app.py

import spaces
import os
# import shlex
# import subprocess

# if os.getenv('SYSTEM', "") == 'spaces' and os.getenv('USE_PRIVATE_PACKAGE', False):
#     GITHUB_TOKEN = os.getenv('GITHUB_TOKEN')
#     GITHUB_USER = os.getenv('GITHUB_USER')
#     git_repo = f"https://{GITHUB_TOKEN}@github.com/{GITHUB_USER}/VIBE.git"
#     subprocess.call(shlex.split(f'pip install git+{git_repo}'))

from functools import partial
from gradio.components import Image, Textbox
import gradio as gr

from PIL import Image as PILImage
from huggingface_hub import snapshot_download
import os
import random
import torch
import numpy as np
import pathlib

from vibe.editor import ImageEditor

MAX_SEED = np.iinfo(np.int32).max


def load_pipeline():
    HF_TOKEN = os.getenv('HF_TOKEN')
    model_path = snapshot_download(
        repo_id="iitolstykh/VIBE-Image-Edit",
        repo_type="model",
        token=HF_TOKEN,
    )

    # Load model
    editor_pipeline = ImageEditor(
        checkpoint_path=model_path,
        image_guidance_scale=1.2,
        guidance_scale=4.5,
        num_inference_steps=20,
        device="cuda",
    )

    print(f"Model loaded. Model device: {editor_pipeline.pipe.device}")

    return editor_pipeline


pipeline = load_pipeline()


def set_env(seed=0):
    torch.manual_seed(seed)
    torch.set_grad_enabled(False)


def randomize_seed_fn(seed: int, randomize_seed: bool) -> int:
    if randomize_seed:
        seed = random.randint(0, MAX_SEED)
    return seed


@spaces.GPU(duration=180)
def generate_img(
    pil_image,
    edit_prompt: str,
    sample_steps,
    scale,
    image_guidance_scale,
    seed,
    progress=gr.Progress(track_tqdm=True),
):    
    edited_image = pipeline.generate_edited_image(
        instruction=edit_prompt,
        conditioning_image=pil_image,
        num_images_per_prompt=1,
        num_inference_steps=sample_steps,
        guidance_scale=scale,
        image_guidance_scale=image_guidance_scale,
        seed=seed,
    )
    return edited_image[0]


if __name__ == "__main__":

    DESCRIPTION = f"""DEMO for VIBE-Image-Edit model: https://huggingface.co/iitolstykh/VIBE-Image-Edit"""

    image_dir = pathlib.Path('images')
    examples = [[path.as_posix(), "let this case swim in the river", 20, 4.5, 1.2, 42] for path in sorted(image_dir.glob('*.png'))]

    demo = gr.Interface(
        fn=generate_img,
        inputs=[
            gr.Image(label="Input", type="pil"),
            Textbox(label="Prompt", placeholder="Please enter your prompt. \n"),
            gr.Slider(label="Sample Steps", minimum=1, maximum=100, value=20, step=1),
            gr.Slider(
                label="Guidance Scale", minimum=0.1, maximum=30.0, value=4.5, step=0.1
            ),
            gr.Slider(
                label="Image Guidance Scale",
                minimum=0.1,
                maximum=30.0,
                value=1.2,
                step=0.1,
            ),
            gr.Slider(
                label="Seed",
                minimum=0,
                maximum=MAX_SEED,
                step=1,
                value=42,
            ),
        ],
        # outputs = [gr.Gallery(label="Result", show_label=False, type="pil")],
        outputs = [gr.Image(label="Result", show_label=False, type="pil")],
        title="",
        description=DESCRIPTION,
        examples=examples,
    )

    demo.queue(max_size=100).launch()