GitLab

- it works, but it needs further analysis, so far it did not outperform the LSTM.



git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@5183 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

trunk: standardizing all perl and python scripts to start with a shebang /usr/bin/env perl|python, and to be executable.  Minor script change to train_mono.sh to print better info in case of error.

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@5110 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4966 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

- lstm : implemented gradient-clipping, extended gradient print to show stats of forward / backward pass.
- nnet-loss.cc : bugfix in Xent::Eval (avoiding log(0) in cross-entropy computation)
- nnet-pdf-priors.{h,cc},nnet-forward.cc : refactored, behavior stays the same
- train-transitions : updates transition probabilities in transition model (but there was no WER improvment)




git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4904 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4902 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4862 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8