search

manojpamk / pytorch_xvectors

Deep speaker embeddings in PyTorch, including x-vectors. Code used in this work: https://arxiv.org/abs/2007.16196
MIT License
304 stars 65 forks source link

tdnn layers #16

Open groadabike opened 3 years ago

groadabike commented 3 years ago

Hi, First of all, very nice work. well done. I want to ask you about your xvector implementation. I am not sure if I have a misunderstanding of how The Conv1D with dilation > 1 works.

Why did you use kernel_size=5 for the second and kernel_size=7 for the third TDNN layers? I would use kernel_size=3 for both of them with the same dilation you used.

Thank you Gerardo

clockw commented 3 years ago

Same question here. Looking forward to your answer. Thank you!