Procedures

Accumulate corner gradients for 2D padding - raw array version

Accumulate corner gradients for 3D padding - raw array version

Accumulate edge gradients for 1D padding - raw array version

Accumulate edge gradients for 2D padding - raw array version

Accumulate edge gradients for 3D padding - raw array version

Accumulate face gradients for 3D padding - raw array version

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Setup the desired activation function

Interface for the activation layer type

Interface for setting up the Adagrad optimiser

Interface for setting up the Adam optimiser

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Add two 3D convolutional layers without whole-object allocatable copy

Create one GNN or NOP layer from metadata and append it to the network.

Add one graph input tensor declaration to the ONNX input list.

Interface for setting up the add layer

Create standard (non-GNN) layers for a given layer_id using the registered ONNX creator framework (list_of_onnx_layer_creators).

Interface for allocating and assigning a vector to a variable

Allocate and assign an arbitrary length vector of reals to variable

Allocate and populate the list of ONNX activation creation functions

Append one integer value stored as a quoted JSON string.

Append one string element from a multiline JSON string array.

Append a layer id parsed from athena_gnn_node_ or athena_nop_node_ if not already present.

Append a /layerN prefix to a list if it is not already present.

Append a layer id parsed from a primary node name node_.

Append a value to the history of the metric

Load ONNX attributes into Gaussian activation function

Load ONNX attributes into leaky ReLU activation function

Load ONNX attributes into linear activation function

Load ONNX attributes into none activation function

Load ONNX attributes into piecewise activation function

Load ONNX attributes into ReLU activation function

Load ONNX attributes into SELU activation function

Load ONNX attributes into sigmoid activation function

Load ONNX attributes into softmax activation function

Load ONNX attributes into swish activation function

Load ONNX attributes into tanh activation function

Function to apply clipping to gradients

Apply Gaussian activation to array

Apply linear activation to 1D array

Apply identity activation to 1D array

Apply piecewise activation to 1D array

Apply ReLU activation to 1D array

Apply SELU activation to array

Apply sigmoid activation to 1D array

Apply softmax activation to 1D array

Apply swish activation to 1D array

Apply tanh activation to 1D array

Interface for assigning a value to a variable

Interface for assigning a vector to a variable

Assign an integer to variable

Assign an arbitrary length vector of integers to variable

Assign a logical to variable (T/t/1 and F/f/0 accepted)

Assign a real to variable

Assign an arbitrary length vector of reals to variable

Assign a string to variable

Interface for setting up the 1D average pooling layer

Interface for setting up the 2D average pooling layer

Interface for setting up the 3D average pooling layer

Core base64 decoder

Core base64 encoder (allocatable output)

Core base64 encoder (fixed-length output)

Interface for base learning rate decay type

Interface for setting up the base optimiser

Interface for setting up the 1D batch normalisation layer

Interface for setting up the 2D batch normalisation layer

Interface for setting up the 3D batch normalisation layer

Interface for binary cross entropy loss function

Build JSON string for layer attributes

Build a Duvenaud layer from an expanded-ONNX cluster.

Build one dynamic LNO layer from an expanded-ONNX node cluster.

Build one fixed LNO layer from an expanded-ONNX node cluster.

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Read ONNX attributes for activation layer

Read ONNX attributes for 1D average pooling layer

Read ONNX attributes for 2D average pooling layer

Read ONNX attributes for 3D average pooling layer

Read ONNX attributes for 1D batch normalisation layer

Read ONNX attributes for 2D batch normalisation layer

Read ONNX attributes for 3D batch normalisation layer

Read ONNX attributes for 1D convolutional layer

Read ONNX attributes for 2D convolutional layer

Read ONNX attributes for 3D convolutional layer

Read ONNX attributes for 2D dropblock layer

Read ONNX attributes for 3D dropblock layer

Read ONNX attributes for dropout layer

Read ONNX attributes for flattening layer

Read ONNX attributes for fully connected layer

Read ONNX attributes for fully connected layer

Read ONNX attributes for 1D max pooling layer

Read ONNX attributes for 2D max pooling layer

Read ONNX attributes for 3D max pooling layer

Read ONNX attributes for 1D padding layer

Read ONNX attributes for 2D padding layer

Read ONNX attributes for 3D padding layer

Build reshape layer from ONNX node and initialiser

Build the metadata entry required to reconstruct a GNN layer.

