Abstract
In this paper, a new low-precision hardware-software codesign framework is presented, to optimally select the numerical formats and bit-precision for TinyML models and benchmarks. The selection is performed by integer linear programming using constraints mandated by tiny edge devices. Practitioners can use the proposed framework to reduce design costs in the early stages of designing accelerators for TinyML models. The efficacy of various numerical formats is studied within a new low-precision framework, ACTION. Results assert that generalized posit and tapered fixed are suitable numerical formats for TinyML when the trade-off between accuracy and hardware complexity is desired.
H.F. Langroudi and V. Karia—Equal contribution.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
Banbury, C., et al.: Mlperf tiny benchmark. In: Proceedings of the Neural Information Processing Systems Track on Datasets and Benchmarks (2021)
Curnick, D.J., et al.: Smallsats: a new technological frontier in ecology and conservation? Remote Sens. Ecol. Conserv. 8(2), 139–150 (2021)
Reddi, V.J., et al.: Widening access to applied machine learning with tinyml. arXiv preprint arXiv:2106.04008 (2021)
Banbury, C.R., et al.: Benchmarking tinyml systems: Challenges and direction. arXiv preprint arXiv:2003.04821 (2020)
Wang, K., Liu, Z., Lin, Y., Lin, J., Han, S.: Haq: hardware-aware automated quantization with mixed precision. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 8612–8620 (2019)
Rusci, M., Fariselli, M., Capotondi, A., Benini, L.: Leveraging automated mixed-low-precision quantization for tiny edge microcontrollers. In: Gama, J., et al. (eds.) ITEM/IoT Streams -2020. CCIS, vol. 1325, pp. 296–308. Springer, Cham (2020). https://doi.org/10.1007/978-3-030-66770-2_22
Langroudi, H.F., Karia, V., Pandit, T., Kudithipudi, D.: Tent: Efficient quantization of neural networks on the tiny edge with tapered fixed point. arXiv preprint arXiv:2104.02233 (2021)
Ghamari, S., et al.: Quantization-guided training for compact tinyml models. arXiv preprint arXiv:2103.06231 (2021)
Fahim, F., et al.: hls4ml: an open-source codesign workflow to empower scientific low-power machine learning devices. arXiv preprint arXiv:2103.05579 (2021)
Ravaglia, L., Rusci, M., Nadalini, D., Capotondi, A., Conti, F., Benini, L.: A tinyml platform for on-device continual learning with quantized latent replays. IEEE J. Emerg. Sel. Top. Circuits Syst. 11(4), 789–802 (2021)
Lin, J., Chen, W.M., Cai, H., Gan, C., Han, S.: Mcunetv2: memory-efficient patch-based inference for tiny deep learning. arXiv e-prints pp. arXiv-2110 (2021)
Langroudi, H.F., Carmichael, Z., Pastuch, D., Kudithipudi, D.: Cheetah: mixed low-precision hardware & software co-design framework for dnns on the edge. arXiv preprint arXiv:1908.02386 (2019)
Hashemi, S., Anthony, N., Tann, H., Bahar, R.I., Reda, S.: Understanding the impact of precision quantization on the accuracy and energy of neural networks. In: Design, Automation & Test in Europe Conference & Exhibition (DATE), pp. 1474–1479. IEEE (2017)
Gustafson, J.L., Yonemoto, I.T.: Beating floating point at its own game: posit arithmetic. Supercomputing Front. Innov. 4(2), 71–86 (2017)
Brumar, I., Zacharopoulos, G., Yao, Y., Rama, S., Wei, G.Y., Brooks, D.: Early DSE and automatic generation of coarse grained merged accelerators. arXiv preprint arXiv:2111.09222 (2021)
Yao, Z., et al.: Hawq-v3: Dyadic neural network quantization. In: International Conference on Machine Learning, pp. 11875–11886. PMLR (2021)
Hubara, I., Nahshan, Y., Hanani, Y., Banner, R., Soudry, D.: Improving post training neural quantization: layer-wise calibration and integer programming. arXiv preprint arXiv:2006.10518 (2020)
Gustafson, L.J.: A generalized framework for matching arithmetic format to application requirements. https://posithub.org/ (2020)
Langroudi, H.F., et al.: Alps: adaptive quantization of deep neural networks with generalized posits. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 3100–3109 (2021)
Gysel, P., Pimentel, J., Motamedi, M., Ghiasi, S.: Ristretto: a framework for empirical study of resource-efficient inference in convolutional neural networks. IEEE Trans. Neural Netw. Learn. Syst. 29(11), 5784–5789 (2018)
Langroudi, H.F., Karia, V., Gustafson, J.L., Kudithipudi, D.: Adaptive posit: parameter aware numerical format for deep learning inference on the edge. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, pp. 726–727 (2020)
Tambe, T., et al.: Algorithm-hardware co-design of adaptive floating-point encodings for resilient deep learning inference. In: 2020 57th ACM/IEEE Design Automation Conference (DAC), pp. 1–6. IEEE (2020)
Samajdar, A., Zhu, Y., Whatmough, P., Mattina, M., Krishna, T.: Scale-sim: Systolic CNN accelerator simulator. arXiv preprint arXiv:1811.02883 (2018)
Abadi, M., et al.: TensorFlow: large-scale machine learning on heterogeneous systems (2015). https://www.tensorflow.org/, software available from tensorflow.org
Li, S., Romaszkan, W., Graening, A., Gupta, P.: Swis-shared weight bit sparsity for efficient neural network acceleration. arXiv preprint arXiv:2103.01308 (2021)
Zhou, C., et al.: Analognets: Ml-hw co-design of noise-robust tinyml models and always-on analog compute-in-memory accelerator. arXiv preprint arXiv:2111.06503 (2021)
Gousev, E.: Recent progress on tinyml technologies and opportunities. https://www.youtube.com/ (2021)
Banbury, C., et al.: Micronets: neural network architectures for deploying tinyml applications on commodity microcontrollers. In: Proceedings of Machine Learning and Systems, vol. 3 (2021)
Acknowledgement
This research was supported by the Matrix AI Consortium for Human Well-Being at UTSA. The authors would like to thank Dr. John Gustafson, who is the inventor of Posit, Generalized posit and Tapered Fixed-point and has provided valuable insights over the years. The authors would also like to express gratitude to NUAI lab members at RIT and UTSA who supported this research study.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2022 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this paper
Cite this paper
Langroudi, H.F., Karia, V., Pandit, T., Mashaido, B., Kudithipudi, D. (2022). ACTION: Automated Hardware-Software Codesign Framework for Low-precision Numerical Format SelecTION in TinyML. In: Gustafson, J., Dimitrov, V. (eds) Next Generation Arithmetic. CoNGA 2022. Lecture Notes in Computer Science, vol 13253. Springer, Cham. https://doi.org/10.1007/978-3-031-09779-9_4
Download citation
DOI: https://doi.org/10.1007/978-3-031-09779-9_4
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-031-09778-2
Online ISBN: 978-3-031-09779-9
eBook Packages: Computer ScienceComputer Science (R0)