espnet2.svs.singing_tacotron.encoder.encoder_init

source

espnet2.svs.singing_tacotron.encoder.encoder_init(m)

Initialize encoder parameters.

This function initializes the parameters of the encoder module, specifically for convolutional layers. It applies the Xavier uniform initialization to the weights of Conv1d layers to improve convergence during training.

• Parameters:m (torch.nn.Module) – The neural network module to initialize.

NOTE

This function is typically used with the apply method of a PyTorch module to initialize all its sub-modules.

Examples

>>> model = Encoder(idim=256)
>>> model.apply(encoder_init)

• Raises:None –