handler.py

from typing import  Dict, List, Any
import torch
from transformers.pipelines.audio_utils import ffmpeg_read
from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor, pipeline

class EndpointHandler():
    def __init__(self, path=""):
        device = "cuda:0" if torch.cuda.is_available() else "cpu"
        torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32
        model_id = "openai/whisper-large-v3"
        self.model = AutoModelForSpeechSeq2Seq.from_pretrained(
            model_id, torch_dtype=torch_dtype, low_cpu_mem_usage=True, use_safetensors=True
        )
        self.model.to(device)

        processor = AutoProcessor.from_pretrained(model_id)

        self.pipeline = pipeline(
            "automatic-speech-recognition",
            model=self.model,
            tokenizer=processor.tokenizer,
            feature_extractor=processor.feature_extractor,
            max_new_tokens=128,
            chunk_length_s=30,
            batch_size=16,
            return_timestamps=True,
            torch_dtype=torch_dtype,
            device=device,
        )


    def __call__(self, data: Dict[str, bytes]) -> Dict[str, str]:
        """
        Args:
            data (:obj:):
                includes the input data and the parameters for the inference.
        Return:
            A :obj:`list`:. The object returned should be a list of one list like [[{"label": 0.9939950108528137}]] containing :
                - "label": A string representing what the label/class is. There can be multiple labels.
                - "score": A score between 0 and 1 describing how confident the model is for this label/class.
        """
        inputs = data.pop("inputs", data)
        parameters = data.pop("parameters", None)

        # pass inputs with all kwargs in data
        if parameters is not None:
            result  = self.pipeline(inputs, return_timestamps=True, **parameters)
        else:
            result  = self.pipeline(inputs, return_timestamps=True, generate_kwargs={"task": "translate"})
        # postprocess the prediction
        return {"chunks": result["chunks"]}