PyTensorContractionLayerModel

Full Name

neuroptimiser.core.models.PyTensorContractionLayerModel

Description

class PyTensorContractionLayerModel

Bases: PyLoihiProcessModel

Tensor contraction layer model for Loihi-based perturbation-based nheuristics

This model implements a tensor contraction layer that performs a contraction operation on the input spikes and weights, producing an output activation matrix.

See also

neuroptimiser.core.processes.TensorContractionLayer

Process that implements a tensor contraction layer for perturbation-based nheuristics.

__init__(proc_params)

Initialises the tensor contraction layer model with the given parameters.

Parameters:

proc_params (dict) – A dictionary containing additional parameters for the process model.

a_out: PyOutPort = LavaPyType(cls=<class 'lava.magma.core.model.py.ports.PyOutPortVectorDense'>, d_type=<class 'bool'>, precision=1)

implements_process

alias of TensorContractionLayer

implements_protocol

alias of LoihiProtocol

required_resources: ty.List[ty.Type[AbstractResource]] = [<class 'lava.magma.core.resources.CPU'>]

run_spk()

Runs the tensor contraction layer process model.

The process is summarised as follows:

1. Receives the input spikes from the inport.

2. If the weight tensor is not initialised, it sets it to the weight matrix with an additional dimension.

3. Performs the contraction operation using np.einsum to compute the output activation matrix.

4. Sends the output activation matrix to the outport.

s_in: PyInPort = LavaPyType(cls=<class 'lava.magma.core.model.py.ports.PyInPortVectorDense'>, d_type=<class 'bool'>, precision=None)

s_matrix: ndarray = LavaPyType(cls=<class 'numpy.ndarray'>, d_type=<class 'bool'>, precision=None)

weight_matrix: ndarray = LavaPyType(cls=<class 'numpy.ndarray'>, d_type=<class 'float'>, precision=None)