Call for Papers

Foundation models have become the foundation of a new wave of machine learning models. The application of such models spans from natural language understanding into image processing, protein folding, and many more. The main objective of this workshop is to bring the attention of machine learning and system communities to the upcoming architectural and system challenges for the foundational models and drive the productive usage of these models in chip design process and system design. Subject areas of the workshop included (but not limited to):

System for Machine Learning

• System and architecture support of foundational models at scale

• Efficient model compression (e.g. quantization, sparsity) techniques

• Efficient and sustainable training and serving

• Benchmarking and evaluation of foundational models

• System and architecture support for Mixture-of-Experts (MoEs) and alternative model architectures

• Conditional and adaptive computing for machine learning

• Ethical accelerator and system design for AGI

• Evaluation of deployed machine learning systems and architectures

• Data-driven full-stack chip design

Machine Learning for System

• Learned models for computer architecture and systems optimizations

• Machine learning techniques for compiler and code optimization

• Distributed systems and infrastructure design for machine learning workloads

• Machine learning for hardware/software co-design (AutoML for hardware)

• Automated machine learning in EDA tools

• Optimized code generation for hardware and software

• Benchmarking and comparison between machine learning algorithms and heuristics

Architecture 2.0

* Find additional details on the Architecture 2.0 website.

• Datasets

  • What datasets do we need?

  • How should we collect these datasets for architecture research?

  • What metadata should the datasets contain to enable broad usage?

  • How do we create standard data formats from any ML algorithm?

• ML Algorithms

  • How can we learn and apply new ML algorithms to effectively design high-performance/efficient systems?

  • How do we make our community more accessible to ML researchers?

  • How do we embrace ML algorithm design as part of architecture research?

• Tools and Infra

  • How do we reduce the sim2real gap?

  • What instrumentation mechanisms do we need for creating the datasets?

  • What gym environments do we need to enable data-centric AI?

  • How do we define standard data formats for interoperability?

• Best Practices

  • Can we create a systematic playbook for best-known methods?

  • How do we ensure strong baselines and reproducibility?

• Industry Relations

  • How do we share traces/infrastructure without IP leakage?

  • What resources can we get from the industry?

  • How can industry contribute to academia?

  • How can academia tech transfer to the industry?

• Workforce Training

  • What do future architects need to know in addition to conventional architecture background?

  • How should we teach ML to architecture students so that they have the necessary foundations?

Diversity and Inclusion Statement

We are committed to fostering an inclusive and diverse environment for all participants. Our vision for this workshop is to build a diverse community and collectively work towards tackling challenges of foundational models. We recognize the value of diversity in promoting innovation, creativity, and meaningful discussions. Therefore, we have made significant efforts to ensure demographic diversity among our organizers and speakers. We acknowledge that achieving diversity is an ongoing process, and we continuously strive to improve our efforts in this regard. We encourage open feedback from our participants and the broader community to help us identify areas where we can enhance our inclusivity initiatives.

The Use of Large Language Models (LLMs)

The use of LLMs is allowed as a general-purpose writing assist tool. Authors should understand that they take full responsibility for the contents of their papers, including content generated by LLMs that could be construed as plagiarism or scientific misconduct (e.g. fabrication of facts). LLMs are not eligible for authorship.

Withdrawal Policy

Authors have the right to withdraw papers from consideration at any time until paper notification. Before the paper submission deadline, if an author withdraws the appear it will be deleted from the OpenReview hosting site. However, note that after the paper submission deadline, if an author chooses to withdraw a submission, it will remain hosted by OpenReview in a publicly visible "withdrawn papers" section. Withdrawn papers will be de-anonymized.

Authorship/Title/Abstract

Authors can change author order, but not add or remove authors. However, minor changes to titles and abstracts are allowed, if properly justified by the authors.

Submission Instructions

• We welcome submissions of up to 4 pages (not including references). This is not a strict limit, but authors are encouraged to adhere to it if possible.

• All submissions must be in PDF format and should follow the MLArchSys'24 Latex Template (Overleaf).

• Please follow the guidelines provided at ISCA 2024 Paper Submission Guidelines.

• Please submit your paper at OpenReview.   While the review process is not public, we make the accepted papers and their reviews public after the notification deadline. 

• Please carefully read and understand the MLArchSys 2024 Paper Checklist Guidelines.

• Reviewing will be double blind: please do not include any author names on any submitted documents except in the space provided on the submission form.

• We welcome submissions that include parts of ongoing work intended for a future conference submission; however, please ensure that your submitted work has not been previously published at a conference or in a journal.

Organizing Committee

• Bahar Asgari (UMD)

• Qijing Huang (NVIDIA)

• Tushar Krishna (GaTech)

• Yingyan (Celine) Lin (GaTech)

• Mohammad Shoeybi (Nvidia)

• Suvinay Subramanian (Google)

• Tianqi Tang (Meta)

• Neeraja J. Yadwadkar (University of Texas, Austin)

• Amir Yazdanbakhsh (Google DeepMind)

• Full Paper Submission Deadline: May 3rd, 2024, 11:59 AoE (OpenReview)

  • Please select one of the following main topics for your submission:

    • System for Machine Learning

    • Machine Learning for System

    • Architecture 2.0

• Author Notification: May 24th, 2024, 11:59 AoE.

• Workshop: June 29-30, 2024 (Buenos Aires, Argentina).

Contact us at mlarchsys@gmail.com