mindscience.models

GraphCast Models

mindscience.models.GraphCast.GraphCastNet

The GraphCast is based on graph neural networks and a novel high-resolution multi-scale mesh representation autoregressive model.

Attention Modules

mindscience.models.transformer.Attention	Attention implementation base class
mindscience.models.transformer.MultiHeadAttention	Multi Head Attention proposed in Attention Is All You Need.
mindscience.models.transformer.TransformerBlock	TransformerBlock comprises a MultiHeadAttention and a FeedForward layer.

Vision Transformer (ViT)

mindscience.models.transformer.VisionTransformer

This module based on VisionTransformer backbone which including encoder, decoding_embedding, decoder and dense layer.

Activation Functions

mindscience.models.layers.get_activation	Gets the activation function.
mindscience.models.layers.activation.SReLU	Sin rectified Linear Unit activation function.

Basic Blocks

mindscience.models.layers.DropPath	Drop paths (Stochastic Depth) per sample (when applied in main path of residual blocks).
mindscience.models.layers.FCSequential	A sequential container of the dense layers, dense layers are added to the container sequentially.
mindscience.models.layers.InputScale	Scale the input value to specified region based on \((x_i - input\_center)*input\_scale\)
mindscience.models.layers.LinearBlock	The LinearBlock.
mindscience.models.layers.MultiScaleFCSequential	The multi-scale fully conneted network.
mindscience.models.layers.ResBlock	The ResBlock of dense layer.

UNet2D

mindscience.models.layers.UNet2D

The 2-dimensional U-Net model.

Mask Layers

mindscience.models.layers.MaskedLayerNorm

Masked layer normalization.

Fourier Neural Operators (FNO)

mindscience.models.neural_operator.FNO1D	The 1D Fourier Neural Operator, which usually contains a Lifting Layer, a Fourier Block Layer and a Projection Layer.
mindscience.models.neural_operator.FNO2D	The 2D Fourier Neural Operator, which usually contains a Lifting Layer, a Fourier Block Layer and a Projection Layer.
mindscience.models.neural_operator.FNO3D	The 3D Fourier Neural Operator, which usually contains a Lifting Layer, a Fourier Block Layer and a Projection Layer.
mindscience.models.neural_operator.FNOBlocks	The FNOBlock, which usually accompanied by a Lifting Layer ahead and a Projection Layer behind, is a part of Fourier Neural Operator.

Fast Fourier Neural Operators (FFNO)

mindscience.models.neural_operator.FFNO	The FFNO base class, which usually contains a Lifting Layer, a Factorized Fourier Block Layer and a Projection Layer.
mindscience.models.neural_operator.FFNO1D	The 1D Factorized Fourier Neural Operator, which usually contains a Lifting Layer, a Factorized Fourier Block Layer and a Projection Layer.
mindscience.models.neural_operator.FFNO2D	The 2D Factorized Fourier Neural Operator, which usually contains a Lifting Layer, a Factorized Fourier Block Layer and a Projection Layer.
mindscience.models.neural_operator.FFNO3D	The 3D Factorized Fourier Neural Operator, which usually contains a Lifting Layer, a Factorized Fourier Block Layer and a Projection Layer.
mindscience.models.neural_operator.FFNOBlocks	The FFNOBlock, which usually accompanied by a Lifting Layer ahead and a Projection Layer behind, is a part of Factorized Fourier Neural Operator.