Build the ONNX graph input tensor specifications.

Build the ONNX graph output tensor specifications.

Build a Kipf GCN layer from an expanded-ONNX cluster.

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Build a network containing GNN layers from parsed JSON data.

Reconstruct ATHENA GNN layers from an expanded-ONNX graph when metadata is absent.

Reconstruct ATHENA NOP layers from an expanded-ONNX decomposed graph.

Build a standard, non-GNN network from parsed JSON data.

Build one neural operator layer from an expanded-ONNX node cluster.

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Build one spectral filter layer from an expanded-ONNX node cluster.

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Calculate input shape based on shapes of input layers

Calculate input shape based on shapes of input layers

Compute the categorical accuracy of a model

Interface for categorical cross entropy loss function

Return true when the expanded-ONNX node cluster for prefix is a Duvenaud message-passing layer.

Return true when the expanded-ONNX node cluster for prefix is a dynamic LNO.

Return true when the expanded-ONNX node cluster for prefix is a fixed LNO.

Return true when the expanded-ONNX node cluster for prefix is a Kipf GCN layer.

Return true when the expanded-ONNX node cluster for prefix is a neural operator.

Return true when the expanded-ONNX node cluster for prefix is a spectral filter.

Set up the clip dictionary

Interface for the clip type

Convert flat column-major [m,n] to flat row-major [m,n] Fortran stores arrays column-major; ONNX rawData expects row-major.

Build the ONNX nodes, initialisers and GNN metadata.

Forward propagation for 2D input

Forward propagation for 2D input

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Placeholder for compute function in base_loss_type

Compute the binary cross entropy loss of a model

Compute the categorical cross entropy loss of a model

Compute the huber loss of a model

Compute the mean absolute error of a model

Compute the mean squared error of a model

Compute the negative log likelihood of a model

Interface for setting up the concatenate layer

Interface for setting up the 1D convolutional layer

Interface for setting up the 2D convolutional layer

Interface for setting up the 3D convolutional layer

Build activation layer from attributes and return layer

Build avgpool layer from attributes and return layer

Build batchnorm layer from attributes and return layer

Build conv layer from attributes and return layer

Build dropout layer from attributes and return layer

Build Duvenaud message-passing layer from ONNX metadata and return layer

Build dynamic LNO layer from ONNX metadata and return layer

Build fixed LNO layer from ONNX metadata and return layer

Build flattening layer from attributes and return layer

Build fully connected layer from attributes and return layer

Create Gaussian activation function from ONNX attributes

Build fully connected layer from attributes and return layer

Build Kipf GCN layer from ONNX metadata and return layer

Create leaky ReLU activation function from ONNX attributes

Create linear activation function from ONNX attributes

Build maxpool layer from attributes and return layer

Build neural_operator layer from ONNX metadata and return layer

Create none activation function from ONNX attributes

Build orthogonal attention layer from ONNX metadata and return layer

Build orthogonal NOP block from ONNX metadata and return layer

Build pad layer from attributes and return layer

Create piecewise activation function from ONNX attributes

Create ReLU activation function from ONNX attributes

Build reshape layer from ONNX node and initialiser

Create SELU activation function from ONNX attributes

Create sigmoid activation function from ONNX attributes

Create softmax activation function from ONNX attributes

Create swish activation function from ONNX attributes

Create tanh activation function from ONNX attributes

Initialise the weights and biases using the data distribution

Decode base64 string to float32 array

Decode base64 string to integer array (from 8-byte int64 encoding)

Detect the activation op used in a GNN layer cluster.

Detect the active top-level graph section.

Reconstruct the activation name from the tail of an expanded-ONNX NOP cluster.

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Interface for setting up the 2D dropblock layer

Interface for setting up the 3D dropblock layer

Interface for setting up the dropout layer

Interface for setting up the MPNN layer

Emit an activation function node

Emit a Constant node producing a float32 tensor

Emit a Constant node producing an int64 tensor

Emit a ConstantOfShape node

Emit the degree-dependent weight selection and update block.

Emit ONNX nodes for Duvenaud readout

Emit one Duvenaud readout timestep.

Emit ONNX nodes for one Duvenaud message passing time step.

Gather one edge_index row and squeeze it into a vector.

