# NVIDIA/physicsnemo

Open-source deep-learning framework for building, training, and fine-tuning deep learning models using state-of-the-art Physics-ML methods

Repository: https://github.com/NVIDIA/physicsnemo
Canonical: https://ross.abutalabs.com/products/physicsnemo
Homepage: https://developer.nvidia.com/physicsnemo
Language: Python
License: Apache-2.0
License Family: permissive
Topics: deep-learning, machine-learning, nvidia-gpu, physics, pytorch, nvidia-warp
Last push: 2026-08-26T21:28:23+00:00

## Health v2 (maintenance only)
Score: 89/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 75, longevity 93
- inputs: {"age_days": 1315, "days_push": 7, "days_rel": 86, "gap_med": 72.5, "n_releases_24m": 11}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 3198, forks 763 (observed 2026-08-28T04:07:48.406029+00:00)

## What it is
NVIDIA PhysicsNeMo is an open-source Python deep-learning framework for building, training, fine-tuning, and inferring physics AI models using state-of-the-art Physics-ML methods. It provides GPU-optimized modules for scalable training pipelines combining physics knowledge with data, supporting neural operators, GNNs, transformers, and Physics-Informed Neural Networks.

## Use cases
- train physics-informed neural networks on gpu
- build ai surrogate models for cfd simulation
- train neural operators at scale with pytorch
- develop digital twin models for engineering
- accelerate automotive aerodynamics simulation with ai
- fine-tune physics ai models for structural mechanics

## When to choose
- you need GPU-optimized, scalable training of physics-ML models like PINNs, neural operators, or GNNs
- you are building AI surrogate models for engineering simulations such as CFD, structural mechanics, or electromagnetics
- you want a PyTorch-integrated framework with prebuilt model architectures and examples for AI4Science

## When to avoid
- your problem has no physics or simulation component and standard deep learning frameworks suffice
- you need a lightweight CPU-only solution, since the stack is optimized for NVIDIA GPUs
- you require a turnkey commercial simulation product rather than a framework for building custom models

## Facets
- artifact type: framework
- maturity: active
- function: machine-learning, deep-learning, llm-training, simulation, gpu-computing
- domain: machine-learning, deep-learning, simulation, artificial-intelligence
- platform: python, cloud
- tags: physics-informed-neural-networks, neural-operators, gnn, surrogate-models, cfd, digital-twins, pytorch, sciml, ai4science, physics, engineering, gpu, docker, linux

## Member repositories
- NVIDIA/physicsnemo (main) score 89

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:48.406029+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T07:24:51.692174+00:00, confidence not recorded.
  - readme: https://github.com/NVIDIA/physicsnemo (fetched 2026-08-28T04:07:48.406029+00:00, sha cffd7d315b4f)
  - homepage: https://developer.nvidia.com/physicsnemo (fetched 2026-08-29T09:38:36.783287+00:00, sha 5f5611930f01)
  - site_page: https://docs.nvidia.com/nim/physicsnemo/domino-automotive-aero/latest/overview.html (fetched 2026-08-29T09:38:36.798233+00:00, sha 4e1fb452593f)
  - site_page: https://docs.nvidia.com/physicsnemo/index.html (fetched 2026-08-29T09:38:36.792781+00:00, sha 1b1b649f649e)
  - site_page: https://docs.nvidia.com/deeplearning/physicsnemo/getting-started/index.html (fetched 2026-08-29T09:38:36.794594+00:00, sha 2571ed8878d8)
- Data as of 2026-08-30T08:39:29.467469+00:00.