GPU multiprocess error

[deryaguler95]

I got below error. raise ProcessRaisedException(msg, error_index, failed_process.pid) torch.multiprocessing.spawn.ProcessRaisedException: -- Process 1 terminated with the following error: raise exception KeyError: Caught KeyError in DataLoader worker process 0. Original Traceback (most recent call last): sequence = [_symbol_to_id[symbol] for symbol in cleaned_text] KeyError: '1'

O could not fix that. Are there any idea?

[daniilrobnikov]

It seems that your training dataset includes digits, "1" specifically.

• The optimal way to solve it would be to convert all numbers in dataset from digit representation into letter representation (f.e. "1" -> "one"). This is tedious, but will give the optimal inference results.

• Alternatively, you can just add numbers to text/symbols.py file. Something like that:

  
  """ from https://github.com/keithito/tacotron """

''' Defines the set of symbols used in text input to the model. ''' pad = '' _punctuation = ';:,.!?¡¿—…"«»“” ' _letters = 'ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz' _letters_ipa = "ɑɐɒæɓʙβɔɕçɗɖðʤəɘɚɛɜɝɞɟʄɡɠɢʛɦɧħɥʜɨɪʝɭɬɫɮʟɱɯɰŋɳɲɴøɵɸθœɶʘɹɺɾɻʀʁɽʂʃʈʧʉʊʋⱱʌɣɤʍχʎʏʑʐʒʔʡʕʢǀǁǂǃˈˌːˑʼʴʰʱʲʷˠˤ˞↓↑→↗↘'̩'ᵻ" _numbers = "1234567890"

Export all symbols:

symbols = [_pad] + list(_punctuation) + list(_letters) + list(_letters_ipa) + list(_numbers)

Special symbol ids

SPACE_ID = symbols.index(" ")