Emit a float32 initialiser with explicit dimensions.

Emit Identity nodes that rename GNN inputs to the expected convention.

Emit initialisers for a learnable layer

Emit ONNX nodes for one Kipf GCN time step.

Emit a 2D float32 initialiser after converting to row-major order.

Emit the standard graph input tensors used by message-passing layers.

Emit a simple ONNX node (individual string interface) Avoids gfortran array constructor issues

Emit the common NOP input transpose.

Build the metadata entry required to reconstruct a NOP layer.

Emit the common transpose and optional activation at the end of a NOP.

Emit graph input tensor declarations for Duvenaud GNN layer

Emit ONNX JSON nodes for Duvenaud GNN layer

Emit decomposed standard ONNX nodes for a Dynamic LNO layer.

Emit decomposed standard ONNX nodes for a Fixed LNO layer.

Emit ONNX JSON nodes for Kipf GCN layer

Emit decomposed standard ONNX nodes for a Neural Operator layer.

Emit decomposed standard ONNX nodes for a Spectral Filter layer.

Emit a final Identity node using the standard ATHENA output naming.

Emit the zero-initialise, expand, and scatter-add aggregation block.

Emit a Squeeze node (ONNX opset 13+: axes as input)

Emit ONNX node records for a standard, non-GNN layer.

Store a 2D weight matrix as an ONNX initialiser in row-major order.

Store a stacked bank of 2D weight matrices as one ONNX tensor.

Encode float32 array as base64 string (fixed-length output)

Encode float32 array as base64 string (allocatable output)

Encode integer array as base64 int64 string (fixed-length output)

Encode integer array as base64 int64 string (allocatable output)

Encode a string as base64

Interface for exponential learning rate decay type

Export Gaussian activation function attributes as ONNX attributes

Export leaky ReLU activation function attributes as ONNX attributes

Export linear activation function attributes as ONNX attributes

Export none activation function attributes as ONNX attributes

Export piecewise activation function attributes as ONNX attributes

Export ReLU activation function attributes as ONNX attributes

Export SELU activation function attributes as ONNX attributes

Export sigmoid activation function attributes as ONNX attributes

Export softmax activation function attributes as ONNX attributes

Export swish activation function attributes as ONNX attributes

Export tanh activation function attributes as ONNX attributes

Extract the subtype=... token from one metadata value string.

Extract an integer value from a JSON key-value pair.

Extract a string value from a JSON key-value pair.

Extract the node prefix from an athena_nop_node_X metadata key.

Extract the layerN prefix from an expanded-ONNX node name.

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Fill corner region for 2D padding

Fill corner region for 3D padding

Fill edge region for 1D padding

Fill edge region for 2D padding

Fill edge region for 3D padding

Fill face region for 3D padding

Finalise 1D batch normalisation layer

Finalise 2D batch normalisation layer

Finalise 3D batch normalisation layer

Finalise 1D convolutional layer

finalise layer ! finalise layer Finalise 2D convolutional layer

Finalise 3D convolutional layer

Finalise the message passing layer

Finalise the dynamic Laplace neural operator layer

Finalise the fixed-basis Laplace neural operator layer

Finalise fully connected layer

Finalise neural operator layer

Finalise the orthogonal neural operator block

Finalise the orthogonal attention layer

Finalise the spectral filter layer

Return node index for activation attached to node_, or 0.

Return the index of a node with exact name match, or zero.

Return the index of a named initialiser, or zero when not found.

Return metadata index for a given layer id, or 0 if absent.

Return the first initialiser referenced by a node's inputs.

Return the node index matching one /layerN/suffix name, or zero.

Return node index for primary node_, or 0 if not found.

Interface for setting up the flattening layer

Format a training metric with a configurable number of decimal places.

