Build a Physical AI model factory with NVIDIA Cosmos 3 on SageMaker HyperPod
Building a Physical AI system takes a continuous pipeline, not a single training job. This post shows how to run that model factory (synthetic data generation, post-training, and closed-loop evaluation with NVIDIA Cosmos 3) on a persistent, resilient Amazon SageMaker HyperPod cluster on Amazon EKS, with GPU goodput as the metric that matters.
Building a Physical AI system requires a continuous pipeline that generates synthetic data, trains perception and policy models, and evaluates them in closed-loop simulation. This continuous process is facilitated by a Physical AI model factory. One way to implement such a factory is by using NVIDIA Cosmos 3 on Amazon SageMaker HyperPod.
Cosmos 3 is a Mixture-of-Transformers (MoT) design with per-layer joint attention and a deliberate train-versus-inference asymmetry. The architecture integrates video, image, action, and sound into a single token stream, treating them all as a unified sequence. This integration enables Cosmos 3 to function both as a video generator for synthetic data and as an action policy for deployment. The model utilizes a single transformer trunk to manage both generation and evaluation tasks.
Written by urgent.news from AWS Machine Learning's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
Also reported by 1 other outlet
- Nvidia lets you build your own AI clusters locally with PAIR software computerworld.com
