Update app.py

app.py

@@ -2,43 +2,36 @@ import gradio as gr
 from diffusers import StableDiffusionImg2ImgPipeline
 import torch
 from PIL import Image
-import numpy as np
 
 # Load the model
 model_id = "nitrosocke/Ghibli-Diffusion"
 pipe = StableDiffusionImg2ImgPipeline.from_pretrained(model_id, torch_dtype=torch.float32)
 
+# Move pipeline to GPU if available
+device = "cuda" if torch.cuda.is_available() else "cpu"
+pipe = pipe.to(device)
+
 # Define the inference function
 def ghibli_transform(input_image, prompt="ghibli style", strength=0.75, guidance_scale=7.5):
-    print(f"Input received: {input_image is not None}")
-    print(f"Input type: {type(input_image)}")
     if input_image is None:
         raise gr.Error("No image uploaded! Please upload an image before clicking Transform.")
     
-    # Process the input image
+    # Process the input image (keep it as PIL Image)
     try:
-        print(f"Input size: {input_image.size}, mode: {input_image.mode}")
-        init_image = input_image.resize((768, 768)).convert("RGB")
-        print(f"Converted init_image size: {init_image.size}, mode: {init_image.mode}")
-    except AttributeError as e:
-        raise gr.Error(f"Input is not a valid image: {str(e)}")
+        # Ensure the image is in RGB and resized
+        init_image = input_image.convert("RGB").resize((768, 768))
     except Exception as e:
         raise gr.Error(f"Failed to process image: {str(e)}")
     
-    # Convert to NumPy array as a workaround
-    init_image_np = np.array(init_image)
-    print(f"Converted to NumPy: {type(init_image_np)}, shape: {init_image_np.shape}")
-    
     # Generate the Ghibli-style image
     try:
         output = pipe(
             prompt=prompt,
-            init_image=init_image_np,  # Pass NumPy array instead
+            image=init_image,  # Pass PIL Image directly
             strength=strength,
             guidance_scale=guidance_scale,
             num_inference_steps=50
         ).images[0]
-        print("Pipeline executed successfully")
     except Exception as e:
         raise gr.Error(f"Pipeline error: {str(e)}")
     
@@ -66,5 +59,5 @@ with gr.Blocks(title="Ghibli Diffusion Image Transformer") as demo:
         outputs=output_img
     )
 
-# Launch the Space
-demo.launch()
+# Launch the Space with share=True for public link
+demo.launch(share=True)