Live Demo — powered by Hugging Face ZeroGPU
Model Details
🧠
Architecture
FNO · 4 layers · 64 modes · 64 channels
📊
Training Data
10,000 trajectories · ν = 1 · T = 100
📐
IC Normalization
‖u₀‖_H² = ‖θ₀‖_H² = ε/2
🎚️
Amplitude ε
trained range 0.021 – 0.476
How It Works
1
Choose ε & generate ICs
Pick the amplitude ε. Gaussian Random Fields are drawn and normalized so that ‖u₀‖_H² = ‖θ₀‖_H² = ε/2.
2
Run the FNO
The operator takes (ω₀, θ₀, ε) and returns the vorticity field at t = 100 in milliseconds.
3
Visualize
All three fields displayed side by side with symmetric per-panel colormaps.
4
Download
Export the raw NumPy arrays as an .npz file for further analysis.