NGC | Catalog

Modulus

For copy image paths and more information, please view on a desktop device.
Logo for Modulus

Description

NVIDIA Modulus is a toolkit for developing AI enabled physics-ML applications.

Publisher

NVIDIA

Latest Tag

23.05

Modified

June 1, 2023

Compressed Size

9.35 GB

Multinode Support

Yes

Multi-Arch Support

No

23.05 (Latest) Scan Results

Linux / amd64

What is Modulus?

NVIDIA Modulus is a toolkit for developing AI enabled physics-ML applications.

With NVIDIA Modulus, we aim to provide researchers and industry specialists, various tools that will help accelerate your development of such models for the scientific discipline of your need.

Visit the Nvidia Modulus for more information.

Modulus Documentation

Running Modulus Using Docker

If you have Docker 19.03 or later, a typical command to launch the container with a interactive bash terminal is:

docker run --gpus all --shm-size=1g --ulimit memlock=-1 --ulimit stack=67108864 --runtime nvidia --rm -it nvidia/modulus/modulus:xx.xx bash

Where,

  • xx.xx is the container version. For example, 23.05.

Once inside the container, you can clone the Modulus repositories from GitHub and use the samples and examples provided to get started with Modulus.

Running Modulus using Base Command Platform

Jobs using the Pytorch NGC Container on Base Command Platform clusters can be launched either by using the NGC CLI tool or by using the Base Command Platform Web UI. To use the NGC CLI tool, configure the Base Command Platform user, team, organization, and cluster information using the ngc config command as described here.

An example command to launch the container on a single-GPU instance is:

ngc batch run --name "My-1-GPU-Modulus-job" --instance dgxa100.80g.1.norm --commandline "sleep 30" --result /results --image "nvidia/modulus/modulus:23.05"

For details on running Modulus in Multi-GPU/Multi-Node configuration, refer this Technical Blog and Modulus Documentation

Key Features v23.05

  • Support for GNNs starting with MeshGaphNet and GraphCast models
  • Support for Convolutional RNN-based models
  • Modulus has been rearchitected into modules:
    • Modulus Core is the base module that consists of the core components of the framework for developing Physics ML models
    • Modulus Sym provides an abstraction layer for using PDE-based symbolic loss functions
    • Modulus Launch provides optimized training recipes for data driven physics ML models
  • Expanded feature set for AI weather and climate models
    • SOTA models including FourCastNet and GraphCast
    • Climate and weather evaluation metrics
    • Efficient datapipe for loading historical weather datasets using NVIDIA DALI.
  • Fast utilities and kernels for producing training data on-the-fly using NVIDIA’s Warp library.
  • Cugraph-Ops support for GraphCast that reduces the training time by 30% compared to DGL. Cugraph-Ops is a GNN library of highly optimized and performant primitives.

Key Features v22.09

  • Enhancements to FNO, PINO, and DeepONet architectures to enable more customizability and configure new networks
  • Modeling enhancements such as Selective Equations Term Suppression (SETS), Causal weighting scheme and criteria-based training termination APIs
  • Performance and usability enhancements such as FunTorch integration, more example-guided workflows for beginners

Key Features v22.07

  • Performance enhancements such as Meshless Finite Differentiation, leveraging CUDA Graphs and Tiny Cuda NN networks
  • Usability enhancements such as support for map style datasets, improved point cloud sampling for continuous and tessellated geometries
  • Support for generalized DeepOnet and FourCastNet

Key Features v22.03

  • FNO/AFNO support to create physics-ML models from data
  • Modulus Omniverse extension to visualize the outputs of the physics-ML model and interactively infer in real-time for new conditions defined by parameters
  • Support for DeepOnet architecture
  • Support for 2-eqn. turbulence models for modeling fully developed turbulent flow

Benefits

  • Broad Applicability - Models multiple physics types in forward and inverse simulations with accuracy and convergence.
  • Fast Turnaround Time - Provides parameterized system representation that solves for multiple scenarios simultaneously.
  • Easy to Adopt - Provides application programming interfaces (APIs) for implementing new physics and geometry and detailed user guide examples.

Modulus Forum

Please visit the Modulus Forum for :

  • Latest news and announcements on Modulus
  • Technical support
  • Report a bug
  • Customer success stories

License

By pulling and using the container, you accept the terms and conditions of this SOFTWARE DEVELOPER KITS, SAMPLES AND TOOLS LICENSE AGREEMENT.