espnet2.asr_transducer.utils.get_convinput_module_parameters

source

espnet2.asr_transducer.utils.get_convinput_module_parameters(input_size: int, last_conv_size, subsampling_factor: int, is_vgg: bool = True) → Tuple[Tuple[int, int] | int, int]

Return the convolution module parameters.

This function calculates the parameters for a convolutional module based on the input size, last convolution size, subsampling factor, and whether the architecture follows a VGG-like structure.

• Parameters:
  • input_size – Module input size.
  • last_conv_size – Last convolution size for module output size computation.
  • subsampling_factor – Total subsampling factor.
  • is_vgg – Whether the module type is VGG-like. Defaults to True.
• Returns:
  • First MaxPool2D kernel size or second Conv2D kernel size and stride.
  • output_size: Convolution module output size.
• Return type: A tuple containing

Examples

>>> get_convinput_module_parameters(64, 32, 2, True)
(1, 15)

>>> get_convinput_module_parameters(64, 32, 4, False)
((3, 2), 12)

NOTE

The output size is computed based on the input size and the specified subsampling factor. For VGG-like architectures, the calculation may differ from standard convolutional networks.