|
from typing import Dict, List, Any |
|
import torch |
|
from diffusers import StableDiffusionXLImg2ImgPipeline, DiffusionPipeline, AutoencoderKL |
|
from PIL import Image |
|
import base64 |
|
from io import BytesIO |
|
from diffusers.image_processor import VaeImageProcessor |
|
|
|
|
|
|
|
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu') |
|
|
|
if device.type != 'cuda': |
|
raise ValueError("need to run on GPU") |
|
|
|
class EndpointHandler(): |
|
def __init__(self, path=""): |
|
|
|
self.smooth_pipe = StableDiffusionXLImg2ImgPipeline.from_pretrained( |
|
"stabilityai/stable-diffusion-xl-refiner-1.0", torch_dtype=torch.float16 |
|
) |
|
self.smooth_pipe.to("cuda") |
|
|
|
|
|
|
|
def __call__(self, data: Any) -> List[List[Dict[str, float]]]: |
|
""" |
|
:param data: A dictionary contains `inputs` and optional `image` field. |
|
:return: A dictionary with `image` field contains image in base64. |
|
""" |
|
encoded_image = data.pop("image", None) |
|
|
|
prompt = data.pop("prompt", "") |
|
|
|
if encoded_image is not None: |
|
image = self.decode_base64_image(encoded_image) |
|
|
|
image_processor = VaeImageProcessor(); |
|
latents = image_processor.preprocess(image) |
|
latents = latents.to(device="cuda") |
|
|
|
vae = AutoencoderKL.from_pretrained("stabilityai/stable-diffusion-xl-base-1.0", |
|
subfolder="vae", use_safetensors=True, |
|
).to("cuda") |
|
|
|
with torch.no_grad(): |
|
latents_dist = vae.encode(latents).latent_dist.sample() * vae.config.scaling_factor |
|
|
|
self.smooth_pipe.enable_xformers_memory_efficient_attention() |
|
out = self.smooth_pipe(prompt, image=latents_dist).images[0] |
|
|
|
return out |
|
|
|
|
|
def decode_base64_image(self, image_string): |
|
base64_image = base64.b64decode(image_string) |
|
buffer = BytesIO(base64_image) |
|
image = Image.open(buffer) |
|
return image |
|
|
