en.Wedoany.com Reported - NVIDIA has partnered with the open-source artificial intelligence community Hugging Face to introduce its open reasoning vision-language action model for humanoid robots, Isaac GR00T 1.7, and its data collection framework, Isaac Teleop, into Hugging Face's open-source robotics library LeRobot. It also plans to soon introduce the physical AI frontier model Cosmos 3. These integrations aim to provide developers with a more accessible and standardized end-to-end robotics development path, fostering collaboration and innovation within the open robotics community.

Hugging Face co-founder and Chief Science Officer Thomas Wolf stated that open source is how a field transforms advanced research into something people can learn from, adapt, and build upon. With Isaac GR00T 1.7 and Isaac TeleOp integrated into LeRobot, developers can train and evaluate robots in an open environment using shared models, data, and workflows. The upcoming addition of Cosmos 3 will provide the community with a pathway to incorporate frontier world models into the same collaborative loop.

Hugging Face LeRobot is an open-source library for training, running, and sharing robotics datasets, models, policies, and workflows. This collaboration connects NVIDIA's 3 million robotics developers with Hugging Face's 16 million AI builders, expanding access to frontier physical AI tools. Specific integration details include: Isaac Teleop as an open-source data collection framework, helping developers capture high-quality human demonstrations in a standardized, interoperable format and directly expand shared datasets within LeRobot; Isaac GR00T 1.7 as the first open and commercially viable robotics foundation model, simplifying post-training and deployment through LeRobot workflows, supporting model adaptation to new robot embodiments and tasks, and providing baseline performance; and Cosmos 3 as a physical AI frontier world foundation model, soon to be available in LeRobot for generating and augmenting robot data, simulating scenarios, and supporting policy development when real-world data is limited.

These integrations build upon existing NVIDIA resources to support the complete robotics development loop. These resources include: the largest open-source physical AI dataset, downloaded over 15 million times, containing more than 350,000 real and simulated trajectories and 57 million grasp actions; simulation frameworks based on Isaac Sim and Isaac Lab; the Isaac Lab-Arena within the LeRobot Environment Hub, enabling developers to rapidly prototype complex simulation environments, register them in LeRobot EnvHub, and use them to train and evaluate general-purpose robot policies such as GR00T, Pi, and SmolVLA; and the NVIDIA Jetson Thor integration with LeRobot's Reachy 2, supporting the deployment of VLA models on open-source humanoid robots.