PhysicsNeMo: Neural-Operator-Based Physics Simulation Framework

Paper summary

Title: PhysicsNeMo: A Framework for Neural Operator-based Physics-ML Models Authors: NVIDIA Physics Simulation & AI team

Contributions

1. Unified neural-operator framework (FNO / DeepONet / AFNO / Kapoor Transformer)
2. End-to-end differentiable pipeline (geometry -> BC -> solve -> post)
3. Multi-GPU/TPU distributed training (Megatron-style)
4. Production deployment (ONNX Export + Triton)

Key results

• FNO accuracy 30% higher than traditional RANS surrogate on NASA Clap benchmark
• 8-GPU linear speedup ~= 7.5x
• End-to-end inference < 50 ms/sample (A100)

Use cases

• Industrial physics-AI products (auto, aerospace, energy)
• Large physics foundation-model pretraining
• Integration with NVIDIA Modulus (now PhysicsNeMo)

Notes

• Apache-2.0, commercial-friendly
• Needs CUDA 12+, A100/H100 recommended