A 95.6-TOPS/W Deep Learning Inference Accelerator With Per-Vector Scaled 4-bit Quantization in 5 nm-Reference-Cited by-同舟云学术

A 95.6-TOPS/W Deep Learning Inference Accelerator With Per-Vector Scaled 4-bit Quantization in 5 nm

Published:2023-04 Issue:4 Volume:58 Page:1129-1141
ISSN:0018-9200
Container-title:IEEE Journal of Solid-State Circuits
language:
Short-container-title:IEEE J. Solid-State Circuits

Author:

Keller Ben¹^ORCID,Venkatesan Rangharajan¹,Dai Steve¹^ORCID,Tell Stephen G.²,Zimmer Brian¹^ORCID,Sakr Charbel¹,Dally William J.¹^ORCID,Gray C. Thomas²^ORCID,Khailany Brucek³^ORCID

Affiliation:

1. NVIDIA, Inc., Santa Clara, CA, USA

2. NVIDIA, Inc., Durham, NC, USA

3. NVIDIA, Inc., Austin, TX, USA

Publisher

Institute of Electrical and Electronics Engineers (IEEE)

Subject

Electrical and Electronic Engineering

Link

http://xplorestaging.ieee.org/ielx7/4/10082904/10019275.pdf?arnumber=10019275

Reference65 articles.

1. SCNN

2. PyTorch: An imperative style, high-performance deep learning library;paszke;Proc Conf Neural Inf Process Syst (NIPS),2019

3. A 7 nm 4-core AI chip with 25.6 TFLOPS hybrid FP8 training, 102.4 TOPS INT4 inference and workload-aware throttling;agrawal;IEEE Int Solid-State Circuits Conf (ISSCC) Dig Tech Papers,2021

4. Estimating or propagating gradients through stochastic neurons for conditional computation;bengio;arXiv 1308 3432,2013

5. DIANA: An End-to-End Energy-Efficient Digital and ANAlog Hybrid Neural Network SoC

Cited by 13 articles. 订阅此论文施引文献订阅此论文施引文献，注册后可以免费订阅5篇论文的施引文献，订阅后可以查看论文全部施引文献

1. BSViT: A Bit-Serial Vision Transformer Accelerator Exploiting Dynamic Patch and Weight Bit-Group Quantization;IEEE Transactions on Circuits and Systems I: Regular Papers;2024-09

2. High-performance deep spiking neural networks with 0.3 spikes per neuron;Nature Communications;2024-08-09

3. MECLA: Memory-Compute-Efficient LLM Accelerator with Scaling Sub-matrix Partition;2024 ACM/IEEE 51st Annual International Symposium on Computer Architecture (ISCA);2024-06-29

4. Language Modeling on a SpiNNaker2 Neuromorphic Chip;2024 IEEE 6th International Conference on AI Circuits and Systems (AICAS);2024-04-22

5. Novel adaptive quantization methodology for 8-bit floating-point DNN training;Design Automation for Embedded Systems;2024-02-16