handler.py

import requests
from typing import Dict, Any
from PIL import Image
import torch
import base64
import io
from transformers import BlipForConditionalGeneration, BlipProcessor
import logging
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
# Configure logging
logging.basicConfig(level=logging.DEBUG)
# Configure logging
logging.basicConfig(level=logging.ERROR)
# Configure logging
logging.basicConfig(level=logging.WARNING)
class EndpointHandler():
    def __init__(self, path=""):
        self.processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-large")
        self.model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-large").to(device)
        self.model.eval()

    def __call__(self, data: Dict[str, Any]) -> Dict[str, Any]:
        logging.error(f"----------This is an error message {str(data)}")
        input_data = data.get("inputs", {})
        logging.warning(f"------input_data-- {str(input_data)}")
       # encoded_images = input_data.get("images")
        #logging.warning(f"---encoded_images----- {str(encoded_images)}")
        if not input_data:
            #logging.warning(f"---encoded_images--not provided in if block--- {str(encoded_images)}")
            return {"captions": [], "error": "No images provided"} 
        try:
            logging.warning(f"---encoded_images-- provided in try block--- {str(input_data)}")
            byteImgIO = io.BytesIO()
            
            byteImg = Image.open(input_data)
            byteImg.save(byteImgIO, "PNG")
            byteImgIO.seek(0)
            byteImg = byteImgIO.read()


            # Non test code
            dataBytesIO = io.BytesIO(byteImg)
            raw_images =[Image.open(dataBytesIO)]
            logging.warning(f"----raw_images----0--- {str(raw_images)}")
            # Check if any images were successfully decoded
            if not raw_images:
                print("No valid images found.")
            processed_inputs = [
                self.processor(image, return_tensors="pt") for image in zip(raw_images)
            ]
            processed_inputs = {
                "pixel_values": torch.cat([inp["pixel_values"] for inp in processed_inputs], dim=0).to(device),
                "max_new_tokens":40
            }
            with torch.no_grad():
                out = self.model.generate(**processed_inputs)

            captions = self.processor.batch_decode(out, skip_special_tokens=True)
            logging.warning(f"----captions---- {str(captions)}")
            print("caption is here-------",captions)
            return {"captions": captions}
        except Exception as e:
            print(f"Error during processing: {str(e)}")
            logging.error(f"Error during processing: ----------------{str(e)}")
            return {"captions": [], "error": str(e)}