Inference-endpoint-for-jina-clip-v1

Fork of jinaai/jina-clip-v1 for a multimodal-multilanguage-embedding Inference endpoint.

This repository implements a custom task for multimodal-multilanguage-embedding for 🤗 Inference Endpoints. The code for the customized handler is in the handler.py.

To use deploy this model a an Inference Endpoint you have to select Custom as task to use the handler.py file.

The repository contains a requirements.txt to download the einops, timm and pillow library.

Call to endpoint example

import json
from typing import List
import requests as r
import base64

ENDPOINT_URL = "endpoint_url"
HF_TOKEN = "token_key"

def predict(path_to_image: str = None, text : str = None):
    with open(path_to_image, "rb") as i:
        b64 = base64.b64encode(i.read())

    payload = {"inputs": 
            {
            "image": b64.decode("utf-8"),
            "text": text
            }
        }

    response = r.post(
        ENDPOINT_URL, headers={"Authorization": f"Bearer {HF_TOKEN}"}, json=payload
    )
    return response.json()


prediction = predict(
    path_to_image="image/accidentdevoiture.webp", text="An image of a cat and a remote control"
)

print(json.dumps(prediction, indent=2))

Expected result

{
  "text_embedding": [-0.009289545938372612,
    -0.03686045855283737,
    ...
    0.038627129048109055,
    -0.01346363127231597]
  "image_embedding": [-0.009289545938372612,
    -0.03686045855283737,
    ...
    0.038627129048109055,
    -0.01346363127231597]
}