handler.py

from typing import Dict, List, Any

import torch as torch
from transformers import pipeline, WhisperProcessor

from scipy.io.wavfile import read



class EndpointHandler():



    def __init__(self, path=""):
        device = 0 if torch.cuda.is_available() else "cpu"
        self.pipe = pipeline(
            task="automatic-speech-recognition",
            model="openai/whisper-large",
            chunk_length_s=30,
            device=device,
        )
        processor = WhisperProcessor.from_pretrained("openai/whisper-large")
        self.pipe.model.config.forced_decoder_ids = processor.get_decoder_prompt_ids(language="nl", task="transcribe")

    def __call__(self, data: Dict[str, Any]) -> List[Dict[str, Any]]:
        """
       data args:
            inputs (:obj: `str`)
            date (:obj: `str`)
      Return:
            A :obj:`list` | `dict`: will be serialized and returned
        """
        #print request
        print("request")
        print(data.inputs)
        # audio_data = read(io.BytesIO(data))
        # get inputs, inputs in request body is possible equal to wav or mp3 file
        inputs = data.pop("inputs", data)
        print("here comes text")
        print(self.pipe(inputs))
        text = self.pipe(inputs)["text"]
        print(text)
        return text