Datasetpublic

camlab-ethz/NS-SVS

Short Description Based on the incompressible Navier-Stokes equations, this dataset contains trajectories starting from sinusoidal vortex sheet initial conditions, see https://arxiv.org/abs/2405.19101. It also contains a passive tracer that is carried by the flow. Dimensions The assembled NetCDF file has a single variable called velocity with dimensionality 20000 (number of trajectories) 21 (time steps) 3 (horizontal velocity, vertical velocity, passive tracer)… See the full description on the dataset page: https://huggingface.co/datasets/camlab-ethz/NS-SVS.

sourceHugging Facecc-by-nc-4.0updated 2y agoView on Hugging Face

1likes601downloads

Dataset Card

Short Description

Based on the incompressible Navier-Stokes equations, this dataset contains trajectories starting from sinusoidal vortex sheet initial conditions, see https://arxiv.org/abs/2405.19101. It also contains a passive tracer that is carried by the flow.

Dimensions

The assembled NetCDF file has a single variable called velocity with dimensionality

—20000 (number of trajectories)
—21 (time steps)
—3 (horizontal velocity, vertical velocity, passive tracer)
—128 (x-dim)
—128 (y-dim)

It was simulated on the unit square up to T=1 and saved as uniformly spaced in space and time.

Train/Val/Test-split

19640/120/240 trajectories

Download & Assembly

The dataset can be downloaded, e.g., via huggingface-cli download.

After download, the chunked data can be assembled into a single NetCDF file using the provided assemble_data.py script. Use it as follows:

bash

python assemble_data.py --input_dir . --output_file NS-SVS.nc