CODAI’23: Workshop on Compilers, Deployment, and Tooling for Edge AI
Workshop Webpage
Please visit the CODAI’23 website https://codai-workshop.com/ for more details.
About
The goal of the CODAI’23 workshop is to bring together the emerging energy in the AI compiler communities and AI accelerator communities focusing on Edge AI in both academic and industrial research. These realms of research have the opportunity to deliver a pervasive and seamless end-to-end tooling that connects hardware and software development methodologies. This workshop mainly focuses on:
- Discussing new impulses for deployment of neural networks on edge devices.
- Cooperative development of (open source) toolchains/frameworks for neural network deployment + cooperation between industry and academia.
Workshop Program
| Introduction & Keynote | |
|---|---|
| 10:00 – 10:30 | Welcome Coffee |
| 10:30 – 10:35 | Opening |
| 10:35 – 11:15 | Keynote: Next-generation Compilers for Emerging Systems |
| Session 1: Deployment and Optimization Techniques | |
| 11:15 – 11:40 | Scaling Up Quantization-Aware Neural Architecture Search for Efficient Deep Learning on the Edge |
| 11:40 – 12:05 | Tiny Machine Learning: Enabling Intelligence on Constrained Devices |
| 12:05 – 12:30 | Hardware-Aware Network Compression: From Data to Silicon |
| 12:30 – 13:30 | Lunch Break |
| Session 2: Compilation Frameworks and Techniques | |
| 13:30 – 13:55 | Accelerating Edge AI with Morpher: An Integrated Design, Compilation and Simulation Framework for CGRAs |
| 13:55 – 14:20 | Towards Rapid Exploration of Heterogeneous TinyML Systems using Virtual Platforms and TVM’s UMA |
| 14:20 – 14:45 | ART: An Actor transition systems RunTime for enabling efficient partitioning of neural network graphs |
| 14:45 – 15:10 | SYCL – A Modern C++ Programming Model for Accelerators |
| 15:10 – 15:40 | Coffee Break |
| Session 3: Applications | |
| 15:40 – 16:05 | Temporal Patience: Efficient Adaptive Deep Learning for Embedded Radar Data Processing |
| 16:05 – 16:30 | Pros and Cons of Executable Neural Networks for Deeply Embedded Systems |
| 16:30 – 16:55 | Software and Hardware for Sparse ML |
| 16:55 – 17:00 | Closing remarks |
Organizers
Michael J. Klaiber
EnCharge AI, USA
Sebastian Vogel
NXP Semiconductors, The Netherlands
Dayane Reis
University of South Florida, USA
Andreas Bytyn
Axelera AI, Germany
Dennis Rieber
Bosch Research, Germany
Miguel Aguilar
Aptiv, Germany
