Closed Jendker closed 5 years ago
I suggest to start with disabling the post processing (comment out the lines at the yaml file), and report if you see any change.
That's how easy it was, thank you! This line is every yaml configuration file.
Would be nice if someone could mention this solution in the README file :)
I struggled with this myself. It seems that the postprocessor mechanism doesn't work with Tornado 5. It works with Tornado 4.5.3.
Is it then worth downgrading to Tornado 4.5.3 and leave postprocessor running?
It depends whether you need the post-processor. Post-processor is required only when you need to rewrite recognition results somehow (it's application specific).
Thank you for your input!
I am starting with Kaldi-Gstreamer-Server and trying out some MP3 Ted talks with different models and I am getting to the point (if the recording is longer then ~1 min), that the client stops delivering any output.
Here is the output from the worker, which I am getting on the beginning
but then it gets into:
and finaly the only output is
or later
With the transition between both:
and the master output at the end in this endless loop is:
And it goes like this forever and it happens for every configuration file which I tested.
The model configuration file, which I am using in this case is the default librispeech_english_nnet2.yaml.
For the debugging purpose I am starting the elements in the following way:
Every command executed in separate terminal of course.
I tried also the Docker image to avoid this problem, but there I wasn't able to start the worker for some reason, so I am trying to concentrate on this installation.
I will be thankful for any help! If I should provide anything more to investigate this, I will be happy to do so.