handler.py

from typing import Dict, Any
from transformers import AutoProcessor, Qwen2VLForConditionalGeneration
from PIL import Image
import io
import base64
import requests

class EndpointHandler():
    def __init__(self, path=""):
        self.processor = AutoProcessor.from_pretrained(path)
        self.model = Qwen2VLForConditionalGeneration.from_pretrained(path)

    def __call__(self, data: Any) -> Dict[str, Any]:

        image_input = data.get('image', None)
        text_input = data.get('text', None)
        

        if isinstance(data, dict):
            if image_input.startswith('http'):
                image = Image.open(requests.get(image_input, stream=True).raw).convert('RGB')
            else:
                image_data = base64.b64decode(image_input)
                image = Image.open(io.BytesIO(image_data)).convert('RGB')
        else:
            return {"error": "Invalid input data. Expected binary image data or a dictionary with 'image' key."}

        messages = [
            {
                "role": "user",
                "content": [
                    {"type": "image", "image": image},
                    {"type": "text", "text": text_input},
                ],
            }
        ]

        text = self.processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
        inputs = self.processor(
            text=[text],
            images=[image],
            padding=True,
            return_tensors="pt",
        )

        generate_ids = self.model.generate(inputs.input_ids, max_length=30)
        output_text = self.processor.batch_decode(
            generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False
        )[0]

        return {"generated_text": output_text}