Deploy a Space as inference Endpoint

_This is a fork of the naver-clova-ix/donut-base-finetuned-cord-v2 Space.

This repository implements a custom container for 🤗 Inference Endpoints using a Gradio space.

To deploy this model as an Inference Endpoint, you have to select Custom as task and a custom image.

CPU image: philschmi/gradio-api:cpu
GPU image: philschmi/gradio-api:gpu
PORT: 7860
~~Health Route: /~~-> is default

Also make sure to add server_name="0.0.0.0" in your launch() call to make sure the request is correct proxied.

If you want to use the UI with the Inference Endpoint, you have to select as endpoint type public and add auth through gradio

Example API Request Payload

Get an image you want to use, e.g.

!wget https://datasets-server.huggingface.co/assets/naver-clova-ix/cord-v2/--/naver-clova-ix--cord-v2/train/0/image/image.jpg

run inference

import requests as r
import base64

ENDPOINT_URL = ""
HF_TOKEN = ""

def predict(path_to_image: str = None):
    ext = path_to_image.split('.')[-1]
    prefix = f'data:image/{ext};base64,'
    with open(path_to_image, 'rb') as f:
        img = f.read()

    payload = {"data": [prefix + base64.b64encode(img).decode('utf-8')]}
    response = r.post(
        f"{ENDPOINT_URL}/api/predict", headers={"Authorization": f"Bearer {HF_TOKEN}"}, json=payload
    )
    if response.status_code == 200:
        return response.json()
    else:
        raise Exception(f"Error: {response.status_code}")


prediction = predict(path_to_image="image.jpg")