GitLab

trunk: changing the lattice-align-words algorithm from breadth first to depth first, so that it can output partial results when it hits the maximum required states. This will likely make lattice-align-words slower. Also changed the default --max-expend value in steps/make_index.sh from 30 to 180. This will also make the indexing a bit slower, but will give better results since it reduces the number of failed utterances.

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4537 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4536 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4534 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4533 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

trunk: various small unrelated minor improvements: include copying tree to online-nnet2 directory; remove outdated assert

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4532 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4531 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4530 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

- nnet-loss.cc - simplicfications of old code
- lattice-to-ctm-conf.cc - print average confidence per-sentence,
- steps/nnet/train.sh - use /tmp for local caching of features, or override by copy_feats_rootdir,



git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4528 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4525 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

trunk: changes to babel scripts: 1. using train_pnorm_fast.sh instead of train_pnorm.sh; 2. removing unused script for extended lexicon. Both tested.

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4524 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4522 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4521 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4520 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4519 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

Includes fixes for
- compiler errors due to unqualified type names,
- compiler warnings related to const correctness and uninitialized variables,
- and a few other bugs.

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4518 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4517 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4516 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4515 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4514 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4513 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4512 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4510 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4509 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

trunk: merging sandbox/dan back to trunk.  Includes addition of recipe for the LibriSpeech corpus, and the capability to rescore lattices using ARPA language models that are too big to convert into FSTs.

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4504 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4502 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

Prevents crash on 2nd input file


git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4497 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

trunk: Removing some inessential nnet components from nnet2(PowerExpandComponent, PiecewiseLinearComponent, AffineComponentModified, AffineComponentA, AffinePreconInputComponent, MixtureProbComponent).

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4496 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

several nnet2-online changes: make it easier to get the feature extraction options right in cross-system training; add train_pnorm_simple.sh script (simplified learning-rate schedule and improved combination at the end, supersedes train_pnorm_fast.sh); modifying big-data online-nnet2 recipes to use 40-dimensional MFCC rather than 13 as input (will add results soon, but they are improved).  Modified filter_scp.pl to have one-based, not zero-based, field index.

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4493 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4492 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4490 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4489 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4483 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

trunk: adding multithreaded and scoring options to steps/decode_basis_fmllr.sh. Also adding the thread_str suffix to an invocation of lattice-determinize-pruned in steps/decode_fmllr.sh

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4481 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4480 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

(babel/s5b) More changes towards finalizing the recipe. The scripts were developed during the eval period and not comitted until now 

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4474 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4473 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

trunk,nnet1: some cosmetic changes, adding stubs of minor experimental features (block-softmax,frame-pooling) 



git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4472 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4471 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

- updating the CNN recipe (results not better than DNN)
- removing old and unused initialization scripts



git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4470 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

(trunk/babel/s5c) Creating directory for further recipe development, so that s5b recipe can be finalized

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4465 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8