日本語 | English | 한국어

Physical AI Scaffolding Kit

This is a sample repository introduced in the Physical AI Development Support Program by AWS Japan.

This repository contains the following samples.

Environment Setup

When working with limited local resources, efficiently iterating on training can become a challenge. Amazon SageMaker HyperPod eliminates the undifferentiated heavy lifting associated with building generative AI models, enabling you to build models efficiently with abundant resources.

• Building a Slurm Cluster with Amazon SageMaker HyperPod
  • If you only want to build a standard Slurm cluster, please refer to this sample to set up your Amazon SageMaker HyperPod environment
• Physical AI Pipeline Platform SDK
  • If you want to use a pipeline that utilizes the SDK developed for Amazon SageMaker HyperPod, please refer to this sample to set up your environment
• NVIDIA IsaacSim Development Workstation
  • A sample for using NVIDIA Isaac Sim on an EC2 instance
  • Provides access to NVIDIA Isaac Sim / Lab environments

Samples for Amazon SageMaker HyperPod

1. π0 Sample
2. NVIDIA Isaac GR00T Sample
3. NVIDIA Isaac Lab Newton RL Sample