nvidia/xmgn_drivaerml_surface

XMeshGraphNet DrivAerML

XMeshGraphNet-DrivAerML is a pre-trained AI model for automotive external aerodynamics. This model has been trained using the DrivAerML dataset, that are LES simulations of road-cars of varying geometries. This pre-trained model works by taking the input from a single DrivAerML STL (Standard Tessellation Language) geometry and evaluates a solution on the surface of the vehicle.

This model is ready for commercial use.

License/Terms of Use:

Use of this model is governed by the NVIDIA Open Model Agreement.

Deployment Geography:

Global

Use Case:

Computational Fluid Dynamics (CFD) engineers accelerating automotive external aerodynamics with AI.

Release Date:

05/01/2026

Hugging Face: https://huggingface.co/nvidia/xmgn_drivaerml_surface

Reference(s):

• Codebase
  

• Paper
  

Model Architecture:

Architecture Type: Graph Neural Network with message passing blocks, fully connected blocks, and partitioning with halo.

Network Architecture: The X-MeshGraphNet (X-MGN) is a scalable, multi-scale extension of MeshGraphNet designed for fast physics simulation. Its architecture features three technical pillars: Custom Graph Construction directly from CAD files (e.g., STLs) via point clouds and $k$-nearest neighbors (KNN); Scalable Partitioning of large graphs with halo regions, where gradient aggregation ensures the training is mathematically equivalent to processing the full graph; and a Multi-Scale approach that refines graph resolution to efficiently capture long-range interactions.

Number of model parameters: 12M

Input:

Input Type(s): Surface mesh (STL nodes and face connectivities) (3D)
Input Format(s): PyTorch Tensor / NumPy array
Input Parameters:

• Surface mesh (STL) coordinates (M, 3), where M is the number of cells in the surface mesh
• Surface mesh normals (M, 3)
  

Other Properties Related to Input: None

Output:

Output Type(s): Point cloud
Output Format: PyTorch Tensor / NumPy array
Output Parameters:

• Surface pressure (M, 1)
• Wall shear stress (M, 3)
  

Other Properties Related to Output: The outputs are non-dimensionalized, and then normalized using mean and standard deviation calculated from the training dataset.

Our AI models are designed and/or optimized to run on NVIDIA GPU-accelerated systems. By leveraging NVIDIA’s hardware (e.g. GPU cores) and software frameworks (e.g., CUDA libraries), the model achieves faster training and inference times compared to CPU-only solutions.

Software Integration

Runtime Engine(s): PyTorch
Supported Hardware Microarchitecture Compatibility:

• NVIDIA Ampere
  
• NVIDIA Blackwell
  
• NVIDIA Hopper
  
• NVIDIA Turing
  

Supported Operating System(s):

• Linux
  

Model Version(s):

Model version: 1.0.0

Training and Evaluation Datasets:

The DrivAerML dataset is used for training and evaluation, which is a publicly available, high-fidelity dataset comprising aerodynamic data for 500 parametrically morphed variants of the DrivAer notchback vehicle. The dataset was generated using hybrid RANSLES (HRLES), a scale-resolving CFD method, which provides time-averaged quantities for each variant. The available data includes surface pressure, wall shear stress, and flow-field quantities, provided in formats compatible with mesh-based analysis (.vtp for surface data and .vtu for flow-field data).10% of the samples are used as the test set, with 20% of the test set consisting of out-of-distribution samples based on drag coefficients. These samples represent extreme cases with the lowest and highest drag coefficients in the entire dataset, which remain unseen by the model during training.

Training Dataset:

Data Modality:

• Other: Mesh

Training Data Size:

• 436 files in VTP format that contain meshes and corresponding physical quantities

Link: DrivAerML Dataset

Data Collection Method by dataset:

• Synthetic CFD Simulation

Labeling Method by dataset:

• Automated

Properties: The data is a simulation/synthetic dataset generated using the OpenFOAM CFD solver to generate flow fields such as velocity and pressure for different car geometries for the same boundary condition configuration as used to generate the training set.

Evaluation Dataset:

Link: DrivAerML Dataset

Data Collection Method by dataset:

• Synthetic CFD Simulation

Labeling Method by dataset:

• Automated

Properties: Validation split from DrivAerML dataset with vehicle geometries held out from training. The full DrivAerML dataset is split as 90% for training and 10% for validation.

Inference:

Engine: PyTorch

Test Hardware:

• A100
  
• H100
  
• L40S
  
• RTX PRO 6000 Blackwell
  

Ethical Considerations:

NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications. When downloaded or used in accordance with our terms of service, developers should work with their supporting model team to ensure this model meets requirements for the relevant industry and use case and addresses unforeseen product misuse.

For more detailed information on ethical considerations for this model, please see the Model Card++ subcards: Bias, Explainability, Privacy, and Safety & Security.

Please report model quality, risk, security vulnerabilities or NVIDIA AI Concerns here.