search

mravanelli / pytorch-kaldi

pytorch-kaldi is a project for developing state-of-the-art DNN/RNN hybrid speech recognition systems. The DNN part is managed by pytorch, while feature extraction, label computation, and decoding are performed with the kaldi toolkit.
2.36k stars 447 forks source link

In training quantization of weights (parameters) and activations #16

Closed razor1179 closed 5 years ago

razor1179 commented 5 years ago

Have there been any attempts at implementing in training quantization, like that in "Binarized Neural Networks: Training Deep Neural Networks with Weights and Activations Constrained to +1 or -1" using pytorch-kaldi? If not can the admin provide direction on how to proceed to obtain in-training quantization?

TParcollet commented 5 years ago

To my knowledge, we haven't tried to implement binarized NN in pytorch-Kaldi. Nonetheless, following the "How can I plug-in my model?" section of the readme, you can find everything you need to know to plug your own model based on other pytorch implementations of BNN such as https://github.com/itayhubara/BinaryNet.pytorch.

mravanelli commented 5 years ago

This model is not implemented in the current version. However, if someone implements it we can definitely add it into the toolkit!