TY - GEN
T1 - Compressing DNN Parameters for Model Loading Time Reduction
AU - Yeh, Yang Ming
AU - Hu, Jennifer Shueh Inn
AU - Lin, Yen Yu
AU - Lu, Yi Chang
N1 - Publisher Copyright:
© 2019 IEEE.
PY - 2019/6
Y1 - 2019/6
N2 - Deep neural network (DNN) has been applied to a variety of computer vision tasks these days. However, DNN often suffers from its enormous execution time even with the aid of GPU. In this paper, we argue that the bandwidth bottleneck between GPU and GDRAM has to be addressed. To reduce loading time, we propose a DNN acceleration approach which compresses DNN parameters before loading model information to GPU and performs decompressing on GPU. Using JPEG compression as an example, the loss of the test accuracy can be kept within 4%, while an 8 × parameter-size reduction is achieved for VGG16.
AB - Deep neural network (DNN) has been applied to a variety of computer vision tasks these days. However, DNN often suffers from its enormous execution time even with the aid of GPU. In this paper, we argue that the bandwidth bottleneck between GPU and GDRAM has to be addressed. To reduce loading time, we propose a DNN acceleration approach which compresses DNN parameters before loading model information to GPU and performs decompressing on GPU. Using JPEG compression as an example, the loss of the test accuracy can be kept within 4%, while an 8 × parameter-size reduction is achieved for VGG16.
KW - DNN
KW - architecture
KW - compression
UR - http://www.scopus.com/inward/record.url?scp=85077999948&partnerID=8YFLogxK
U2 - 10.1109/ICCE-Asia46551.2019.8942192
DO - 10.1109/ICCE-Asia46551.2019.8942192
M3 - Conference contribution
AN - SCOPUS:85077999948
T3 - 2019 IEEE International Conference on Consumer Electronics - Asia, ICCE-Asia 2019
SP - 78
EP - 79
BT - 2019 IEEE International Conference on Consumer Electronics - Asia, ICCE-Asia 2019
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 4th IEEE International Conference on Consumer Electronics - Asia, ICCE-Asia 2019
Y2 - 12 June 2019 through 14 June 2019
ER -