Forward propagation

Forward propagation

Forward propagation

Forward propagation

Forward propagation

Forward propagation

Forward propagation

Forward propagation

Forward propagation

Forward propagation

Forward propagation

Forward propagation

Forward propagation

Forward propagation for the Laplace Neural Operator layer

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Forward propagation for the Laplace Neural Operator layer

Forward propagation

Forward propagation

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Forward propagation for an input layer

Forward propagation

Forward propagation

Forward propagation

Forward propagation for the neural operator layer

Forward propagation for the Orthogonal Neural Operator layer

Forward propagation for the Orthogonal Attention layer

Forward propagation

Forward propagation

Forward propagation

Forward propagation

Forward propagation derived type handler

Forward propagation for the spectral filter layer

Interface for setting up the fully connected layer

Initialise the weights and biases using the Gaussian distribution

Generate Bernoulli mask

Generate Bernoulli mask Apply threshold to create binary mask

Generate dropout mask

Get the attributes of the Duvenaud message passing layer (for ONNX export)

Return list of dynamic LNO attributes for ONNX export

Return list of fixed LNO attributes for ONNX export

Get the attributes of the Kipf GCN layer (for ONNX export)

Return list of neural operator attributes for ONNX export

Return list of ONO attributes for ONNX export

Return list of orthogonal attention attributes for ONNX export

Rebuild the dynamic Laplace encoder/decoder bases from the current learnable pole values (params(1)).

Orthogonalise the basis matrix B using modified Gram-Schmidt

Orthogonalise the basis matrix B using modified Gram-Schmidt

Get the default initialiser based on the activation function

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Get the number of parameters for the message passing layer

Return the number of learnable parameters for the layer

Return the number of learnable parameters for the layer

Get the number of parameters for fully connected layer

Get the number of learnable parameters

Get the number of parameters for the message passing layer

Interfaces for handling learnable parameters and gradients

Get the number of parameters for the neural operator layer

Return the number of learnable parameters for the block

Return the number of learnable parameters for the layer

Return the number of learnable parameters for the layer

Compute max(|Phi^T @ Phi - I|) as a measure of basis orthogonality

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Function to return half the padding width

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Get partial derivative with respect to left operand

Get partial derivative with respect to bias operand

Get partial derivative with respect to left operand

Get the partial derivative for average pooling

Optimised backward pass for 1D average pooling

Get the partial derivative for average pooling

Optimised backward pass for 2D average pooling

Get the partial derivative for 3D average pooling

Optimised backward pass for 3D average pooling

Get partial derivative wrt input for batchnorm (subroutine version)

Get partial derivative wrt params for batchnorm (subroutine version)

Get partial derivative wrt input for 1D convolution

Get partial derivative wrt input for 1D convolution (subroutine version)

Get partial derivative wrt kernel for 1D convolution

Get partial derivative wrt kernel for 1D convolution (subroutine version)

Get partial derivative wrt input for 2D convolution

Get partial derivative wrt input for 2D convolution (subroutine version)

Get partial derivative wrt kernel for 2D convolution

Get partial derivative wrt kernel for 2D convolution (subroutine version)

Get partial derivative wrt input for 3D convolution

Get partial derivative wrt input for 3D convolution (subroutine version)

Get partial derivative wrt kernel for 3D convolution

Get partial derivative wrt kernel for 3D convolution (subroutine version)

Gradient of duvenaud_propagate with respect to vertex_features.

In-place value gradient for duvenaud_propagate left operand.

Gradient of duvenaud_propagate with respect to edge_features.

In-place value gradient for duvenaud_propagate right operand.

Gradient of duvenaud_update with respect to input features.

In-place value gradient for duvenaud_update input features.

Gradient of duvenaud_update with respect to packed weights.

In-place value gradient for duvenaud_update packed weights.

d(out)/d(input): upstream * scale (broadcast scale along samples)

d(out)/d(scale): upstream * input (element-wise, per sample)

Gradient of gno_aggregate w.r.t. features (left operand)

In-place gradient w.r.t. features

Gradient of gno_aggregate w.r.t. edge_kernels (right operand)

In-place gradient w.r.t. edge_kernels

Gradient of gno_kernel_eval w.r.t. edge features (left operand)

In-place gradient w.r.t. edge features

Gradient of gno_kernel_eval w.r.t. kernel_params (right operand)

In-place gradient w.r.t. packed kernel params

Get partial derivative of huber loss

Get partial derivative of huber loss (in-place version)

Gradient of kipf_propagate with respect to vertex features.

In-place value gradient for kipf_propagate left operand.

