handler.py

from typing import Dict, List, Any
from diffusers import AutoPipelineForInpainting
from PIL import Image
from io import BytesIO
import base64
import torch


class EndpointHandler():
    def __init__(self, path=""):
        self.pipeline = AutoPipelineForInpainting.from_pretrained("OzzyGT/RealVisXL_V4.0_inpainting", torch_dtype=torch.float16, variant="fp16").to("cuda")

    def __call__(self, data: Dict[str, Any]):
        """
        data args:
            image: b64 string
            mask: b64 string
            prompt string
        returns:
            image
        """
        inputs = data.pop("inputs", data)

        # decode base64 image to PIL
        image = Image.open(BytesIO(base64.b64decode(inputs['image'])))
        mask = Image.open(BytesIO(base64.b64decode(inputs['mask'])))
        prompt = inputs['prompt']
        # blur mask for smooth transition
        mask = self.pipeline.mask_processor.blur(mask, blur_factor=20)
        
        # fix the seed
        generator = torch.Generator(device="cuda").manual_seed(0)

        image = self.pipeline(
            prompt=prompt,
            image=image,
            mask_image=mask,
            guidance_scale=8.0,
            num_inference_steps=20,  # steps between 15 and 30 work well for us (from model card)
            strength=0.99,  # make sure to use `strength` below 1.0
            generator=generator,
        ).images[0]
        
        return image