Create app.py

app.py

@@ -0,0 +1,247 @@
+import gradio as gr
+import numpy as np
+import random
+import torch
+import spaces
+
+from PIL import Image
+from diffsynth.pipelines.qwen_image import QwenImagePipeline, ModelConfig
+
+import os
+
+from huggingface_hub import  hf_hub_download
+
+
+
+
+pipe = QwenImagePipeline.from_pretrained(
+    torch_dtype=torch.bfloat16,
+    device="cuda",
+    model_configs=[
+        ModelConfig(model_id="Qwen/Qwen-Image-Edit-2509", origin_file_pattern="transformer/diffusion_pytorch_model*.safetensors"),
+        ModelConfig(model_id="Qwen/Qwen-Image", origin_file_pattern="text_encoder/model*.safetensors"),
+        ModelConfig(model_id="Qwen/Qwen-Image", origin_file_pattern="vae/diffusion_pytorch_model.safetensors"),
+    ],
+    tokenizer_config=None,
+    processor_config=ModelConfig(model_id="Qwen/Qwen-Image-Edit", origin_file_pattern="processor/"),
+)
+
+
+
+
+
+
+speedup = hf_hub_download(repo_id="witcherderivia/Qwen-Image-Style-Transfer", filename="diffsynth_Qwen-Image-Edit-2509-Lightning-4steps-V1.0-bf16.safetensors")
+qwenstyle= hf_hub_download(repo_id="witcherderivia/Qwen-Image-Style-Transfer", filename="diffsynth_Qwen-Image-Edit-2509-Style-Transfer-V1.safetensors")
+
+
+pipe.load_lora(pipe.dit, qwenstyle)
+pipe.load_lora(pipe.dit,speedup)
+
+
+
+dtype = torch.bfloat16
+device = "cuda" if torch.cuda.is_available() else "cpu"
+
+
+
+
+MAX_SEED = np.iinfo(np.int32).max
+
+
+@spaces.GPU
+def infer(
+    content_ref,
+    style_ref,
+    prompt,
+    seed=123,
+    randomize_seed=False,
+    true_guidance_scale=1.0,
+    num_inference_steps=4,
+    minedge=1024,
+    progress=gr.Progress(track_tqdm=True),
+    
+):
+    
+    
+
+    
+
+    content_ref=Image.fromarray(content_ref)
+    style_ref=Image.fromarray(style_ref)
+    
+    if randomize_seed:
+        seed = random.randint(0, MAX_SEED)
+
+    
+    
+    
+    
+    
+    
+    w,h=content_ref.size
+
+
+
+    #minedge=1024
+    if w>h:
+        r=w/h
+        h=minedge
+        w=int(h*r)-int(h*r)%16
+        
+    else:
+        r=h/w
+        w=minedge
+        h=int(w*r)-int(w*r)%16
+
+
+    
+    print(f"Calling pipeline with prompt: '{prompt}'")
+    
+    print(f"Seed: {seed}, Steps: {num_inference_steps}, Guidance: {true_guidance_scale}, Size: {w}x{h}")
+    
+    images = [
+        content_ref.resize((w, h)),
+        style_ref.resize((minedge, minedge)) ,
+    ]
+    
+
+    # Generate the image
+    
+    image = pipe(prompt, edit_image=images, seed=seed, num_inference_steps=num_inference_steps, height=h, width=w,edit_image_auto_resize=False,cfg_scale=true_guidance_scale)#ligtning
+
+
+    return image, seed
+
+# --- Examples and UI Layout ---
+examples = []
+
+
+
+_HEADER_ = '''
+<div style="text-align: center; max-width: 650px; margin: 0 auto;">
+    <h1 style="font-size: 2.5rem; font-weight: 700; margin-bottom: 1rem; display: contents;">QwenStyle V1</h1>
+    
+</div>
+
+
+<p style="font-size: 1rem; margin-bottom: 1.5rem;">Paper: <a href='https://openreview.net/forum?id=Cgb7JpOA5Q&referrer=%5Bthe%20profile%20of%20Shiwen%20Zhang%5D(%2Fprofile%3Fid%3D~Shiwen_Zhang1)' target='_blank'>QwenStyle: Content-Preserving Style Transfer with Qwen-Image-Edit</a> | Codes: <a href='https://github.com/witcherofresearch/Qwen-Image-Style-Transfer' target='_blank'>GitHub</a></p>
+'''  
+
+with gr.Blocks() as demo:
+
+    with gr.Column(elem_id="col-container"):
+        gr.HTML('<img src="https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-Image/qwen_image_edit_logo.png" alt="Qwen-Image Logo" width="400" style="display: block; margin: 0 auto;">')
+        gr.Markdown(_HEADER_)
+        gr.Markdown("This is a demo of QwenStyle v1, the first Content-Preserving Style Transfer Lora on Qwen-Image-Edit-2509.")
