Open
Description
Hello hello, thanks for the great work!
I have successfully been running this on many audios, but today I've hit one that resulted in error when doing alignment.
I'm using docker version which packages the binary but I think it shouldn't be relevant.
Any idea what this could be about?
Thanks!
>>Performing alignment...
Traceback (most recent call last):
File "/home/1001/.local/bin/whisperx", line 8, in <module>
sys.exit(cli())
^^^^^
File "/home/1001/.local/lib/python3.11/site-packages/whisperx/transcribe.py", line 221, in cli
result: AlignedTranscriptionResult = align(
^^^^^^
File "/home/1001/.local/lib/python3.11/site-packages/whisperx/alignment.py", line 266, in align
trellis = get_trellis(emission, tokens, blank_id)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/1001/.local/lib/python3.11/site-packages/whisperx/alignment.py", line 401, in get_trellis
trellis[t, :-1] + get_wildcard_emission(emission[t], tokens[1:], blank_id),
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/1001/.local/lib/python3.11/site-packages/whisperx/alignment.py", line 426, in get_wildcard_emission
regular_scores = frame_emission[tokens.clamp(min=0)] # clamp to avoid -1 index
~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^
IndexError: tensors used as indices must be long, int, byte or bool tensors
Error executing transcription: Command '['docker', 'run', '-it', '-v', '.:/app', '-v', 'whisper_cache:/.cache', 'ghcr.io/jim60105/whisperx:no_model', '--', '--model', 'small', '--language', 'es', '--output_format', 'srt', '--compute_type', 'int8', 'audio.mp3']' returned non-zero exit status 1.
Metadata
Metadata
Assignees
Labels
No labels