ALCF AI Testbed’s Cerebras and SambaNova systems are now available to the research community

May 3, 2022 – The Argonne Leadership Computing Facility (ALCF), a U.S. Department of Energy (DOE) Office of Science User Facility located at the DOE’s Argonne National Laboratory, is now accepting research proposals for the access to his AI testbeda growing collection of some of the world’s most advanced artificial intelligence (AI) accelerators available to science.

The ALCF AI Testbed is designed to enable researchers to explore next-generation machine learning applications and workloads to advance the use of AI for science. The AI ​​platforms will complement the facility’s current and next-generation supercomputers to provide a state-of-the-art environment that supports pioneering research at the intersection of AI, big data, and high-performance computing (HPC). ).

ALCF’s SambaNova DataScale system. Credit: Argonne National Laboratory

“It is clear that AI will play an important role in the future of scientific computing,” said ALCF Director Michael Papka. “With the ALCF AI testbed, our goal is to understand the role that AI accelerators can play in advancing data-driven discoveries, and how these systems can be combined with supercomputers to scale. to extremely large and complex scientific problems.”

Researchers interested in using the AI ​​testbed Cerebra CS-2 and SambaNova Data Scale can now submit project proposals through the ALCF Director’s Discretionary Program. Access to additional testbed resources, including Graphcore, Groqand Habana Acceleratorswill be announced later.

ALCF AI Testbed systems are designed to support machine learning and data-centric workloads, making them well suited to meet challenges involving the ever-increasing amounts of data produced by supercomputers, light sources and particle accelerators, among other powerful research tools. Additionally, the test bed will allow researchers to explore new workflows that combine AI methods with simulation and experimental science to accelerate the pace of discovery.

Cerebra CS-2 system.

“New AI technologies are largely designed for enterprise workloads and applications, such as e-commerce and social media,” said Venkat Vishwanath, head of ALCF’s Data Science group. “By making the latest AI accelerators available to the open science community, we provide a testing ground for innovative research campaigns focused on machine learning and HPC. We are really looking forward to seeing how the community uses these accelerators for different types of scientific applications and workflows.

Prior to opening the ALCF AI testbed to the wider scientific community, Argonne researchers led several collaborative efforts to use AI accelerators for a variety of data-centric studies. Check out some of the early successes below.

Edge Computing

To keep pace with the growing amount of data produced at DOE light source facilities, researchers are turning to machine learning methods to help with tasks such as data reduction and provide insights to guide future research. experiences. Using the ALCF’s Cerebras system, researchers from Argonne, the University of Chicago, SLAC National Accelerator Laboratory, and Stanford University demonstrated how specialized AI systems can be used to Rapidly train machine learning models through a geographically distributed workflow. To gain actionable insights in real time, the team trained the models on the remote AI system and then deployed them to edge computing devices near the experimental data source. Their work has been recognized by the Best Paper Award at last year’s Extreme Scale Experiment-in-the-Loop Computing (XLOOP) workshop at SC21.

COVID-19 Research

Using a combination of AI and supercomputing resources, a team led by Argonne has achieved a study of the replication mechanism of SARS-CoV-2 which was nominated for the Gordon Bell Special Prize for HPC-Based COVID-19 Research at SC21. The team used cryo-electron microscopy data to explore the molecular machinery, but static images alone did not provide high enough resolution to capture the inner workings of the process. To take a closer look at the mechanism of replication, the team developed an innovative workflow to improve resolution using a hierarchy of AI methods that continuously learn and infer features to maintain consistency between different types of simulations. The researchers used the Balsam workflow engine to orchestrate AI and simulation campaigns on four of the nation’s top supercomputers and the ALCF’s Cerebras CS-2 system. The method allowed the team to study the transcription process of SARS-CoV-2 replication at an unprecedented level of detail, while demonstrating a generalized, multi-scale computational toolkit to explore biomolecular machinery. dynamic.

Neutrino physics

Scientists use liquid argon time projection chambers (LArTPCs) to detect neutrinos, but the resulting images are sensitive to background particles induced by cosmic interactions. To improve neutrino signal efficiency, scientists use image segmentation to label each input pixel as one of three classes: cosmic-induced noise, neutrino-induced noise, or background noise. Deep learning has been a useful tool for accelerating this classic image segmentation task, but it has been limited by the image size that available GPU-based platforms can effectively train on. Leveraging ALCF’s SambaNova System, the researchers were able to improve this method to establish a new level of peak accuracy of 90.23% using images at their original resolution without the need to downsample. Their work demonstrates capabilities that can be used to improve model quality for a variety of important and difficult image processing problems.

drug discovery

A team of Argonne researchers leveraged the ALCF’s Groq system to speed up the process of searching through large numbers of small molecules to find antiviral drugs to fight COVID-19. With billions and billions of potential drug candidates to sort through, scientists needed a way to dramatically accelerate their research. When testing a large dataset of molecules, the team found they could make 20 million predictions, or inferences, per second, dramatically reducing the time needed for each search from days to minutes. Once the best candidates were found, the researchers identified those that could be obtained commercially and had them tested on human cells at the Howard T. Ricketts Laboratory at the University of Chicago.

To request time, submit a proposal: Allocation request form

Learn more about using our AI testbed: AI Testbed User Guides

please contact [email protected] with all questions.

The Argonne Leadership Computing Facility provides supercomputing capabilities to the scientific and engineering community to advance fundamental discovery and understanding across a wide range of disciplines. Supported by the Advanced Scientific Computing Research (ASCR) program of the U.S. Department of Energy’s (DOE) Office of Science, the ALCF is one of two DOE advanced computing facilities dedicated to open science.

Argonne National Laboratory seeks solutions to pressing national problems in science and technology. The nation’s first national laboratory, Argonne conducts cutting-edge basic and applied scientific research in virtually every scientific discipline. Argonne researchers work closely with researchers from hundreds of companies, universities, and federal, state, and municipal agencies to help them solve their specific problems, advance American scientific leadership, and prepare the nation for a better future. With employees from more than 60 nations, Argonne is led by UChicago Argonne, LLC for the U.S. Department of Energy Office of Science.

U.S. Department of Energy Office of Science is the largest supporter of basic physical science research in the United States and strives to address some of the most pressing challenges of our time. For more information, visit https://​ener​gy​.gov/​s​c​ience.


Source: Jim Collins, Argonne Leadership Computing Facility

Sam D. Gomez