Method set_batch_size

bindsnet/network/nodes.py:1487–1517

Sets mini-batch size. Called when layer is added to a network. :param batch_size: Mini-batch size.

(self, batch_size)

        )  # Adaptive threshold decay (per timestep).

    def set_batch_size(self, batch_size) -> None:
        # language=rst
        """
        Sets mini-batch size. Called when layer is added to a network.

        :param batch_size: Mini-batch size.
        """
        super().set_batch_size(batch_size=batch_size)
        self.v = self.rest * torch.ones(batch_size, *self.shape, device=self.v.device)
        self.last_spikes = torch.zeros(batch_size, self.ref_window_size, *self.shape)

        resKernels = {
            "AlphaKernel": self.AlphaKernel,
            "AlphaKernelSLAYER": self.AlphaKernelSLAYER,
            "LaplacianKernel": self.LaplacianKernel,
            "ExponentialKernel": self.ExponentialKernel,
            "RectangularKernel": self.RectangularKernel,
            "TriangularKernel": self.TriangularKernel,
        }

        if self.responseKernel not in resKernels.keys():
            raise Exception(" The given response Kernel is not implemented")

        self.resKernel = resKernels[self.responseKernel](self.dt)

        refKernels = {"EtaKernel": self.EtaKernel}

        if self.refractoryKernel not in refKernels.keys():
            raise Exception(" The given refractory Kernel is not implemented")

        self.refKernel = refKernels[self.refractoryKernel](self.dt)

    def AlphaKernel(self, dt):
        t = torch.arange(0, self.res_window_size, dt)