argument 'ids': 'float' object cannot be interpreted as an integer

[jhj0517]

  File "C:\Whisper_Project\Whisper-WebUI\venv\Lib\site-packages\transformers\models\whisper\tokenization_whisper_fast.py", line 565, in _decode_asr
    return _decode_asr(
           ^^^^^^^^^^^^
  File "C:\Whisper_Project\Whisper-WebUI\venv\Lib\site-packages\transformers\models\whisper\tokenization_whisper.py", line 985, in _decode_asr
    text = tokenizer.decode([token])
           ^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Whisper_Project\Whisper-WebUI\venv\Lib\site-packages\transformers\models\whisper\tokenization_whisper_fast.py", line 354, in decode
    text = super().decode(
           ^^^^^^^^^^^^^^^
  File "C:\Whisper_Project\Whisper-WebUI\venv\Lib\site-packages\transformers\tokenization_utils_base.py", line 4004, in decode
    return self._decode(
           ^^^^^^^^^^^^^
  File "C:\Whisper_Project\Whisper-WebUI\venv\Lib\site-packages\transformers\models\whisper\tokenization_whisper_fast.py", line 380, in _decode
    text = super()._decode(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Whisper_Project\Whisper-WebUI\venv\Lib\site-packages\transformers\tokenization_utils_fast.py", line 654, in _decode
    text = self._tokenizer.decode(token_ids, skip_special_tokens=skip_special_tokens)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
TypeError: argument 'ids': 'float' object cannot be interpreted as an integer

This only happens for a few audio files when I use insanely_fast_whisper.
This probably happens when the transcription ends without a proper 'end token'.

[kadmy]

I'm getting this error on any audio files when I use insanely_fast_whisper. What should I do to add a proper 'end token' to the transcription?

[jhj0517]

@kadmy This bug is related to transformers, somehow the token_ids is the list of float types sometimes.

Quick fix would be to follow line 654 of transformers\tokenization_utils_fast.py as the stack trace says,
and edit token_ids to cast to int:

token_ids = [int(_id) for _id in token_ids]
text = self._tokenizer.decode(token_ids, skip_special_tokens=skip_special_tokens)

I made a PR in transformers for this : huggingface/transformers#34663