Mix and Match: A Novel FPGA-Centric Deep Neural Network Quantization Framework
Open accessThe DOI matches its Crossref record; 29 of the 29 references it cites were found in a registry or an index.
Abstract
Deep Neural Networks (DNNs) have achieved extraordinary performance in various application domains. To support diverse DNN models, efficient implementations of DNN inference on edge-computing platforms, e.g., ASICs, FPGAs, and embedded systems, are extensively investigated. Due to the huge model size and computation amount, model compression is a critical step to deploy DNN models on edge devices. This paper focuses on weight quantization, a hardware-friendly model compression approach that is complementary to weight pruning.Unlike existing methods that use the same quantization scheme for all weights, we propose the first solution that applies different quantization schemes for different rows of the weight matrix. It is motivated by (1) the distribution of the weights in the different rows are not the same; and (2) the potential of achieving better utilization of heterogeneous FPGA hardware resources. To achieve that, we first propose a hardware-friendly quantization scheme named sum-of-power-of-2 (SP2) suitable for Gaussian-like weight distribution, in which the multiplication arithmetic can be replaced with logic shifter and adder, thereby enabling highly efficient implementations with the FPGA LUT resources. In contrast, the existing fixed-point quantization is suitable for Uniform-like weight distribution and can be implemented efficiently by DSP. Then to fully explore the resources, we propose an FPGA-centric mixed scheme quantization (MSQ) with an ensemble of the proposed SP2 and the fixed-point schemes. Combining the two schemes can maintain, or even increase accuracy due to better matching with weight distributions.For the FPGA implementations, we develop a parameterized architecture with heterogeneous Generalized Matrix Multiplication (GEMM) cores-one using LUTs for computations with SP2 quantized weights and the other utilizing DSPs for fixed-point quantized weights. Given the partition ratio among the two schemes based on resource characterization, MSQ quantization training algorithm derives an optimally quantized model for the FPGA implementation. We evaluate our FPGA-centric quantization framework across multiple application domains. With optimal SP2/fixed-point ratios on two FPGA devices, i.e., Zynq XC7Z020 and XC7Z045, we achieve performance improvement of 2.1 × -4.1 × compared to solely exploiting DSPs for all multiplication operations. In addition, the CNN implementations with the proposed MSQ scheme can achieve higher accuracy and comparable hardware utilization efficiency compared to the state-of-the-art designs.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Deep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.90 · Cited by 235,474
ImageNet classification with deep convolutional neural networks
Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017) · Communications of the ACM · doi:10.1145/3065386 · Cited by 112,155
Hochreiter, S., & Schmidhuber, J. (1997) · Neural Computation · doi:10.1162/neco.1997.9.8.1735 · Cited by 102,318
Feature Pyramid Networks for Object Detection
Lin, T.-Y., Dollar, P., Girshick, R., He, K., Hariharan, B., & Belongie, S. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.106 · Cited by 30,939
MobileNetV2: Inverted Residuals and Linear Bottlenecks
Sandler, M., Howard, A., Zhu, M., Zhmoginov, A., & Chen, L.-C. (2018) · 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2018.00474 · Cited by 28,755
A survey on deep learning in medical image analysis
Litjens, G., Kooi, T., Bejnordi, B. E., Setio, A. A. A., Ciompi, F., Ghafoorian, M., van der Laak, J. A. W. M., van Ginneken, B., & Sánchez, C. I. (2017) · Medical Image Analysis · doi:10.1016/j.media.2017.07.005 · Cited by 15,332
Hinton, G., Deng, L., Yu, D., Dahl, G., Mohamed, A.-R., Jaitly, N., Senior, A., Vanhoucke, V., Nguyen, P., Sainath, T., et al. (2012) · IEEE Signal Processing Magazine · doi:10.1109/msp.2012.2205597 · Cited by 10,260
Building a Large Annotated Corpus of English: The Penn Treebank
Marcus, M. (1993) · doi:10.21236/ada273556 · Cited by 7,405
A unified architecture for natural language processing: deep neural networks with multitask learning
Collobert, R., & Weston, J. (2008) · Proceedings of the 25th international conference on Machine learning - ICML '08 · doi:10.1145/1390156.1390177 · Cited by 5,058
XNOR-Net: ImageNet Classification Using Binary Convolutional Neural Networks
Rastegari, M., Ordonez, V., Redmon, J., & Farhadi, A. (2016) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-46493-0_32 · Cited by 4,408
Clinically applicable deep learning for diagnosis and referral in retinal disease
De Fauw, J., Ledsam, J. R., Romera-Paredes, B., Nikolov, S., Tomasev, N., Blackwell, S., Askham, H., Glorot, X., O’Donoghue, B., Visentin, D., et al. (2018) · Nature Medicine · doi:10.1038/s41591-018-0107-6 · Cited by 2,639
Microsoft COCO: Common Objects in Context
Lin, T.-Y., Maire, M., Belongie, S., Bourdev, L., Girshick, R., Hays, J., Perona, P., Ramanan, D., Zitnick, C. L., & Dollár, P. (2014) · arXiv · doi:10.48550/arxiv.1405.0312 · Cited by 2,579
BinaryConnect: Training Deep Neural Networks with binary weights during propagations
Courbariaux, M., Bengio, Y., & David, J.-P. (2015) · arXiv · doi:10.48550/arxiv.1511.00363 · Cited by 2,180
Optimizing FPGA-based Accelerator Design for Deep Convolutional Neural Networks
Zhang, C., Li, P., Sun, G., Guan, Y., Xiao, B., & Cong, J. (2015) · Proceedings of the 2015 ACM/SIGDA International Symposium on Field-Programmable Gate Arrays · doi:10.1145/2684746.2689060 · Cited by 2,158 · cited 2×
“Extensive research efforts have been dedicated to the efficient implementations of DNN inference models on various edgecomputing platforms, such as ASICs [8]–[14], FPGAs [15]– [18], and embedded CPUs/GPUs [19]–[23].”
BackgroundCourbariaux, M., Hubara, I., Soudry, D., El-Yaniv, R., & Bengio, Y. (2016) · arXiv · doi:10.48550/arxiv.1602.02830 · Cited by 2,106
Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computation
Bengio, Y., Léonard, N., & Courville, A. (2013) · arXiv · doi:10.48550/arxiv.1308.3432 · Cited by 1,910 · cited 6×
“In this work, we use a combination of ADMM and STE, as shown in Algorithm 1, which in general follows the ADMM algorithm for weight quantization and where the STE is only applied for activation quantization.”
MethodInfluentialDoReFa-Net: Training Low Bitwidth Convolutional Neural Networks with Low Bitwidth Gradients
Zhou, S., Wu, Y., Ni, Z., Zhou, X., Wen, H., & Zou, Y. (2016) · arXiv · doi:10.48550/arxiv.1606.06160 · Cited by 1,690 · cited 7×
“[38] first explored the potential of fixed-point quantization by introducing hyperbolic tangent transformation to weights and activations, with scaling factors to minimize quantization error.”
BackgroundMethodInfluentialRethinking the Value of Network Pruning
Liu, Z., Sun, M., Zhou, T., Huang, G., & Darrell, T. (2018) · arXiv · doi:10.48550/arxiv.1810.05270 · Cited by 1,625
Self-citationWeight Uncertainty in Neural Networks
Blundell, C., Cornebise, J., Kavukcuoglu, K., & Wierstra, D. (2015) · arXiv · doi:10.48550/arxiv.1505.05424 · Cited by 1,342 · cited 1×
“Power-of-2 quantization replaces multiplications by bit shifting operations and this number system also possesses higher precision around the mean, which fits the Gaussian distribution of DNN weights better [48], [49].”
BackgroundTVM: an automated end-to-end optimizing compiler for deep learning
(2018) · arXiv (Cornell University) · doi:10.5555/3291168.3291211 · Cited by 1,170
Zhu, C., Han, S., Mao, H., & Dally, W. J. (2016) · arXiv · doi:10.48550/arxiv.1612.01064 · Cited by 865
Li, F., Liu, B., Wang, X., Zhang, B., & Yan, J. (2016) · arXiv · doi:10.48550/arxiv.1605.04711 · Cited by 732
PACT: Parameterized Clipping Activation for Quantized Neural Networks
Choi, J., Wang, Z., Venkataramani, S., Chuang, P. I.-J., Srinivasan, V., & Gopalakrishnan, K. (2018) · arXiv · doi:10.48550/arxiv.1805.06085 · Cited by 703 · cited 6×
“Table III shows that Dorefa, PACT, DSQ, µL2Q, and LQNETS have up to 3.84% accuracy degradation, only QIL reports lossless accuracy performance.”
