transcribe output number

[ZhikangNiu]

Thanks for this meangingful repo and I have a question.
My transcription result is listed as follows and how can I change the transcription result don't output number and generation the english number? It will get higher wer

{"truth": "ninety five lines and no more thats it", "hypo": " 95 lines and no more thats it", "wer": 0.25}

{"truth": "my grandmother has type one diabetes", "hypo": " my grandmother has type 1 diabetes", "wer": 0.16666666666666666}

{"truth": "ford is approximately two hundred years old as supported by the books", "hypo": " ford is approximately 200 years old as supported by the books", "wer": 0.16666666666666666}

[nonnoxer]

def get_suppress_tokens() -> list[int]:
        """Get list of all tokens with numerics characters.

        Store this list in the `suppress_tokens` field in whisper parameters.

        Returns:
            list[int]
                List of all tokens with numeric characters.
        """
        tokenizer = Tokenizer(
            tokenizer=model.hf_tokenizer,
            task="transcribe",
            language="en",
            multilingual=True
        )
        number_tokens = [
            i 
            for i in range(tokenizer.eot)
            if all(c in "0123456789" for c in tokenizer.decode([i]).strip())
        ]
        suppress_tokens = [-1] + number_tokens
        return suppress_tokens

where model is an instance of faster_whisper.WhisperModel

Pass the suppress_tokens argument into transcription parameters