Planned intervention: On Wednesday June 26th 05:30 UTC Zenodo will be unavailable for 10-20 minutes to perform a storage cluster upgrade.
Published November 1, 2023 | Version v1
Book chapter Open

Inside the AI Accelerators: From High Performance to Energy Efficiency

Creators

  • 1. TIMA Laboratory, Univ. Grenoble Alpes, CNRS, Grenoble INP
  • 1. TIMA Laboratory, Univ. Grenoble Alpes, CNRS, Grenoble INP

Description

This paper overviews current technologies for high-performance, low-power neural networks. To cope with the high computational and storage resources, hardware optimisation techniques are proposed: Deep Learning (DL) compilers and frameworks, DL hardware coupled with hardware-specific code generators. More specifically, we explore the quantization mechanism in deep learning, based on a deep-CNN classification model. We highlight the accuracy of quantized models and explore their efficiency on a variety of hardware platforms. Through experiments, we show the performance achieved using general-purpose hardware (CPU and GPU) and a custom ASIC (TPU), as well as the simulated performance for a reduced bit-width representation of 4 bits, 2 bits (ternary) down to 1-bit heterogeneous quantization (FPGA).

Files

Inside the AI Accelerators: From High Performance to Energy Efficiency.pdf

Files (597.8 kB)

Additional details

Funding

Edge AI Technologies for Optimised Performance Embedded Processing 101097300
European Commission
MIAI – MIAI @ Grenoble Alpes ANR-19-P3IA-0003
Agence Nationale de la Recherche

Dates

Other
2023-11-01