Problem training acoustic model with LDA and MLLT

[Ernesto Pariona Díaz]

Hello how are you

I have problems trying to embed my acoustic model following this tutorial: https://cmusphinx.github.io/wiki/ldamllt/, install the requirements and they work, I made the changes in the file sphinx_train.cfg; at the moment of running the process I skipped several problems with the file lda.py for example: on line 39, reading the error was a syntax problem, something I could correct and so several similar problems. I am using python 3.6

raise Exception, "Please re-run bw with '-2passvar yes'"

Until I get to this problem I do not understand how to correct it, I hope you can guide me. Attached image of the problem.

Thank you

[Ernesto Pariona Díaz]

Please can you help me with that problem. Thank you

[Nickolay V. Shmyrev]

Try with python 2.7

[Ernesto Pariona Díaz]

Hello and thank you for your time

I could solve the problem using python 2.7 with virtualenv. Now I have several doubts:

• According to the tutorial at the end of doing the training when the option LDA and MLLT is activated you must have a folder called: feature_transform, but I do not see it (attached capture) and how do I use the resulting files?

• The WER and SER are the same obtained without activating the LDA and MLLT option. What is this about?

• I do not understand these warning messages that come to me during training. (Attached image)

Thank you.