Download handler.py from unitminer/urlslab_inference: direct link, hf CLI and curl.
- Browser
- Download file 2.54 kB
-
https://huggingface.co/unitminer/urlslab_inference/resolve/main/handler.py
- Command line
-
hf download hf://unitminer/urlslab_inference/handler.py
-
curl -L -o handler.py https://huggingface.co/unitminer/urlslab_inference/resolve/main/handler.py
2.54 kB
| import os | |
| from typing import Any | |
| from sentence_transformers import SentenceTransformer | |
| import pytube | |
| from whisper_jax import FlaxWhisperPipline | |
| import jax.numpy as jnp | |
| import mimetypes | |
| class EndpointHandler(): | |
| def __init__(self, path=""): | |
| self.model = SentenceTransformer('sentence-transformers/paraphrase-multilingual-mpnet-base-v2') | |
| self.pipeline = FlaxWhisperPipline("openai/whisper-tiny", dtype=jnp.float16, batch_size=16) | |
| def __call__(self, data: Any) -> Any: | |
| """ | |
| Args: | |
| data (:obj:): | |
| includes the input data and the parameters for the inference. | |
| Return: | |
| A :obj:`list`:. The list contains the embeddings of the inference inputs | |
| """ | |
| if data.get("task") == "transcribe": | |
| yt_id = data.get("inputs") | |
| if yt_id is None: | |
| return {"error": "inputs is required, should contain youtube id or youtube video url"} | |
| yt_url = 'https://youtu.be/' + yt_id | |
| filename = yt_id + ".m4a" | |
| # first try to load captions from YT | |
| yt = pytube.YouTube(yt_url) | |
| captions = yt.captions | |
| if captions: | |
| for c in captions: | |
| caption = yt.captions.get_by_language_code(c.code) | |
| return {"transcript": caption.generate_srt_captions()} | |
| if not os.path.isfile(filename): | |
| try: | |
| stream = yt.streams.filter(only_audio=True)[0] | |
| extension = mimetypes.guess_extension(stream.mime_type) | |
| if extension: | |
| filename = yt_id + extension | |
| stream.download(filename=filename) | |
| except KeyError: | |
| return {"error":"An error occurred while loading the YouTube video. Please try again."} | |
| try: | |
| if os.path.isfile(filename): | |
| text = self.pipeline(filename, task="transcribe", return_timestamps=True) | |
| else: | |
| text = "file not found" | |
| except Exception as e: | |
| return {"error": "An error occurred while transcribing the audio. Please try again."+ str(e)} | |
| finally: | |
| if os.path.isfile(filename): | |
| os.remove(filename) | |
| return { "transcript": text} | |
| else: | |
| inputs = data.get("inputs") | |
| embeddings = self.model.encode(inputs) | |
| return {"embeddings": embeddings.tolist()} |