app.py

import gradio as gr
import torch
import random
import numpy as np
from PIL import Image
from diffusers import StableDiffusionInstructPix2PixPipeline
import spaces

# ==============================
# Device (CPU ONLY)
# ==============================
device = "cpu"
dtype = torch.float32

print("Loading InstructPix2Pix pipeline...")

pipe = StableDiffusionInstructPix2PixPipeline.from_pretrained(
    "timbrooks/instruct-pix2pix",
    torch_dtype=dtype,
    safety_checker=None
).to(device)

# CPU optimizations
pipe.enable_attention_slicing()

print("Model loaded successfully.")

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


# ==============================
# Inference Function
# ==============================
@spaces.GPU()  # Safe even on CPU Basic
def infer(
    image,
    prompt,
    seed=0,
    randomize_seed=True,
    guidance_scale=7.5,
    num_inference_steps=20,
):
    if image is None:
        return None, seed

    if randomize_seed:
        seed = random.randint(0, MAX_SEED)

    generator = torch.Generator(device=device).manual_seed(seed)

    image = image.convert("RGB").resize((512, 512))

    result = pipe(
        prompt=prompt,
        image=image,
        guidance_scale=guidance_scale,
        num_inference_steps=min(num_inference_steps, 30),
        generator=generator,
    ).images[0]

    return result, seed


# ==============================
# UI
# ==============================
with gr.Blocks() as demo:
    gr.Markdown("# 🖼️ Image Edit (CPU Version)")
    gr.Markdown("Stable Diffusion InstructPix2Pix – works on 16GB CPU Basic")

    with gr.Row():
        input_image = gr.Image(type="pil", label="Input Image")
        output_image = gr.Image(type="pil", label="Edited Image")

    prompt = gr.Textbox(
        label="Edit Instruction",
        placeholder="e.g. make the sky pink"
    )

    with gr.Row():
        seed = gr.Slider(0, MAX_SEED, value=0, step=1, label="Seed")
        randomize_seed = gr.Checkbox(value=True, label="Randomize Seed")

    with gr.Row():
        guidance_scale = gr.Slider(1.0, 15.0, value=7.5, step=0.5, label="Guidance Scale")
        num_inference_steps = gr.Slider(1, 40, value=20, step=1, label="Steps")

    run_button = gr.Button("Edit Image")

    run_button.click(
        fn=infer,
        inputs=[
            input_image,
            prompt,
            seed,
            randomize_seed,
            guidance_scale,
            num_inference_steps
        ],
        outputs=[output_image, seed],
    )

if __name__ == "__main__":
    demo.launch()