pierreguillou

/

whisper-medium-portuguese

like 31

I'm using the generic code from Whisper-medium to generate long-form transcription with timestamps, so that I can create subtitles

prediction = pipe(audio_path, batch_size=8, return_timestamps=True)["chunks"]

But when I use this fine-tuned model, I get the following error:

1411 if return_timestamps and not hasattr(generation_config, "no_timestamps_token_id"):
1412 raise ValueError(
1413 "You are trying to return timestamps, but the generation config is not properly set. "
1414 "Make sure to initialize the generation config with the correct attributes that are needed such as "