espnet2.asr.encoder.avhubert_encoder.conv3x3

source

espnet2.asr.encoder.avhubert_encoder.conv3x3(in_planes, out_planes, stride=1)

Create a 3x3 convolutional layer with padding.

This function returns a 2D convolution layer that uses a kernel size of 3x3 and includes padding to maintain the spatial dimensions of the input. The layer does not include a bias term.

• Parameters:
  • in_planes (int) – Number of input channels.
  • out_planes (int) – Number of output channels.
  • stride (int , optional) – Stride of the convolution. Default is 1.
• Returns: A 2D convolution layer configured with the specified parameters.
• Return type: nn.Conv2d

Examples

>>> conv_layer = conv3x3(16, 32)
>>> input_tensor = torch.randn(1, 16, 64, 64)  # (N, C, H, W)
>>> output_tensor = conv_layer(input_tensor)
>>> output_tensor.shape
torch.Size([1, 32, 64, 64])  # Output shape remains the same as input