| # import io | |
| from typing import Dict, List, Any | |
| import requests | |
| # import librosa | |
| from transformers import ClapModel, ClapProcessor | |
| # import gc | |
| # import base64 | |
| class EndpointHandler: | |
| def __init__(self, path=""): | |
| model_name = "laion/larger_clap_general" | |
| self.model = ClapModel.from_pretrained(model_name) | |
| self.processor = ClapProcessor.from_pretrained(model_name) | |
| def __call__(self, data: Dict[str, Any]): | |
| """ | |
| data args: | |
| inputs (:obj: `str`) | |
| Return: | |
| A :obj:`list` | `dict`: will be serialized and returned | |
| """ | |
| # print(type(data)) | |
| if 'inputs' in data: | |
| query = data['inputs'] | |
| text_inputs = self.processor(text=query, return_tensors="pt") | |
| text_embed = self.model.get_text_features(**text_inputs)[0] | |
| return text_embed.detach().numpy() | |
| # return requests.get('https://api.ipify.org?format=json').text | |
| # if 'audio' in data: | |
| # # Load the audio data into librosa | |
| # audio_buffer = io.BytesIO(base64.b64decode(data['audio'])) | |
| # y, sr = librosa.load(audio_buffer, sr=48000) | |
| # inputs = self.processor(audios=y, sampling_rate=sr, return_tensors="pt") | |
| # embedding = self.model.get_audio_features(**inputs)[0] | |
| # gc.collect() | |