QuantNAS for Super Resolution: Searching for Efficient Quantization-Friendly Architectures Against Quantization Noise
Open Access
- 19 August 2024
- journal article
- research article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 117008-117025
Abstract
This work aims to develop an automated procedure for discovering new, efficient solutions that can be effectively quantized in mixed-precision mode with minimal degradation. While our primary focus is on Super-Resolution (SR), our proposed procedure is applicable beyond this domain. To achieve our goals, we first develop an efficient Neural Architecture Search (NAS) procedure for full-precision a models, surpassing existing NAS solutions for SR.We then adapt this procedure for quantization-aware search. By introducing Quantization Noise (QN) during the search phase, we approximate the model degradation after quantization. Additionally, we improve search performance by implementing entropy regularization, which prioritizes operations and its precision within each search space block. Our experiments confirm the superiority of quantization-aware NAS compared to the two-step process: NAS followed by quantization. Furthermore, approximating quantization with QN offers a 30% speed improvement over direct weight quantization. We validate our approach by developing and applying it to two search spaces inspired by state-of-the-art SR models. Our code is publicly available b .Keywords
All Related Versions
Funding Information
- Analytical Center under the RF Government (70-2021-00145 02.11.2021)
This publication has 32 references indexed in Scilit:
- Neural Architecture Search Benchmarks: Insights and SurveyIEEE Access, 2023
- Discretization-aware architecture searchPattern Recognition, 2021
- Rethinking Differentiable Search for Mixed-Precision Neural Networks2020
- Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition2020
- FBNet: Hardware-Aware Efficient ConvNet Design via Differentiable Neural Architecture Search2019
- Fast and Accurate Single Image Super-Resolution via Information Distillation Network2018
- Quantization and Training of Neural Networks for Efficient Integer-Arithmetic-Only Inference2018
- Deep Learning with Low Precision by Half-Wave Gaussian Quantization2017
- Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network2017
- Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network2016