app.py

import os
import gradio as gr
import numpy as np
import spaces
import torch
import random
from PIL import Image
from typing import Iterable

# --- Gradio Theme ---
from gradio.themes import Soft
from gradio.themes.utils import colors, fonts, sizes

colors.blue_ish = colors.Color(
    name="blue_ish",
    c50="#F0F5FF",
    c100="#E0EBFF",
    c200="#C2D7FF",
    c300="#A3C2FF",
    c400="#85AFFF",
    c500="#4A8DFF",
    c600="#3374E6",
    c700="#1A5CCC",
    c800="#0043B3",
    c900="#002B80",
    c950="#00144D",
)

class QwenTheme(Soft):
    def __init__(
        self,
        *,
        primary_hue: colors.Color | str = colors.gray,
        secondary_hue: colors.Color | str = colors.blue_ish,
        neutral_hue: colors.Color | str = colors.slate,
        text_size: sizes.Size | str = sizes.text_lg,
        font: fonts.Font | str | Iterable[fonts.Font | str] = (
            fonts.GoogleFont("Outfit"), "Arial", "sans-serif",
        ),
        font_mono: fonts.Font | str | Iterable[fonts.Font | str] = (
            fonts.GoogleFont("IBM Plex Mono"), "ui-monospace", "monospace",
        ),
    ):
        super().__init__(
            primary_hue=primary_hue,
            secondary_hue=secondary_hue,
            neutral_hue=neutral_hue,
            text_size=text_size,
            font=font,
            font_mono=font_mono,
        )
        super().set(
            body_background_fill="linear-gradient(135deg, *primary_100, *primary_50)",
            body_background_fill_dark="linear-gradient(135deg, *primary_900, *primary_800)",
            button_primary_background_fill="linear-gradient(90deg, *secondary_500, *secondary_600)",
            button_primary_background_fill_hover="linear-gradient(90deg, *secondary_600, *secondary_700)",
            button_primary_text_color="white",
            slider_color="*secondary_500",
            slider_color_dark="*secondary_600",
            block_title_text_weight="600",
            block_border_width="2px",
            block_shadow="*shadow_drop_lg",
        )

qwen_theme = QwenTheme()


# --- Model Loading ---
from diffusers import FlowMatchEulerDiscreteScheduler
from optimization import optimize_pipeline_
from qwenimage.pipeline_qwenimage_edit_plus import QwenImageEditPlusPipeline
from qwenimage.transformer_qwenimage import QwenImageTransformer2DModel
from qwenimage.qwen_fa3_processor import QwenDoubleStreamAttnProcessorFA3

dtype = torch.bfloat16
device = "cuda" if torch.cuda.is_available() else "cpu"

pipe = QwenImageEditPlusPipeline.from_pretrained(
    "Qwen/Qwen-Image-Edit-2509",
    transformer=QwenImageTransformer2DModel.from_pretrained(
        "linoyts/Qwen-Image-Edit-Rapid-AIO",
        subfolder='transformer',
        torch_dtype=dtype,
        device_map='cuda'
    ),
    torch_dtype=dtype
).to(device)

# Load all LoRA adapters
pipe.load_lora_weights("autoweeb/Qwen-Image-Edit-2509-Photo-to-Anime",
                       weight_name="Qwen-Image-Edit-2509-Photo-to-Anime_000001000.safetensors",
                       adapter_name="anime")
pipe.load_lora_weights("dx8152/Qwen-Edit-2509-Multiple-angles",
                       weight_name="镜头转换.safetensors",
                       adapter_name="multiple-angles")
pipe.load_lora_weights("dx8152/Qwen-Image-Edit-2509-Light_restoration",
                       weight_name="移除光影.safetensors",
                       adapter_name="light-restoration")
pipe.load_lora_weights("dx8152/Qwen-Image-Edit-2509-Relight",
                       weight_name="Qwen-Edit-Relight.safetensors",
                       adapter_name="relight")

pipe.transformer.set_attn_processor(QwenDoubleStreamAttnProcessorFA3())
# It's recommended to run optimization after loading all weights
# optimize_pipeline_(pipe, image=[Image.new("RGB", (1024, 1024)), Image.new("RGB", (1024, 1024))], prompt="prompt")

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

# --- Helper Functions ---
def update_dimensions_on_upload(image):
    if image is None:
        return 1024, 1024
    
    original_width, original_height = image.size
    
    # Cap max dimension to 1024 while preserving aspect ratio
    if original_width > original_height:
        new_width = 1024
        aspect_ratio = original_height / original_width
        new_height = int(new_width * aspect_ratio)
    else:
        new_height = 1024
        aspect_ratio = original_width / original_height
        new_width = int(new_height * aspect_ratio)
        
    # Ensure dimensions are multiples of 8 for model compatibility
    new_width = (new_width // 8) * 8
    new_height = (new_height // 8) * 8
    
    return new_width, new_height

# --- Main Inference Function ---
@spaces.GPU
def infer(
    input_image,
    prompt,
    lora_adapter,
    seed,
    randomize_seed,
    guidance_scale,
    steps,
    width,
    height,
    progress=gr.Progress(track_tqdm=True)
):
    if input_image is None:
        raise gr.Error("Please upload an image to edit.")

