Whisper Fine Tuned model cannot used on WhisperX

Models
1

Hi guys,

I couldn’t use my model on Whisperx. It gives me an error of “AttributeError: ‘WhisperForConditionalGeneration’ object has no attribute ‘hf_tokenizer’”. What should I do ?
I did my model from Fine-Tune Whisper For Multilingual ASR with 🤗 Transformers. My transformer version is 4.37.0. I’m lost could you help me ?

3 Likes
2

did you find the answer? i am also searching the answer of same question? if you find it please share with me it will help me in my project.

3

Hi AdnanBajwa,

Before you pass your model into WhisperX, you need to convert your model in CTranslate2.

ct2-transformers-converter --model openai/whisper-large-v3 --output_dir blabla
–copy_files tokenizer.json preprocessor_config.json --quantization float16

You can do this using above.

Have a nice day.

1 Like
4

Thank you for your response. If i do not training model with quantisation techniques that time i neet to convert model in CTranslate2?

5

Sorry for late reply, if you want you don’t use quantization or --copy files.