Skip to content

Available Model Architectures¤

If you would like your architecture added, please submit a pull request.

Standard¤

Architecture Module Description Based on
FNO ResNet with FourierBasicBlock Fourier Neural Operator implementation with support for deeper architectures (8 and 4 layers)
ResNet ResNet with BasicBlock ResNet architectures using 8 residual blocks, no downscaling
DilResNet ResNet with DilatedBasicBlock ResNet where each block individually consists of 7 dilated CNN layers with dilation rates of [1, 2, 4, 8, 4, 2, 1], no downscaling
U-Net2015 Unet2015 Original U-Net implementation
U-Netbase Unetbase Our interpretation of original U-Net implementation without bottleneck layer and using GroupNorm
U-Netmod Unet Modern U-Nets with Wide ResNet blocks, as used in various diffusion modeling applications
U-F[*]Net FourierUnet Modern U-Nets with [*] Fourier blocks in the downsampling path
UNO UNO Original U-shaped Neural Operator Implementation

Conditioned¤

Note

Currently only scalar parameter conditioning is available.

Architecture Model Name Description
FNO ResNet with FourierBasicBlock Addition based conditioning in both spatial and spectral domain.
U-Net-modern Unet Addition and AdaGN style conditioning in the spatial domain.
UF-Net FourierUnet Addition and AdaGN style conditioning in the spatial domain, Addition in the spectral domain.

Model Architecture Registry Philosophy¤

While in principle we can make every architecture fully configurable via configuration files, we find it can affect the readability of the code quite a bit. Feel free to open issues or pull-requests for further configuration ability or any other suggestions for managing the configurability-readability tradeoffs.

Other Projects of Interest¤