handler.py

from diffusers import AutoPipelineForImage2Image
import torch
from typing import Dict, Any
from PIL import Image
from io import BytesIO
import base64


class EndpointHandler():

    def __init__(self, path="."):
        if torch.cuda.is_available():
            device = "cuda"
        else:
            device = "cpu"
        self._pipe = AutoPipelineForImage2Image.from_pretrained(path, torch_dtype=torch.float32) #.to(device)

    def __call__(self, data: Dict[str, Any]) -> list[Dict[str, Any]]:
        inputs = data.pop("inputs", data)

        params = {"prompt": inputs.get("prompt", ""),
                  "image": Image.open(BytesIO(base64.b64decode(inputs['image']))),
                  "strength": inputs.get("strength", 0.3),
                  "guidance_scale": inputs.get("guidance_scale", 10),
                  "height": 768,
                  "width": 768}

        return self._pipe(**params).images[0]