Gradient of lno_decode with respect to poles.

dL/dmu_m per sample: output[m,s] = sum_i upstream[i,s](-tau_i)exp(-mu_mtau_i)x[m,s]

Gradient of lno_decode with respect to spectral input.

dL/dx = D^T @ upstream [M, batch]

Gradient of lno_encode with respect to input.

dL/du = E^T @ upstream [n_in, batch]

Gradient of lno_encode with respect to poles.

dL/dmu_m per sample: output[m,s] = upstream[m,s] * sum_j (-t_j) * exp(-mu_m*t_j) * u[j,s]

Get the partial derivative for max pooling

Optimised backward pass for 1D max pooling

Get the partial derivative for max pooling

Get the partial derivative for 3D max pooling

Optimised backward pass for 3D max pooling

Gradient of ono_decode with respect to basis weights.

dL/dB per sample through Gram-Schmidt backward.

Gradient of ono_decode with respect to mixed input.

dL/dx = Q^T @ upstream [k, batch]

Gradient of ono_encode with respect to basis weights.

dL/dB per sample through Gram-Schmidt backward.

Gradient of ono_encode with respect to input.

dL/du = Q @ upstream [n, batch]

Get the partial derivative for the pad1d operation

Get the partial derivative for the pad1d operation - raw array version

Get the partial derivative for the pad2d operation

Get the partial derivative for the pad2d operation - raw array version

Get the partial derivative for the pad3d operation

Get the partial derivative for the pad3d operation - raw array version

Get partial derivative of piecewise activation

Get partial derivative of piecewise activation (in-place version)

Get partial derivative of softmax activation

Get partial derivative of softmax reverse operation

Get partial derivative of softmax reverse operation (in-place version)

Get partial derivative of softmax reverse operation

Get partial derivative of softmax reverse operation (in-place version)

Get partial derivative of softmax activation (in-place version)

Get partial derivative of softmax activation (in-place version, summed over samples)

Get partial derivative of swish activation

Get partial derivative of swish activation (in-place version)

Build the canonical ONNX output name for one exported timestep.

Extract the section of buffer that occurs after the field separator fs

Get the line from a grep and assign it to buffer

Initialise the weights and biases using the Glorot normal distribution

Initialise the weights and biases using the Glorot uniform distribution

Initialise the weights and biases using the He normal distribution

Initialise the weights and biases using the He uniform distribution

Interface for huber loss function

Initialise the weights and biases using the identity matrix

Reconstruct dynamic LNO poles from exported encoder/decoder arguments.

Initialise activation layer

Initialise add layer

Initialise base optimiser

Initialise concatenate layer

Initialise 2D dropblock layer

Initialise 3D dropblock layer

Initialise dropout layer

Initialise the message passing layer

Initialise parameter storage and output buffers for the layer

Initialise parameter storage, fixed bases and output buffers

Initialise flattening layer

Initialise fully connected layer

Initialise the Graph Neural Operator layer

Initialise gradients for Adagrad optimiser

Initialise gradients for Adam optimiser

Initialise gradients for base optimiser

Initialise gradients for RMSprop optimiser

Initialise gradients for SGD optimiser

Initialise an input layer

Initialise the message passing layer

Interfaces for handling graphs and outputs, and initialising the layer

Initialise neural operator layer

Initialise parameter storage and output buffers for the block

Initialise parameter storage and output buffers for the layer

Initialise the recurrent layer

Initialise reshape layer

Initialise parameter storage, fixed bases and output buffers

Initialise a ReLU activation function

Initialise a sigmoid activation function

Initialise a tanh activation function

Initialise a leaky ReLU activation function

Initialise a piecewise activation function

Initialise a none (no-op) activation function

Initialise a softmax activation function

Initialise a SELU activation function

Initialise a swish activation function

Initialise a Gaussian activation function

Initialise a linear activation function

Allocate the working arrays used during ONNX export.

Initialise the reusable parser state objects.

Interface for the data initialiser

Interface for the Gaussian initialiser

Interface for the Identity initialiser

Interface for the LeCun normal initialiser

Interface for the LeCun uniform initialiser

Interface for the Ones initialiser

Set up the initialiser function

Interface for the Zeros initialiser

Interface for an input layer

Interface for inverse learning rate decay type

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Return true for ONNX activation nodes emitted by ATHENA export.

Return true for section object lines like {.

Return true when the parsed ONNX graph contains expanded-ONNX GNN patterns that ATHENA can collapse back into native message passing layers.

