def normalize_payload_audio()

in api_inference_community/validation.py [0:0]

• 10 lines of code
• 7 McCabe index (conditional complexity)

def normalize_payload_audio(bpayload: bytes, sampling_rate: int) -> Tuple[Any, Dict]:
    if os.path.isfile(bpayload) and bpayload.startswith(DATA_PREFIX.encode("utf-8")):
        # XXX:
        # This is necessary for batch jobs where the datasets can contain
        # filenames instead of the raw data.
        # We attempt to sanitize this roughly, by checking it lives on the data
        # path (hardcoded in the deployment and in all the dockerfiles)
        # We also attempt to prevent opening files that are not obviously
        # audio files, to prevent opening stuff like model weights.
        filename, ext = os.path.splitext(bpayload)
        if ext.decode("utf-8")[1:] in AUDIO:
            with open(bpayload, "rb") as f:
                bpayload = f.read()
    inputs = ffmpeg_read(bpayload, sampling_rate)
    if len(inputs.shape) > 1:
        # ogg can take dual channel input -> take only first input channel in this case
        inputs = inputs[:, 0]
    return inputs, {}