nnf_conv_tbc function

Conv_tbc

Applies a 1-dimensional sequence convolution over an input sequence. Input and output dimensions are (Time, Batch, Channels) - hence TBC.

nnf_conv_tbc(input, weight, bias, pad = 0)

Arguments

• input: input tensor of shape c("$(\\mbox{sequence length} \\times\n$", "$batch \\times \\mbox{in\\_channels})$")
• weight: filter of shape (c("$\\mbox{kernel width} \\times \\mbox{in\\_channels}\n$", "$\\times \\mbox{out\\_channels}$"))
• bias: bias of shape ($\mbox{out\_channels}$)
• pad: number of timesteps to pad. Default: 0