GitLab

trunk: finish example script for nnet1->nnet2 model conversion, see demo in egs/rm/s5/local/run_dnn.sh

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4257 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

trunk: checking in some partial work towards nnet1->nnet2 conversion (still compiles but the program doesn't run.).

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4246 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

trunk: committing changes to support 'perturbed' training (see train-nnet-perturbed.h).  This seems to only be subsantially helpful for small datasets like RM, and I may remove it in future.  Also adding a script that could be helpful for data cleanup, find_bad_utts.sh; this detects utterances where the transcript might not match with the audio.  Plus misc small fixes.

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4241 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

trunk: changes to improve efficiency of new preconditioning method for nnet training; bug-fix in QR algorithm (for symmetric eigenvalue problem)

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4080 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

trunk: modifications to recently-added new preconditioning method (to specify rank of input and output preconditioners separately)

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4079 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

trunk: changes to Dan's neural net setup, with new preconditioning method (speed roughly doubled if you use train_pnorm_online.sh, which uses the new preconditioning method).  Various bug-fixes, optimizations and cleanups in matrix code, cuda-matrix code and thread code.  Still tuning this so recipes not checked in yet.

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4077 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@4074 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@3974 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

trunk: make vector-sum and matrix-sum more general; add matrix-sum-rows; minor and cosmetic code changes; add code (but not yet script) to adjust nnet priors based on computed posteriors

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@3788 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@3678 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@3404 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@3337 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

trunk: change to NormalizeComponent that improves results of pnorm recipe (does not hurt already-trained models)

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@3314 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/sandbox/dan2@3260 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/sandbox/dan2@3154 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/sandbox/dan2@3148 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/sandbox/dan2@3143 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/sandbox/dan2@3142 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

sandbox/dan2: merge various changes from trunk, plus some changes to enable efficient GPU-based training after 'mix-up'.

git-svn-id: https://svn.code.sf.net/p/kaldi/code/sandbox/dan2@3108 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

sandbox/dan2: cudamatrix changes including bug fixes and moving AssertEqual, ApproxEqual to headers; add different form of preconditioned update (doesn't work great, may remove).

git-svn-id: https://svn.code.sf.net/p/kaldi/code/sandbox/dan2@3008 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

sandbox/dan2: various cudamatrix optimizations and fixes.  neural net training job now running with cudamatrix (training script not yet fully tested).

git-svn-id: https://svn.code.sf.net/p/kaldi/code/sandbox/dan2@2931 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

sandbox/dan2: got my neural net code switched over to CUDA (it compiles); small extensions  to CUDA library to enable this.

git-svn-id: https://svn.code.sf.net/p/kaldi/code/sandbox/dan2@2893 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/sandbox/dan2@2844 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/sandbox/dan2@2843 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

sandbox/dan2: committing a bunch of changes to neural-network training setup which are not yet ready to merge to trunk.

git-svn-id: https://svn.code.sf.net/p/kaldi/code/sandbox/dan2@2777 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/sandbox/sharedlibs@2661 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@2659 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

sandbox/tanel: Created interface OptionsItf, implemented by ParseOptions and SimpleOptions. SimpleOptions is a getter/setter based options holder that can provide info about registered options, useful in apps where options are not provided directly over command line but throughthe API

git-svn-id: https://svn.code.sf.net/p/kaldi/code/sandbox/tanel@2636 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@2554 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

Committing a lot of minor changes to Ondra's sandbox (will soon merge to trunk, I hope): fix issues with creating dynamic libraries by reorganizing some code and moving some files; fix a lot of minor issues such as wrong usage messages and include guards; adding Ondra's name to copyright on files he modified.

git-svn-id: https://svn.code.sf.net/p/kaldi/code/sandbox/oplatek@2533 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

Committing changes to support momentum and the --max-change option in my neural net training scripts.  Will ask Thang to integrate into the recipe (need to tune)

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@2508 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

Committing various changes: changes to pitch-interpolation code; a few extra things in neural net setup;some mostly minor script changes and extensions.

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@2478 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

Code changes: adding some pitch-processing code (depends on SAcC tool, see BABEL scripts), and new nnet nonlinearities.

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@2419 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

A lot of changes: script changes RE neural nets (more efficient IO, slightly better WERs); various new functionality for nnets and improving some feature-related binaries' interfaces.

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@1976 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/trunk@1803 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/sandbox/dan@1796 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

A bunch of code and script changes; will next revert some changes with the l2 in AffineComponentPreconditioned.

git-svn-id: https://svn.code.sf.net/p/kaldi/code/sandbox/dan@1780 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/sandbox/dan@1737 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/sandbox/dan@1734 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8

git-svn-id: https://svn.code.sf.net/p/kaldi/code/sandbox/dan@1712 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8