search

gooofy / py-kaldi-asr

Some simple wrappers around kaldi-asr intended to make using kaldi's (online) decoders as convenient as possible.
Apache License 2.0
170 stars 56 forks source link

ATLAS issue [Mac OS X] #30

Open lylyhan opened 5 years ago

lylyhan commented 5 years ago

What is a quick way to get atlas installed on Max OSX? I cloned the repo and encountered the "can't find atlas package" error while making. I realized that apt install command doesn't work on OSX and couldn't find an equivalent package as libatlas-dev on homebrew, which leads to me following a tutorial to install ATLAS package on my MacBook. It took me a whole day(around 8 hours) to build ATLAS, yet for some reasons it still lacks the ".so" files (e.g libatlas.so.3). I couldn't find out where I did wrong during build, and really don't want to waste another whole day building/installing atlas...

If there's any quick and clean way to get the atlas package needed to make this repo, please let me know!! Thanks so much

gooofy commented 5 years ago

While I admire your effort and would welcome pull requests that enable this package to run on Mac OS X, please note that right now that it is not a platform this package supports.

lylyhan commented 5 years ago

Oh I see... thanks for telling me that. Does that mean that the entire zamia-speech repo is not supporting OSX either? I just wanted to transcribe some audio files and the pre-trained models on kaldi site was not working great. Would appreciate it if you know of some other ways to find more accurate models. I only have a MacBook..

gooofy commented 5 years ago

If you just want to transcribe audio files (i.e. not train your own models) and have kaldi running on your system you can just use the pre-trained models from zamia-speech themselves which are (of course) platform agnostic. No need to port any of the scripts or any of the python modules zamia-speech offers, just run the models in kaldi.