Pypi installable TDNN and TDNN-F layers for PyTorch based acoustic model training

A keras layer implementation of Peddinti's paper "A time delay neural network architecture for efficient modeling of long temporal contexts". [Work in progress]

Layer-aware TDNN: Speaker Recognition Using Multi-Layer Features from Pre-Trained Models, to appear in ICAIIC 2026