Fix download location in install_liblbfgs.sh

* Fix matrix data offset for large matrices

* Fix overflow in cudamatrix too

This continues the work started in #4809.

function is not used anyway

Fix -Wdeprecated-copy from c++11

SRILM: allow bypassing download/extraction during automated installation

Do not use ADL to invoke std::binary_search

- the audio data no longer exist in that path
- the beamformer config was missing in 'ami/s5b', it's taken from 'ami/s5'

This is to fix a CI error.

It appears that this is from using "ubuntu-latest" in the CI
workflow. It got upgraded to ubuntu 22.04 automatically, and this
doesn't have python2.7 by default.

Fix "glossaries_opt" variable name at line number 39. It's misspelled due to which words in the glossaries weren't reserved while creating BPE.

* Remove unused variable.

* cudadecoder: Make word alignment optional.

For CTC models using word pieces or graphemes, there is not enough
positional information to use the word alignment.

I tried marking every unit as "singleton" word_boundary.txt, but this
explodes the state space very, very often. See:

https://github.com/nvidia-riva/riva-asrlib-decoder/issues/3

With the "_" character in CTC models predicting word pieces, we at the
very least know which word pieces begin a word and which ones are
either in the middle of the word or the end of a word, but the
algorithm would still need to be rewritten, especially since "blank"
is not a silence phoneme (it can appear between).

I did look into using the lexicon-based word alignment. I don't have a
specific complaint about it, but I did get a weird error where it
couldn't create a final state at all in the output lattice, which
caused Connect() to output an empty lattice. This may be because I
wasn't quite sure how to handle the blank token. I treat it as its own
phoneme, bcause of limitations in TransitionInformation, but this
doesn't really make any sense.

Needless to say, while the CTM outputs of the cuda decoder will be
correct from a WER point of view, their time stamps won't be correct,
but they probably never were in the first place, for CTC models.

* Update run_blstm.sh

fix bug aspire run_blstm.sh

* Update egs/aspire/s5/local/nnet3/run_blstm.sh
Co-authored-by: NCy 'kkm' Katsnelson <kkm@pobox.com>
Co-authored-by: NCy 'kkm' Katsnelson <kkm@pobox.com>

Provide provision to pass subword separator as argument to make_position_dependent_subword_lexicon.py (#4794)

Co-authored-by: Nnpovey <you@example.com>

Coauthored-By: NJonghwan Hyeon <hyeon0145@gmail.com>

The example for the post-to-tacc fails , but with the correct of `ark:- |`  there is no piping error

* [infra] docker images automatically using gh
* minor change

* Set PYTHONUNBUFFERED=TRUE for wsj,tedlium

To force stdout, stderr to be unbuffered for python scripts.
Without this setting parts of the stream might be lost in a few cases.
Relevant mainly for python versions <3.7.

* Set PYTHONUNBUFFERED=TRUE for all recipes

Add to path.sh in all except the already updated wsj, tedlium.
To force stdout, stderr to be unbuffered for python scripts.
Relevant mainly for python versions <3.7.

* Set PYTHONUNBUFFERED=TRUE in remaining path.sh

These files were missed by mistake in earlier commits.

* Use 1 instead of TRUE for PYTHONUNBUFFERED

* Fix missing cblas and lapack external symbols for netlib

* Remaining Conda-forge changes

* Fix indent

* usleep using C+11 constructs

* make codefactor happy
Co-authored-by: NMichael McAuliffe <michael.e.mcauliffe@gmail.com>

* Github action for kaldi
* add ccache
* remove spaces

* Kaldi recipe for SPGISpeech

* adding readme to spgispeech

* fixing some style issues

* more syntax checks cleared

* more style fixes

* yaml format fix

* fix issues reported by @desh2608

* remove the conda calls

* remove more shellcheck errors

* fix one more grumpy shellcheck

fixed portaudio archive location
fixed case when the directory portaudio exists
fixed whitespace formatting

* Fix build for Windows

* Remove double cctype