Denoising Pre-trained Operator Transformer for Navier-Stokes Equations#
This example demonstrates how to set up a Denoising Pre-trained Operator Transformer for solving 2D Naiver-Stokes equation in PhysicsNeMo.
Note: This example runs on a single GPU. Multi-GPU training and pre-trained weights loading are in development.
Note: This example is a work in progress and will be updated in the next release.
Prerequisites#
Install the required dependencies by running below:
pip install -r requirements.txt
Getting Started#
To train the model, run
python train_dpot.py
training data will be generated on the fly.
Set train to False in config file to inference the model.
TODO#
Add utility for plotting the inference results.
Add support for multi-GPU training.
Better handling of the dataset download.
More comprehensive unit tests.