espnet2.gan_codec.shared.discriminator.msstft_discriminator.get_2d_padding

source

espnet2.gan_codec.shared.discriminator.msstft_discriminator.get_2d_padding(kernel_size: Tuple[int, int], dilation: Tuple[int, int] = (1, 1))

Calculate the 2D padding required for a convolutional layer given the kernel

size and dilation.

The padding is computed to ensure that the output dimensions remain the same as the input dimensions when using the specified kernel size and dilation in a convolution operation.

• Parameters:
  • kernel_size (tuple of int) – The size of the convolution kernel in the format (height, width).
  • dilation (tuple of int , optional) – The dilation rate for the kernel in the format (dilation_height, dilation_width). Defaults to (1, 1).
• Returns: A tuple containing the calculated padding values in the format : (padding_height, padding_width).
• Return type: tuple

Examples

>>> get_2d_padding((3, 3))
(1, 1)

>>> get_2d_padding((5, 5), dilation=(2, 2))
(4, 4)

NOTE

The padding values are computed using the formula: padding = ((kernel_size - 1) * dilation) // 2