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.
Sure,
if you take a look into config file
"cfg/TIMIT_baselines/TIMIT_LSTM_fmllr.cfg you can see that the lstm
architecture has a flag called "lstm_bidir=True". When true, you have a
bidirectional rnn, when false a unidirectional one. Note that a similar
flag is available for the other rnn models that we have implemented such as
the GRU and light GRU models. I suggest to derive your own application
starting from these simple examples (see librispeech or timit cfg files).
Sure, if you take a look into config file "cfg/TIMIT_baselines/TIMIT_LSTM_fmllr.cfg you can see that the lstm architecture has a flag called "lstm_bidir=True". When true, you have a bidirectional rnn, when false a unidirectional one. Note that a similar flag is available for the other rnn models that we have implemented such as the GRU and light GRU models. I suggest to derive your own application starting from these simple examples (see librispeech or timit cfg files).
Mirco
On Feb 16, 2019 17:39, "narcise" notifications@github.com wrote: