MagicSource

@ga642381 hi, does multilane tts performant can compatible with single lan? isn't the phone space would be very large?

@csukuangfj Which file should using? doc is very confusing. Also the bin doesn't seems have a token-filename

@csukuangfj this part I get confusing: ``` params.blank_id = sp.piece_to_id("") params.vocab_size = sp.get_piece_size() ``` HOw can I get these using token_table then?

Yes, this code comes from icefall. I get very miss around by swithcing your guys tools.... I just take a workable version nomatter where it comes from. At least, it...

@csukuangfj I am confused...... I don't need Lexicon.... I just want token..... HOw to use this: `token_table = k2.SymbolTable.from_file(args.token_file)`........

@csukuangfj ``` params.blank_id = sp[''] params.vocab_size = max(sp) + 1 logger.info(f'vocab size: {params.vocab_size}') 2/symbol_table.py", line 199, in get return self._id2sym[k] KeyError: 5539 ```

@csukuangfj I directly using token txt SymbolTable, but get above error.

@csukuangfj I error comes from ` params.vocab_size = max(sp) + 1 ` not run into model yet.... I think the only difference is the vocabsize ??

@csukuangfj I can not get your suggest code work: ``` token_table = k2.SymbolTable.from_file(args.bpe_model) vc = max(token_table) + 1 print(vc) ``` Please don't make me miss around... It should be simple...

how does model can get blank_id and vocabsize /?