Return true when the parsed ONNX graph is a supported expanded-ONNX NOP decomposition that ATHENA can collapse back into native NOP layers.

Interface for setting up the MPNN layer

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Set up the 2D convolutional layer

Set up the message passing layer

Set up the 2D average pooling layer

Set up the 3D average pooling layer

Set up the dropout layer

Set up the 1D batch normalisation layer

Set up the 1D convolutional layer

Setup a concatenate layer

Set up layer

Set up the 3D max pooling layer

Set up the 3D convolutional layer

Set up the activation layer

Set up the 1D padding layer

Set up the reshape layer

Set up the 2D padding layer

Setup a fully connected layer

Set up the 2D dropblock layer

Setup a recurrent layer

Set up the 1D average pooling layer

Setup a add layer

Set up the 1D max pooling layer

Setup a neural operator layer

Set up the 3D batch normalisation layer

Set up the 3D padding layer

Set up the 3D dropblock layer

Set up the message passing layer

Set up the 2D batch normalisation layer

!############################################################################# set up layer ! set up layer

Set up the flattening layer

Apply leaky ReLU activation to 1D array

Reset leaky ReLU activation function attributes and variables

Initialise the weights and biases using the LeCun normal distribution

Initialise the weights and biases using the LeCun uniform distribution

Renormalise input data to a specified range

Load a parameter from ONNX initialisers into a diffstruc array.

Copy a row-major ONNX matrix initialiser into a diffstruc parameter.

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Get the learning rate for the exponential decay type

Get the learning rate for the inverse decay type

Get the learning rate for the base decay type

Get the learning rate for the step decay type

Interface for mean absolute error loss function

Compute the mean absolute error of a model

Interface for setting up the 1D max pooling layer

Interface for setting up the 2D max pooling layer

Interface for setting up the 3D max pooling layer

Operation to add two metric_dict_type together

Allocate memory for a metric_dict_type

Check if the metric has converged

Apply gradients to parameters to minimise loss using Adagrad optimiser

Apply gradients to parameters to minimise loss using Adam optimiser

Apply gradients to parameters to minimise loss using base optimiser

Apply gradients to parameters to minimise loss using RMSprop optimiser

Apply gradients to parameters to minimise loss using SGD optimiser Adaptive learning method

Move current position in file based on relative change

Interface for mean squared error loss function

Compute the mean squared error of a model

Interface for setting up the MPNN layer

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Interface for setting up the network (network initialisation)

Interface for setting up the neural operator layer

Interface for negative log likelihood loss function

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Initialise the weights and biases using the Ones distribution

Convert an ONNX activation op_type string to the Athena activation name

Set up the Adagrad optimiser

Set up the Adam optimiser

Set up the base optimiser

Set up the RMSprop optimiser

Set up the SGD optimiser

Interface for setting up the 1D padding layer

Interface for setting up the 2D padding layer

Interface for setting up the 3D padding layer

Pad data for convolutional layers

Parse layer id from node_X or node_X_* names.

Parse one line from the initialiser section.

Parse one or more JSON attribute objects from a line.

Parse a JSON array of string-encoded integers.

Parse a JSON string array from one line.

Parse athena_gnn_node_ or athena_nop_node_ metadata key layer id.

Parse one metadataProps line.

Parse one line from the node section.

Parse common NOP hyperparameters from metadata value string.

Parse node_ names and mark true only for primary layer nodes.

Parse a single JSON attribute object.

Parse space-separated integers from a string into an allocatable array

Parse one line from the input or output tensor section.

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Print the build information of the program.

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Print activation layer to unit

Print add layer to unit

Print base optimiser information

Print concatenate layer to unit

Print 2D dropblock layer to unit

Print 3D dropblock layer to unit

Print dropout layer to unit

Print kipf message passing layer to unit

Print dynamic LNO settings and parameters to a unit

Print fixed LNO settings and parameters to a unit

Print flatten layer to unit

Print fully connected layer to unit

Print graph neural operator settings and parameters to a unit

Print input layer to unit

Print kipf message passing layer to unit

Print neural operator layer to unit

Print orthogonal neural operator settings and parameters to a unit

Print orthogonal attention layer settings and parameters to a unit

Print recurrent layer to unit

Print reshape layer to unit

Print spectral filter settings and parameters to a unit

Print the version number of the program.

