Thanks to Nickolay's help on this topic I can now get all the way through to the validation step.
Unfortunately, the decode step doesn't even really get started. This time training with 11 hours worth of audio data.
I've checked my test.transcription vs the generated mi_test.transcription395 file and they are both the same length.
Is this just a case of the model not being accurate enough to generate a Word Error Rate? It's just that the same data and configruation when used with SphinxTrain on my Mac was able to run to completion just fine?
Hey all,
Thanks to Nickolay's help on this topic I can now get all the way through to the validation step.
Unfortunately, the decode step doesn't even really get started. This time training with 11 hours worth of audio data.
I've checked my test.transcription vs the generated mi_test.transcription395 file and they are both the same length.
Is this just a case of the model not being accurate enough to generate a Word Error Rate? It's just that the same data and configruation when used with SphinxTrain on my Mac was able to run to completion just fine?
Thanks in advance for any help!
Once again all the details including log files, etc directory and Dockerfile def'n etc are here
https://www.dropbox.com/sh/njy6wqaiz3tsena/AAAaJJ_p5a-zvsicoiMRGt80a?dl=0
You need to install pocketsphinx in docker and make sure it works from command line.