically achieved through parallel training across multiple GPUs. However, due to the inherent communication overhead and synchronization delays in traditional model parallelism methods, seamle

Externí odkaz: http://arxiv.org/abs/2411.12780

Zobrazit plný text záznamu

Report

Momentum Auxiliary Network for Supervised Local Learning

Autor: Su, Junhao, Cai, Changpeng, Zhu, Feiyu, He, Chenghao, Xu, Xiaojie, Guan, Dongzhi, Si, Chenyang

Deep neural networks conventionally employ end-to-end backpropagation for their training process, which lacks biological credibility and triggers a locking dilemma during network parameter updates, leading to significant GPU memory use. Supervised lo

Externí odkaz: http://arxiv.org/abs/2407.05623

Zobrazit plný text záznamu

Report

MLAAN: Scaling Supervised Local Learning with Multilaminar Leap Augmented Auxiliary Network

Autor: Zhang, Yuming, Zhang, Shouxin, Wang, Peizhe, Zhu, Feiyu, Guan, Dongzhi, Su, Junhao, Liu, Jiabin, Cai, Changpeng

Deep neural networks (DNNs) typically employ an end-to-end (E2E) training paradigm which presents several challenges, including high GPU memory consumption, inefficiency, and difficulties in model parallelization during training. Recent research has

Externí odkaz: http://arxiv.org/abs/2406.16633

Zobrazit plný text záznamu

Report

Advancing Supervised Local Learning Beyond Classification with Long-term Feature Bank

Autor: Zhu, Feiyu, Zhang, Yuming, Cai, Changpeng, He, Chenghao, Guo, Xiuyuan, Li, Jiao, Wang, Peizhe, Su, Junhao, Gao, Jialin

Local learning offers an alternative to traditional end-to-end back-propagation in deep neural networks, significantly reducing GPU memory usage. While local learning has shown promise in image classification tasks, its application to other visual ta

Externí odkaz: http://arxiv.org/abs/2406.00446

Zobrazit plný text záznamu

Report

SPEAK: Speech-Driven Pose and Emotion-Adjustable Talking Head Generation

Autor: Cai, Changpeng, Guo, Guinan, Li, Jiao, Su, Junhao, Shen, Fei, He, Chenghao, Xiao, Jing, Chen, Yuanxu, Dai, Lei, Zhu, Feiyu

Most earlier researches on talking face generation have focused on the synchronization of lip motion and speech content. However, head pose and facial emotions are equally important characteristics of natural faces. While audio-driven talking face ge

Externí odkaz: http://arxiv.org/abs/2405.07257

Zobrazit plný text záznamu