Journal
2021 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, CVPR 2021
Volume -, Issue -, Pages 983-992Publisher
IEEE COMPUTER SOC
DOI: 10.1109/CVPR46437.2021.00104
Keywords
-
Funding
- Qualcomm through a Taiwan University Research Collaboration Project
- Ministry of Science and Technology, Taiwan [108-2221-E-006-227-MY3, 107-2923-E-006-009-MY3, 109-2218-E-002-015]
Ask authors/readers for more resources
The study introduces a novel search strategy called architecture generator to efficiently and flexibly search for sub-networks. With the architecture generator and unified supernet, a flexible and efficient one-shot NAS framework is proposed.
In one-shot NAS, sub-networks need to be searched from the supernet to meet different hardware constraints. However, the search cost is high and N times of searches are needed for N different constraints. In this work, we propose a novel search strategy called architecture generator to search sub-networks by generating them, so that the search process can be much more efficient and flexible. With the trained architecture generator, given target hardware constraints as the input, N good architectures can be generated for N constraints by just one forward pass without re-searching and supernet retraining. Moreover, we propose a novel single-path supernet, called unified supernet, to further improve search efficiency and reduce GPU memory consumption of the architecture generator. With the architecture generator and the unified supernet, we propose a flexible and efficient one-shot NAS framework, called Searching by Generating NAS (SGNAS). With the pre-trained supernt, the search time of SGNAS forN different hardware constraints is only 5 GPU hours, which is 4N times faster than previous SOTA single-path methods. After training from scratch, the top1-accuracy of SGNAS on ImageNet is 77.1%, which is comparable with the SOTAs.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available