jzlianglu / pykaldi2

Yet another speech toolkit based on Kaldi and PyTorch
MIT License
173 stars 33 forks source link

clean up simulation and data pipeline #1

Closed singaxiong closed 5 years ago

singaxiong commented 5 years ago

This PR mainly clean up the simulation folder:

  1. Remove unnecessary classes and functions.
  2. Reduce the number of public interface by making many files, classes, functions, and variables as private by using the "_" prefix.
  3. Added more docstring to the public interface.
  4. Simplified the _Simulator class. It now does not accept data streams, but only accept waveforms. It is the caller's responsibility to manage the data streams, so the Simulator's job is more clearer.
  5. Revised the data pipeline classes such as SpeechDataset and DataGeneratorTrain.