RunPod•9mo ago

Connection aborted for Faster-Whisper endpoint when using "large-v2" model (Pyhton & NodeJS)

I tried to hit the Faster-Whisper endpoint using large-v2 model. I got ('Connection aborted.', RemoteDisconnected('Remote end closed connection without response')) when using Python and ConnectionClosed: The socket connection was closed unexpectedly. For more information, pass verbose: true in the second argument to fetch() when using NodeJS. With Python, I managed to get proper response from the endpoint when using the medium model. I've also tried using "large-v1" but got the connection aborted error too. With NodeJS, I got the same error with the medium, large-v1, and large-v2 model. Audio URL: https://dts.podtrac.com/redirect.mp3/api.spreaker.com/download/episode/58610513/ca7f9009_d40c_6b5a_05bb_fb5ee92fdfef.mp3 Python Code used:

def transcribe(file_path, audio, language="fr", initial_prompt=None):
    url = "https://api.runpod.ai/v2/faster-whisper/runsync"
    headers = {
        "Accept": "application/json",
        "Content-Type": "application/json",
        "Authorization": os.environ['RUNPOD_API_KEY'],
    }
    data = {
        "input": {
            "audio": audio,
            "model": "large-v2",
            "transcription": "plain_text",
            "translate": False,
            "language": language,
            "temperature": 0,
            "best_of": 5,
            "beam_size": 5,
            "patience": 1,
            "suppress_tokens": "-1",
            "condition_on_previous_text": False,
            "temperature_increment_on_fallback": 0.2,
            "compression_ratio_threshold": 2.4,
            "logprob_threshold": -1,
            "no_speech_threshold": 0.6,
            "word_timestamps": False,
            "initial_prompt": initial_prompt,
        },
        "enable_vad": False,
    }
    try:
        response = requests.post(url, json=data, headers=headers)
        response.raise_for_status()
        data = response.json()
    except requests.exceptions.RequestException as error:
        print(f"Error: {error}")

def transcribe(file_path, audio, language="fr", initial_prompt=None):
    url = "https://api.runpod.ai/v2/faster-whisper/runsync"
    headers = {
        "Accept": "application/json",
        "Content-Type": "application/json",
        "Authorization": os.environ['RUNPOD_API_KEY'],
    }
    data = {
        "input": {
            "audio": audio,
            "model": "large-v2",
            "transcription": "plain_text",
            "translate": False,
            "language": language,
            "temperature": 0,
            "best_of": 5,
            "beam_size": 5,
            "patience": 1,
            "suppress_tokens": "-1",
            "condition_on_previous_text": False,
            "temperature_increment_on_fallback": 0.2,
            "compression_ratio_threshold": 2.4,
            "logprob_threshold": -1,
            "no_speech_threshold": 0.6,
            "word_timestamps": False,
            "initial_prompt": initial_prompt,
        },
        "enable_vad": False,
    }
    try:
        response = requests.post(url, json=data, headers=headers)
        response.raise_for_status()
        data = response.json()
    except requests.exceptions.RequestException as error:
        print(f"Error: {error}")

2 Replies

Madiator2011•9mo ago

runsyc have timeout For longer jobs you might want to use /run

yiuhowardOP•9mo ago

Thank you so much @Papa Madiator , it works now.

Gaming

Programming

Connection aborted for Faster-Whisper endpoint when using "large-v2" model (Pyhton & NodeJS)