Compute the R^2 score of a model

Initialise the random number generator no need to initialise if already initialised

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Read activation function from input file

Read activation layer from file don't look for "e" due to scientific notation of numbers ... i.e. exponent (E+00)

Read activation layer from file

Read add layer from file

Read add layer from file and return layer

Read 1D average pooling layer from file

Read 1D average pooling layer from file and return layer

Read 2D average pooling layer from file

Read 2D average pooling layer from file and return layer

Read 3D average pooling layer from file

Read 3D average pooling layer from file and return layer

Read base optimiser information

Read 1D batch normalisation layer from file

Read 1D batch normalisation layer from file and return layer

Read 2D batch normalisation layer from file

Read 3D batch normalisation layer from file

Read 3D batch normalisation layer from file and return layer

Read clipping information

Read concatenate layer from file

Read concatenate layer from file and return layer

Read 1D convolutional layer from file

Read 1D convolutional layer from file and return layer

Read 2D convolutional layer from file

Read 2D convolutional layer from file and return layer

Read 3D convolutional layer from file

Read 3D convolutional layer from file and return layer

Read 2D dropblock layer from file

Read 2D dropblock layer from file and return layer

Read 3D dropblock layer from file

Read 3D dropblock layer from file and return layer

Read dropout layer from file

Read dropout layer from file and return layer

Read the message passing layer

Read duvenaud message passing layer from file and return layer

Read a dynamic LNO layer from file and return it

Read a fixed LNO layer from file and return it

Read flattening layer from file

Read flattening layer from file and return layer

Read fully connected layer from file

Read fully connected layer from file and return layer

Read a graph NOP layer from file and return

Read an input layer from a file

Read an input layer from a file

Read the message passing layer

Read kipf message passing layer from file and return layer

Read 1D max pooling layer from file

Read 1D max pooling layer from file and return layer

Read 2D max pooling layer from file

Read 2D max pooling layer from file and return layer

Read 3D max pooling layer from file

Read 3D max pooling layer from file and return layer

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Read neural operator layer from file

Read neural operator layer from file and return as base_layer_type

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Read an orthogonal attention layer from file and return it

Read an orthogonal neural operator block from file and return it

Read 1D padding layer from file

Read 1D padding layer from file and return layer

Read 2D padding layer from file

Read 2D padding layer from file and return layer

Read 3D padding layer from file

Read 3D padding layer from file and return layer

Read recurrent layer from file

Read recurrent layer from file and return layer

Read reshape layer from file

Read reshape layer from file

Read a spectral filter layer from file and return it

Merge two 3D convolutional layers via parameter summation

Regularise the parameters using L1 regularisation

Regularise the parameters using L1 and L2 regularisation

Regularise the parameters using L2 regularisation

Renormalise input data to a unit norm

Renormalise input data to a unit sum

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Reset Gaussian activation function attributes and variables

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Reset the reusable initialiser parser state.

Reset linear activation function attributes and variables

Reset the reusable node parser state.

Reset none activation function attributes and variables

Reset piecewise activation function attributes and variables

Reset ReLU activation function attributes and variables

Reset SELU activation function attributes and variables

Reset sigmoid activation function attributes and variables

Reset softmax activation function attributes and variables

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Reset the hidden state of the recurrent layer

Reset swish activation function attributes and variables

Reset tanh activation function attributes and variables

Reset the reusable tensor parser state.

Interface for setting up the reshape layer

Resolve the ONNX export format into the internal integer selector.

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Reverse propagate values from one autodiff array to another

Remove comment from a string (anything after ! or #)

Compute the root mean squared error of a model

Interface for setting up the RMSprop optimiser

Convert flat row-major [m,n] to flat column-major [m,n] Inverse of col_to_row_major_2d.

