sciai.architecture

Cover functions such as activation functions, basic blocks, multi-task loss, etc.

API Name	Description	Supported Platforms
sciai.architecture.AdaptActivation	Adaptive activation function with trainable Parameter and fixed scale.	GPU CPU Ascend
sciai.architecture.FirstOutputCell	Network that return the first output of given network.	GPU CPU Ascend
sciai.architecture.FNO1D	The 1-dimensional Fourier Neural Operator (FNO1D) contains a lifting layer, multiple Fourier layers and a decoder layer.	Ascend GPU
sciai.architecture.FNO2D	The 2-dimensional Fourier Neural Operator (FNO2D) contains a lifting layer, multiple Fourier layers and a decoder layer.	Ascend GPU
sciai.architecture.FNO3D	The 3-dimensional Fourier Neural Operator (FNO3D) contains a lifting layer, multiple Fourier layers and a decoder layer.	Ascend GPU
sciai.architecture.get_activation	Get the activation function according to its name.	Ascend GPU CPU
sciai.architecture.MLP	Multi-layer perceptron.	Ascend GPU CPU
sciai.architecture.MLPAAF	Multi-layer perceptron with adaptive activation function.	GPU CPU Ascend
sciai.architecture.MLPShortcut	Multi-layer perceptron with shortcuts.	GPU CPU Ascend
sciai.architecture.MSE	Mean square error with 0.	GPU CPU Ascend
sciai.architecture.NoArgNet	Convert a net with inputs into a net without inputs in construct.	GPU CPU Ascend
sciai.architecture.Normalize	Normalize inputs with given lower bound and upper bound.	GPU CPU Ascend
sciai.architecture.PDENet	The PDE-Net model.	Ascend GPU
sciai.architecture.SReLU	Sin rectified Linear Unit activation function.	Ascend
sciai.architecture.SSE	Sum square error with 0.	GPU CPU Ascend
sciai.architecture.Swish	Swish(Silu) activation function with backward propagation.	GPU CPU Ascend
sciai.architecture.ViT	This module based on ViT backbone which including encoder, decoding_embedding, decoder and dense layer.	Ascend GPU