Nvidia launches Modulus, a framework for developing “physics-based” AI models


Nvidia today launched Modulus, a framework for the development of “physics-based” machine learning models in industries that require a high level of physical precision. Modulus trains AI systems to use the laws of physics to model the behavior of systems in various fields, according to Nvidia, including climate science and protein engineering.

“Digital twin” simulation approaches have gained ground in many areas. For example, London-based SenSat helps clients in the construction, mining, energy and other industries create location models for the projects they are working on. GE offers technology that allows companies to model digital twins of real machines and closely monitor performance. And Microsoft provides Azure digital twins and Project Bonsai, which models the relationships and interactions between people, places and devices in simulated environments.

Gartner predicted that 50% of large manufacturers would have had at least one digital twin initiative launched by 2020, and that the number of organizations using digital twins would triple from 2018 to 2022. Markets and markets estimates that the global digital twin technology market will reach $ 48.2 billion by 2026, up from $ 3.1 billion in 2020.

Above: A physics simulation run with Nvidia Modulus.

Image Credit: Nvidia

“Digital twins have become powerful tools for solving problems ranging from the molecular level like drug discovery to global challenges like climate change,” Jay Gould, product marketing manager at Nvidia, said in a blog post. “Modulus gives scientists a [toolkit] to create highly accurate digital reproductions of complex and dynamic systems that will enable the next generation of breakthroughs in a wide range of industries.

Physics framework

Nvidia describes Modulus – which was announced at the company’s fall 2021 GPU Technology Conference (GTC) – as a framework to provide engineers, scientists and researchers with tools to create AI models of twins. digital. As in most AI-based approaches, Modulus includes a data preparation module that helps manage observed or simulated data, taking into account the geometry of the systems it models and the explicit parameters of the space. represented by the input geometry.

Modulus includes a sampling planner that allows users to select an approach, such as quasi-random sampling or importance sampling, to improve model accuracy. The framework also comes with APIs for taking governing symbolic partial differential equations and building physical models, as well as layers and organized network architectures suitable for physics-based problems.

Additionally, Modulus offers a “physics-machine learning” engine that uses inputs to train models using machine learning frameworks, including Facebook’s PyTorch and Google’s TensorFlow. The TensorFlow-based implementation of the toolkit optimizes performance by taking advantage of XLA, a domain-specific compiler for linear algebra that accelerates TensorFlow models. By leveraging the Horovod distributed deep learning training framework for multi-GPU scaling, Modulus can perform near real-time or interactive inference once a model is trained.

Module includes tutorials to get started with computational fluid dynamics, heat transfer, turbulence modeling, transient wave equations, and other multiphysics issues. It is now available for free download through the Nvidia Developer Zone.

“The GPU-accelerated toolkit offers fast turnaround time in addition to traditional analysis, enabling faster insights. Modulus allows users to explore different configurations and scenarios of a system by evaluating the impact of changing its settings, ”Gould wrote. “The module is customizable and easy to adopt. It offers APIs for the implementation of new physics and geometries. It is designed so that those new to AI-based digital twin applications can implement it quickly. “

Companies, including Alphabet’s DeepMind, have studied the application of AI systems to physical simulations. Last April, DeepMind described a model that predicts the movement of glass molecules as they transition between liquid and solid states. Beyond glass, the researchers said the work could lead to advancements in sectors such as manufacturing and medicine, including soluble glass structures for drug delivery and renewable polymers.


VentureBeat’s mission is to be a digital public place for technical decision-makers to learn about transformative technology and conduct transactions. Our site provides essential information on data technologies and strategies to guide you in managing your organizations. We invite you to become a member of our community, to access:

  • up-to-date information on the topics that interest you
  • our newsletters
  • Closed thought leader content and discounted access to our popular events, such as Transform 2021: Learn more
  • networking features, and more

Become a member


Leave A Reply