We now save the pronunciation dictionary if the user edits it, but if user inputs phoneme symbols that aren't covered by the L2S map, we will get an error when running training.
ERROR: phone "R" is not in {, non}silence.txt (line 3)
To fix this, we could parse the save_lexicon payload to get all the phone symbols, and write them to a new file. Then use that to build the nonsilence.txt file in wrappers/objects/model.py > prepare_for_training rather than using the l2s file.
We now save the pronunciation dictionary if the user edits it, but if user inputs phoneme symbols that aren't covered by the L2S map, we will get an error when running training.
To fix this, we could parse the save_lexicon payload to get all the phone symbols, and write them to a new file. Then use that to build the nonsilence.txt file in
wrappers/objects/model.py > prepare_for_trainingrather than using the l2s file.