Update app.py

app.py

@@ -101,7 +101,8 @@ pipe.load_lora_weights("dx8152/Qwen-Image-Edit-2509-Relight",
                        adapter_name="relight")
 
 pipe.transformer.set_attn_processor(QwenDoubleStreamAttnProcessorFA3())
-optimize_pipeline_(pipe, image=[Image.new("RGB", (1024, 1024)), Image.new("RGB", (1024, 1024))], prompt="prompt")
+# 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
 
@@ -112,6 +113,7 @@ def update_dimensions_on_upload(image):
     
     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
@@ -121,7 +123,7 @@ def update_dimensions_on_upload(image):
         aspect_ratio = original_width / original_height
         new_width = int(new_height * aspect_ratio)
         
-    # Ensure dimensions are multiples of 8
+    # Ensure dimensions are multiples of 8 for model compatibility
     new_width = (new_width // 8) * 8
     new_height = (new_height // 8) * 8
     
@@ -159,9 +161,13 @@ def infer(
     
     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,
@@ -172,12 +178,13 @@ def infer(
 
     return result, seed, gr.Button(visible=True)
 
-# Wrapper for examples
+# Wrapper for examples to handle file paths
 @spaces.GPU
-def infer_example(input_image, prompt, lora_adapter):
-    input_pil = Image.open(input_image).convert("RGB")
+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)
-    result, seed, _ = infer(input_pil, prompt, lora_adapter, 0, True, 1.0, 4, width, height)
+    # Set default values for example inference
+    result, seed, _ = infer(input_pil, prompt, lora_adapter, 0, True, 4.0, 20, width, height)
     return result, seed
 
 # --- UI Layout ---
@@ -215,11 +222,11 @@ with gr.Blocks(css=css, theme=qwen_theme) as demo:
                 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=40, step=1, value=4)
+                    guidance_scale = gr.Slider(label="Guidance Scale", minimum=1.0, maximum=10.0, step=0.1, value=4.0)
+                    steps = gr.Slider(label="Inference Steps", minimum=1, maximum=50, step=1, value=20)
                     # Hidden sliders to hold image dimensions
-                    height = gr.Slider(label="Height", minimum=256, maximum=2048, step=8, value=1024, visible=False)
-                    width = gr.Slider(label="Width", minimum=256, maximum=2048, step=8, value=1024, visible=False)
+                    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)
 
             with gr.Column():
                 output_image = gr.Image(label="Output Image", show_label=True, interactive=False, format="png", height=480)
@@ -235,7 +242,7 @@ with gr.Blocks(css=css, theme=qwen_theme) as demo:
             inputs=[input_image, prompt, lora_adapter],
             outputs=[output_image, seed],
             fn=infer_example,
-            cache_examples="lazy",
+            cache_examples=False,
             label="Examples"
         )