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Set clipping information

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Set the graph structure of the input data

Interfaces for handling learnable parameters and gradients

Set hyperparameters for activation layer

Set the hyperparameters for add layer

Set hyperparameters for 1D average pooling layer

Set hyperparameters for 2D average pooling layer

Set hyperparameters for 3D average pooling layer

Set hyperparameters for 1D batch normalisation layer

Set hyperparameters for 2D batch normalisation layer

Set hyperparameters for 3D batch normalisation layer

Set the hyperparameters for concatenate layer

Set hyperparameters for 1D convolutional layer

Set hyperparameters for 2D convolutional layer

Set hyperparameters for 3D convolutional layer

Set hyperparameters for 2D dropblock layer

Set hyperparameters for 3D dropblock layer

Set hyperparameters for dropout layer

Set the hyperparameters for the message passing layer

Set hyperparameters for flattening layer

Set the hyperparameters for fully connected layer

Set hyperparameters for an input layer

Set the hyperparameters for the message passing layer

Set hyperparameters for 1D max pooling layer

Set hyperparameters for 2D max pooling layer

Set hyperparameters for 3D max pooling layer

Set the hyperparameters for the neural operator layer

Set hyperparameters for 1D padding layer

Set hyperparameters for 2D padding layer

Set hyperparameters for 3D padding layer

Set the hyperparameters for fully connected layer

Set hyperparameters for reshape layer

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Set input values for an input layer

Set input values for an input layer

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Set padding for convolutional layers

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Set the input and output ranks of the activation layer

Set the input and output ranks of the layer

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Interface for setting up bounds

Set up binary cross entropy loss function

Set up categorical cross entropy loss function

Set up huber loss function

Set up mean absolute error loss function

Set up mean squared error loss function

Set up negative log likelihood loss function

Set up base learning rate decay type

Set up exponential learning rate decay type

Set up inverse learning rate decay type

Set up step learning rate decay type

Interface for setting up the SGD optimiser

Shuffle an array along one dimension

Shuffle a 1D array along one dimension

Shuffle a 2D array along one dimension

Shuffle a 3D array along one dimension

Shuffle a 3D array along one dimension

Shuffle a 3D array along one dimension

Shuffle a 3D array along one dimension

Shuffle a 4D array along one dimension

Shuffle a 4D array along one dimension

Shuffle a 5D array along one dimension

Shuffle a 5D array along one dimension

Shuffle a 5D array along one dimension

Softmax function for reverse mode autodiff

Sort an integer array in ascending order.

Split an array into train and test sets

Split a 2D array along one dimension

Split a 3D array along one dimension

Split a 5D array along one dimension

Split a 5D array along one dimension

Interface for step learning rate decay type

Logical check for stop file Check if file exists Read line-by-line

Copy the current initialiser state into the parsed result collection.

Copy the current node state into the parsed result collection.

Copy the current tensor state into the parsed result collection.

Interface for setting up swish activation function

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Interface for printing the network to file Interface for printing a summary of the network Interface for reading the network from a file Interface for reading network settings from a file Interface for reading optimiser settings from a file Interface for building network from ONNX nodes and initialisers Interface for adding a layer to the network Interface for resetting the network Interface for compiling the network Interface for setting batch size Interface for setting network metrics Interface for setting network loss method Interface for setting network accuracy method Interface for resetting state of recurrent layers Interface for saving input to network Interface for saving output to network Interface for training the network Interface for testing the network Interface for returning predicted results from supplied inputs using the trained network Interface for returning predicted results from supplied inputs using the trained network (graph input) Interface for updating the learnable parameters of the network based on gradients Interface for generating vertex order Interface for depth first search Interface for calculating root vertices Interface for calculating output vertices Interface for reducing two networks down to one (i.e. add two networks - parallel) Interface for copying a network Interface for getting number of learnable parameters in the network Interface for getting learnable parameters Interface for setting learnable parameters Interface for getting gradients of learnable parameters Interface for setting learnable parameter gradients Interface for resetting learnable parameter gradients Interface for forward pass

Update square-bracket nesting depth for multiline JSON arrays.

Update the message

Update message for the Graph Neural Operator layer

Update the message

interface for the message forward and backward passes

Update a nested object depth counter from one JSON line.

Resolve the downstream tensor name after emitting one PyTorch-format NOP.

Update the readout

No graph-level readout needed — GNO produces node-level output

Update the readout (empty for node-level output)

interface for the readout forward and backward passes

Called automatically at the end of each training epoch. Logs "loss" and "accuracy" to the current W&B run.

Initialise a Weights & Biases run for this network.

Write initialisers array to JSON with base64-encoded rawData

Write nodes array to JSON

Write input/output tensor specifications to JSON

Write the collected export data to disk.

Initialise the weights and biases using the Zeros distribution