XA9

/

Belle-faster-whisper-large-v3-zh-punct-int8

like 0

I run the Example code and get this error

Traceback (most recent call last):
  File "d:\Program\VLM\whisper\whisper_int8.py", line 56, in <module>
    transcribe_audio("output_1737024377.wav", "audio.srt")
  File "d:\Program\VLM\whisper\whisper_int8.py", line 19, in transcribe_audio
    for segment in segments:
  File "D:\Miniconda3\envs\asr\Lib\site-packages\faster_whisper\transcribe.py", line 1148, in generate_segments
    encoder_output = self.encode(segment)
                     ^^^^^^^^^^^^^^^^^^^^
  File "D:\Miniconda3\envs\asr\Lib\site-packages\faster_whisper\transcribe.py", line 1358, in encode
    return self.model.encode(features, to_cpu=to_cpu)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ValueError: Conv1D: input dtype is float32 but expected dtype int8