Amazon Braket helps organizations get access to quantum computing hardware and simulators so they can speed up their scientific research and software development for quantum computing. This session shows short tutorials and shares how to run quantum circuits using real gate-based devices and simulators. Learn how to run your first quantum machine learning algorithm using Amazon Braket Hybrid Jobs. Along the way, explore Amazon Braket’s features, see examples from the AWS quantum algorithm library, and get your questions answered.

Training machine learning (ML) models requires setting up clusters that enable many GPUs to talk to each other using low-latency networking that is capable of driving massive throughput. This session shares in real time how you can quickly and easily set up an ML training cluster using AWS ParallelCluster, NVIDIA GPUs, Elastic Fabric Adapter (EFA), and Amazon FSx for Lustre. Learn how to train a GPT model using Megatron-LM, store the results, and collapse the infrastructure when you’re finished.

HPC systems deployed on AWS often rely on various resources like file systems, networking, and directory services. Despite AWS ParallelCluster’s automation capabilities, setting up and integrating these dependencies can become complex if you have a lot of custom requirements. In this session, walk through how the new HPC Recipes for AWS library simplifies this process. Then, learn how to create a multi-user environment, configure shared storage, set up a budget, and deploy a benchmarking cluster with just a few clicks in the AWS Management Console. Finally, discover how to combine, modify, and reuse these recipes to meet your particular needs—without needing to be an AWS expert.

*Special session on Monday, November 13, 8PM-9PM during the SC23 opening gala.

Join this session to learn how organizations across different fields—from EDA to drug design—have been using AWS to scale and improve their R&D. Learn how to create a real, large, and very complete cluster in a few minutes using AWS HPC services, including AWS ParallelCluster, Amazon FSx for Lustre, and visualization with NICE DCV. Find out how to customize compute images to include applications and development tools, and discover how to integrate with Spack for reliable deployment of open source packages.

The generative capability of AI holds significant promise across a diverse range of industries fueled by engineering design, such as automotive, motor sports, and aerospace. In this session, discover how to create a pipeline that uses generative AI designs to feed conventional physics-based simulations, and learn how to loop all of this to create a converging, rapid design process for exploring new design concepts starting from a single image. Learn how to use open source frameworks to create digital twins, deploy OpenFOAM in containers for the simulations, and use

Amazon FSx for Lustre helps you deploy high-performance Lustre systems in just a few clicks and has capabilities that most traditional storage systems lack—think of processing hundreds of gigabytes per second of throughput without months of planning, logistics, and testing. In this session, learn how to spin up a large-scale and fast Lustre file system in less time than it takes to make a coffee. Then, discover how it can synchronize with massive datasets in object storage and how you can choose from different classes of storage performance and price to match your organization’s needs. Finally, learn how you can use Amazon File Cache to deploy Lustre as a massively parallel cloud cache for storage systems in your own data centers and how this can be a useful tool for creating hybrid facilities that speed up your users’ time to results.

HPC systems and workloads have their own unique set of security challenges. Running these workloads on AWS can help you address these challenges in new ways and with this freedom comes the ability to choose different solutions based on your organization’s unique needs. In this session, walk through a traditional HPC security scenario and discover different approaches you can take to address these challenges. Find out how the AWS secure-by-design architecture helps take care of common security challenges, and learn best practices for building secure HPC environments on AWS. When you’re done, you’ll understand how your cloud infrastructure can be even more secure than your data centers.

AWS Trainium purpose-built accelerators can be used to train large language models (LLMs) such as Llama 2. AWS Neuron is the SDK used to run deep learning workloads on AWS Trainium–based instances. It integrates the AI framework, such as PyTorch or TensorFlow, with the hardware to enable end-to-end ML development, including building new models and training and scaling these models. The AWS Neuron SDK supports multiple distributed libraries using tensor parallelism, pipeline parallelism, and sequence parallelism for large-scale training. In this session, find out how to use AWS ParallelCluster to launch Amazon EC2 Trn1 instances and train a Llama 2 model (up to 70B) on more than 1,000 chips.