Edge TPU

Edge TPUs are specialized hardware accelerators designed to perform deep learning inference efficiently at the edge, minimizing the need for cloud connectivity. Current research focuses on optimizing TPU architectures for various applications, including medical image analysis and large language model processing, through techniques like runtime reconfigurable dataflows and heterogeneous integration with other computing architectures. This work aims to improve energy efficiency, reduce latency, and enhance the performance of on-device AI, impacting fields ranging from healthcare to robotics and satellite technology. The development of efficient scheduling algorithms and optimized model architectures for Edge TPUs is a key area of ongoing investigation.

Papers

July 11, 2024

Flex-TPU: A Flexible TPU with Runtime Reconfigurable Dataflow Architecture
Mohammed Elbtity, Peyton Chandarana, Ramtin Zand
Machine Learning Accelerator Tensor Processing Unit Dataflow Accelerator Edge TPU

November 20, 2023

Energy efficiency in Edge TPU vs. embedded GPU for computer-aided medical imaging segmentation and classification
José María Rodríguez Corral, Javier Civit-Masot, Francisco Luna-Perejón, Ignacio Díaz-Cano, Arturo Morgado-Estévez, Manuel Domínguez-Morales
Classification Code Medical Image Segmentation Single GPU Energy Efficiency Color Fundus Optic Disc Edge TPU

October 27, 2023

Is Scaling Learned Optimizers Worth It? Evaluating The Value of VeLO's 4000 TPU Months
Fady Rezk, Antreas Antoniou, Henry Gouk, Timothy Hospedales
Net Present Value Better Optimizers Superior Optimizer Learned Optimizers Machine Learning Benchmark Edge TPU

October 25, 2023

RedCoast: A Lightweight Tool to Automate Distributed Training of LLMs on Any GPU/TPUs
Bowen Tan, Yun Zhu, Lijuan Liu, Hongyi Wang, Yonghao Zhuang, Jindong Chen, Eric Xing, Zhiting Hu
Large Language Model Machine Learning Training Data Single GPU Decomposed Automation Correction Machine Learning Pipeline Model Parallelism Edge TPU

September 16, 2023

Exploration of TPUs for AI Applications
Diego Sanmartín Carrión, Vera Prohaska
Environment Exploration Edge Computing Hardware Accelerator AI Application Tensor Processing Unit Edge TPU

May 26, 2023

InstaGrasp: An Entirely 3D Printed Adaptive Gripper with TPU Soft Elements and Minimal Assembly Time
Xin Zhou, Adam J. Spiers
Assembly Task Gripper Design Single Motor Robotic Gripper Edge TPU Gripper Finger

April 18, 2023

Heterogeneous Integration of In-Memory Analog Computing Architectures with Tensor Processing Units
Mohammed E. Elbtity, Brendan Reidy, Md Hasibul Amin, Ramtin Zand
Analog in Memory Computing Tensor Processing Unit Heterogeneous System Edge TPU Mobile CNN

April 10, 2023

RESPECT: Reinforcement Learning based Edge Scheduling on Pipelined Coral Edge TPUs
Jiaqi Yin, Yingjie Li, Daniel Robinson, Cunxi Yu
Reinforcement Learning Deep Neural Network Optimal Scheduling Edge TPU Edge Server Scheduling Multi DNN Scheduling

April 4, 2023

TPU v4: An Optically Reconfigurable Supercomputer for Machine Learning with Hardware Support for Embeddings
Norman P. Jouppi, George Kurian, Sheng Li, Peter Ma, Rahul Nagarajan, Lifeng Nai, Nishant Patil, Suvinay Subramanian, Andy Swing, Brian Towles, Cliff Young, Xiang Zhou, Zongwei Zhou, David Patterson
Machine Learning Jina Embeddings Optical Computing Hardware Error Dataflow Accelerator Edge TPU

July 26, 2022

A Study on the Use of Edge TPUs for Eye Fundus Image Segmentation
Javier Civit-Masot, Francisco Luna-Perejon, Jose Maria Rodriguez Corral, Manuel Dominguez-Morales, Arturo Morgado-Estevez, Anton Civit
Deep Learning Study Feature Greater Public Use Efficient Segmentation Edge TPU Fundus Image Segmentation

April 13, 2022

Scalable Training of Language Models using JAX pjit and TPUv4
Joanna Yoo, Kuba Perlin, Siddhartha Rao Kamalakara, João G. M. Araújo
Language Model Technical Challenge Training Strategy Scalable Training Edge TPU JAX Primitive

April 9, 2022

Searching for Efficient Neural Architectures for On-Device ML on Edge TPUs
Berkin Akin, Suyog Gupta, Yun Long, Anton Spiridonov, Zhuo Wang, Marie White, Hao Xu, Ping Zhou, Yanqi Zhou
Neural Architecture Search Machine Learning Accelerator Tensor Processing Unit Efficient Neural Network Architecture Device Machine Learning Edge TPU

April 7, 2022

Deep Learning for Real Time Satellite Pose Estimation on Low Power Edge TPU
Alessandro Lotti, Dario Modenini, Paolo Tortora, Massimiliano Saponara, Maria A. Perino
Deep Learning Pose Estimation Pose Estimator Tensor Processing Unit Photorealistic Dataset Edge TPU Satellite Pose Estimation Spacecraft Pose Estimation

December 3, 2021

ALX: Large Scale Matrix Factorization on TPUs
Harsh Mehta, Steffen Rendle, Walid Krichene, Li Zhang
Link Prediction Matrix Factorization Prime Factorization Edge TPU

Edge TPU

Papers

Flex-TPU: A Flexible TPU with Runtime Reconfigurable Dataflow Architecture

Energy efficiency in Edge TPU vs. embedded GPU for computer-aided medical imaging segmentation and classification

Is Scaling Learned Optimizers Worth It? Evaluating The Value of VeLO's 4000 TPU Months

RedCoast: A Lightweight Tool to Automate Distributed Training of LLMs on Any GPU/TPUs

Exploration of TPUs for AI Applications

InstaGrasp: An Entirely 3D Printed Adaptive Gripper with TPU Soft Elements and Minimal Assembly Time

Heterogeneous Integration of In-Memory Analog Computing Architectures with Tensor Processing Units

RESPECT: Reinforcement Learning based Edge Scheduling on Pipelined Coral Edge TPUs

TPU v4: An Optically Reconfigurable Supercomputer for Machine Learning with Hardware Support for Embeddings

A Study on the Use of Edge TPUs for Eye Fundus Image Segmentation

Scalable Training of Language Models using JAX pjit and TPUv4

Searching for Efficient Neural Architectures for On-Device ML on Edge TPUs

Deep Learning for Real Time Satellite Pose Estimation on Low Power Edge TPU

ALX: Large Scale Matrix Factorization on TPUs