Arguments for repetition_penalty and no_repeat_ngram_size about whisperx

[JuanPZuluaga]

Hi, I think something like this would work:

        model = whisperx.load_model(
            asr_options={
                "suppress_numerals": True,
                "repetition_penalty": 1.5
            },
            task="transcribe",
        )

[dgoryeo]

Thanks @JuanPZuluaga .
When I use the asr_options I get this unexpected error. Would you have any insight in that:

whisperx\utils.py", line 285, in iterate_result
    if result["language"] in LANGUAGES_WITHOUT_SPACES:
KeyError: 'language'

I have no idea how to work around that. I have supplied the language argument:

    model = whisperx.load_model("large-v2", 
                                    device, 
                                    compute_type=compute_type,
                                    asr_options={
                                        "beam_size": 10,
                                        "best_of": 10,
                                        "patience": 3,
                                        "length_penalty": 1,
                                        "repetition_penalty": 1.5,
                                        "no_repeat_ngram_size": 2,
                                        "temperatures": [0.0, 0.2],
                                        "compression_ratio_threshold": 2.4,
                                        "log_prob_threshold": -1.0,
                                        "no_speech_threshold": 0.6,
                                        "condition_on_previous_text": False,
                                        "prompt_reset_on_temperature": 0.5,
                                        "initial_prompt": None,
                                        "prefix": None,
                                        "suppress_blank": True,
                                        "suppress_tokens": [-1],
                                        "without_timestamps": True,
                                        "max_initial_timestamp": 0.0,
                                        "word_timestamps": False,
                                        "prepend_punctuations": "\"'“¿([{-",
                                        "append_punctuations": "\"'.。,，!！?？:：”)]}、",
                                        "suppress_numerals": False,
                                    },
                                    language="ja",
                                    task="translate",
                                )