This international workshop is on the software and hardware co-design of deep learning systems on accelerators. To efficiently deploy the state-of-the-art hardware accelerators to meet the diverse demands from scientific domains, software-hardware co-design has become a critical research issue. Our workshop will highlight the latest research and software tools that can deliver higher performance and resource efficiency of deep learning systems, by taking advantage of new architectures and hardware available on the next generation accelerators. Through this workshop, we will provide a platform for researchers to showcase their preliminary results, inspire ideas, explore novel directions, promote collaborations, and enlarge the community. Topics include, but are not limited to:

• Performance modeling of deep learning applications,
• Scalable hyper-parameter optimization,
• Hardware-aware neural architecture search,
• Testing, debugging, and profiling deep learning systems in accelerators,
• Hardware-efficient training and inference,
• Scalability of large language models,
• Automating parallelism for deep learning systems,
• Efficient accelerators for deep learning,
• Quantum machine learning.

The organizers will review extendend abstracts and invite authors of accepted abstracts to give a 20 minute presentation at this half-day workshop. If you are interested to submit, please follow the guidelines below.

Deadlines

• Submission deadline: March 31, 2025 (AoE) April 11, 2025 (AoE)
• Author notification: April 7, 2025 (AoE) May 2, 2025 (AoE)

Submission Guildlines

Extended abstracts should be submitted electronically via EasyChair.

• Extended abstracts should be at most two pages, formatted in a single-column layout with an 11-point font size on letter-sized paper, and peer-reviewed by the organizers.
• Each abstract will have at least three reviewers two reviews. The program committee will execute a fair and uniform selection process based on the novelty and quality of the submission and within the scope of the workshop. The peer review process is single-blind. Reviewers have access to the names of paper submitters.