| from typing import Dict, List, Any | |
| from transformers import pipeline | |
| class EndpointHandler: | |
| def __init__(self, path=""): | |
| self.model = pipeline("text-to-speech", "suno/bark") | |
| 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 | |
| """ | |
| # get inputs | |
| text_prompt = data.pop("inputs", data) | |
| # run normal prediction | |
| speech_array = self.model(text_prompt,forward_params={"do_sample": True}) | |
| return speech_array | |