all,

while working on my next iteration of the german voxforge model I ran into this problem when running sphinxtrain:

This step had 54 ERROR messages and 0 WARNING messages.  Please check the log file for details.
lda.py failed to create LDA transform with status 0

complete logs here:

http://goofy.zamia.org/voxforge/de/sphinxtrain_fail_log/logs/

there are basically two things I changed since the last time I built the model:

- added a lot of new submissions, many from them pretty noisy

- I am no longer restricting my dict to the words used in the transcripts but use the whole dict (all ~30k words I have pronounciations for in my DB, not just the ~15k words that actually occur in the submissions)

anybody seen this error (lda.py failing) before? there are quite a few submissions where alignment fails (that accounts for the 54 errors reported) - do I need to reduce that number to 0 first (maybe by skipping those submissions) ? or is this a bug? :)

sorry for those maybe silly questions, I am still a newbie in the field.

thanks,

   guenter

--- (Edited on 4/2/2014 5:40 am [GMT-0500] by guenter) ---