+        with gr.Row():
+            with gr.Column():
+                with gr.Row():
+                    content_ref = gr.Image(label="content ref", type="numpy", )
+                    style_ref = gr.Image(label="style ref", type="numpy", )
+                    #print(f"type(content_ref)={type(content_ref)}")
+                    
+                #input_images = gr.Gallery(label="Input Images", show_label=False, type="pil", interactive=True)
+            
+
+            result = gr.Image(label="Result", show_label=True, type="pil")
+            #result = gr.Gallery(label="Result", show_label=True, type="pil")
+        with gr.Row():
+            prompt = gr.Text(
+                    label="Prompt",
+                    value='Style Transfer the style of Figure 2 to Figure 1, and keep the content and characteristics of Figure 1.',
+                    show_label=True,
+                    placeholder='Style Transfer the style of Figure 2 to Figure 1, and keep the content and characteristics of Figure 1.',
+                    container=True,
+            )
+            run_button = gr.Button("Edit!", variant="primary")
+
+        with gr.Accordion("Advanced Settings", open=True):
+            # Negative prompt UI element is removed here
+
+            seed = gr.Slider(
+                label="Seed",
+                minimum=0,
+                maximum=MAX_SEED,
+                step=1,
+                value=123,
+            )
+
+            randomize_seed = gr.Checkbox(label="Randomize seed", value=False)
+
+            with gr.Row():
+
+                true_guidance_scale = gr.Slider(
+                    label="CFG should be 1.0",
+                    minimum=0,
+                    maximum=10.0,
+                    step=0.1,
+                    value=1.0
+                )
+
+                num_inference_steps = gr.Slider(
+                    label="Number of inference steps should be 4",
+                    minimum=1,
+                    maximum=50,
+                    step=1,
+                    value=4,
+                )
+                
+                minedge = gr.Slider(
+                    label="Min Edge of the generated image",
+                    minimum=256,
+                    maximum=2048,
+                    step=8,
+                    value=1024,
+                )
+        with gr.Row(), gr.Column():
+            gr.Markdown("## Examples")
+            gr.Markdown("changing the minedge could lead to different style similarity.")
+            default_prompt='Style Transfer the style of Figure 2 to Figure 1, and keep the content and characteristics of Figure 1.'
+            gr.Examples(examples=[
+                ['./qwenstyleref/pulpfiction_2.jpg','./qwenstyleref/styleref=6_style_ref.png',default_prompt],
+                ['./qwenstyleref/styleref=0_content_ref.png','./qwenstyleref/110.png',default_prompt],
+                ['./qwenstyleref/romanholiday_1.jpg','./qwenstyleref/s0099____1113_01_query_1_img_000146_1682705733350_08158389675901344.jpg.jpg',default_prompt],
+                ['./qwenstyleref/styleref=0_content_ref.png','./qwenstyleref/125.png',default_prompt],
+                ['./qwenstyleref/fallenangle.jpg','./qwenstyleref/styleref=s0038.png',default_prompt],
+                ['./qwenstyleref/styleref=0_content_ref.png','./qwenstyleref/styleref=s0572.png',default_prompt],
+                ['./qwenstyleref/startrooper1.jpg','./qwenstyleref/david-face-760x985.jpg','Style Transfer Figure  1 into marble material.'],
+                ['./qwenstyleref/possession.png','./qwenstyleref/s0026____0907_01_query_0_img_000194_1682674358294_041656249089406583.jpeg.jpg',default_prompt],
+                ['./qwenstyleref/styleref=0_content_ref.png','./qwenstyleref/Jotarokujo.webp',default_prompt],
+                ['./qwenstyleref/wallstreet1.jpg','./qwenstyleref/034.png',default_prompt],
+                ['./qwenstyleref/bird.jpeg','./qwenstyleref/styleref=s0539.png',default_prompt],
+                
+
+                
+                ],
+                inputs=[content_ref,style_ref, prompt], 
+                outputs=[result, seed], 
+                fn=infer, 
+                cache_examples=False
+                )        
+                
+                
+                
+                
+
+        # gr.Examples(examples=examples, inputs=[prompt], outputs=[result, seed], fn=infer, cache_examples=False)
+
+    gr.on(
+        triggers=[run_button.click],
+        fn=infer,
+        inputs=[
+            content_ref,
+            style_ref,
+            prompt,
+            seed,
+            randomize_seed,
+            true_guidance_scale,
+            num_inference_steps,
+            minedge,
+            
+        ],
+        outputs=[result, seed],
+    )
+
+if __name__ == "__main__":
+    demo.launch(server_name='0.0.0.0')