    # Dynamically set the adapter
    if lora_adapter == "Photo-to-Anime":
        pipe.set_adapters(["anime"], adapter_weights=[1.0])
    elif lora_adapter == "Multiple-Angles":
        pipe.set_adapters(["multiple-angles"], adapter_weights=[1.0])
    elif lora_adapter == "Light-Restoration":
        pipe.set_adapters(["light-restoration"], adapter_weights=[1.0])
    elif lora_adapter == "Relight":
        pipe.set_adapters(["relight"], adapter_weights=[1.0])
        
    if randomize_seed:
        seed = random.randint(0, MAX_SEED)
    
    generator = torch.Generator(device=device).manual_seed(seed)
    
    # *** FIX: Added a negative prompt to enable classifier-free guidance ***
    negative_prompt = "worst quality, low quality, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, jpeg artifacts, signature, watermark, username, blurry"
    
    result = pipe(
        image=input_image.convert("RGB"),
        prompt=prompt,
        negative_prompt=negative_prompt, # This line enables CFG
        height=height,
        width=width,
        num_inference_steps=steps,
        generator=generator,
        true_cfg_scale=guidance_scale,
        num_images_per_prompt=1,
    ).images[0]

    return result, seed, gr.Button(visible=True)

# Wrapper for examples to handle file paths
@spaces.GPU
def infer_example(input_image_path, prompt, lora_adapter):
    input_pil = Image.open(input_image_path).convert("RGB")
    width, height = update_dimensions_on_upload(input_pil)
    # Set default values for example inference
    result, seed, _ = infer(input_pil, prompt, lora_adapter, 0, True, 1.0, 4, width, height)
    return result, seed

# --- UI Layout ---
css="""
#col-container {
    margin: 0 auto;
    max-width: 960px;
}
#main-title h1 {font-size: 2.1em !important;}
"""

with gr.Blocks(css=css, theme=qwen_theme) as demo:
    with gr.Column(elem_id="col-container"):
        gr.Markdown("# **Qwen-Image-Edit-2509-LoRAs-Fast**", elem_id="main-title")
        gr.Markdown("Perform diverse image edits using specialized LoRA adapters for the Qwen-Image-Edit model.")
        
        with gr.Row():
            with gr.Column():
                input_image = gr.Image(label="Upload Image", type="pil")
                
                lora_adapter = gr.Dropdown(
                    label="Choose Editing Style",
                    choices=["Photo-to-Anime", "Multiple-Angles", "Light-Restoration", "Relight"],
                    value="Photo-to-Anime"
                )
                
                prompt = gr.Text(
                    label="Edit Prompt",
                    show_label=True,
                    placeholder="e.g., transform into anime",
                )
                
                run_button = gr.Button("Run", variant="primary")
                


            with gr.Column():
                output_image = gr.Image(label="Output Image", interactive=False)
                #reuse_button = gr.Button("Reuse this image", visible=False)
                
                with gr.Accordion("⚙️ Advanced Settings", open=False):
                    seed = gr.Slider(label="Seed", minimum=0, maximum=MAX_SEED, step=1, value=0)
                    randomize_seed = gr.Checkbox(label="Randomize Seed", value=True)
                    guidance_scale = gr.Slider(label="Guidance Scale", minimum=1.0, maximum=10.0, step=0.1, value=1.0)
                    steps = gr.Slider(label="Inference Steps", minimum=1, maximum=50, step=1, value=4)
                    # Hidden sliders to hold image dimensions
                    height = gr.Slider(label="Height", minimum=256, maximum=1024, step=8, value=1024, visible=False)
                    width = gr.Slider(label="Width", minimum=256, maximum=1024, step=8, value=1024, visible=False)
                    
        gr.Examples(
            examples=[
                ["examples/anime_example.jpg", "transform into anime", "Photo-to-Anime"],
                ["examples/car_example.jpg", "view from the side", "Multiple-Angles"],
                ["examples/shadow_example.jpg", "Remove shadows and relight the image using soft lighting.", "Light-Restoration"],
                ["examples/relight_example.jpg", "Relight the image using soft, diffused lighting that simulates sunlight filtering through curtains.", "Relight"],
            ],
            inputs=[input_image, prompt, lora_adapter],
            outputs=[output_image, seed],
            fn=infer_example,
            cache_examples=False,
            label="Examples"
        )
            
    # --- Event Handlers ---
    run_button.click(
        fn=infer,
        inputs=[input_image, prompt, lora_adapter, seed, randomize_seed, guidance_scale, steps, width, height],
        outputs=[output_image, seed]
    )
    
    reuse_button.click(
        fn=lambda img: img,
        inputs=[output_image],
        outputs=[input_image]
    )
    
    input_image.upload(
        fn=update_dimensions_on_upload,
        inputs=[input_image],
        outputs=[width, height]
    )

demo.launch(mcp_server=True, ssr_mode=False, show_error=True)