pytorch-tdnn

Implementation of Time Delay Neural Network (TDNN) and Factorized TDNN (TDNN-F) in PyTorch, available as layers which can be used directly.

Setup

For using (no development required)

pip install pytorch-tdnn

To install for development, clone the repository, and then run the following from within the root directory.

pip install -e .

Usage

Using the TDNN layer

from pytorch_tdnn.tdnn import TDNN as TDNNLayer

tdnn = TDNNLayer(
  512, # input dim
  512, # output dim
  [-3,0,3], # context
)

y = tdnn(x)

Here, x should have the shape (batch_size, input_dim, sequence_length).

Note: The context list should follow these constraints:

• The length of the list should be 2 or an odd number.
• If the length is 2, it should be of the form [-1,1] or [-3,3], but not [-1,3], for example.
• If the length is an odd number, they should be evenly spaced with a 0 in the middle. For example, [-3,0,3] is allowed, but [-3,-1,0,1,3] is not.

Using the TDNNF layer

from pytorch_tdnn.tdnnf import TDNNF as TDNNFLayer

tdnnf = TDNNFLayer(
  512, # input dim
  512, # output dim
  256, # bottleneck dim
  1, # time stride
)

y = tdnnf(x, semi_ortho_step=True)

The argument semi_ortho_step determines whether to take the step towards semi- orthogonality for the constrained convolutional layers in the 3-stage splicing. If this call is made from within a forward() function of an nn.Module class, it can be set as follows to approximate Kaldi-style training where the step is taken once every 4 iterations:

import random
semi_ortho_step = self.training and (random.uniform(0,1) < 0.25)

Note: Time stride should be greater than or equal to 0. For example, if the time stride is 1, a context of [-1,1] is used for each stage of splicing.

Credits

• The TDNN implementation is based on: https://github.com/jonasvdd/TDNN and https://github.com/m-wiesner/nnet_pytorch.
• Semi-orthogonal convolutions used in TDNN-F are based on: https://github.com/cvqluu/Factorized-TDNN.
• Thanks to Matthew Wiesner for helpful discussions about the implementations.

This repository aims to wrap up these implementations in easy-installable PyPi packages, which can be used directly in PyTorch based neural network training.

Issues

If you find any bugs in the code, please raise an Issue, or email me at r.desh26@gmail.com.