MethodInfluentialBit Fusion: Bit-Level Dynamically Composable Architecture for Accelerating Deep Neural Network
Sharma, H., Park, J., Suda, N., Lai, L., Chau, B., Kim, J. K., Chandra, V., & Esmaeilzadeh, H. (2018) · 2018 ACM/IEEE 45th Annual International Symposium on Computer Architecture (ISCA) · doi:10.1109/isca.2018.00069 · Cited by 543
The Microsoft 2017 Conversational Speech Recognition System
Xiong, W., Wu, L., Alleva, F., Droppo, J., Huang, X., & Stolcke, A. (2018) · 2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) · doi:10.1109/icassp.2018.8461870 · Cited by 472
Towards Accurate Binary Convolutional Neural Network
Lin, X., Zhao, C., & Pan, W. (2017) · arXiv · doi:10.48550/arxiv.1711.11294 · Cited by 453 · cited 2×
“There are different types of quantization schemes including binary [31]–[34], ternary [35]–[37], low-bit-width fixed-point [38]–[43], and power-of-2 [44]–[47].”
BackgroundMethodSelf-citationLearned Step Size Quantization
Esser, S. K., McKinstry, J. L., Bablani, D., Appuswamy, R., & Modha, D. S. (2019) · arXiv · doi:10.48550/arxiv.1902.08153 · Cited by 415
Learning to Quantize Deep Networks by Optimizing Quantization Intervals With Task Loss
Jung, S., Son, C., Lee, S., Son, J., Han, J.-J., Kwak, Y., Hwang, S. J., & Choi, C. (2019) · 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2019.00448 · Cited by 385 · cited 6×
“QIL [41] parameterized the quantization interval and trained it with task loss, avoiding access to the original training data.”
MethodInfluentialSlimYOLOv3: Narrower, Faster and Better for Real-Time UAV Applications
Zhang, P., Zhong, Y., & Li, X. (2019) · 2019 IEEE/CVF International Conference on Computer Vision Workshop (ICCVW) · doi:10.1109/iccvw.2019.00011 · Cited by 266 · cited 1×
“To provide an idea about the mAP degradation by our MSQ, we can compare with the weight pruning method on YOLO with also 8× compression rate [64], which decreases the mAP@0.”
Method
Cited by
Distributed Artificial Intelligence Empowered by End-Edge-Cloud Computing: A Survey
Duan, S., Wang, D., Ren, J., Lyu, F., Zhang, Y., Wu, H., & Shen, X. (2023) · IEEE Communications Surveys & Tutorials · doi:10.1109/comst.2022.3218527 · Cited by 408
SPViT: Enabling Faster Vision Transformers via Latency-Aware Soft Token Pruning
Kong, Z., Dong, P., Ma, X., Meng, X., Niu, W., Sun, M., Shen, X., Yuan, G., Ren, B., Tang, H., et al. (2022) · Lecture Notes in Computer Science · doi:10.1007/978-3-031-20083-0_37 · Cited by 169
“Pruning has been proved as the one of the most effective methods to reduce network dimensions in convolution-based neural networks [70,101,57,47,107,62,72,52,108,11,36,55,54,4].”
MethodSelf-citationLi, Z., Sun, M., Lu, A., Ma, H., Yuan, G., Xie, Y., Tang, H., Li, Y., Leeser, M., Wang, Z., et al. (2022) · 2022 32nd International Conference on Field-Programmable Logic and Applications (FPL) · doi:10.1109/fpl57034.2022.00027 · Cited by 85
“However, as pointed out in [11], this type of quantization is incompatible with layer-by-layer inference execution on hardware accelerators since it introduces nonuniformality among layers.”
BackgroundMethodInfluentialSelf-citationOnline continual streaming learning for embedded space applications
Mazouz, A. E., & Nguyen, V.-T. (2024) · Journal of Real-Time Image Processing · doi:10.1007/s11554-024-01438-4 · Cited by 3
Metadata
- Authors
- Chang, S.-E., Li, Y., Sun, M., Shi, R., So, H. K.-H., Qian, X., Wang, Y., & Lin, X.
- Venue
- 2021 IEEE International Symposium on High-Performance Computer Architecture (HPCA)
- Publisher
- IEEE
- Published
- 2021-02
- Pages
- 208-220
- ISSN
- 1530-0897 (ISSN-L)
- OpenAlex
- W3158020960
- Cited by
- 126
- References
- 29
- FWCI
- 7.62
- Type
- conference paper
- Language
- en
- Open access
- green
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.